11785_HW4_submission/HW4P2_S24_Student_Starter_V3.ipynb

{"cells":[{"cell_type":"markdown","metadata":{"id":"8XpNMS7Vk6Df"},"source":["# HW4P2: Attention-based Speech Recognition\n","\n","<img src=\"https://cdn.shopify.com/s/files/1/0272/2080/3722/products/SmileBumperSticker_5400x.jpg\" alt=\"A cute cat\" width=\"600\">\n","\n","\n","Welcome to the final assignment in 11785. In this HW, you will work on building a speech recognition system with <i>**attention**</i>. <br> <br>\n","\n","<center>\n","<img src=\"https://popmn.org/wp-content/uploads/2020/03/pay-attention.jpg\" alt=\"A cute cat\" height=\"100\">\n","</center>\n","\n","<h2>Learning Objectives:\n","\n","*  **Understand the Transformer Architecture**: Gain a comprehensive understanding of the transformer model architecture.\n","\n","*  **Develop the  Encoder and Decoder Modules**: Understand the individual components of the transformer's encoder and decoder modules, including positional encoding, multi-head attention, and feed-forward projections.\n","\n","*  **Integrate the Complete Transformer Model**:  Combine the encoder and decoder modules to construct the full transformer model, and learn how to manage the flow of information between these components.\n","\n","* **Understand the Value of Multi-Head Attention**: Learn how to use the multi-head attention mechanism and understand its significance in improving the model's ability to focus on different parts of the input sequence simultaneously.\n","\n","* **Understand and Apply Pretraining Techniques**: Grasp the concept of pretraining by employing LSTMs as an initial embedding mechanism in a transformer's encoder. Learn the strategic process of incrementally training different components of a model — starting with a simple architecture (LSTM), then progressing to more complex structures (transformers). This objective underscores the role of pretraining and its gradual integration to enhance the overall efficacy of neural network models, culminating in superior performance metrics.\n","\n","* **Enhance Training Efficiency Through Strategic Freezing**: Master the technique of selectively freezing and unfreezing model components at different stages of training. This approach aims to direct the model's learning capacity more effectively, concentrating on specific modules at a time. Understand how this method facilitates a more efficient training process, allowing various parts of the model to adapt incrementally. This objective highlights the value of phased learning, especially beneficial when integrating model components with varying learning dynamics.\n","\n","*  **Tune Model Hyperparameters**: Experiment with different configurations of model hyperparameters, such as the number of layers, model dimensionality, and the number of attention heads, to help you understand the tradeoff between model complexity and dataset size.\n","\n","<br>\n","\n","\n","[Kaggle Competition Link](https://www.kaggle.com/competitions/automatic-speech-recognition-s24)<br>\n","\n","[Kaggle Dataset Link](https://www.kaggle.com/competitions/automatic-speech-recognition-s24/data)<br>\n","\n","[Speech Transformer Paper](https://ieeexplore.ieee.org/document/8462506)<br>\n","\n","[Attention Is All You Need Paper](https://arxiv.org/pdf/1706.03762.pdf)<br>"]},{"cell_type":"markdown","metadata":{"id":"vwIdDTTmmZVe"},"source":["# Libraries"]},{"cell_type":"markdown","metadata":{"id":"y9qsVrRemgh7"},"source":["1. By now, we believe that you are already a great deep learning practitioner, Congratulations. 🎉\n","\n","2. There are a lot of resources provided in this notebook, that will help you check if you are running your implementations correctly."]},{"cell_type":"code","execution_count":null,"metadata":{"id":"8UK7J-dp5iN5","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713328301953,"user_tz":240,"elapsed":270,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"bf201d7c-8585-4d62-926e-defc5bb029a8"},"outputs":[{"output_type":"stream","name":"stdout","text":["Wed Apr 17 04:31:41 2024       \n","+---------------------------------------------------------------------------------------+\n","| NVIDIA-SMI 535.104.05             Driver Version: 535.104.05   CUDA Version: 12.2     |\n","|-----------------------------------------+----------------------+----------------------+\n","| GPU  Name                 Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |\n","| Fan  Temp   Perf          Pwr:Usage/Cap |         Memory-Usage | GPU-Util  Compute M. |\n","|                                         |                      |               MIG M. |\n","|=========================================+======================+======================|\n","|   0  Tesla T4                       Off | 00000000:00:04.0 Off |                    0 |\n","| N/A   49C    P0              29W /  70W |   1741MiB / 15360MiB |      0%      Default |\n","|                                         |                      |                  N/A |\n","+-----------------------------------------+----------------------+----------------------+\n","                                                                                         \n","+---------------------------------------------------------------------------------------+\n","| Processes:                                                                            |\n","|  GPU   GI   CI        PID   Type   Process name                            GPU Memory |\n","|        ID   ID                                                             Usage      |\n","|=======================================================================================|\n","+---------------------------------------------------------------------------------------+\n"]}],"source":["!nvidia-smi"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"nYgaLmgy5iqR"},"outputs":[],"source":["''' Installing some required libraries. '''\n","!pip install python-levenshtein torchsummaryX wandb kaggle pytorch-nlp --quiet"]},{"cell_type":"markdown","source":["##  Imports"],"metadata":{"id":"xLXQgnj7E8BM"}},{"cell_type":"code","execution_count":4,"metadata":{"id":"0mkii-6Dsjr8","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713552924432,"user_tz":240,"elapsed":5885,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"64eb3576-90d8-49b3-c2f6-8609b2c4c54a"},"outputs":[{"output_type":"stream","name":"stdout","text":["Device:  cuda\n"]}],"source":["import torch\n","import torch.nn as nn\n","import torch.nn.functional as F\n","import torchaudio.transforms as tat\n","\n","from torchsummaryX import summary\n","from torch.autograd import Variable\n","from torch.nn.utils.rnn import pad_sequence, pack_padded_sequence, pad_packed_sequence\n","\n","import gc\n","import os\n","import math\n","import yaml\n","import random\n","import zipfile\n","import datetime\n","\n","import glob\n","import wandb\n","import numpy as np\n","import pandas as pd\n","from tqdm import tqdm\n","\n","import seaborn as sns\n","import matplotlib.pyplot as plt\n","\n","''' Imports for decoding and distance calculation. '''\n","import Levenshtein\n","\n","import warnings\n","warnings.filterwarnings(\"ignore\")\n","\n","device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n","print(\"Device: \", device)"]},{"cell_type":"markdown","metadata":{"id":"AIOBPQjzrx5n"},"source":["# Configurations"]},{"cell_type":"code","source":["%%writefile config.yaml\n","\n","###### Dataset -----------------------------------------------------------------\n","train_dataset   : \"train-clean-100\"     # train-clean-50 (primarily for debugging purposes), train-clean-100\n","cepstral_norm   : True\n","input_dim       : 27\n","batch_size      : 64                    # decrease this as you modify the network architecture\n","\n","###### Encoder Parameters ------------------------------------------\n","## Universal (Part I and II)\n","enc_dropout     : 0.2                   # [0.1, 0.4]\n","## Transformer-related (Part II)\n","enc_num_layers  : 2                     # [1, 3]\n","enc_num_heads   : 2                     # [1, 4]\n","\n","###### Decoder Parameters ------------------------------------------\n","## Transformer-related (Part I and II)\n","dec_dropout     : 0.2                   # [0.1, 0.4]\n","dec_num_layers  : 2                     # [1, 3]\n","dec_num_heads   : 2                     # [1, 4]\n","\n","###### Network Parameters ------------------------------------------------------\n","d_model         : 512                   # [256, 1024]\n","d_ff            : 2048                  # [512, 4096]\n","\n","###### Learning Rate ---------------------------------------------------------------\n","learning_rate   : 1E-4                  # [1E-3, 1E-4], this will depend on the specified optimizer\n","\n","###### Optimizer ---------------------------------------------------------------\n","optimizer       : \"AdamW\"               # Adam, AdamW\n","\n","## if SGD\n","momentum        : 0.0\n","nesterov        : True\n","\n","###### Scheduler ---------------------------------------------------------------\n","scheduler       : \"ReduceLR\"     # CosineAnnealing, ReduceLR\n","\n","## if ReduceLR\n","\n","## we are validating every 2 epochs but scheduler acts on every epoch. Set patience accordingly\n","## patience less than validation frquency can mean learning rate always dropping after patience epochs\n","## specify a suitable threshold too\n","factor          : 0.9\n","patience        : 6\n","\n","###### Training Parameters -----------------------------------------------------\n","epochs          : 100\n","\n","###### Name --------------------------------------------------------------------\n","Name: \"\"                                # write your name here for study group"],"metadata":{"id":"qkzzA5l2VbXZ","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713636042137,"user_tz":240,"elapsed":157,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"86c4fe97-a99d-4a38-a667-410496c22dc5"},"execution_count":91,"outputs":[{"output_type":"stream","name":"stdout","text":["Overwriting config.yaml\n"]}]},{"cell_type":"code","source":["with open(\"config.yaml\") as file:\n","    config = yaml.safe_load(file)"],"metadata":{"id":"xcQEffYWYOVm","executionInfo":{"status":"ok","timestamp":1713636042513,"user_tz":240,"elapsed":2,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":92,"outputs":[]},{"cell_type":"code","source":["config"],"metadata":{"id":"rgkkGyEa4TMS","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713636043360,"user_tz":240,"elapsed":3,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"7ceae6b3-fcb3-4cc5-92f6-2cea03ceabd4"},"execution_count":93,"outputs":[{"output_type":"execute_result","data":{"text/plain":["{'train_dataset': 'train-clean-100',\n"," 'cepstral_norm': True,\n"," 'input_dim': 27,\n"," 'batch_size': 64,\n"," 'enc_dropout': 0.2,\n"," 'enc_num_layers': 2,\n"," 'enc_num_heads': 2,\n"," 'dec_dropout': 0.2,\n"," 'dec_num_layers': 2,\n"," 'dec_num_heads': 2,\n"," 'd_model': 512,\n"," 'd_ff': 2048,\n"," 'learning_rate': '1E-4',\n"," 'optimizer': 'AdamW',\n"," 'momentum': 0.0,\n"," 'nesterov': True,\n"," 'scheduler': 'ReduceLR',\n"," 'factor': 0.9,\n"," 'patience': 6,\n"," 'epochs': 100,\n"," 'Name': ''}"]},"metadata":{},"execution_count":93}]},{"cell_type":"markdown","metadata":{"id":"-njBvl2Opd6I"},"source":["# Kaggle Dataset Download"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"PTyWR2sIp0Ns","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1712625145680,"user_tz":240,"elapsed":1621,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"f4cf85a7-38f3-4e3c-ebe8-a0877a9a2b3a"},"outputs":[{"output_type":"stream","name":"stdout","text":["Collecting kaggle==1.5.8\n","  Using cached kaggle-1.5.8-py3-none-any.whl\n","Installing collected packages: kaggle\n","  Attempting uninstall: kaggle\n","    Found existing installation: kaggle 1.5.8\n","    Uninstalling kaggle-1.5.8:\n","      Successfully uninstalled kaggle-1.5.8\n","Successfully installed kaggle-1.5.8\n","mkdir: cannot create directory ‘/root/.kaggle’: File exists\n"]}],"source":["!pip install --upgrade --force-reinstall --no-deps kaggle==1.5.8\n","!mkdir /root/.kaggle\n","\n","with open(\"/root/.kaggle/kaggle.json\", \"w+\") as f:\n","\n","    # Enter your kaggle username <USERNAME> and key <KEY> here.\n","    f.write('{\"username\":\"<USERNAME>\",\"key\":\"<KEY>\"}')\n","\n","!chmod 600 /root/.kaggle/kaggle.json"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"F581gjfnqE2C"},"outputs":[],"source":["''' Downloading the dataset. '''\n","!kaggle competitions download -c automatic-speech-recognition-s24\n","\n","''' Unzipping quickly and quietly. '''\n","!unzip -q automatic-speech-recognition-s24.zip -d ./data"]},{"cell_type":"markdown","metadata":{"id":"zUJyBBwIqQs6"},"source":["# Character-based LibriSpeech (HW4P2)\n","\n","In terms of the dataset, the dataset structure for HW3P2 and HW4P2 dataset are very similar.\n","\n","Can you spot out the differences? What all will be required??\n","\n","Hints:\n","\n","- Check how big is the dataset? (do you require memory efficient loading techniques??)\n","- How do we load MFCCs? Do we need to normalize them?\n","- Does the data have \\<SOS> and \\<EOS> tokens in each sequences? Do we remove them or do we not remove them? (Read writeup)\n","- Would we want a collating function? Ask yourself: Why did we need a collate function last time?\n","- Observe the VOCAB -- is the dataset same as HW3P2?\n","- Should you add augmentations, if yes which augmentations? When should you add augmentations? (Check bootcamp for answer)\n"]},{"cell_type":"code","execution_count":8,"metadata":{"id":"MBMLGYX-kZcd","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713552925873,"user_tz":240,"elapsed":150,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"0af96180-b208-496c-b471-ffd9e213ef04"},"outputs":[{"output_type":"stream","name":"stdout","text":["Length of Vocabulary    : 31\n","VOCAB                   : ['<pad>', '<sos>', '<eos>', 'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z', \"'\", ' ']\n","PAD_TOKEN               : 0\n","SOS_TOKEN               : 1\n","EOS_TOKEN               : 2\n"]}],"source":["VOCAB = [\n","    \"<pad>\", \"<sos>\", \"<eos>\",\n","    \"A\",   \"B\",    \"C\",    \"D\",\n","    \"E\",   \"F\",    \"G\",    \"H\",\n","    \"I\",   \"J\",    \"K\",    \"L\",\n","    \"M\",   \"N\",    \"O\",    \"P\",\n","    \"Q\",   \"R\",    \"S\",    \"T\",\n","    \"U\",   \"V\",    \"W\",    \"X\",\n","    \"Y\",   \"Z\",    \"'\",    \" \",\n","]\n","\n","VOCAB_MAP = {VOCAB[i]:i for i in range(0, len(VOCAB))}\n","\n","PAD_TOKEN = VOCAB_MAP[\"<pad>\"]\n","SOS_TOKEN = VOCAB_MAP[\"<sos>\"]\n","EOS_TOKEN = VOCAB_MAP[\"<eos>\"]\n","\n","print(f\"Length of Vocabulary    : {len(VOCAB)}\")\n","print(f\"VOCAB                   : {VOCAB}\")\n","print(f\"PAD_TOKEN               : {PAD_TOKEN}\")\n","print(f\"SOS_TOKEN               : {SOS_TOKEN}\")\n","print(f\"EOS_TOKEN               : {EOS_TOKEN}\")"]},{"cell_type":"markdown","source":["## Dataset"],"metadata":{"id":"kWiWhqBHNXt-"}},{"cell_type":"code","source":["''' This class is very similar to HW3P2 except for targets (now sequence of characters). '''\n","\n","class SpeechDataset(torch.utils.data.Dataset):\n","    ''' memory inefficient : loading data in __init__ for simplicty\n","\n","        You may decide to load data in __getitem__ if you wish.\n","        However, doing this memory inefficiently makes __init__ function takes\n","        the load of loading the data, and shifts it away from during the training process.\n","    '''\n","\n","    def __init__(self, root_dir, partition, cepstral=True, transforms=True):\n","\n","        self.transforms = transforms\n","        self.mfcc_dir           = f\"{root_dir}/{partition}*/mfcc/*\"\n","        self.transcript_dir     = f\"{root_dir}/{partition}*/transcripts/*\"\n","\n","        self.mfcc_files         = sorted(glob.glob(self.mfcc_dir))\n","        self.transcript_files   = sorted(glob.glob(self.transcript_dir))\n","\n","        assert len(self.mfcc_files) == len(self.transcript_files)\n","\n","        self.mfcc_files         = self.mfcc_files\n","        self.transcript_files   = self.transcript_files\n","        self.length             = len(self.transcript_files)\n","        self.vocab              = VOCAB\n","\n","        print(\"Loaded Path: \", partition)\n","\n","        self.mfccs, self.transcripts_shifted, self.transcripts_golden  = [], [], []\n","        for file in self.mfcc_files:\n","            mfcc = np.load(file)\n","\n","            if cepstral:\n","                mfcc = (mfcc - np.mean(mfcc, axis=0)) / (np.std(mfcc, axis=0) + 1E-8)\n","            self.mfccs.append(mfcc)\n","\n","        for file in self.transcript_files:\n","            temp = np.load(file)[1:-1]\n","\n","            # Why do we have two different types of targets?\n","            # How do we want our decoder to know the start of sequence <SOS> and end of sequence <EOS>?\n","\n","            # @TODO: Uncomment the code below after answering the above questions.\n","            self.transcripts_shifted.append(np.array([SOS_TOKEN] + [self.vocab.index(i) for i in temp]))\n","            self.transcripts_golden.append(np.array([self.vocab.index(i) for i in temp] + [EOS_TOKEN]))\n","\n","        assert len(self.mfccs) == len(self.transcripts_shifted)\n","\n","\n","    def __len__(self): return self.length\n","\n","\n","    def __getitem__(self, ind):\n","        mfcc                = torch.FloatTensor(self.mfccs[ind])\n","        shifted_transcript  = torch.tensor(self.transcripts_shifted[ind])\n","        golden_transcript   = torch.tensor(self.transcripts_golden[ind])\n","\n","        return mfcc, shifted_transcript, golden_transcript\n","\n","\n","    def collate_fn(self, batch):\n","        # @NOTE: batch corresponds to output from __getitem__ for a minibatch\n","\n","        '''\n","        1.  Extract the features and labels from 'batch'.\n","        2.  We will additionally need to pad both features and labels,\n","            look at PyTorch's documentation for pad_sequence.\n","        3.  This is a good place to perform transforms, if you so wish.\n","            Performing them on batches will speed the process up a bit.\n","        4.  Return batch of features, labels, lengths of features, and lengths of labels.\n","\n","        '''\n","\n","        # Batch of input mfcc coefficients.\n","        batch_mfcc              = [i[0] for i in batch]\n","\n","        # Batch of output characters (shifted and golden).\n","        batch_transcript        = [i[1] for i in batch]\n","        batch_golden            = [i[2] for i in batch]\n","\n","        lengths_mfcc            = [len(i) for i in batch_mfcc]\n","        lengths_transcript      = [len(i) for i in batch_transcript]\n","\n","        # HINT: CHECK OUT -> pad_sequence (imported above)\n","        # Also be sure to check the input format (batch_first).\n","        batch_mfcc_pad          = pad_sequence(batch_mfcc, batch_first=True, padding_value=PAD_TOKEN)\n","        batch_transcript_pad    = pad_sequence(batch_transcript, batch_first=True, padding_value=PAD_TOKEN)\n","        batch_golden_pad        = pad_sequence(batch_golden, batch_first=True, padding_value=PAD_TOKEN)\n","\n","\n","        # You may apply some transformations, Time and Frequency masking, here in the collate function:\n","        # Food for Thought -> Why are we applying the transformation here and not in the __getitem__?\n","        #                  -> Would we apply transformation on the validation set as well?\n","        #                  -> Is the order of axes / dimensions as expected for the transform functions?\n","        if self.transforms == True:\n","            time_mask = tat.TimeMasking(time_mask_param=80,iid_masks=True)\n","            frequency_mask = tat.FrequencyMasking(freq_mask_param=2,iid_masks=True)\n","            batch_mfcc_pad = torch.permute(batch_mfcc_pad, (0, 2, 1))\n","            batch_mfcc_pad = time_mask(batch_mfcc_pad)\n","            batch_mfcc_pad = frequency_mask(batch_mfcc_pad)\n","            batch_mfcc_pad = torch.permute(batch_mfcc_pad, (0, 2, 1))\n","\n","        # Return the following values:\n","        # padded features, padded shifted labels, padded golden labels, actual length of features, actual length of the shifted labels\n","        return batch_mfcc_pad, batch_transcript_pad, batch_golden_pad, torch.tensor(lengths_mfcc), torch.tensor(lengths_transcript)"],"metadata":{"id":"QfYjnJ6VZCBp","executionInfo":{"status":"ok","timestamp":1713552926711,"user_tz":240,"elapsed":157,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":9,"outputs":[]},{"cell_type":"code","source":["class SpeechTestDataset(torch.utils.data.Dataset):\n","\n","    def __init__(self, root_dir, partition, cepstral=True):\n","\n","        self.mfcc_dir       = f\"{root_dir}/{partition}*/mfcc/*\"\n","        self.mfcc_files     = sorted(glob.glob(self.mfcc_dir))\n","        self.length         = len(self.mfcc_files)\n","\n","        self.mfccs = []\n","        for file in self.mfcc_files:\n","            mfcc = np.load(file)\n","            if cepstral: mfcc = (mfcc - np.mean(mfcc, axis=0)) / (np.std(mfcc, axis=0) + 1E-8)\n","            self.mfccs.append(mfcc)\n","\n","        print(\"Loaded Path: \", partition)\n","\n","\n","    def __len__(self): return self.length\n","\n","\n","    def __getitem__(self, ind):\n","        mfcc = self.mfccs[ind]\n","\n","        return torch.FloatTensor(mfcc)\n","\n","\n","    def collate_fn(self, batch):\n","        # @NOTE: batch corresponds to output from __getitem__ for a minibatch\n","\n","        '''\n","        TODO:\n","        1.  Extract the features and labels from 'batch'.\n","        2.  We will additionally need to pad the features,\n","            look at PyTorch's documentation for pad_sequence.\n","        3.  Return batch of features, lengths of features.\n","\n","        '''\n","\n","        batch_mfcc      = batch\n","        batch_mfcc_pad  = pad_sequence(batch_mfcc, batch_first=True)\n","        lengths_mfcc    = [len(i) for i in batch_mfcc]\n","\n","        return batch_mfcc_pad, torch.tensor(lengths_mfcc)"],"metadata":{"id":"SeehjDLGNJz2","executionInfo":{"status":"ok","timestamp":1713552926999,"user_tz":240,"elapsed":2,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":10,"outputs":[]},{"cell_type":"code","source":["# @NOTE: use the config file to specify PARTITION and CEPSTRAL\n","DATA_DIR        =     '/content/data/11785-HW4P2'      # Enter your data directory here.\n","PARTITION       = config['train_dataset']             # Enter your training directory here.\n","CEPSTRAL        = config['cepstral_norm']           # Enter your cepstral_norm parameter here.\n","\n","train_dataset   = SpeechDataset(\n","    root_dir    = DATA_DIR,\n","    partition   = PARTITION,\n","    cepstral    = CEPSTRAL,\n","    transforms = True\n",")\n","\n","val_dataset     = SpeechDataset(\n","    root_dir    = DATA_DIR,\n","    partition   = \"dev-clean\",\n","    cepstral    = CEPSTRAL,\n","    transforms = False\n",")\n","\n","test_dataset    = SpeechTestDataset(\n","    root_dir    = DATA_DIR,\n","    partition   = \"test-clean\",\n","    cepstral    = CEPSTRAL,\n",")\n","\n","gc.collect()"],"metadata":{"id":"GTXbTFReehR3","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553030600,"user_tz":240,"elapsed":103003,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"5fbc7728-8b1e-4bf8-e9ed-68d1603247e8"},"execution_count":11,"outputs":[{"output_type":"stream","name":"stdout","text":["Loaded Path:  train-clean-100\n","Loaded Path:  dev-clean\n","Loaded Path:  test-clean\n"]},{"output_type":"execute_result","data":{"text/plain":["735"]},"metadata":{},"execution_count":11}]},{"cell_type":"markdown","source":["## DataLoader"],"metadata":{"id":"vQRQ_nQVNaFx"}},{"cell_type":"code","source":["train_loader    = torch.utils.data.DataLoader(\n","    dataset     = train_dataset,\n","    batch_size  = config[\"batch_size\"],\n","    shuffle     = True,\n","    num_workers = 4,\n","    pin_memory  = True,\n","    collate_fn  = train_dataset.collate_fn\n",")\n","\n","val_loader      = torch.utils.data.DataLoader(\n","    dataset     = val_dataset,\n","    batch_size  = config[\"batch_size\"],\n","    shuffle     = False,\n","    num_workers = 2,\n","    pin_memory  = True,\n","    collate_fn  = val_dataset.collate_fn,\n",")\n","\n","test_loader     = torch.utils.data.DataLoader(\n","    dataset     = test_dataset,\n","    batch_size  = config[\"batch_size\"],\n","    shuffle     = False,\n","    num_workers = 2,\n","    pin_memory  = True,\n","    collate_fn  = test_dataset.collate_fn,\n",")\n","\n","print(\"No. of Train MFCCs   : \", train_dataset.__len__())\n","print(\"Batch Size           : \", config[\"batch_size\"])\n","print(\"Train Vatches        : \", train_loader.__len__())\n","print(\"Val Batches          : \", val_loader.__len__())\n","print(\"Test Batches         : \", test_loader.__len__())"],"metadata":{"id":"mtYEwlZhfUbT","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553231253,"user_tz":240,"elapsed":158,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"489eefa8-fff4-4c6d-ac13-30473b70d307"},"execution_count":12,"outputs":[{"output_type":"stream","name":"stdout","text":["No. of Train MFCCs   :  28539\n","Batch Size           :  64\n","Train Vatches        :  446\n","Val Batches          :  43\n","Test Batches         :  41\n"]}]},{"cell_type":"code","source":["''' Sanity Check '''\n","\n","print(\"Checking the Shapes of the Data --\\n\")\n","\n","for batch in train_loader:\n","    x_pad, y_shifted_pad, y_golden_pad, x_len, y_len, = batch\n","\n","    print(f\"x_pad shape:\\t\\t{x_pad.shape}\")\n","    print(f\"x_len shape:\\t\\t{x_len.shape}\\n\")\n","\n","    print(f\"y_shifted_pad shape:\\t{y_shifted_pad.shape}\")\n","    print(f\"y_golden_pad shape:\\t{y_golden_pad.shape}\")\n","    print(f\"y_len shape:\\t\\t{y_len.shape}\\n\")\n","\n","    # print(y_shifted_pad)\n","\n","    break"],"metadata":{"id":"eJo1BOHUPjpv","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553233066,"user_tz":240,"elapsed":766,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"64582d37-a90d-4898-befe-f23409bcfd3f"},"execution_count":13,"outputs":[{"output_type":"stream","name":"stdout","text":["Checking the Shapes of the Data --\n","\n","x_pad shape:\t\ttorch.Size([64, 1684, 27])\n","x_len shape:\t\ttorch.Size([64])\n","\n","y_shifted_pad shape:\ttorch.Size([64, 263])\n","y_golden_pad shape:\ttorch.Size([64, 263])\n","y_len shape:\t\ttorch.Size([64])\n","\n"]}]},{"cell_type":"markdown","source":["## DataLoader Verification\n","\n","- Train Dataset (train-clean-50)\n","```\n","Loaded Path:  train-clean-50\n","Max MFCC Length          :  1982\n","Avg MFCC Length          :  1256.06069100848\n","Max Transcript Length    :  354\n","Max Transcript Length    :  182.2676431424767\n","```\n","\n","- Train Dataset (train-clean-100)\n","```\n","Loaded Path:  train-clean-100\n","Max MFCC Length          :  2448\n","Avg MFCC Length          :  1264.6258453344547\n","Max Transcript Length    :  399\n","Max Transcript Length    :  186.65321139493324\n","```\n","\n","- Validation Dataset\n","```\n","Loaded Path:  dev-clean\n","Max MFCC Length          :  3260\n","Avg MFCC Length          :  713.3570107288198\n","Max Transcript Length    :  517\n","Max Transcript Length    :  107.71698113207547\n","```\n","\n","- Test Dataset\n","```\n","Loaded Path:  test-clean\n","Max MFCC Length          :  27\n","Avg MFCC Length          :  27.0\n","```\n","\n","- Population Stats\n","```\n","Max Length:  3260\n","```\n","\n","If your values is not matching, read hints and think what could have gone wrong. Then approach TAs.\n","\n"],"metadata":{"id":"F9nl0yf9Q_-2"}},{"cell_type":"code","source":["def verify_dataset(dataset, partition=\"train-clean-100\"):\n","    print(\"Loaded Path: \", partition)\n","\n","    if partition != \"test-clean\":\n","        print(\"Max MFCC Length          : \", np.max([data[0].shape[0] for data in dataset]))\n","        print(\"Avg MFCC Length          : \", np.mean([data[0].shape[0] for data in dataset]))\n","        print(\"Max Transcript Length    : \", np.max([data[1].shape[0] for data in dataset]))\n","        print(\"Max Transcript Length    : \", np.mean([data[1].shape[0] for data in dataset]))\n","    else:\n","        print(\"Max MFCC Length          : \", np.max([data[0].shape[0] for data in dataset]))\n","        print(\"Avg MFCC Length          : \", np.mean([data[0].shape[0] for data in dataset]))\n","\n","    print()\n","\n","verify_dataset(train_dataset, partition=\"train-clean-100\")\n","verify_dataset(val_dataset, partition=\"dev-clean\")\n","verify_dataset(test_dataset, partition=\"test-clean\")\n","\n","dataset_max_len = max(\n","    np.max([data[0].shape[0] for data in train_dataset]),\n","    np.max([data[0].shape[0] for data in val_dataset]),\n","    np.max([data[0].shape[0] for data in test_dataset]),\n",")\n","\n","print(\"Max Length: \", dataset_max_len)"],"metadata":{"id":"4t7hgbxfRDT3","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553240551,"user_tz":240,"elapsed":2879,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"52172272-634f-481f-9e26-79f941ba4ed9"},"execution_count":14,"outputs":[{"output_type":"stream","name":"stdout","text":["Loaded Path:  train-clean-100\n","Max MFCC Length          :  2448\n","Avg MFCC Length          :  1264.6258453344547\n","Max Transcript Length    :  399\n","Max Transcript Length    :  185.65321139493324\n","\n","Loaded Path:  dev-clean\n","Max MFCC Length          :  3260\n","Avg MFCC Length          :  713.3570107288198\n","Max Transcript Length    :  517\n","Max Transcript Length    :  107.71698113207547\n","\n","Loaded Path:  test-clean\n","Max MFCC Length          :  27\n","Avg MFCC Length          :  27.0\n","\n","Max Length:  3260\n"]}]},{"cell_type":"markdown","source":["# Introduction\n","\n","Transformers are highly effective for speech recognition tasks because they utilize self-attention mechanisms, enabling them to analyze entire sequences of audio data in one shot. This allows for a nuanced understanding of long-range dependencies and the complex temporal dynamics inherent in speech. Compared to traditional models, transformers offer improved accuracy and efficiency, adeptly handling variations in speech patterns, accents, and intonations; thus, making them exceptionally suitable for advanced speech recognition applications."],"metadata":{"id":"6QiV7F8hhS0c"}},{"cell_type":"markdown","source":["![image.png](data:image/png;base64,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)"],"metadata":{"id":"AF6R4uUfrm5j"}},{"cell_type":"markdown","source":["Given the complicated nature of these models, training in this homework will be a little different from other homeworks.<br>\n","We will employ the following to effectively train the transformer with fewer resources (data and compute power).\n","<br>\n","* **Data Complexity and Model Capacity**:  Starting with a CNN-LSTM for input sequences captures temporal dynamics and dependencies early on, simplifying initial learning. This is effective for data with strong temporal or sequential patterns, which LSTMs excel at handling. We will use an already trained model (trained in HW3P2).\n","\n","\n","* **Gradual (Incremental) Learning**: Starting with a simpler model (like an LSTM) allows the neural network to first learn the basic structures and patterns in the data. When we later introduce more complex mechanisms (like a transformer encoder), the model already has a foundational understanding of the data. This foundation can lead to a smoother and more effective learning process when the complexity of the model increases. In other words, we are using a simple, pretrained LSTM encoder as the basis for our transformer decoder and incrementally adding complexity!\n","\n","* **Learning Efficiency**: By freezing certain parts of the model during different stages of training, we focus the learning capacity of the network on specific modules, making the training process more efficient. This stepwise unfreezing and training allows different components of the model to adapt gradually, which can be very useful when integrating parts of the model that learn at different rates.\n","\n","* **Adaptation to Data Characteristics**: This strategy implicitly takes advantage of the data's characteristics by adjusting the model's architecture and training procedure to better suit the data's structure. For instance, if the data contains intricate sequential patterns, starting with an LSTM can harness these patterns effectively. Then, by introducing a transformer encoder, the model can further learn to capture complex, long-range dependencies within the data, something transformers are exceptionally good at.\n"],"metadata":{"id":"l6mtisVqr4FF"}},{"cell_type":"markdown","metadata":{"id":"wTZ-lv47XOj0"},"source":["## Utilities\n","\n","To make make building the transformer easy, we will define various utility functions and classes which corresponds to the various modules of the transformer. All Components are provided for you but your task will be too understand the modules as you will need  them in answering Canvas quizzes as well as implementing other parts of this notebook. Please refer to the write-up for more information."]},{"cell_type":"code","source":["def create_mask_1(padded_input, input_lengths=None, pad_idx=None):\n","    \"\"\" Create a mask to identify non-padding positions.\n","\n","    Args:\n","        padded_input: The input tensor with padding, shape (N, T, ...) or (N, T).\n","        input_lengths: Optional, the actual lengths of each sequence before padding, shape (N,).\n","        pad_idx: Optional, the index used for padding tokens.\n","\n","    Returns:\n","        A mask tensor with shape (N, T, 1), where non-padding positions are marked with 1 and padding positions are marked with 0.\n","    \"\"\"\n","\n","    assert input_lengths is not None or pad_idx is not None\n","\n","    # Create a mask based on input_lengths\n","    if input_lengths is not None:\n","        N = padded_input.size(0)        # padded_input : (N x T x ...)\n","        non_pad_mask = padded_input.new_ones(padded_input.size()[:-1])  # (N x T)\n","\n","        # Set the mask to 0 for padding positions\n","        for i in range(N):\n","          non_pad_mask[i, input_lengths[i]:] = 0\n","\n","    if pad_idx is not None:             # padded_input : N x T\n","\n","        assert padded_input.dim() == 2\n","\n","        # Create a mask where non-padding positions are marked with 1 and padding positions are marked with 0\n","        non_pad_mask = padded_input.ne(pad_idx).float()\n","\n","    return non_pad_mask.unsqueeze(-1)   # unsqueeze(-1) for broadcasting"],"metadata":{"id":"xPKEiQYszRjm","executionInfo":{"status":"ok","timestamp":1713553243742,"user_tz":240,"elapsed":173,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":15,"outputs":[]},{"cell_type":"code","source":["def create_mask_2(seq, pad_idx=None):\n","    \"\"\" Create a mask to prevent positions from attending to subsequent positions.\n","\n","    Args:\n","        seq: The input sequence tensor, shape (batch_size, sequence_length).\n","\n","    Returns:\n","        A mask tensor with shape (batch_size, sequence_length, sequence_length),\n","            where positions are allowed to attend to previous positions but not to subsequent positions.\n","    \"\"\"\n","\n","    sz_b, len_s = seq.size()\n","\n","    # Create an upper triangular matrix with zeros on the diagonal and below (indicating allowed positions)\n","    #   and ones above the diagonal (indicating disallowed positions)\n","    subsequent_mask = torch.triu(\n","        torch.ones((len_s, len_s), device=seq.device, dtype=torch.uint8), diagonal=1)\n","\n","    # Expand the mask to match the batch size, resulting in a mask for each sequence in the batch.\n","    mask = subsequent_mask.unsqueeze(0).expand(sz_b, -1, -1)  # b x ls x ls\n","\n","\n","    ''' Create a mask to ignore padding positions in the key sequence during attention calculation. '''\n","\n","    # Expanding to fit the shape of key query attention matrix.\n","    if pad_idx != None:\n","      len_q = seq.size(1)\n","\n","      # Create a mask where padding positions in the key sequence are marked with 1.\n","      padding_mask  = seq.eq(pad_idx)\n","\n","      # Expand the mask to match the dimensions of the key-query attention matrix.\n","      padding_mask  = padding_mask.unsqueeze(1).expand(-1, len_q, -1)  # b x lq x lk\n","      mask          = (padding_mask + mask).gt(0)\n","\n","    else:\n","      mask = mask.gt(0)\n","\n","    return mask"],"metadata":{"id":"OS7su_eQzWp5","executionInfo":{"status":"ok","timestamp":1713553244438,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":16,"outputs":[]},{"cell_type":"code","source":["def create_mask_3(padded_input, input_lengths, expand_length):\n","    \"\"\" Create an attention mask to ignore padding positions in the input sequence during attention calculation.\n","\n","    Args:\n","        padded_input: The input tensor with padding, shape (N, Ti, ...).\n","        input_lengths: The actual lengths of each sequence before padding, shape (N,).\n","        expand_length: The length to which the attention mask should be expanded,\n","            usually equal to the length of the sequence that the attention scores will be applied to.\n","\n","    Returns:\n","        An attention mask tensor with shape (N, expand_length, Ti),\n","            where padding positions in the input sequence are marked with 1 and other positions are marked with 0.\n","    \"\"\"\n","\n","    # Create a mask to identify non-padding positions, shape (N, Ti, 1)\n","    # (N x Ti x 1)\n","    non_pad_mask    = create_mask_1(padded_input, input_lengths=input_lengths)\n","\n","    # Invert the mask to identify padding positions, shape (N, Ti)\n","    # N x Ti, lt(1) like-not operation\n","    pad_mask        = non_pad_mask.squeeze(-1).lt(1)\n","\n","\n","    # Expand the mask to match the dimensions of the attention matrix, shape (N, expand_length, Ti)\n","    attn_mask       = pad_mask.unsqueeze(1).expand(-1, expand_length, -1)\n","\n","    return attn_mask"],"metadata":{"id":"VoU14igzzbk3","executionInfo":{"status":"ok","timestamp":1713553245474,"user_tz":240,"elapsed":2,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":17,"outputs":[]},{"cell_type":"code","source":["class ScaledDotProductAttention(torch.nn.Module):\n","    ''' Scaled Dot-Product Attention '''\n","\n","    def __init__(self, temperature, attn_dropout=0.1):\n","        super().__init__()\n","        self.temperature    = temperature                       # Scaling factor for the dot product\n","        self.dropout        = torch.nn.Dropout(attn_dropout)    # Dropout layer for attention weights\n","        self.softmax        = torch.nn.Softmax(dim=2)           # Softmax layer along the attention dimension\n","\n","    def forward(self, q, k, v, mask=None):\n","\n","        # Calculate the dot product between queries and keys.\n","        attn = torch.bmm(q, k.transpose(1, 2))\n","\n","        # Scale the dot product by the temperature.\n","        attn = attn / self.temperature\n","\n","        if mask is not None:\n","            # Apply the mask by setting masked positions to a large negative value.\n","            # This ensures they have a softmax score close to zero.\n","            mask_value = -1e+30 if attn.dtype == torch.float32 else -1e+4\n","            attn = attn.masked_fill(mask, mask_value)\n","\n","        # Apply softmax to obtain attention weights.\n","        attn    = self.softmax(attn)\n","\n","        # Apply dropout to the attention weights.\n","        attn    = self.dropout(attn)\n","\n","        # Compute the weighted sum of values based on the attention weights.\n","        output  = torch.bmm(attn, v)\n","\n","        return output, attn # Return the attention output and the attention weights."],"metadata":{"id":"fb5-BhwgZydi","executionInfo":{"status":"ok","timestamp":1713553245787,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":18,"outputs":[]},{"cell_type":"code","source":["def save_attention_plot(attention_weights, epoch=0):\n","    ''' function for saving attention weights plot to a file\n","\n","        @NOTE: default starter code set to save cross attention\n","    '''\n","\n","    plt.clf()  # Clear the current figure\n","    sns.heatmap(attention_weights, cmap=\"GnBu\")  # Create heatmap\n","\n","    # Save the plot to a file. Specify the directory if needed.\n","    plt.savefig(f\"cross_attention-epoch{epoch}.png\")"],"metadata":{"id":"iuJntqzGzssH","executionInfo":{"status":"ok","timestamp":1713553246436,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":19,"outputs":[]},{"cell_type":"code","source":["class MultiHeadAttention(torch.nn.Module):\n","    ''' Multi-Head Attention Module '''\n","\n","    def __init__(self, n_head, d_model, dropout=0.1):\n","        super().__init__()\n","\n","        self.n_head = n_head # Number of attention heads\n","        self.d_k    = d_model // n_head\n","        self.d_v    = d_model // n_head\n","\n","\n","        # Linear layers for projecting the input query, key, and value to multiple heads\n","        self.w_qs   = torch.nn.Linear(d_model, n_head * self.d_k)\n","        self.w_ks   = torch.nn.Linear(d_model, n_head * self.d_k)\n","        self.w_vs   = torch.nn.Linear(d_model, n_head * self.d_v)\n","\n","        torch.nn.init.normal_(self.w_qs.weight, mean=0, std=np.sqrt(2.0 / (d_model + self.d_k)))\n","        torch.nn.init.normal_(self.w_ks.weight, mean=0, std=np.sqrt(2.0 / (d_model + self.d_k)))\n","        torch.nn.init.normal_(self.w_vs.weight, mean=0, std=np.sqrt(2.0 / (d_model + self.d_v)))\n","\n","        # Initialize the weights of the linear layers\n","        self.attention = ScaledDotProductAttention(\n","            temperature=np.power(self.d_k, 0.5), attn_dropout=dropout)\n","\n","        # Final linear layer to project the concatenated outputs of the attention heads back to the model dimension\n","        self.fc = torch.nn.Linear(n_head * self.d_v, d_model)\n","\n","        torch.nn.init.xavier_normal_(self.fc.weight)\n","\n","        self.dropout = torch.nn.Dropout(dropout)\n","\n","    def forward(self, q, k, v, mask=None):\n","\n","        # following key, value, query standard computation\n","        d_k, d_v, n_head    = self.d_k, self.d_v, self.n_head\n","        sz_b, len_q, _      = q.size()\n","        sz_b, len_k, _      = k.size()\n","        sz_b, len_v, _      = v.size()\n","\n","        # Project the input query, key, and value to multiple heads\n","        q = self.w_qs(q).view(sz_b, len_q, n_head, d_k)\n","        k = self.w_ks(k).view(sz_b, len_k, n_head, d_k)\n","        v = self.w_vs(v).view(sz_b, len_v, n_head, d_v)\n","\n","        # Rearrange the dimensions to group the heads together for parallel processing\n","        q = q.permute(2, 0, 1, 3).contiguous().view(-1, len_q, d_k) # (n*b) x lq x dk\n","        k = k.permute(2, 0, 1, 3).contiguous().view(-1, len_k, d_k) # (n*b) x lk x dk\n","        v = v.permute(2, 0, 1, 3).contiguous().view(-1, len_v, d_v) # (n*b) x lv x dv\n","\n","        # Repeat the mask for each attention head if a mask is provided\n","        if mask is not None:\n","            mask = mask.repeat(n_head, 1, 1)\n","\n","        # Apply scaled dot-product attention to the projected query, key, and value\n","        output, attn    = self.attention(q, k, v, mask=mask)\n","\n","        # Rearrange the output back to the original order and concatenate the heads\n","        output          = output.view(n_head, sz_b, len_q, d_v)\n","        output          = output.permute(1, 2, 0, 3).contiguous().view(sz_b, len_q, -1) # b x lq x (n*dv)\n","\n","        output          = self.dropout(self.fc(output))\n","\n","        return output, attn"],"metadata":{"id":"wOCsC5z7qvZ5","executionInfo":{"status":"ok","timestamp":1713553247085,"user_tz":240,"elapsed":141,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":20,"outputs":[]},{"cell_type":"markdown","source":["## Transformer Modules"],"metadata":{"id":"RCwIlOrXvxGO"}},{"cell_type":"markdown","source":["### Positional Encoding"],"metadata":{"id":"4Lt4-VJerMY3"}},{"cell_type":"code","source":["class PositionalEncoding(torch.nn.Module):\n","    ''' Position Encoding from Attention Is All You Need Paper '''\n","\n","    def __init__(self, d_model, max_len=5000):\n","        super().__init__()\n","\n","        # Initialize a tensor to hold the positional encodings\n","        pe          = torch.zeros(max_len, d_model)\n","\n","        # Create a tensor representing the positions (0 to max_len-1)\n","        position    = torch.arange(0, max_len, dtype=torch.float).unsqueeze(1)\n","\n","        # Calculate the division term for the sine and cosine functions\n","        # This term creates a series of values that decrease geometrically, used to generate varying frequencies for positional encodings\n","        div_term    = torch.exp(torch.arange(0, d_model, 2).float() * (-math.log(10000.0) / d_model))\n","\n","        # Compute the positional encodings using sine and cosine functions\n","        pe[:, 0::2] = torch.sin(position * div_term)\n","        pe[:, 1::2] = torch.cos(position * div_term)\n","\n","        # Reshape the positional encodings tensor and make it a buffer\n","        pe = pe.unsqueeze(0)\n","        self.register_buffer(\"pe\", pe)\n","\n","    def forward(self, x):\n","\n","      return x + self.pe[:, :x.size(1)]"],"metadata":{"id":"Pn9zEhZQrPyH","executionInfo":{"status":"ok","timestamp":1713553248461,"user_tz":240,"elapsed":135,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":21,"outputs":[]},{"cell_type":"markdown","source":["### Feed Forward Networks"],"metadata":{"id":"WfFLc2WCtEjr"}},{"cell_type":"code","source":["class FeedForward(torch.nn.Module):\n","    ''' Projection Layer (Fully Connected Layers) '''\n","\n","    def __init__(self, d_model, d_ff=2048, dropout=0.1):\n","        super().__init__()\n","\n","        self.linear_1   = torch.nn.Linear(d_model, d_ff)\n","        self.dropout    = torch.nn.Dropout(dropout)\n","        self.linear_2   = torch.nn.Linear(d_ff, d_model)\n","\n","    def forward(self, x):\n","\n","        # Apply the first linear layer, GeLU activation, and then dropout\n","        x = self.dropout(torch.nn.functional.gelu(self.linear_1(x)))\n","\n","         # Apply the second linear layer to project the dimension back to d_model\n","        x = self.linear_2(x)\n","\n","        return x"],"metadata":{"id":"nGfM9LObtC8k","executionInfo":{"status":"ok","timestamp":1713553251313,"user_tz":240,"elapsed":207,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":22,"outputs":[]},{"cell_type":"markdown","source":["# PART #1 (Preliminary Submission)\n","\n","In this part we will use a simple CNN-LSTM as an encoder with a Transformer Decoder (with Self-Attention and Cross-Attention)."],"metadata":{"id":"tkbco9zxxbnc"}},{"cell_type":"markdown","source":["## CNN-LSTM Encoder"],"metadata":{"id":"euflRQmX4fBC"}},{"cell_type":"code","source":["class Permute(nn.Module):\n","    def forward(self, x):\n","        return x.transpose(1, 2)"],"metadata":{"id":"cOg9-ZuWkT_n","executionInfo":{"status":"ok","timestamp":1713553254611,"user_tz":240,"elapsed":149,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":23,"outputs":[]},{"cell_type":"code","source":["class pBLSTM(torch.nn.Module):\n","    def __init__(self, input_size, hidden_size):\n","        super(pBLSTM, self).__init__()\n","        self.blstm = torch.nn.LSTM(input_size=2*input_size, hidden_size=hidden_size, bidirectional=True, batch_first=True)\n","\n","    def forward(self, x_packed):\n","        x_padded, x_lens = pad_packed_sequence(x_packed, batch_first=True)\n","        x_padded, x_lens = self.trunc_reshape(x_padded, x_lens)\n","        x_packed = pack_padded_sequence(x_padded, x_lens, batch_first=True, enforce_sorted=False)\n","        output_packed, hidden = self.blstm(x_packed)\n","        return output_packed\n","\n","    def trunc_reshape(self, x, x_lens):\n","        if x.shape[1]%2==1:\n","            x = x[:,:-1,:]\n","        x = x.reshape((x.shape[0], int(x.shape[1]//2), x.shape[2]*2))\n","        x_lens = (x_lens/2).type(torch.int)\n","        return x, x_lens"],"metadata":{"id":"RHIIFL0WV7P-","executionInfo":{"status":"ok","timestamp":1713553255424,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":24,"outputs":[]},{"cell_type":"code","source":["from torch.autograd import Variable\n","\n","class LockedDropout(nn.Module):\n","    def __init__(self, dropout):\n","        super().__init__()\n","        self.dropout = dropout\n","\n","    def forward(self, x):\n","        if not self.training or not self.dropout:\n","            return x\n","        x_unpacked, x_lens = pad_packed_sequence(x, batch_first=True)\n","        m = x.data.new(x_unpacked.size(0), 1, x_unpacked.size(2)).bernoulli_(1 - self.dropout)\n","        mask = Variable(m, requires_grad=False) / (1 - self.dropout)\n","        mask = mask.expand_as(x_unpacked)\n","        out = mask * x_unpacked\n","        return pack_padded_sequence(out, lengths=x_lens, batch_first=True, enforce_sorted=False)"],"metadata":{"id":"gNhnIDjJWBM6","executionInfo":{"status":"ok","timestamp":1713553257003,"user_tz":240,"elapsed":152,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":25,"outputs":[]},{"cell_type":"code","source":["class CNN_LSTM_Encoder(torch.nn.Module):\n","    ''' The Encoder takes utterances as inputs and returns latent feature representations '''\n","\n","    def __init__(self, input_size, encoder_hidden_size, enc_dropout=0.2):\n","        super(CNN_LSTM_Encoder, self).__init__()\n","\n","        self.embedding = torch.nn.Sequential(\n","            torch.nn.Conv1d(input_size, 2*input_size, kernel_size=5, stride=1, padding=2),\n","            torch.nn.BatchNorm1d(2*input_size),\n","            torch.nn.ReLU(inplace=True),\n","            torch.nn.Conv1d(2*input_size, 4*input_size, kernel_size=5, stride=1, padding=2),\n","            torch.nn.BatchNorm1d(4*input_size),\n","            torch.nn.ReLU(inplace=True)\n","        )\n","\n","        self.bilstm = torch.nn.LSTM(input_size=4*input_size, hidden_size=encoder_hidden_size, bidirectional=True, batch_first=True, dropout=enc_dropout)\n","\n","        self.pBLSTMs = torch.nn.Sequential(\n","            pBLSTM(2 * encoder_hidden_size, encoder_hidden_size),\n","            LockedDropout(dropout=2*enc_dropout),\n","            pBLSTM(2*encoder_hidden_size, 2*encoder_hidden_size),\n","            LockedDropout(dropout=enc_dropout),\n","        )\n","\n","        self.permute = Permute()\n","\n","    def forward(self, x, x_lens):\n","        x = self.permute(x)\n","        x = self.embedding(x)\n","        x = self.permute(x)\n","\n","        x_packed        = pack_padded_sequence(x, x_lens, batch_first= True, enforce_sorted=False)\n","        x_packed, _     = self.bilstm(x_packed)\n","        x_packed        = self.pBLSTMs(x_packed)\n","        encoder_outputs, encoder_lens = pad_packed_sequence(x_packed, batch_first=True)\n","\n","        return encoder_outputs, encoder_lens"],"metadata":{"id":"eL2vCjRfw82o","executionInfo":{"status":"ok","timestamp":1713553257928,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":26,"outputs":[]},{"cell_type":"markdown","source":["## HW3P2 model\n","\n"],"metadata":{"id":"NfeEl35NDZvo"}},{"cell_type":"markdown","metadata":{"id":"EdJbO38yDZvu"},"source":["### Pyramid Bi-LSTM (pBLSTM)"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"KKe-M31ODZvu"},"outputs":[],"source":["# Utils for network\n","torch.cuda.empty_cache()\n","# from torch.autograd import Variable\n","\n","class PermuteBlock(torch.nn.Module):\n","    def forward(self, x):\n","        return x.transpose(1, 2)\n","\n","class LockedDropout(nn.Module):\n","    \"\"\" LockedDropout applies the same dropout mask to every time step.\n","\n","    **Thank you** to Sales Force for their initial implementation of :class:`WeightDrop`. Here is\n","    their `License\n","    <https://github.com/salesforce/awd-lstm-lm/blob/master/LICENSE>`__.\n","\n","    Args:\n","        p (float): Probability of an element in the dropout mask to be zeroed.\n","    \"\"\"\n","\n","    def __init__(self, p=0.25):\n","        self.p = p\n","        super().__init__()\n","\n","    def forward(self, x):\n","        \"\"\"\n","        Args:\n","            x (:class:`torch.FloatTensor` [sequence length, batch size, rnn hidden size]): Input to\n","                apply dropout too.\n","        \"\"\"\n","        if not self.training or not self.p:\n","            return x\n","        x = x.clone()\n","        mask = x.new_empty(1, x.size(1), x.size(2), requires_grad=False).bernoulli_(1 - self.p)\n","        mask = mask.div_(1 - self.p)\n","        mask = mask.expand_as(x)\n","        return x * mask\n","\n","\n","    def __repr__(self):\n","        return self.__class__.__name__ + '(' \\\n","            + 'p=' + str(self.p) + ')'\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"oO5XoniqDZvu"},"outputs":[],"source":["class pBLSTM(torch.nn.Module):\n","\n","    '''\n","    Pyramidal BiLSTM\n","    Read the write up/paper and understand the concepts and then write your implementation here.\n","\n","    At each step,\n","    1. Pad your input if it is packed (Unpack it)\n","    2. Reduce the input length dimension by concatenating feature dimension\n","        (Tip: Write down the shapes and understand)\n","        (i) How should  you deal with odd/even length input?\n","        (ii) How should you deal with input length array (x_lens) after truncating the input?\n","    3. Pack your input\n","    4. Pass it into LSTM layer\n","\n","    To make our implementation modular, we pass 1 layer at a time.\n","    '''\n","\n","    def __init__(self, input_size, hidden_size):\n","        super(pBLSTM, self).__init__()\n","\n","        self.blstm = torch.nn.LSTM(input_size=2*input_size, hidden_size=hidden_size, num_layers=3, bidirectional=True, dropout = 0.3)  # TODO: Initialize a single layer bidirectional LSTM with the given input_size and hidden_size\n","        #could drop out\n","    def forward(self, x_packed): # x_packed is a PackedSequence\n","\n","        # TODO: Pad Packed Sequence\n","        seq_unpacked, lens_unpacked = pad_packed_sequence(x_packed, batch_first=True)\n","        # print(f'in the plstm section after pad {seq_unpacked.shape},{lens_unpacked.shape}')\n","        # Call self.trunc_reshape() which downsamples the time steps of x and increases the feature dimensions as mentioned above\n","        x, x_lens = self.trunc_reshape(x=seq_unpacked, x_lens=lens_unpacked)\n","        # print(f'in the plstm section after truncate {x.shape},{x_lens.shape}')\n","        # self.trunc_reshape will return 2 outputs. What are they? Think about what quantites are changing.\n","        # TODO: Pack Padded Sequence. What output(s) would you get?\n","        PackedSequence = pack_padded_sequence(input=x,lengths=x_lens,batch_first=True,enforce_sorted=False)\n","        # TODO: Pass the sequence through bLSTM\n","        packed_output, (_, _) = self.blstm(PackedSequence)\n","        # What do you return?\n","\n","        return packed_output\n","\n","    def trunc_reshape(self, x, x_lens):\n","        # TODO: If you have odd number of timesteps, how can you handle it? (Hint: You can exclude them)\n","        # TODO: Reshape x. When reshaping x, you have to reduce number of timesteps by a downsampling factor while increasing number of features by the same factor\n","        # TODO: Reduce lengths by the same downsampling factor\n","        # The shape of x is [batch_size, frames, features]\n","        batch_size = x.shape[0]\n","        frames = x.shape[1]\n","        features = x.shape[2]\n","        if frames % 2 != 0:\n","            x = x[:, :-1, :]\n","        x = x.reshape((batch_size, x.shape[1]//2, features*2))\n","        x_lens = x_lens/2\n","        return x, x_lens"]},{"cell_type":"markdown","metadata":{"id":"z12P7GytDZvu"},"source":["### Encoder"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"3T3GLRikDZvu"},"outputs":[],"source":["class Encoder_HW3(torch.nn.Module):\n","    '''\n","    The Encoder takes utterances as inputs and returns latent feature representations\n","    '''\n","    def __init__(self, input_size, encoder_hidden_size):\n","        super(Encoder_HW3, self).__init__()\n","\n","        self.PermuteBlock = PermuteBlock()\n","        self.embedding = torch.nn.Sequential(torch.nn.Conv1d(in_channels=input_size,out_channels=64,kernel_size=3,padding=1,stride=1),\n","                                             torch.nn.BatchNorm1d(64),\n","                                             torch.nn.GELU(),\n","                                             torch.nn.Conv1d(in_channels=64, out_channels=128, kernel_size=3, padding=1, stride=1),\n","                                             torch.nn.BatchNorm1d(128),\n","                                             torch.nn.GELU(),\n","                                             torch.nn.Conv1d(in_channels=128, out_channels=256, kernel_size=3, padding=1, stride=1),\n","                                             torch.nn.BatchNorm1d(256),\n","                                             torch.nn.GELU()\n","                                             )\n","        self.lstm_layers = torch.nn.ModuleList()\n","        for i in range(2):\n","            input_size_pblstm = 256 if i == 0 else encoder_hidden_size * 2\n","            self.lstm_layers.append(pBLSTM(input_size=input_size_pblstm, hidden_size=encoder_hidden_size))\n","            self.lstm_layers.append(LockedDropout())\n","\n","    def forward(self, x, x_lens):\n","        # Where are x and x_lens coming from? The dataloader\n","        #TODO: Call the permute layer and embedding layer\n","        x = self.PermuteBlock(x)\n","        x = self.embedding(x)\n","        x = self.PermuteBlock(x)\n","        # # TODO: Pack Padded Sequence\n","        # packed = pack_padded_sequence(input=x,lengths=x_lens,batch_first=True,enforce_sorted=False)\n","        # # TODO: Pass Sequence through the pyramidal Bi-LSTM layer\n","        # x = self.pBLSTM1(packed)\n","        # # TODO: Pad Packed Sequence\n","        # encoder_outputs, encoder_lens= pad_packed_sequence(sequence=x,batch_first=True)\n","        # print(f'The output from plstm, {encoder_outputs.shape},{encoder_lens.shape}')\n","        for layer in self.lstm_layers:\n","            if isinstance(layer, pBLSTM):\n","                x_packed = pack_padded_sequence(x, x_lens, batch_first=True, enforce_sorted=False)\n","                x_packed = layer(x_packed)\n","                x, x_lens = pad_packed_sequence(x_packed, batch_first=True)\n","            else:\n","                x = torch.permute(x, (1, 0, 2))\n","                x = layer(x)\n","                x = torch.permute(x,(1,0,2))\n","        # Remember the number of output(s) each function returns\n","\n","        return x, x_lens\n"]},{"cell_type":"markdown","metadata":{"id":"s6WtcU0LDZvu"},"source":["### Decoder"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"b0BLLCHxDZvu"},"outputs":[],"source":["class HW3P2_encoder(torch.nn.Module):\n","\n","    def __init__(self, input_size, embed_size= 192):\n","        super().__init__()\n","\n","        self.encoder        = Encoder_HW3(input_size=input_size, encoder_hidden_size=embed_size*2)\n","\n","    def forward(self, x, lengths_x):\n","\n","        encoder_out, encoder_lens   = self.encoder(x, lengths_x)\n","        return encoder_out, encoder_lens"]},{"cell_type":"markdown","source":["## Transformer Decoder\n","\n","![image.png](data:image/png;base64,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)"],"metadata":{"id":"bdJsTVEvZIwe"}},{"cell_type":"markdown","source":["Tips to complete the Decoder below.\n","\n","* Can you seperate the decoder layers from the whole decoder?\n","* How many multi-head attention do we have in the figure above.\n","* How many layer normalizations?\n","* How do we make the residual connections?\n","\n","For Forward Method:\n","* What are the inputs to various multi-head attention modules?\n","* How do we make the residual connection (do we need dropout)?\n"],"metadata":{"id":"sdzUsFHRyT6J"}},{"cell_type":"code","source":["class DecoderLayer(torch.nn.Module):\n","    def __init__(self, d_model, num_heads, d_ff, dropout=0.1):\n","        super().__init__()\n","\n","        # @TODO: fill in the blanks appropriately (given the modules above)\n","        self.mha1       = MultiHeadAttention(n_head = num_heads, d_model = d_model, dropout=dropout)\n","        self.mha2       = MultiHeadAttention(n_head = num_heads, d_model=d_model, dropout=dropout)\n","        self.ffn        = FeedForward(d_model=d_model, d_ff=d_ff, dropout=dropout) # this d_ff might be different from dcoder layer\n","\n","        self.layernorm1 = torch.nn.LayerNorm(normalized_shape=d_model)\n","        self.layernorm2 = torch.nn.LayerNorm(normalized_shape=d_model)\n","        self.layernorm3 = torch.nn.LayerNorm(normalized_shape=d_model)\n","\n","        self.dropout1   = torch.nn.Dropout(p=dropout)\n","        self.dropout2   = torch.nn.Dropout(p=dropout)\n","        self.dropout3   = torch.nn.Dropout(p=dropout)\n","\n","\n","    def forward(self, padded_targets, enc_output, enc_input_lengths, dec_enc_attn_mask, pad_mask, slf_attn_mask):\n","\n","        # Masked Multi-Head Attention\n","        #   (1) apply MHA with the lookahead mask\n","        ''' TODO '''\n","        attn_output1, attention_weights1  = self.mha1(q=padded_targets, k=padded_targets, v=padded_targets, mask=slf_attn_mask) # here the padded_targets is (batch, length, d_model)\n","        # Skip (Residual) Connections\n","        #   (1) perform dropout on padded attention output\n","        #   (2) add the true outputs (padded_targets) as a skip connection\n","        ''' TODO '''\n","        out1 = padded_targets + self.dropout1(attn_output1)  # or out1 = padded_targets + self.dropout1(attn_output1)\n","        # Layer Normalization\n","        #   (1) call layernorm on this resulting value\n","        ''' TODO '''\n","        normalized1 = self.layernorm1(out1)\n","        # Masked Multi-Head Attention on Encoder Outputs and Targets\n","        #   (1) apply MHA with the self-attention mask\n","        ''' TODO '''\n","        attn_output2, attention_weights2 = self.mha2(q=normalized1, k=enc_output, v=enc_output, mask=dec_enc_attn_mask) # this a question\n","        # Skip (Residual) Connections\n","        #   (1) perform dropout on this second padded attention output\n","        #   (2) add the output of first MHA block as a skip connection\n","        ''' TODO '''\n","        out2 = normalized1 + self.dropout2(attn_output2)\n","        # Layer Normalization\n","        #   (1) call layernorm on this resulting value\n","        ''' TODO '''\n","        normalized2 = self.layernorm2(out2)\n","        # Feed Forward Network\n","        #   (1) pass through the FFN\n","        ''' TODO '''\n","        FFN_out = self.ffn(normalized2)\n","        # Skip (Residual) Connections\n","        #   (1) perform dropout on the output\n","        #   (2) add the output of second MHA block as a skip connection\n","        ''' TODO '''\n","        out3 = self.dropout3(FFN_out) + normalized2\n","        # apply Layer Normalization onto output of feed-forward network\n","        ''' TODO '''\n","        normalized3 = self.layernorm3(out3)\n","        return normalized3, attention_weights1, attention_weights2"],"metadata":{"id":"80KuP9_-1AMw","executionInfo":{"status":"ok","timestamp":1713553265714,"user_tz":240,"elapsed":152,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":27,"outputs":[]},{"cell_type":"code","source":["class Decoder(torch.nn.Module):\n","    def __init__(self, num_layers, d_model, num_heads, d_ff, dropout,\n","            target_vocab_size, max_seq_length, eos_token, sos_token, pad_token):\n","        super().__init__()\n","\n","        self.EOS_TOKEN      = eos_token\n","        self.SOS_TOKEN      = sos_token\n","        self.PAD_TOKEN      = pad_token\n","\n","        self.max_seq_length = max_seq_length\n","        self.num_layers     = num_layers\n","\n","        # use torch.nn.ModuleList() with list comprehension looping through num_layers\n","        # @NOTE: think about what stays constant per each DecoderLayer (how to call DecoderLayer)\n","        # @HINT: We've implemented this for you.\n","        self.dec_layers = torch.nn.ModuleList(\n","            [DecoderLayer(d_model, num_heads, d_ff, dropout) for _ in range(num_layers)]\n","        )\n","\n","        self.target_embedding = torch.nn.Embedding(target_vocab_size, d_model, padding_idx=self.PAD_TOKEN)\n","        self.positional_encoding    = PositionalEncoding(d_model=d_model, max_len=max_seq_length)\n","        self.final_linear           = torch.nn.Sequential(torch.nn.Linear(d_model, target_vocab_size))\n","        self.dropout                = torch.nn.Dropout(p=dropout)\n","        self.layernorm        = torch.nn.LayerNorm(d_model)\n","\n","    def forward(self, padded_targets, enc_output, enc_input_lengths):\n","\n","        # create a padding mask for the padded_targets with <PAD_TOKEN>\n","        ''' TODO '''\n","        # print(f'The shape for padded_target: {padded_targets.shape}')\n","        # print(f'The shape for encoder output: {enc_output.shape}')\n","        # print(f'The shape for encoder input length: {enc_input_lengths.shape}')\n","        # The shape for padded_target: torch.Size([64, 332])\n","        # The shape for encoder output: torch.Size([64, 1625, 512])\n","        # The shape for encoder input length: torch.Size([64])\n","        padding_mask = create_mask_1(padded_targets, pad_idx=self.PAD_TOKEN)\n","        # print(f'The shape after padding mask: {padding_mask.shape}')\n","        \"\"\" Create a mask to identify non-padding positions.\n","\n","        Args:\n","          padded_input: The input tensor with padding, shape (N, T, ...) or (N, T).\n","          input_lengths: Optional, the actual lengths of each sequence before padding, shape (N,).\n","          pad_idx: Optional, the index used for padding tokens.\n","\n","        Returns:\n","          A mask tensor with shape (N, T, 1), where non-padding positions are marked with 1 and padding positions are marked with 0.\n","        \"\"\"\n","\n","        # creating an attention mask for the future subsequences (look-ahead mask)\n","        ''' TODO '''\n","        look_ahead_mask =  create_mask_2(padded_targets, pad_idx=self.PAD_TOKEN)\n","        # print(f'The shape after look ahead mask: {look_ahead_mask.shape}')\n","        \"\"\" Create a mask to prevent positions from attending to subsequent positions.\n","\n","        Args:\n","          seq: The input sequence tensor, shape (batch_size, sequence_length).\n","\n","        Returns:\n","            A mask tensor with shape (batch_size, sequence_length, sequence_length),\n","            where positions are allowed to attend to previous positions but not to subsequent positions.\n","        \"\"\"\n","        # creating attention mask to ignore padding positions in the input sequence during attention calculation\n","        ''' TODO '''\n","        attn_mask = create_mask_3(padded_input=enc_output, input_lengths=enc_input_lengths, expand_length=padded_targets.size(1))\n","        # print(f'The shape after attn mask: {attn_mask.shape}')\n","        \"\"\" Create an attention mask to ignore padding positions in the input sequence during attention calculation.\n","\n","        Args:\n","          padded_input: The input tensor with padding, shape (N, Ti, ...).\n","          input_lengths: The actual lengths of each sequence before padding, shape (N,).\n","          expand_length: The length to which the attention mask should be expanded,\n","            usually equal to the length of the sequence that the attention scores will be applied to.\n","\n","        Returns:\n","          An attention mask tensor with shape (N, expand_length, Ti),\n","            where padding positions in the input sequence are marked with 1 and other positions are marked with 0.\n","        \"\"\"\n","        # computing embeddings for the target sequence\n","        embedding = self.target_embedding.forward(padded_targets)  # the shape is [batch, target_time_length, dmodel]\n","        # print(f'The shape after embedding: {embedding.shape}')\n","        # computing Positional Encodings with the embedded targets\n","        out = self.positional_encoding.forward(embedding)   # positional encoding doesn't change the shape the shape is [batch, target_time_length, dmodel]\n","        # out = self.dropout(out)\n","        # print(f'The shape after positional encoding: {out.shape}')\n","        # passing through decoder layers\n","        ''' TODO '''\n","        # i=0\n","        all_attention_weights = {}\n","        for i, layer in enumerate(self.dec_layers):\n","            out, attention_weights1, attention_weights2 = layer(padded_targets = out, enc_output = enc_output, enc_input_lengths = enc_input_lengths, dec_enc_attn_mask = attn_mask, pad_mask = padding_mask, slf_attn_mask = look_ahead_mask)\n","            all_attention_weights[f'layer_{i+1}_weight1'] = attention_weights1   # I Think shape is (batch_size, target_time_length, target_time_length)\n","            all_attention_weights[f'layer_{i+1}_weight2'] = attention_weights2   # I Think shape is (batch_size, target_time_length, input time length)\n","        # linear layer (Final Projection) for next character prediction\n","        ''' TODO '''\n","        # out = self.layernorm(out)\n","        out = self.final_linear(out)\n","        return out, all_attention_weights\n","\n","\n","    def recognize_greedy_search(self, enc_outputs, enc_input_lengths):\n","        ''' passes the encoder outputs and its corresponding lengths through autoregressive network\n","\n","            @NOTE: You do not need to make changes to this method.\n","        '''\n","\n","        batch_size = enc_outputs.size(0)\n","\n","        # start with the <SOS> token for each sequence in the batch\n","        target_seq = torch.full((batch_size, 1), self.SOS_TOKEN, dtype=torch.long).to(enc_outputs.device)\n","\n","        finished = torch.zeros(batch_size, dtype=torch.bool).to(enc_outputs.device)\n","\n","        for _ in range(self.max_seq_length):\n","\n","            # preparing attention masks\n","            # filled with ones becaues we want to attend to all the elements in the sequence\n","            pad_mask = torch.ones_like(target_seq).float().unsqueeze(-1)  # (batch_size x i x 1)\n","            slf_attn_mask_subseq = create_mask_2(target_seq)\n","\n","            x = self.positional_encoding(self.target_embedding(target_seq))\n","\n","            for i in range(self.num_layers):\n","                x, block1, block2 = self.dec_layers[i](\n","                    x, enc_outputs, enc_input_lengths, None, pad_mask, slf_attn_mask_subseq)\n","\n","            seq_out = self.final_linear(x[:, -1])\n","            logits = torch.nn.functional.log_softmax(seq_out, dim=1)\n","\n","            # selecting the token with the highest probability\n","            # @NOTE: this is the autoregressive nature of the network!\n","            next_token = logits.argmax(dim=-1).unsqueeze(1)\n","\n","            # appending the token to the sequence\n","            target_seq = torch.cat([target_seq, next_token], dim=-1)\n","\n","            # checking if <EOS> token is generated\n","            eos_mask = next_token.squeeze(-1) == self.EOS_TOKEN\n","            # or opration, if both or one of them is true store the value of the finished sequence in finished variable\n","            finished |= eos_mask\n","\n","            # end if all sequences have generated the EOS token\n","            if finished.all(): break\n","\n","        # remove the initial <SOS> token and pad sequences to the same length\n","        target_seq = target_seq[:, 1:]\n","        max_length = target_seq.size(1)\n","        target_seq = torch.nn.functional.pad(target_seq,\n","            (0, self.max_seq_length - max_length), value=self.PAD_TOKEN)\n","\n","        return target_seq"],"metadata":{"id":"pl09au7xdMFg","executionInfo":{"status":"ok","timestamp":1713553267235,"user_tz":240,"elapsed":317,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":28,"outputs":[]},{"cell_type":"markdown","source":["## CNN-LSTM Encoder + Speech Transformer Decoder\n","\n","In this section, let's build the first (simplest) Transformer network by combining the CNN-LSTM encoder and our Transformer Decoder."],"metadata":{"id":"_JrVsaskeH1G"}},{"cell_type":"code","source":["class SpeechTransformer(torch.nn.Module):\n","\n","    def __init__(self, input_dim, dec_num_layers, dec_num_heads,\n","                    d_model, d_ff, target_vocab_size, eos_token, sos_token,\n","                    pad_token, enc_dropout, dec_dropout, max_seq_length=512):\n","\n","        super(SpeechTransformer, self).__init__()\n","\n","        # self.encoder    = CNN_LSTM_Encoder(input_dim, 256, enc_dropout)\n","        self.encoder    = HW3P2_encoder(input_size=input_dim)\n","\n","\n","        # self.proj       = torch.nn.Linear(1024, d_model)\n","        self.proj       = torch.nn.Linear(192*4, d_model)\n","\n","        # @NOTE: layernorm here is meant to normalize the dynamic ranges of\n","        #   the cross attention weights such that the model doesn't bias\n","        #   itself to only using the cross attention weights and neglect using\n","        #   the multi-head self attention weights.\n","        #   This is to enfoce the Language Model constraints in the Decoder.\n","        #   This is an experimental change -- you may try it if you want.\n","        self.layernorm  = torch.nn.LayerNorm(d_model)\n","\n","        self.decoder    = Decoder(dec_num_layers, d_model, dec_num_heads, d_ff,\n","                dec_dropout, target_vocab_size, max_seq_length, eos_token, sos_token, pad_token)\n","\n","        # You can experiment with different weight initialization schemes or no initialization here\n","        for p in self.parameters():\n","            if p.dim() > 1:\n","                torch.nn.init.xavier_uniform_(p)\n","\n","    def forward(self, padded_input, input_lengths, padded_target, target_lengths):\n","\n","        # passing through Encoder\n","        encoder_output, encoder_lens = self.encoder(padded_input, input_lengths)\n","\n","        # adding projection layer to change dimension of encoder_outputs to match d_model\n","        encoder_output = self.proj(encoder_output)\n","\n","        # @NOTE: adding layernorm to adjust dynamic ranges of the encoder_outputs\n","        # @NOTE: uncomment this if you choose to use layernorm described in __init__ above\n","        encoder_output = self.layernorm(encoder_output)\n","\n","        # passing Encoder output and Attention masks through Decoder\n","        output, attention_weights = self.decoder(padded_target, encoder_output, encoder_lens)\n","\n","        return output, attention_weights\n","\n","    def recognize(self, inp, inp_len):\n","        \"\"\" sequence-to-sequence greedy search -- decoding one utterence at a time \"\"\"\n","\n","        encoder_outputs, encoder_lens  = self.encoder(inp, inp_len)\n","        encoder_outputs                = self.proj(encoder_outputs)\n","        out                            = self.decoder.recognize_greedy_search(encoder_outputs, encoder_lens)\n","\n","        return out"],"metadata":{"id":"oqiAstL22fBV","executionInfo":{"status":"ok","timestamp":1713553271919,"user_tz":240,"elapsed":155,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":29,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"bPZD3vqdUisj"},"source":["## Model Setup\n"]},{"cell_type":"code","execution_count":30,"metadata":{"id":"p2uX2P9YVnbk","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553275450,"user_tz":240,"elapsed":234,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"dd4555a8-c928-48a7-b611-bc832ff7961f"},"outputs":[{"output_type":"execute_result","data":{"text/plain":["0"]},"metadata":{},"execution_count":30}],"source":["torch.cuda.empty_cache()\n","gc.collect()"]},{"cell_type":"code","execution_count":46,"metadata":{"id":"a9LN0l5VUk_s","scrolled":true,"colab":{"base_uri":"https://localhost:8080/","height":343},"executionInfo":{"status":"error","timestamp":1713554135845,"user_tz":240,"elapsed":159,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"217c3e7b-d378-47a4-edca-37c703ea7994"},"outputs":[{"output_type":"error","ename":"NameError","evalue":"name 'HW3P2_encoder' is not defined","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)","\u001b[0;32m<ipython-input-46-9093bcbabd00>\u001b[0m in \u001b[0;36m<cell line: 3>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;34m''' Please refer to the config file and top sections to fill in the following '''\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m model = SpeechTransformer(\n\u001b[0m\u001b[1;32m      4\u001b[0m \u001b[0minput_dim\u001b[0m                   \u001b[0;34m=\u001b[0m \u001b[0;36m27\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mdec_num_layers\u001b[0m              \u001b[0;34m=\u001b[0m \u001b[0mconfig\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m'dec_num_layers'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m<ipython-input-29-c736d063374f>\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, input_dim, dec_num_layers, dec_num_heads, d_model, d_ff, target_vocab_size, eos_token, sos_token, pad_token, enc_dropout, dec_dropout, max_seq_length)\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      9\u001b[0m         \u001b[0;31m# self.encoder    = CNN_LSTM_Encoder(input_dim, 256, enc_dropout)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 10\u001b[0;31m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mencoder\u001b[0m    \u001b[0;34m=\u001b[0m \u001b[0mHW3P2_encoder\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minput_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minput_dim\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     11\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     12\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mNameError\u001b[0m: name 'HW3P2_encoder' is not defined"]}],"source":["''' Please refer to the config file and top sections to fill in the following '''\n","\n","model = SpeechTransformer(\n","input_dim                   = 27,\n","dec_num_layers              = config['dec_num_layers'],\n","dec_num_heads               = config['dec_num_heads'],\n","\n","d_model                     = config['d_model'],\n","d_ff                        = config['d_ff'],\n","\n","target_vocab_size           = len(VOCAB),\n","eos_token                   = EOS_TOKEN,\n","sos_token                   = SOS_TOKEN,\n","pad_token                   = PAD_TOKEN,\n","\n","enc_dropout                 = config['enc_dropout'],\n","dec_dropout                 = config['dec_dropout'],\n","\n","# decrease to a small number if you are just trying to implement the network\n","max_seq_length              = 550 , # Max sequence length for transcripts. Check data verification.\n",").to(device)\n","\n","def num_parameters(mode):\n","    total_params = sum(p.numel() for p in model.parameters() if p.requires_grad)\n","    return total_params / 1E6\n","\n","para = num_parameters(model)\n","print(\"#\"*10)\n","print(f\"Model Parameters:\\n {para}\")\n","print(\"#\"*10)\n","\n","print(model)"]},{"cell_type":"markdown","source":["### Loading the CNN-LSTM Checkpoint\n","\n","---\n","For ease of training we use a pretrained LSTM -- load the .pth file.\n","\n","Download the weights from here - https://drive.google.com/file/d/1wRcHwCvxksxiPv6WQykOkY9i4bip-ooa/view?usp=sharing"],"metadata":{"id":"oRUBrF1R78pm"}},{"cell_type":"code","source":["# model.encoder.load_state_dict(torch.load(\"/content/pblstm_fast_training_weights.pth\"))\n"],"metadata":{"id":"oVhuszyX8D13","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713394769385,"user_tz":240,"elapsed":241,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"5a7edda9-345f-485d-c851-0fff42a30399"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<All keys matched successfully>"]},"metadata":{},"execution_count":242}]},{"cell_type":"markdown","metadata":{"id":"23DMfXsaU6kj"},"source":["## Loss, Optimizer, and Scheduler Definitions"]},{"cell_type":"code","source":["loss_func   = nn.CrossEntropyLoss(ignore_index = PAD_TOKEN)\n","scaler      = torch.cuda.amp.GradScaler()"],"metadata":{"id":"zGxaiyESLpwI","executionInfo":{"status":"ok","timestamp":1713554115071,"user_tz":240,"elapsed":143,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":43,"outputs":[]},{"cell_type":"code","source":["''' defining optimizer '''\n","\n","if config[\"optimizer\"] == \"SGD\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.SGD(model.parameters(),\n","                              lr=config[\"learning_rate\"],\n","                              momentum=config[\"momentum\"],\n","                              weight_decay=1E-4,\n","                              nesterov=config[\"nesterov\"])\n","\n","elif config[\"optimizer\"] == \"Adam\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.Adam(model.parameters(),\n","                               lr=float(config[\"learning_rate\"]),\n","                               weight_decay=1e-4)\n","\n","elif config[\"optimizer\"] == \"AdamW\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.AdamW(model.parameters(),\n","                                lr=float(config[\"learning_rate\"]),\n","                                weight_decay=0.01)\n","\n","''' defining scheduler '''\n","\n","if config[\"scheduler\"] == \"ReduceLR\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,\n","                factor=config[\"factor\"], patience=config[\"patience\"], min_lr=1E-8, verbose=True)\n","\n","elif config[\"scheduler\"] == \"CosineAnnealing\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer,\n","                T_max = config[\"epochs\"], eta_min=1E-8)"],"metadata":{"id":"W-jth-6HLPLr"},"execution_count":null,"outputs":[]},{"cell_type":"code","execution_count":47,"metadata":{"id":"216ukmHbU-ol","executionInfo":{"status":"ok","timestamp":1713554235540,"user_tz":240,"elapsed":166,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"outputs":[],"source":["def save_model(model, optimizer, scheduler, metric, epoch, path):\n","    torch.save(\n","        {\"model_state_dict\"         : model.state_dict(),\n","         \"optimizer_state_dict\"     : optimizer.state_dict(),\n","         \"scheduler_state_dict\"     : scheduler.state_dict() if scheduler is not None else {},\n","         metric[0]                  : metric[1],\n","         \"epoch\"                    : epoch},\n","         path\n","    )\n","\n","def load_model(path, model, metric= \"valid_acc\", optimizer= None, scheduler= None):\n","\n","    checkpoint = torch.load(path)\n","    model.load_state_dict(checkpoint[\"model_state_dict\"])\n","\n","    if optimizer != None:\n","        optimizer.load_state_dict(checkpoint[\"optimizer_state_dict\"])\n","    if scheduler != None:\n","        scheduler.load_state_dict(checkpoint[\"scheduler_state_dict\"])\n","\n","    epoch   = checkpoint[\"epoch\"]\n","    metric  = checkpoint[metric]\n","\n","    return [model, optimizer, scheduler, epoch, metric]"]},{"cell_type":"markdown","source":["## Training and Validation Modules"],"metadata":{"id":"lWUHzWOEs_ky"}},{"cell_type":"markdown","source":["### Levenshtein Distance"],"metadata":{"id":"iWWO5vM4_RGC"}},{"cell_type":"code","source":["''' utility function which takes a sequence of indices and converts them to a list of characters '''\n","def indices_to_chars(indices, vocab):\n","    tokens = []\n","    for i in indices:   # looping through all indices\n","\n","        if int(i) == SOS_TOKEN:     # If SOS is encountered, don't add it to the final list\n","            continue\n","        elif int(i) == EOS_TOKEN:   # If EOS is encountered, stop the decoding process\n","            break\n","        else:\n","            tokens.append(vocab[i])\n","\n","    return tokens"],"metadata":{"id":"QXNHNbnh5w5z","executionInfo":{"status":"ok","timestamp":1713554239022,"user_tz":240,"elapsed":174,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":48,"outputs":[]},{"cell_type":"code","source":["''' utility function for Levenshtein Distantce quantification '''\n","def calc_edit_distance(predictions, y, y_len, vocab=VOCAB, print_example=False):\n","\n","    dist = 0.0\n","    batch_size, seq_len = predictions.shape\n","\n","    for batch_idx in range(batch_size):\n","\n","        y_sliced    = indices_to_chars(y[batch_idx, 0 : y_len[batch_idx]], vocab)\n","        pred_sliced = indices_to_chars(predictions[batch_idx], vocab)\n","\n","        # strings - when you are using characters from the SpeechDataset\n","        y_string    = \"\".join(y_sliced)\n","        pred_string = \"\".join(pred_sliced)\n","\n","        dist        += Levenshtein.distance(pred_string, y_string)\n","\n","    if print_example:\n","        print(\"\\nGround Truth : \", y_string)\n","        print(\"Prediction   : \", pred_string)\n","\n","    dist /= batch_size\n","\n","    return dist"],"metadata":{"id":"Lbv5sJSIFB8b","executionInfo":{"status":"ok","timestamp":1713554239578,"user_tz":240,"elapsed":2,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":49,"outputs":[]},{"cell_type":"markdown","source":["### Training and Validation Functions"],"metadata":{"id":"b0RXXoJk_WU5"}},{"cell_type":"code","source":["def train_model(model, train_loader, optimizer):\n","\n","    model.train()\n","    batch_bar = tqdm(total=len(train_loader), dynamic_ncols=True, leave=False, position=0, desc=\"Train\")\n","\n","    total_loss          = 0\n","    running_loss        = 0.0\n","    running_perplexity  = 0.0\n","\n","    for i, (inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths) in enumerate(train_loader):\n","\n","        optimizer.zero_grad()\n","\n","        inputs          = inputs.to(device)\n","        targets_shifted = targets_shifted.to(device)\n","        targets_golden  = targets_golden.to(device)\n","\n","        with torch.cuda.amp.autocast():\n","            # passing the minibatch through the model\n","            raw_predictions, attention_weights = model(inputs, inputs_lengths, targets_shifted, targets_lengths)\n","\n","            padding_mask = torch.logical_not(torch.eq(targets_shifted, PAD_TOKEN))\n","\n","            # cast the mask to float32\n","            padding_mask = padding_mask.float()\n","            loss = loss_func(raw_predictions.transpose(1,2), targets_golden)*padding_mask\n","            loss = loss.sum() / padding_mask.sum()\n","\n","        scaler.scale(loss).backward()   # This is a replacement for loss.backward()\n","        scaler.step(optimizer)          # This is a replacement for optimizer.step()\n","        scaler.update()                 # This is something added just for FP16\n","\n","        running_loss        += float(loss.item())\n","        perplexity          = torch.exp(loss)\n","        running_perplexity  += perplexity.item()\n","\n","        # online training monitoring\n","        batch_bar.set_postfix(\n","            loss = \"{:.04f}\".format(float(running_loss / (i + 1))),\n","            perplexity = \"{:.04f}\".format(float(running_perplexity / (i + 1)))\n","        )\n","\n","        batch_bar.update()\n","\n","        del inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths\n","        torch.cuda.empty_cache()\n","\n","    running_loss        = float(running_loss / len(train_loader))\n","    running_perplexity  = float(running_perplexity / len(train_loader))\n","\n","    batch_bar.close()\n","\n","    return running_loss, running_perplexity, attention_weights"],"metadata":{"id":"VJHSLMCmp8l5","executionInfo":{"status":"ok","timestamp":1713554241820,"user_tz":240,"elapsed":157,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":50,"outputs":[]},{"cell_type":"code","source":["def validate_fast(model, dataloader):\n","    model.eval()\n","\n","    # progress bar\n","    batch_bar = tqdm(total=len(dataloader), dynamic_ncols=True, leave=False, position=0, desc=\"Val\", ncols=5)\n","\n","    running_distance = 0.0\n","\n","    for i, (inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths) in enumerate(dataloader):\n","\n","        inputs  = inputs.to(device)\n","        targets_golden = targets_golden.to(device)\n","\n","        with torch.inference_mode():\n","            greedy_predictions = model.recognize(inputs, inputs_lengths)\n","\n","        # calculating Levenshtein Distance\n","        # @NOTE: modify the print_example to print more or less validation examples\n","        running_distance += calc_edit_distance(greedy_predictions, targets_golden, targets_lengths, VOCAB, print_example=True)\n","\n","        # online validation distance monitoring\n","        batch_bar.set_postfix(\n","            running_distance = \"{:.04f}\".format(float(running_distance / (i + 1)))\n","        )\n","\n","        batch_bar.update()\n","\n","        del inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths\n","        torch.cuda.empty_cache()\n","\n","        if i==4: break      # validating only upon first five batches\n","\n","    batch_bar.close()\n","    running_distance /= 5\n","\n","    return running_distance"],"metadata":{"id":"tuYsBMoQ30US","executionInfo":{"status":"ok","timestamp":1713554243584,"user_tz":240,"elapsed":150,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":51,"outputs":[]},{"cell_type":"code","source":["def validate_full(model, dataloader):\n","    model.eval()\n","\n","    # progress bar\n","    batch_bar = tqdm(total=len(dataloader), dynamic_ncols=True, leave=False, position=0, desc=\"Val\", ncols=5)\n","\n","    running_distance = 0.0\n","\n","    for i, (inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths) in enumerate(dataloader):\n","\n","        inputs  = inputs.to(device)\n","        targets_golden = targets_golden.to(device)\n","\n","        with torch.inference_mode():\n","            greedy_predictions = model.recognize(inputs, inputs_lengths)\n","\n","        # calculating Levenshtein Distance\n","        # @NOTE: modify the print_example to print more or less validation examples\n","        running_distance += calc_edit_distance(greedy_predictions, targets_golden, targets_lengths, VOCAB, print_example=True)\n","\n","        # online validation distance monitoring\n","        batch_bar.set_postfix(\n","            running_distance = \"{:.04f}\".format(float(running_distance / (i + 1)))\n","        )\n","\n","        batch_bar.update()\n","\n","        del inputs, targets_shifted, targets_golden, inputs_lengths, targets_lengths\n","        torch.cuda.empty_cache()\n","\n","\n","    batch_bar.close()\n","    running_distance /= len(dataloader)\n","\n","    return running_distance"],"metadata":{"id":"_fvRunxh_36z","executionInfo":{"status":"ok","timestamp":1713554245647,"user_tz":240,"elapsed":135,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":52,"outputs":[]},{"cell_type":"markdown","source":["## WandB"],"metadata":{"id":"gP7QK6rD_mT9"}},{"cell_type":"code","source":["# using WandB? resume training?\n","\n","USE_WANDB = True\n","RESUME_LOGGING = False\n","\n","\n","# creating your WandB run\n","run_name = \"{}_Transformer_ENC-{}/{}_DEC-{}/{}_{}_{}_{}_{}_Lr:{}\".format(\n","    config[\"Name\"],\n","    config[\"enc_num_layers\"],       # only used in Part II with the Transformer Encoder\n","    config[\"enc_num_heads\"],        # only used in Part II with the Transformer Encoder\n","    config[\"dec_num_layers\"],\n","    config[\"dec_num_heads\"],\n","    config[\"d_model\"],\n","    config[\"d_ff\"],\n","    config[\"optimizer\"],\n","    config[\"scheduler\"],\n","    config['learning_rate'])\n","\n","if USE_WANDB:\n","\n","    wandb.login(key=\"b9fec08c02d03c41bdf5fe1f5268589277270fbe\", relogin=True) # TODO enter your key here\n","\n","    if RESUME_LOGGING:\n","        run_id = \"f0bqyz1j\"\n","        run = wandb.init(\n","            id     = run_id,        ### Insert specific run id here if you want to resume a previous run\n","            resume = True,          ### You need this to resume previous runs, but comment out reinit=True when using this\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","        )\n","\n","    else:\n","        run = wandb.init(\n","            name    = run_name,     ### Wandb creates random run names if you skip this field, we recommend you give useful names\n","            reinit  = True,         ### Allows reinitalizing runs when you re-run this cell\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","            config  = config        ### Wandb Config for your run\n","        )\n","\n","        ### Save your model architecture as a string with str(model)\n","        model_arch  = str(model)\n","\n","        ### Save it in a txt file\n","        arch_file   = open(\"model_arch.txt\", \"w\")\n","        file_write  = arch_file.write(model_arch)\n","        arch_file.close()\n","\n","        ### Log it in your wandb run with wandb.save()\n","        wandb.save(\"model_arch.txt\")"],"metadata":{"id":"CUz-15ws79RI","colab":{"base_uri":"https://localhost:8080/","height":227,"referenced_widgets":["472fa7722fc74c09bff6ad00d7cc2451","193f09f303d547328c03c9b7f82cbbb2","c3463b441fda4798b0b10cead972933c","5c9a63aad0fa4bc28af3661b0caf4366","fab7f97206ad4e41b0fc89d04ee4148e","9de0938b74ee4942b39104d04f15fc08","d2bef54bcb974f68b55f273b9e5cb54c","6fe6c18183a04ee8b531b5f0069293ae"]},"executionInfo":{"status":"ok","timestamp":1713442426070,"user_tz":240,"elapsed":6586,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"a106e133-efc4-48c1-ef56-43d760b4245c"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[33mWARNING\u001b[0m Calling wandb.login() after wandb.init() has no effect.\n"]},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Finishing last run (ID:x10oa58p) before initializing another..."]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["VBox(children=(Label(value='0.005 MB of 0.005 MB uploaded\\r'), FloatProgress(value=1.0, max=1.0)))"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"472fa7722fc74c09bff6ad00d7cc2451"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run <strong style=\"color:#cdcd00\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_CosineAnnealing_Lr:1E-4</strong> at: <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24-2/runs/x10oa58p/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24-2/runs/x10oa58p/workspace</a><br/>Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 1 other file(s)"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Find logs at: <code>./wandb/run-20240418_121326-x10oa58p/logs</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Successfully finished last run (ID:x10oa58p). Initializing new run:<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["wandb version 0.16.6 is available!  To upgrade, please run:\n"," $ pip install wandb --upgrade"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Tracking run with wandb version 0.16.5"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Run data is saved locally in <code>/content/wandb/run-20240418_121339-6f6xoc27</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Syncing run <strong><a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/6f6xoc27/workspace' target=\"_blank\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_CosineAnnealing_Lr:1E-4</a></strong> to <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View project at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24</a>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/6f6xoc27/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/6f6xoc27/workspace</a>"]},"metadata":{}}]},{"cell_type":"markdown","source":["## Experiments"],"metadata":{"id":"4vUMtM_dVyrG"}},{"cell_type":"code","source":["e                   = 0\n","best_loss           = 20\n","\n","checkpoint_root = os.path.join(os.getcwd(), \"checkpoints-basic-transformer\")\n","os.makedirs(checkpoint_root, exist_ok=True)\n","wandb.watch(model, log=\"all\")\n","\n","checkpoint_best_loss_model_filename     = 'checkpoint-best-loss-model.pth'\n","checkpoint_last_epoch_filename          = 'checkpoint-epoch-.pth'\n","best_loss_model_path                    = os.path.join(checkpoint_root, checkpoint_best_loss_model_filename)\n","epoch_model_path                        = os.path.join(checkpoint_root, checkpoint_last_epoch_filename)\n","if RESUME_LOGGING:\n","    # change if you want to load best test model accordingly\n","    # checkpoint = torch.load(wandb.restore(checkpoint_best_loss_model_filename, run_path=\"\"+run_id).name)\n","    model, optimizer, scheduler, epoch, metric = load_model(epoch_model_path, model, metric= \"train_loss\", optimizer= optimizer, scheduler= scheduler)\n","    # model.load_state_dict(checkpoint['model_state_dict'])\n","    # optimizer.load_state_dict(checkpoint['optimizer_state_dict'])\n","    e = epoch\n","\n","    print(\"Resuming from epoch {}\".format(epoch+1))\n","    print(\"Epochs left: \", config['epochs']-epoch)\n","    print(\"Optimizer: \\n\", optimizer)\n","    print(\"Scheduler: \\n\", scheduler.state_dict())\n","    print(\"metric: \\n\", metric)\n","    scheduler.T_max = 75\n","    print(\"Scheduler: \\n\", scheduler.state_dict())\n","\n","torch.cuda.empty_cache()\n","gc.collect()\n","\n","epochs = config[\"epochs\"]\n","for epoch in range(e, epochs):\n","\n","    print(\"\\nEpoch {}/{}\".format(epoch+1, config[\"epochs\"]))\n","\n","    curr_lr = float(optimizer.param_groups[0][\"lr\"])\n","\n","    train_loss, train_perplexity, attention_weights = train_model(model, train_loader, optimizer)\n","\n","    print(\"\\nEpoch {}/{}: \\nTrain Loss {:.04f}\\t Train Perplexity {:.04f}\\t Learning Rate {:.09f}\".format(\n","        epoch + 1, config[\"epochs\"], train_loss, train_perplexity, curr_lr))\n","\n","    if (epoch % 2 == 0):    # validate every 2 epochs to speed up training\n","        levenshtein_distance = validate_fast(model, val_loader)\n","        print(\"Levenshtein Distance {:.04f}\".format(levenshtein_distance))\n","\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr,\n","                \"val_distance\"      : levenshtein_distance})\n","\n","    else:\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr})\n","\n","    # plotting the encoder-nearest and decoder-nearest attention weights\n","    attention_keys = list(attention_weights.keys())\n","\n","    attention_weights_decoder_self       = attention_weights[attention_keys[0]][0].cpu().detach().numpy()\n","    attention_weights_decoder_cross      = attention_weights[attention_keys[-1]][0].cpu().detach().numpy()\n","\n","    # saving the cross-attention weights\n","    save_attention_plot(attention_weights_decoder_cross, epoch+100)\n","\n","    # plot_attention_weights((attention_weights[attention_keys[0]][0]).cpu().detach().numpy())\n","    # plot_attention_weights(attention_weights[attention_keys[-1]][0].cpu().detach().numpy())\n","\n","    if config[\"scheduler\"] == \"ReduceLR\":\n","        scheduler.step(levenshtein_distance)\n","    else:\n","        scheduler.step()\n","\n","    ### Highly Recommended: Save checkpoint in drive and/or wandb if accuracy is better than your current best\n","    epoch_model_path = os.path.join(checkpoint_root, (checkpoint_last_epoch_filename + str(epoch) + '.pth'))\n","    save_model(model, optimizer, scheduler, ['train_loss', train_loss], epoch, epoch_model_path)\n","    ## wandb.save(epoch_model_path) ## Can't save on wandb for all epochs, may blow up storage\n","\n","    print(\"Saved epoch model\")\n","\n","    if best_loss >= train_loss:\n","        best_loss = train_loss\n","        save_model(model, optimizer, scheduler, ['train_loss', train_loss], epoch, best_loss_model_path)\n","        # wandb.save(best_loss_model_path)\n","        print(\"Saved best training model\")\n","\n","### Finish your wandb run\n","# run.finish()"],"metadata":{"id":"a1LD0uyPEHTI","colab":{"base_uri":"https://localhost:8080/","height":1000},"outputId":"88e3a0f0-be59-4160-e98c-3df7dfc4950c","executionInfo":{"status":"error","timestamp":1713457814854,"user_tz":240,"elapsed":15164292,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["\n","Epoch 1/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 1/50: \n","Train Loss 2.7904\t Train Perplexity 16.8330\t Learning Rate 0.000100000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:17<12:34, 17.95s/it, running_distance=179.8594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  HE W HE HE HER HER HER HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:32<10:43, 15.69s/it, running_distance=181.2344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  HE W HE HE HER HER HER HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [00:46<10:12, 15.32s/it, running_distance=181.3229]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  HE W HE HE HER HERER HER HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:04<10:26, 16.06s/it, running_distance=180.1953]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  HE W HE HE HER HER HER HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [01:21<10:26, 16.49s/it, running_distance=180.0563]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  HE W HE HE HER HERER HER HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE HE\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 180.0563\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 2/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 2/50: \n","Train Loss 2.3797\t Train Perplexity 10.8161\t Learning Rate 0.000099901\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 3/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 3/50: \n","Train Loss 2.2132\t Train Perplexity 9.1533\t Learning Rate 0.000099606\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:21<14:46, 21.10s/it, running_distance=195.5312]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE TO TO TO TO TO TO TO AND AND AND AND AND ALD AND ALD TO TO AND AND TO TO AND TO TO TO TO TO AND AND AND AND TO TO TO AND AND AND AT AND AT ALD AT ALD AND AT AND TO TO AND TO AND ALD AT AND TO TO AND ALD AND ALD AND AND AND AS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:38<12:51, 18.81s/it, running_distance=195.0469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE TO TO TO TO TO TO TO ALD AND AND AND AND ALD AND ALD TO TO AND AND TO TO AND TO TO TO TO TO ALD AND AND AND TO TO TO AND AND AND AT AND AT ALD AT ALD AND AT AND TO TO AS ALD AND ALD AT AND TO TO TO TO TO TO ALD AND AND AND AS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [00:56<12:15, 18.39s/it, running_distance=194.6719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE TO TO TO TO TO TO TO AND AND AND AND AND ALD AND AND TO TO AND AND AT AND AND TO AND AT TO AND TO AND ALD AT AND AND AND AND AND AT AND AT ALD AT ALD AND AT AT AT AT AND TO AND ALD AT AT AT ALD AT AT AND ALD AND AND AND AND AT\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:16<12:28, 19.20s/it, running_distance=193.7734]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND TO TO TO TO TO TO TO TO TO AND AND AND AND AND AND AND AND TO AND AND AND TO AND TO TO TO TO TO AND AND AND AND TO TO TO TO TO AND AND TO AND ALD AND ALD AND ALD AND TO AND TO TO TO TO TO AND TO TO TO TO AND AND AND AND AND\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [01:37<12:28, 19.71s/it, running_distance=193.3281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE TO TO TO TO TO TO TO AND AND AND AND AND ALD AND AND TO TO AND AND AT AND AND TO TO TO TO TO AND AND AND AND TO TO TO AND AND AND AND AND ALD AT ALD AND ALD AT AND AT AND TO AND AND AT AND TO TO AT ALD AND ALD AND AND AND AND AT\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 193.3281\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 4/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 4/50: \n","Train Loss 2.0967\t Train Perplexity 8.1433\t Learning Rate 0.000099114\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 5/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 5/50: \n","Train Loss 2.0187\t Train Perplexity 7.5306\t Learning Rate 0.000098429\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:30<21:36, 30.87s/it, running_distance=255.7812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE AND THE AND THE THE AND THE AND AND AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE THE AND THE AND THE AND THE AND THE AND AND THE AND THE AND THE AND AND AND THE AND THE THE AND TO THE AND TERED TO TED TO THE ACTERE THE AND THE ACE ASE T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:56<18:53, 27.65s/it, running_distance=253.3359]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE AND THE AND THE AND AND THE AND AND AND THE AND THE AND THE AND AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND AND THE AND THE AND THE AND AND AND THE AND AND THE AND TO THE AND TERED TO TED TO THE ACTED THE AND THE THE ASE T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:58, 26.97s/it, running_distance=249.8698]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE AND THE AND THE AND AND THE AND AND AND THE AND AND AND THE AND AND THE AND THE AND AND THE AND THE AND TO THE AND THE AND AND THE AND THE AND THE AND AND TO THE AND THE AND THE AND AND TO THE AND THE AND AND AND AND THE AND AND TO TO TO THE AND TERED TO TED TO TO THE ACTE THE AND THE ACEASE AS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:51<18:08, 27.92s/it, running_distance=248.2578]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE AND THE AND THE AND AND THE AND AND AND THE AND THE AND THE AND AND THE AND THE AND AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND THE AND AND AND AND TO THE AND AND THE AND THE AND TO THE AND TERED TO TO TE AND THE ACTEAT\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [02:21<18:10, 28.69s/it, running_distance=247.8375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE AND AND THE AND THE AND AND AND AND AND THE AND AND AND THE AND AND AND THE AND AND AND THE AND THE AND THE AND AND AND AND AND THE AND THE AND THE AND AND THE AND THE AND THE AND AND THE AND AND AND AND AND AND AND AND THE AND THE AND TO THE AND TATED TED TED AND AND ACTEAS\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 247.8375\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 6/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 6/50: \n","Train Loss 1.9646\t Train Perplexity 7.1330\t Learning Rate 0.000097553\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 7/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 7/50: \n","Train Loss 1.9233\t Train Perplexity 6.8448\t Learning Rate 0.000096489\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:33<23:19, 33.32s/it, running_distance=258.0000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE ASTERESS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<19:42, 28.85s/it, running_distance=254.4297]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE AND THE THE THE THE THE THE THE THE THE ASTER THE THE THE THE THE AND THE AND THE THE THE THE THE THE THE THE THE THE T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:27<19:07, 28.69s/it, running_distance=251.9010]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE AND THE THE THE THE THE THE THE THE THE ASTEREST THE THE THE THE AND THE AND THE THE TEARD THE THE THE THE THE THE THE T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:59<19:31, 30.03s/it, running_distance=248.2422]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE AND THE THE THE THE THE THE THE THE THE AND THE THE THE THE THE AND THE THE AND THE THE THE THE THE THE THE THE THE THE T\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [02:31<19:31, 30.84s/it, running_distance=247.1594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE THE AND THE THE THE THE THE THE THE THE THE ASTEREST THE THE THE THE AND THE AND THE TEARD THE THE THE THE THE THE THE THE T\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 247.1594\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 8/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 8/50: \n","Train Loss 1.8937\t Train Perplexity 6.6446\t Learning Rate 0.000095242\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 9/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 9/50: \n","Train Loss 1.8677\t Train Perplexity 6.4743\t Learning Rate 0.000093816\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:43, 83.90s/it, running_distance=447.9219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE STER AS AND THE THE AND THE AS AND AS AND THE AND THE AND THE AND THE THE THE THE AND THE AS AND THE AS AND THE AND THE AS AND THE AND THE THE THE AS AND AS AND AND THE AS AND THE AND THE AND AS AND THE AS AND AS AND AS AND THE STAND THE AND THE SEED TER THE THE AND THE TEAND AND THE TEED THE T T T T T T T T T T T T T T TEREASPRATHATASTASPASPASPANOWASOWASOWASINO TSISINOWAS T T T T T AS AS ASTASTSSASINEEEEEEASIST TSIS TREREREREASTAS TOWERERERERIST T T T T T TS T TASISIST T T T T T TASITATATTTOWANOWANOWANOWASTSTHASTTTOTOTOWATTTTTTTTASTTTT\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:36<52:39, 77.07s/it, running_distance=444.4766]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE SAID THE THE AS AND AND THE AS AND AS AND THE AND THE AND THE AND THE T THE THE AND THE ST THE AS AND THE AND THE AND THE AS AND AS AND THE THE AND THE THE AS AND THE AS AND AS AND AND THE AS AND AS AND AS AND AS AND AND THE STER AS AND THE AND THE T T THE THE AND THE THE T THE TO THE AS THE AND T T T T T T T T T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:50<50:36, 75.91s/it, running_distance=432.6615]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE ST THE THE AS AND THE AND AS AND AS AND THE AND AND AND AND THE AND THE AND THE AND THE AND THE AND AS AND THE THE AND THE AND THE AND THE THE AND THE AND AS AND THE AS AND AS AND AS AND AND AS AND AS AND AS AND AS AND THE SEED THE AS AND THE AS AND TER THE THE AND THE THE AND AND THE AS AND T T T TEEEE T T T TTHEAND T T TEASITATATHATASTASPASIST TTOWASEASISIN ASISISISIN AST T T ASIN AS AS ASTASTSSSISIN AS ASIN T T T T T T T TASTAS T T T T T T T T T T T T T T TASISISIST T T T T T TITATATATTTOWANOWAS T TOWANOWASTOTTTTOTOTOWATTTTTTTTTTTTTT\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:12<50:48, 78.16s/it, running_distance=433.1211]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE SEE THE AND THE THE AND AS AND AND AS AND AND AND AND AND AS AND AND THE AND THE AND AND AND AND AND AS AND AS AND THE AS AND AND AND AND AND THE AND AND AS AND AND AS AND AND AND AND AND AND AS AND AS AND THE AS AND AND THE AND THE AND THE AND THE TER TER THE AND THE THE T THE AS AND AS AND TO T T TE T T T T THEAN T T T THE TASATHATHASTASTASIND TINOWASANOWASIN T TANOWASIN AN T T AS AS A AST TISTSASIN AS ASIN T TASIS TANOWASTASTAS TOWE T T T T T T T T T T TS T T TAN T T T T T T T TITATHATTTTOWANOWANOWANOWANOTHASTTTTTOTOTTTTTTTTTTTTTTTTT\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:34<50:21, 79.52s/it, running_distance=435.4781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE ST THE AS AND THE THE AND AS AND AS AND THE AND AND AND AND THE AND THE AND THE AND THE AND THE AND AS AND THE AND THE THE AND AS AND THE AND THE AND THE AS AND AND AS AND AS AND AND THE AND AS AND AS AND AS AND AS AND THE SEED THE AS AND THE AND TAND TER THE AND TO THE TEAND AND THE TEED THE T T T T T T T T T T T T T TO TEREASPREATHASTASPASPASPANOWASOWASOWAN T TSISISIN AST T T T AS AS AS ASTASTSSASINEEEEEEASISISTASIS TRERISTASTAS TOWERERERERIST T T T T T T TS TASISISIST T T T T TISITATATTTTOWANOWANOWANOWANOTHATOTTTOTOTOWATTTTTTTTASTTTT\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 435.4781\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 10/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 10/50: \n","Train Loss 1.8466\t Train Perplexity 6.3391\t Learning Rate 0.000092217\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 11/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 11/50: \n","Train Loss 1.8279\t Train Perplexity 6.2216\t Learning Rate 0.000090452\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:36<25:34, 36.54s/it, running_distance=267.0312]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE WAS T THE STER THE THE ST THE SAID THE STRACE THE THE STER AND THE ST THE ST THE THE ST THE ST T THE ST AND THE ST T THE ST T T T THE ST THE SEE ST THE ST THE ST THE ST AS AND THE ST THE STER THE STER AS AND THE SAID THE THE SEE STER THE STER THE STER THE SEE THE SEE SEE SCERCESS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:06<22:31, 32.96s/it, running_distance=267.1797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  THE WAS T THE T THE THE T T T T THE SAID THE SAID THE THE T THE THE SAID T THE T THE T THE SAID T T T T T T AND THE T T T AND THE T T T T T T THE T T T T T T T THE S AND THE S AND T T T T THE T AND THE AS AND THE SAID THE THE T T T THE SEE SAID THE SAID THE T T T THE SEE THE SEE THE T THE T THE STEE AS T T TEEEEE T T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:21<34:31, 51.79s/it, running_distance=265.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE WAS T THE THE SAID THE T THE SAID THE ST AND THE THE T THE THE AS AND THE T THE THE ST THE ST T T THE T AND THE T T T T T T T T T THE T THE T THE T THE SEE THE ST THE ST AND THE SAID THE T T T T T T THE ST T T T THE THE THE SEE STER THE SAID THE STER THE T T THE THE THE SEE THE T THE T THE STEE AND T T TEEEEE T T T T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:42<41:15, 63.48s/it, running_distance=264.3945]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  THE WAS T THE T T THE THE T T T THE SAID THE SAID THE THE T THE THE AND T T T T THE T T THE ST T T T T T T T AND THE T T T T T T T T T T T T THE T T T T THE T THE S AND THE S AND THE SAID THE T T T T T THE ST T T T THE SAID THE T T THE SEE SAID THE SAID THE T T T THE SEE THE SEE THE T THE T THE STEE AS T T TEEEEE T T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [05:04<44:25, 70.14s/it, running_distance=264.6812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  THE WAS T THE ST THE STRACE AND THE SAID THE SAID THE THE STER AND THE ST THE ST THE THE ST THE ST T THE ST AND THE T T THE S AND T T THE ST THE THE ST THE ST THE ST THE T THE ST T T T THE ST THE ST T T THE STRAND THE SAID THE THE STER THE STER THE STAND THE SEE THE SEE SEE SEE THE T THE T THE STEE THE T T TEEEEAST T T\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 264.6812\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 12/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 12/50: \n","Train Loss 1.8120\t Train Perplexity 6.1237\t Learning Rate 0.000088527\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 13/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 13/50: \n","Train Loss 1.7983\t Train Perplexity 6.0401\t Learning Rate 0.000086450\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:14, 83.19s/it, running_distance=284.9062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE SAID THE SAID THE THE THE THE SAID THE SAID THE THE AND THE AND THE THE THE THE THE THE THE AS THE AND THE THE THE THE THE THE THE THE THE THE THE AND THE STAND THE AND THE THE AND THE THE SAID THE SAID THE SAID THE THE SAID THE STAND THE STAND THE THE THE THE AND THE THE THE THE THE THE THE AND T TEREAND T THAND T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:34<52:11, 76.39s/it, running_distance=289.1484]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE SAID THE AND THE THE THE THE THE THE THE THE THE THE AND THE AND THE THE THE THE AND THE THE AS THE AND THE THE THE THE THE THE AND THE THE THE AND THE THE THE THE AND THE THE AND THE THE THE THE AND THE AND THE THE THE THE SAID THE THE SAID THE SAID THE THE THE AND THE THE THE THE THE THE AS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:48<50:12, 75.32s/it, running_distance=279.6667]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE SAID THE AND THE THE THE THE THE THE THE AND THE THE AND THE AND THE THE THE THE AND THE THE AS THE AND THE THE THE THE THE THE AND THE THE THE AND THE THE THE THE AND THE THE AND THE THE THE THE AND THE AND THE THE THE THE SAID THE THE SAID THE SAID THE THE THE AND THE THE THE THE THE THE AS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:10<50:27, 77.64s/it, running_distance=278.4688]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE SAID THE AND THE THE THE THE THE THE THE AND THE THE AND THE AND THE THE THE THE AND THE THE AS THE AND THE THE THE THE THE THE AND THE THE THE AND THE THE T THE THE AND THE AND THE THE THE THE AND THE AND THE AND THE THE THE SAID THE AND THE SAID THE THE THE AND THE AND THE THE THE THE AS THE AS T TEE T T T T T\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:31<49:58, 78.90s/it, running_distance=281.0750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE SAID THE AND THE THE THE THE THE THE THE AND THE THE AND THE AND THE THE THE THE AND THE THE AS THE AND THE THE THE THE THE THE AND THE THE THE AND THE THE SAID THE AND THE AND THE AND THE THE THE SAID THE SAID THE THE THE SAID THE THE SAID THE STRAND THE THE AND THE THE THE THE THE THE THE AND T T TEEEAND THE T\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 281.0750\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 14/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 14/50: \n","Train Loss 1.7852\t Train Perplexity 5.9618\t Learning Rate 0.000084229\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 15/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 15/50: \n","Train Loss 1.7708\t Train Perplexity 5.8763\t Learning Rate 0.000081873\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:28, 83.53s/it, running_distance=238.3750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S THE S THE STREAT THE SPERSS AND THE SEEMBER AND THE SEEMBER AND THE S THE THE S THE SEEMBER AND THE SAID THE S THE S THE S THE S THE SAID AND THE S THE S THE S THE S TRAND THE S THE S THE S AND THE S THE S THE S THE S THE S THE SAID THE S THE S TRACE THE THE S THE SEEMBLE AND THE THE STERES\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:35<52:22, 76.65s/it, running_distance=234.1641]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE SAID AND THE THE THE THE THE SEEMBER AND THE THE SEE THE SAID AND THE S THE THE SAID AND THE SAID AND THE SEE THE SAID THE THE SAID AND THE SAID THE S THE S THE S T THE T THE SAID AND THE S AND THE SAID AND THE SAID THE S THE SAID THE S THE S AND THE STER\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:06<37:17, 55.93s/it, running_distance=230.8594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE STREAT THE SEEMBER AND THE SEEMBER AND THE SAID AND THE S THE SEEMBER AND THE SEEMBER AND THE SEEMBER AND THE SAID AND THE SAID AND THE S THE S THE S THE S THE T THE S AND THE S AND THE S THE S AND THE S THE S THE SAID THE S THE S AND THE S THE SEEMBER\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:39<30:24, 46.79s/it, running_distance=227.9883]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S THE S THE STREAT THE THE STEAT THE SEEMBER AND THE SEEMBER AND THE S THE SEEMBER AND THE SEEMBER AND THE SEEMBER AND THE S THE SAID THE SEEMBER THE S THE S THE S THE S THE S THE S THE S AND THE S THE S THE S THE S THE S THE SAID THE S THE S AND THE S THE SEEMBER\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:12<26:28, 41.81s/it, running_distance=226.4812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S THE S THE SAID AND THE SEEMBER AND THE SEEMBER AND THE SAID AND THE S THE THE S THE SAID AND THE SAID AND THE S THE S THE S THE S THE SAID THE S THE S THE S THE S THE S THE S THE S THE S T THE S THE S THE S THE SAID THE S THE S THE S THE S THE S TRAND THE THE S THE SAID THE SCCONT THE SAID T T T THE THE TE THE T\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 226.4812\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 16/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 16/50: \n","Train Loss 1.7560\t Train Perplexity 5.7902\t Learning Rate 0.000079391\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 17/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 17/50: \n","Train Loss 1.7458\t Train Perplexity 5.7313\t Learning Rate 0.000076794\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:24<58:51, 84.08s/it, running_distance=464.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S TREATED THE S TRAIN THE S TRAIN THE S THE S TRAIN THE AND THE S THE S TRAND THE S TRAIN THE S TRAIN THE S TREE AND THE S TRAIN THE S THE S TRAIN THE AND THE THE S AND THE AND THE S TRAIN T THE S AND THE AND THE S AND THE S THE S AND T T THE S T T T T T THE S T T TRAIN THE T T THE T T THE S AND THE STEAS STAND TAIS T T T TANOUT TEATERE TATAND T T T T T T T T TANOWANOUREAT T T T T T T A T T T T T T T T T T T T T T T T T T T T T TASTASTASTAS T T T T T T T T T T TASTAS T TASTATATATATATAT THEATHEATATAS TATATATAT T TATATATATATATATTTTATHEATH\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:36<52:53, 77.39s/it, running_distance=441.2969]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE SAID AND THE SAID AND AND AND AND THE S AND AND AND AND AND AND THE WAS AND THE WAS AND AND AND AND AND AND THE S THE THE THE SAID AND AND AND THE S THE AND THE THE AND AND AND THE S THE S AND AND THE S AND AND AND AND AND AND THE S AND AND AND THE S AND THE T T THE AND THE T IMME\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:51<50:46, 76.16s/it, running_distance=425.2031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S TREATED AND THE AND THE AND AND THE S THE AND AND AND AND THE THE S TRAIN THE AND THE AND THE AND AND THE S THE S TREE THE THE S THE AND AND THE AND AND THE THE AND AND THE AND AND THE S AND AND THE S AND AND AND AND THE S THE S AND AND AND AND THE T THE THE AND THE AND AND THE T THE S THE S AND AS TE AS T TAND T T T T TAND TEAT TEAT T TATAND T T T T T T T TATATATAT T T T T T T T T T AINDAS T T T T T T T T T T T T T T T T T TT TASTAS T TAS T T T T T T T T T TAS T T T T T TATAT TATHAT THAT T T T TATATAT TAT T TATATATATATATATHEATATHEY T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:13<50:55, 78.36s/it, running_distance=423.6680]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S TREATED AND THE S THE AND AND AND AND AND AND AND AND AND AND THE S TRAIN THE THE AND AND AND THE AND THE THE S THE THE THE THE S THE S AND THE AND THE THE THE THE AND THE AND THE S AND AND AND THE S AND AND AND AND AND THE S THE S AND AND AND THE T T THE THE AND THE AND AND THE T THE S THE S AND AS T THE T THE T T T T TAND TEAST TEATEAT T T T T T T T T T TATANOWAT T T T T T T T T T ANO T T T T T T T T T T T T T T T T T T T T T TAST T TAS T T T T T T T T T T TAST T T T TATATATATATHAT T THEAT T T TATAT T T T T THE TATATHEATHEATHEATHEAT\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:35<50:29, 79.74s/it, running_distance=427.8687]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S TREATED THE S TRAIN THE S TRAIN THE S THE S TRAIN THE AND THE S THE S TRAND THE S TRAIN THE S TRAIN THE S TREATED THE S TRAIN THE S THE S TREE THE S AND THE THE S AND THE AND THE S TRAIN THE S AND THE S AND THE S AND THE S THE S AND T THE S THE S T T THE S THE S TRAIN THE T THE T THE S THE S AND TEAT THE TUREAND T T T TEAND TEAND THEASTEATAND T TUT TEAT T T T TATANDIS T T T T T T T T ANDAS T T T T T T T T T T T T T T T T T T T TASTAS T TAS T T T T T T T T T T TAST T TASTAS TATATATHEAT THEAT T TAS TANOWATAT T TATAT TATATATATATTTATHEATT\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 427.8687\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 18/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 18/50: \n","Train Loss 1.7289\t Train Perplexity 5.6352\t Learning Rate 0.000074090\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 19/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 19/50: \n","Train Loss 1.7188\t Train Perplexity 5.5786\t Learning Rate 0.000071292\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:25, 83.47s/it, running_distance=366.3750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S TREATED THE S TRAIN THE S TRAIN THE S TRAIN AND THE S TRAIN THE S TRAIN THE S AND THE S TRACE AND AND AND THE S TRAIN AND THE S AND THE S TREESS AND THE S TRAIN THE S TRACE AND THE S TRACE AND AND THE S TREEMENT AND THE S TRAIN THE S TRAIN THE S T T THE S THE S TRAIN THE S T THE S THE S THE S T TREMAND SEE STAND TE TUSE TANDE TEALIGEASEASTEASTUTUTUTEASTEAND ASE TANDEA TE T T TE T A A A AN T T TEARIGEAREAS T T T T T T T T T T T TUREALEAS T T T T T T T T T T T TAN T T TAS T T TAS TATHE TUS TUS TUCOUCOUS TATENOUS THE TUTATUS TATHEATENOUT\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:35<52:26, 76.74s/it, running_distance=322.3359]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE S THE S THE S THE ARE AND AND AND AND AND AND AND AND AND AND AND THE S TRAIN THE S TRAIN THE S TRAIN THE S TREE THE S AID AND AND AND AND AND THE S TRAIN THE S TRAIL AND THE S AID AND THE S AID AND AND AND AND AND THE S TRAIN THE S THE S AID AND S T THE S THE S AID\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:49<50:27, 75.68s/it, running_distance=301.4375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE S TRAIN THE AND AND AND AND AND AND AND AND AND AND THE S THE S TRAIN AND AND THE S TRAIN THE S TREE THE S AND AND AND AND AND AND THE S TRAIN THE S TRAIL AND THE S TRAIN THE S THE S TREE THE AND AND AND THE S TRAIN THE S TRAIN THE S T T THE S THE S AID\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:11<50:39, 77.94s/it, running_distance=290.2773]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S THE S THE S TRAIN THE AND AND AND AND AND AND AND AND AND AND THE S THE S TREE THE S TRAIN THE AND AND AND THE S TRAIN THE THE SAID AND AND AND AND AND AND AND AND AND THE ARE AND AND AND AND AND THE S THE S AID AND THE SAID THE SAID THE S AID THE S THE S THE S AND AND AND THE S AID\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:33<50:14, 79.33s/it, running_distance=290.5719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S THE S THE S TRAIN THE AND AND AND AND AND AND AND AND AND AND THE S TRAIN THE S TRAIN THE S TRAIN THE S AID AND AND AND AND AND AND AND AND AND AND THE S TRAIN THE S TRAIN AND THE S TRAIN THE S TREED AND AND AND AND THE S THE SAID THE S AID AND S T THE S THE S AND AND THE S THE S THE S THE S AND SAID AND WA T T T T T T TALIT T\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 290.5719\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 20/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 20/50: \n","Train Loss 1.7041\t Train Perplexity 5.4972\t Learning Rate 0.000068409\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 21/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 21/50: \n","Train Loss 1.6922\t Train Perplexity 5.4323\t Learning Rate 0.000065454\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:41, 83.83s/it, running_distance=410.6875]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S TRAIN THE S TREATED THE S TRAIN THE S TRAIN THE S TREE THE S TREEMENT AND THE S TREATED THE S TREATED THE S TREATED THE S TRAIN AND S TRAIN THE S TRAIN THE S TRAIN THE S TREATED THE S TRAIN THE S TRAIN THE S TREE ARDERS TRAIN THE S TRAIN THE S T T TRAIN THE S THE S T THE S THE S T T T THE S T READ STAND STAND ST TE TEEE TEAND TEANOWATEAND TAND TUTEAIGEAT TAND TEAND TEAITEAS T TEAS T AS T T T T TEAS T TTT T T TE T T T T TEAT TEAS TANOUREANDANT T T T T TEANDENTEANTANTEAS TAS TAS TAS TEATE TEAND TEANOUS TANOWAS TENENENTE TEATENENEATEATEA\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:35<52:30, 76.85s/it, running_distance=408.0703]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE HAD NOT THE S THE S THE HAD AND AND SAID THE S THE HAD AND AND AND THE HAD AND AND AND AND AND THE SAID AND AND AND AND AND AND AND AND AND AND AND AND AND THE S TRAIN THE HAD AND AND AND AND AND AND AND AT ALL AND AND THE S THE HAD AND S TRAID T T THE S THE S THE S AND S THE T THE S THE DOR S AND STEE DARE SKEEE SEEEE TAREE TAREE THEE THASKID ASKEAD TEATE ASTASTASTE ATE A A AS AS S A A AS T T T TEAT AINOUPEAS T T T T T T T T T TE TARE T T T T T T T T T T T T T TAY T TAY TAS TAY TAS THE THE TAS TAS TASTAS TAS TENONOUCHE TE TENENENEATHEA\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:50<50:35, 75.89s/it, running_distance=399.2969]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE S TRAIN THE S TRANGE THE S TRAIN THE S TRANGER AND AND THE S TRAIN AND THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN AND AND AND THE S TRAIN THE S TRAIN THE ARDER AND AND AND AND AND THE S TRAIN AND AND AT TRAIN THE S TRAIN THE S TRAIN THE S THE S THE S THE S THE S THE S T T THE S T AND SAID WAS W TEAND TE TALLEAND TE T TE TEAD A T A TEALE A A A A T T TEAS A A T T A AS A A ANDA T T T T T T T T T T T T T T T T T T TEAS T T T T T T T T T T T T T T TANY TANY TANONY TAS TATHE T T T T T TAS TANOUS TEAS T TAS TE TE TENENEATEAS T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:12<50:51, 78.23s/it, running_distance=407.3711]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S THE S THE S TREE THE S TRANGE THE S TRAIN THE S TREAT AND AND AND THE S TREAT AND STRANGE THE S TRAIN THE S THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S THE S TRAIN THE S AND THE S TRAIN AND AND THE S TRAIN THE S TRAIN THE S T TRAIN THE S THE S THE S THE S AND S T THE S AND S TEREAD S AD STEEE T T TUTAND TE TE TE TEAD TAND ASKEASTEAIS ANDAS TATEAS AIT T T AS A A A AS T T T T T T T T T T T T T T T T T T T T TEAREAS T T T T T T T T T T T TANY TANY TAY TAS TANOWATAT THE T T TAS TANOUTATATEAS TATAS TE TE TENENEATEAS T\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:34<50:25, 79.62s/it, running_distance=410.3344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S TRAIN THE S TRAIN THE S TRANGE THE S TRAIN THE S TRANGER AND THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S THE S T THE S T THE S T T T THE S T READ STAND STAND ST TE TEEE TEEEEATE TEATEASTAND T TUTEAIS AT T T AS TANDAS T T T T T T T T AS T T T T T T T T T T T T T T T T T T T T TEAS TANOUS T T T T T T T T T T TANOUS TANTAS TANONONONOUS TEAN TEAND TANOUS TANONONO TE TENE TE TATENEATEATEA\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 410.3344\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 22/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 22/50: \n","Train Loss 1.6777\t Train Perplexity 5.3540\t Learning Rate 0.000062438\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 23/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 23/50: \n","Train Loss 1.6649\t Train Perplexity 5.2861\t Learning Rate 0.000059373\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:23<58:41, 83.84s/it, running_distance=385.0312]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN AND S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN TRAIN THE S TREE S TRAIN T T T T THE S T T T T REAS TREMAND STURE TEANDEANDE TEANDEANEAND TEAND TEAND TUTEAIS T T T TEANEANDAIND TE TE T T T T T T T T TEAS T T TT T T TEAND T T TEANT TEAS TUS TEANTANT T T T TEAS TENEANEANEAN TEAS TEAS TAS TEATEAN TEAN TEANOUS TEANEAS TEATAS TE TEAS TEAS TEAS T\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:36<52:37, 77.01s/it, running_distance=367.6719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE S THE S THE S THE S THE S THE SAID AND SHE S THE SKED AND SHE SAID AND AND AND AND SHE SAID AND THE SKY AND THE SAID AND AND AND AND AND AND SHE SAID THE S THE S TRAIL AND SHE SAID AND AND AND AND THE S TRAIN THE SAID AND THE SAID THE S TRAIN THE S TRAIN\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:50<50:38, 75.97s/it, running_distance=361.6094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE S TRAIN THE S TRAIN THE S THE S TRAIN THE S THE SAID AND S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S THE S TRAIN THE S TRAIN THE S TRAIL AND STAND THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE SAID THE S TRAIN THE S TREE THE S THE S THE S T WAS T T T T THE S T WAS TELEAS T TTEAIGE T T WOND T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T ANO T T T T T T T T T T T T T T T T T T T T T TAS T T T T T T T T T T T T TANEANY TANEANOONO TANEAS T T T T T T TANOOUNO TAS TAS TAS TAS TAS TENEAS THEAS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:12<50:49, 78.19s/it, running_distance=360.4102]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S THE S THE S TRAIN THE S TRAIN THE S THE S TRAIN THE S THE S TRAIN THE S TRAIN AND S TRAIL AND S TRAIN THE S THE S TRAIL THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIL THE S TRAIN THE S AND S TRAIL THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TREE S THE S TRAIN T WAS T T T T T THE S T WAS TELL T AND S T T TEEEREANDIT TEAIS T T T T T T TTT T TT T T TANEANEANEAS T T T T T T T T T T T T T T T T T T T T T T T T T T T T T T TANOUREANOUREANT T T T T T T TANEANEANO T T T TANEAS THEAS TE T T TANOUANONOUREAS TEAS TAS TAS TEAS TEAS THE\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:34<50:23, 79.56s/it, running_distance=365.8844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S THE S THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN AND S TRAIN THE S TRAIN THE S THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIN THE S TRAIL THE S TRAIN THE S TRAIN THE S TRAIN THE S TREE THE S TREEMENT TRAIN THE S TRAIN THE S THE S THE S THE S T WAS T T T T T THE S T WAS TREAND WAND S S THE TEANOWEAND TEEAS TEAND TEAND TEAIS T T T TANONO TEANO T T T T T T T T T T T T T T T T T T T T T T T T T T T T TEASTANOUREANONT T T T T T T T TANEANO TAN TAS TAS TAS THE TEAN TEAS TANOUTANOUTEAS TEAS TAS TEAS TEAS TEAS TE\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 365.8844\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 24/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 24/50: \n","Train Loss 1.6529\t Train Perplexity 5.2227\t Learning Rate 0.000056271\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 25/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 25/50: \n","Train Loss 1.6421\t Train Perplexity 5.1666\t Learning Rate 0.000053144\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:24<58:52, 84.10s/it, running_distance=443.3750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S THE S THE S TREE AND S THE S THE S TRAIN AND S THE S THE S TRAIN THE S TRAIN AND S TRAIN THE S TRAIN AND S TREE THE S TRAIN AND S TRAIN AND S THE S TRAIN THE S TRAIL AND S TREE THE S TREE T MAN AND S TRAIN THE S TREEMENT THE S TREE AND S TRAIN T TREEMEN AND S THE S T T T T T T ME T T T THE S T WAS TREAS T T TEANDEEREEREAINDEAIS TEAIS T TEREREAND TTTEAS T TEANDANEAS TEAS T TEAS T T T T T T T T TEANOOOO T T T T TEAS T T T T TEAS TEASTEAILEANDEANT T TEAS TEANDEANEANEANEAS TEAS TEAN TEATEAS TEATEAS TATEANEATEAS TEATEATEAS TEAS TEATEAS TEA\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:36<52:51, 77.34s/it, running_distance=426.6406]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE S THE SKY THE MAN AND AND S THE SKED AND WHEN THE PART AND SHE SAID NOT IMMEN AND THE SAID THE SKED AND AND THE SAID NOT IMPOSITE AND AND AND S THE SKED THE SKY THE SAID NOT IMPOSITED THE S AID NOT IMPOSE T IMPOSITED THE PART IN THE SAID NOT THE S THE SKED THE SKED AND T IMP\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:51<50:45, 76.14s/it, running_distance=409.7083]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE PART AND S THE S TRAIN AND S THE S THE SAID THE POORT OF THE S TRAIN AND S TRAIN THE S TRAIN AND S THE S TRAIN THE S TRANGE THE S TRAIN THE S TRAIN AND S TREE THE S TREE THE S TREE THE S TRAIN THE S TREE AND THE S TREE THE S TRAIN THE S TRACK TO THE S TREE S T T IMPOSEEN T T T THE P AND S THE WAS T T T T THEEREREEEEEEEREREREEEEREREREAND TEAIS ANDANDIS TEEEEEA TERE TE T T T T AN T T T TEEEEANDO T T T T T T T T T TEATEAITHEAILEAILEAILE T T T T T T T TEAN TEAS TEAS AS TAS THEATEAS TEATEANOOOOOLEAS TEATEATEATENEATEAS TEASEASEAS \n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [05:13<50:56, 78.38s/it, running_distance=406.3906]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  AND THE S THE S THE S THE PPOSITING THE S THE S THE S THE S THE S THE S TRAIN THE S TREE T MAN AND THE S TRAIN AND S THE S TRAIN AND S TREE THE S TREE T AND S THE S TRAIN THE S TRAIN AND THE S TREE THE S TREE S TRAIN THE S TREE THE S TREE TREE AND S TRAIN THE S TREE S THE S T T T T IMPOSSIRECENT\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [06:35<50:29, 79.73s/it, running_distance=411.3531]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  AND THE S THE S THE PPOSITING THE S THE S THE S THE S THE S THE S THE POORT OF THE S THE S THE S TREE THE S TRAIN AND S THE S THE S TREE T AND S THE S THE S TRAIN THE S TREE AND S TREET THE S TREE THE S THE S TREE T ME T TREEMENT THE S TREE AND S TRAIN THE S TREEMEN AND S THE S T T T T ME T T T T THE S T WAS T AND S AND S S TEEEEREREEEAITEEEEREATEEREREANDIGOLIS ATEANDANOWEEATEAIND TEAND TEANDA A AN TEAS TEANOOO TEAS T TEAS T T T TEAIS TEAISTEAILEAILEANT T T T T TEANDEANEAN TEAN TEAS TEANOOOOUS TEATEATEAS TEANOUTEAS TEATEATEATEATEAS TEAS TEAS TE\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 411.3531\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 26/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 26/50: \n","Train Loss 1.6319\t Train Perplexity 5.1141\t Learning Rate 0.000050005\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 27/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 27/50: \n","Train Loss 1.6223\t Train Perplexity 5.0654\t Learning Rate 0.000046866\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:24<59:05, 84.42s/it, running_distance=390.2656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  AND THE S TRAIN AND S THE S TRAIN THE S THE S TRAIN AND S THE S TREET THE S THE S TRAIN AND S THE S TRAIN THE S TRAIN THE S TRAIN AND S TREE THE S TRAIN THE S TRAIN THE S TRAIL AND S TRAIN AND S TRAIN AND S TRACE AND S TREE T TREE T MAN AND S TRAIN THE S TRACE T TREATER THE S T T T T T THE S T T THE S T WAS T REND WAS T TEND TEREREEEEAITEAIS TEAS TEREREND TUREAIS T T TEANEANEAND TEA TTTE T T T T T T TEAS TEANDANDAIS T TEAIS T T T T T TEAIS TUREAILEANT T T T T TEANTENEANEANEANTEANEAS TENEAS TEATEATEANOOUTEANOOOOOUS TEAS TENENEATEANENENEATEATEAS \n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [02:37<52:59, 77.54s/it, running_distance=374.2109]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  AND THE SKED THE SAID NOT THE PART AND SHE WAS AND SHE WAS AND SHE WAS AND THE SKY THE SKY THE SAID NOT ALL THE SAID NOT THE SAID NOT THE SAID NOT THE PART AND THE SAID NOT THE SAID NOT THE SAID NOT THE SKY AND S AID NOT THE MAN AND SAID NOT THE S TRUE HAD NOT THE SAID\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [03:51<50:49, 76.25s/it, running_distance=355.8958]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  AND THE S THE S THE S TRANGER THE S THE S THE S THE S THE S THE S THE S TRAIN THE S TRANGE THE S TRAIN THE S TRAIN AND S THE S THE S TRANGER AND S THE S TRANGER AND STRANGER THE S TRAIN THE S TRAIN THE S AND STRANGER AND S TRAIN THE S TRAIN THE S TRAIN THE S TREE S TREATER\n"]},{"output_type":"error","ename":"KeyboardInterrupt","evalue":"","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)","\u001b[0;32m<ipython-input-322-3bca53f806b5>\u001b[0m in \u001b[0;36m<cell line: 32>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     42\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     43\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mepoch\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0;36m2\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m    \u001b[0;31m# validate every 2 epochs to speed up training\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 44\u001b[0;31m         \u001b[0mlevenshtein_distance\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mvalidate_fast\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mval_loader\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     45\u001b[0m         \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"Levenshtein Distance {:.04f}\"\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlevenshtein_distance\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     46\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m<ipython-input-318-ce9c8b9d1969>\u001b[0m in \u001b[0;36mvalidate_fast\u001b[0;34m(model, dataloader)\u001b[0m\n\u001b[1;32m     13\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     14\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0mtorch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minference_mode\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 15\u001b[0;31m             \u001b[0mgreedy_predictions\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrecognize\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs_lengths\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     16\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     17\u001b[0m         \u001b[0;31m# calculating Levenshtein Distance\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m<ipython-input-309-0fb1aff83f45>\u001b[0m in \u001b[0;36mrecognize\u001b[0;34m(self, inp, inp_len)\u001b[0m\n\u001b[1;32m     52\u001b[0m         \u001b[0mencoder_outputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mencoder_lens\u001b[0m  \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mencoder\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0minp\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minp_len\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     53\u001b[0m         \u001b[0mencoder_outputs\u001b[0m                \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mproj\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mencoder_outputs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 54\u001b[0;31m         \u001b[0mout\u001b[0m                            \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdecoder\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrecognize_greedy_search\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mencoder_outputs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mencoder_lens\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     55\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     56\u001b[0m         \u001b[0;32mreturn\u001b[0m \u001b[0mout\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m<ipython-input-305-56458a6700cc>\u001b[0m in \u001b[0;36mrecognize_greedy_search\u001b[0;34m(self, enc_outputs, enc_input_lengths)\u001b[0m\n\u001b[1;32m    140\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    141\u001b[0m             \u001b[0;31m# end if all sequences have generated the EOS token\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 142\u001b[0;31m             \u001b[0;32mif\u001b[0m \u001b[0mfinished\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mall\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mbreak\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    143\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    144\u001b[0m         \u001b[0;31m# remove the initial <SOS> token and pad sequences to the same length\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mKeyboardInterrupt\u001b[0m: "]},{"output_type":"display_data","data":{"text/plain":["<Figure size 640x480 with 2 Axes>"],"image/png":"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\n"},"metadata":{}}]},{"cell_type":"code","source":["valid_actual_distance = validate_full(model, val_loader)"],"metadata":{"id":"i2W3Dh78Adyg"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"hgFYFaBGeBqM"},"source":["## Testing"]},{"cell_type":"code","source":["model.eval()\n","\n","# progress bar\n","batch_bar = tqdm(total=len(test_loader), dynamic_ncols=True, leave=False, position=0, desc=\"Test\", ncols=5)\n","\n","all_predictions = []\n","\n","for i, data in enumerate(test_loader):\n","    inputs, inputs_lengths = data\n","    inputs                 = inputs.to(device)\n","\n","    with torch.inference_mode():\n","        greedy_predictions = model.recognize(inputs, inputs_lengths)\n","\n","    greedy_predictions = greedy_predictions.detach().cpu().numpy()\n","\n","    for batch_idx in range(greedy_predictions.shape[0]):\n","        pred_sliced = indices_to_chars(greedy_predictions[batch_idx], vocab= VOCAB)\n","        pred_string = ''.join(pred_sliced)\n","        all_predictions.append(pred_string)\n","\n","    batch_bar.update()"],"metadata":{"id":"vVKku0SQW8y6","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1712714638670,"user_tz":240,"elapsed":2274339,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"d78b1b5f-acb8-4aee-d8fd-34b3f1678079"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["Test: 100%|██████████| 41/41 [37:54<00:00, 66.53s/it]"]}]},{"cell_type":"code","source":["%cd /content/\n","df = pd.DataFrame({\n","    \"index\" : list(range(len(all_predictions))),\n","    \"label\" : all_predictions\n","})\n","\n","df.to_csv(\"test.csv\", index=False)"],"metadata":{"id":"XH9kvWIz5_9Z","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1712714708418,"user_tz":240,"elapsed":150,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"992c1015-4821-4050-fe34-30b1ce686c91"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["/content\n"]}]},{"cell_type":"code","source":["!kaggle competitions submit -c hw4p2-sp24 -f test.csv -m \"HW4P2 Preliminary Submission\""],"metadata":{"id":"LQP8QgXnAxqP","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1712714712502,"user_tz":240,"elapsed":552,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"c9a9b05b-a716-4cae-fa8a-766414a63c01"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Traceback (most recent call last):\n","  File \"/usr/local/bin/kaggle\", line 8, in <module>\n","    sys.exit(main())\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/cli.py\", line 67, in main\n","    out = args.func(**command_args)\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 561, in competition_submit_cli\n","    submit_result = self.competition_submit(file_name, message,\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 532, in competition_submit\n","    url_result['createUrl'], quiet)\n","KeyError: 'createUrl'\n"]}]},{"cell_type":"code","source":["# TODO Sumbit predictions to Kaggle"],"metadata":{"id":"Gyb5IUOu6Ba3"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["# PART #2 Full Transformer (Encoder-Decoder)\n","\n","Using an LSTM encoder helps to capture some dependencies in the input sequence which helps the transformer decoder to decode. However, we can do more than that. The transformer encoder plays a crucial role by processing the input sequence into a high-dimensional space, capturing complex relationships between elements. It uses self-attention to understand the context around each feature, enabling the model to grasp subtle nuances and dependencies, crucial for our speech recognition task and improving performance."],"metadata":{"id":"eccQFlez6Kr3"}},{"cell_type":"markdown","source":["![image.png](data:image/png;base64,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)"],"metadata":{"id":"RIDOlW7v8BUk"}},{"cell_type":"markdown","source":["Tips in completing this section\n","\n","* We will use the LSTM Encoder as input embedding.\n","* How many multi-head attention modules do we have.\n","* Can you seperate the encoder layer from the whole encoder architecture ?\n","* How do  we implement skip  connections\n","* How many layer norms do we have ?"],"metadata":{"id":"dooU-7ld8GZ4"}},{"cell_type":"markdown","source":["## Transformer Encoder"],"metadata":{"id":"kH7Lz9tPgtb0"}},{"cell_type":"code","source":["class EncoderLayer(torch.nn.Module):\n","    def __init__(self, d_model, num_heads, d_ff, dropout=0.1):\n","        super().__init__()\n","\n","        # @TODO: fill in the blanks appropriately (given the modules above)\n","        self.mha        = MultiHeadAttention(n_head=num_heads, d_model=d_model, dropout=dropout)\n","        self.ffn        = FeedForward(d_model, d_ff, dropout)\n","\n","        self.layernorm1 = torch.nn.LayerNorm(d_model)\n","        self.layernorm2 = torch.nn.LayerNorm(d_model)\n","        self.dropout1   = torch.nn.Dropout(dropout)\n","        self.dropout2   = torch.nn.Dropout(dropout)\n","\n","    def forward(self, inp):  # shape: (batch_size, time_length, d_model)\n","\n","        # Multi-Head Attention\n","        #   (1) perform Multi-Head Attention on inp\n","        ''' TODO '''\n","        # print(f'The shape of inp is {inp.shape}')\n","        attn_output, _ = self.mha(q=inp, k=inp, v=inp)\n","        # Skip (Residual) Connection\n","        #   (1) perform dropout\n","        #   (2) add the input as a skip connection\n","        ''' TODO '''\n","        out1 = self.dropout1(attn_output) + inp\n","        # Layer Normalization\n","        #   (1) call layernorm on this resulting value\n","        ''' TODO '''\n","        layernorm_out1 = self.layernorm1(out1)\n","        # Feed Forward Network\n","        #   (1) apply feed forward layer\n","        ''' TODO '''\n","        ffn_out = self.ffn(layernorm_out1)\n","        # Skip (Residual) Connection\n","        #   (1) perform dropout\n","        #   (2) add the output of the first layernorm as a skip connection\n","        ''' TODO '''\n","        out2 = self.dropout2(ffn_out) + layernorm_out1\n","        # Layer Normalization\n","        #   (1) call layernorm on this resulting value\n","        ''' TODO '''\n","        layernorm_out2 = self.layernorm2(out2)  # shape: (batch_size, time_length, d_ff)\n","        return layernorm_out2"],"metadata":{"id":"DAw0eHIt0gez","executionInfo":{"status":"ok","timestamp":1713553379077,"user_tz":240,"elapsed":132,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":31,"outputs":[]},{"cell_type":"code","source":["class Encoder(torch.nn.Module):\n","    def __init__(self, input_dim, num_layers, d_model, num_heads, d_ff, mfcc_max_seq_length, dropout=0.1):\n","        super(Encoder, self).__init__()\n","\n","        # @TODO: fill in the blanks appropriately (given the modules above and P1 setup)\n","        self.embedding      = CNN_LSTM_Encoder(input_size=input_dim, encoder_hidden_size=256, enc_dropout=dropout)\n","        self.projection     = torch.nn.Linear(1024, d_model)\n","        self.positional_encoding    = PositionalEncoding(d_model=d_model, max_len=mfcc_max_seq_length)\n","\n","        # concatenating all num_layers EncoderLayer blocks\n","        self.enc_layers     = torch.nn.ModuleList([EncoderLayer(d_model, num_heads, d_ff, dropout) for _ in range(num_layers)])\n","\n","        self.dropout        = torch.nn.Dropout(p=dropout)\n","        self.layernorm      = torch.nn.LayerNorm(d_model)\n","\n","    def forward(self, x, input_lengths):\n","\n","        # apply the CNN-LSTM Encoder (embedding)\n","        ''' TODO '''\n","        # print(f'The shape before embedding is {x.shape}')\n","        embedding_out, embedding_out_length = self.embedding(x, input_lengths)  # shape: (batch_size, time_length, features)\n","        # print(f'The shape of embedding_out is {embedding_out.shape}')\n","        # adding projection layer to change dimension of encoder_outputs to match d_model\n","        ''' TODO '''\n","        projection_out = self.projection(embedding_out)       # shape: (batch_size, time_length, d_model)\n","        # print(f'The shape of projection_out is {projection_out.shape}')\n","        # apply Positional Encoding on these extracted features\n","        ''' TODO '''\n","        position_out = self.positional_encoding(projection_out)   # shape: (batch_size, time_length, d_model)\n","        # print(f'The shape of position_out is {position_out.shape}')\n","        # apply dropout as regularization technique\n","        ''' TODO '''\n","        out = self.dropout(position_out) #   shape: (batch_size, time_length, d_model)\n","        # print(f'The shape of dropout is {out.shape}')\n","        # passing inputs through Transformer Encoder blocks\n","        ''' TODO '''\n","        for i, layer in enumerate(self.enc_layers):\n","            out = layer(out)     # shape: (batch_size, time_length, d_ff)\n","            # all_attention_weights[f'layer_{i+1}_weight1'] = attention_weights1   # I Think shape is (batch_size, target_time_length, target_time_length)\n","            # all_attention_weights[f'layer_{i+1}_weight2'] = attention_weights2   # I Think shape is (batch_size, target_time_length, input time length)\n","        return out, embedding_out_length"],"metadata":{"id":"9K0aEZhz0y8v","executionInfo":{"status":"ok","timestamp":1713553379419,"user_tz":240,"elapsed":1,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":32,"outputs":[]},{"cell_type":"code","execution_count":33,"metadata":{"id":"wPL9Xgl50AVr","executionInfo":{"status":"ok","timestamp":1713553393935,"user_tz":240,"elapsed":161,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"outputs":[],"source":["class FullTransformer(torch.nn.Module):\n","    def __init__(self, input_dim, enc_num_layers, dec_num_layers, enc_num_heads, dec_num_heads,\n","                 d_model, d_ff, target_vocab_size, eos_token, sos_token,\n","                 pad_token, dropout=0.1, trans_max_seq_length=550, mfcc_max_seq_length=3260):\n","\n","        super(FullTransformer, self).__init__()\n","\n","        self.encoder = Encoder(input_dim, enc_num_layers, d_model, enc_num_heads, d_ff,\n","                               mfcc_max_seq_length,  dropout)\n","\n","        self.decoder = Decoder(dec_num_layers, d_model, dec_num_heads, d_ff,\n","                               dropout, target_vocab_size, trans_max_seq_length, eos_token, sos_token, pad_token)\n","\n","    def forward(self, padded_input, input_lengths, padded_target, target_lengths):\n","        # passing through Encoder\n","        encoder_output, encoder_lens = self.encoder(padded_input, input_lengths)\n","\n","        # passing Encoder output and Attention masks through Decoder\n","        output, attention_weights = self.decoder(padded_target, encoder_output, encoder_lens)\n","\n","        return output, attention_weights\n","\n","    def recognize(self, inp, inp_len):\n","        \"\"\" sequence-to-sequence greedy search -- decoding one utterance at a time \"\"\"\n","\n","        encoder_outputs, encoder_lens = self.encoder(inp, inp_len)\n","        out = self.decoder.recognize_greedy_search(encoder_outputs, encoder_lens)\n","\n","        return out"]},{"cell_type":"markdown","source":["## Full Model Setup"],"metadata":{"id":"Lam5bYQ8gxke"}},{"cell_type":"code","source":["full_model = FullTransformer(\n","input_dim                   = 27,\n","enc_num_layers              = config['enc_num_layers'],\n","dec_num_layers              = config['dec_num_layers'],\n","enc_num_heads               = config['enc_num_heads'],\n","dec_num_heads               = config['dec_num_heads'],\n","\n","d_model                     = config['d_model'],\n","d_ff                        = config['d_ff'],\n","\n","target_vocab_size           = len(VOCAB),\n","eos_token                   = EOS_TOKEN,\n","sos_token                   = SOS_TOKEN,\n","pad_token                   = PAD_TOKEN,\n","\n","dropout                     = config['enc_dropout'],\n","\n","# decrease to a small number if you are just trying to implement the network\n","# we've computed the max length of the transcripts for you\n","trans_max_seq_length        = 550,\n","mfcc_max_seq_length         = dataset_max_len).to(device)\n","\n","\n","\n","# para_full = num_parameters(full_model)\n","# print(\"#\"*10)\n","# print(f\"Model Parameters:\\n {para_full}\")\n","# print(\"#\"*10)\n","\n","print(full_model)"],"metadata":{"id":"0qKfN3bj9qk7","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713553439535,"user_tz":240,"elapsed":566,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"02a297ec-c701-4a0f-c31c-97f22c37cd2e"},"execution_count":35,"outputs":[{"output_type":"stream","name":"stdout","text":["FullTransformer(\n","  (encoder): Encoder(\n","    (embedding): CNN_LSTM_Encoder(\n","      (embedding): Sequential(\n","        (0): Conv1d(27, 54, kernel_size=(5,), stride=(1,), padding=(2,))\n","        (1): BatchNorm1d(54, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","        (2): ReLU(inplace=True)\n","        (3): Conv1d(54, 108, kernel_size=(5,), stride=(1,), padding=(2,))\n","        (4): BatchNorm1d(108, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","        (5): ReLU(inplace=True)\n","      )\n","      (bilstm): LSTM(108, 256, batch_first=True, dropout=0.2, bidirectional=True)\n","      (pBLSTMs): Sequential(\n","        (0): pBLSTM(\n","          (blstm): LSTM(1024, 256, batch_first=True, bidirectional=True)\n","        )\n","        (1): LockedDropout()\n","        (2): pBLSTM(\n","          (blstm): LSTM(1024, 512, batch_first=True, bidirectional=True)\n","        )\n","        (3): LockedDropout()\n","      )\n","      (permute): Permute()\n","    )\n","    (projection): Linear(in_features=1024, out_features=512, bias=True)\n","    (positional_encoding): PositionalEncoding()\n","    (enc_layers): ModuleList(\n","      (0-1): 2 x EncoderLayer(\n","        (mha): MultiHeadAttention(\n","          (w_qs): Linear(in_features=512, out_features=512, bias=True)\n","          (w_ks): Linear(in_features=512, out_features=512, bias=True)\n","          (w_vs): Linear(in_features=512, out_features=512, bias=True)\n","          (attention): ScaledDotProductAttention(\n","            (dropout): Dropout(p=0.2, inplace=False)\n","            (softmax): Softmax(dim=2)\n","          )\n","          (fc): Linear(in_features=512, out_features=512, bias=True)\n","          (dropout): Dropout(p=0.2, inplace=False)\n","        )\n","        (ffn): FeedForward(\n","          (linear_1): Linear(in_features=512, out_features=2048, bias=True)\n","          (dropout): Dropout(p=0.2, inplace=False)\n","          (linear_2): Linear(in_features=2048, out_features=512, bias=True)\n","        )\n","        (layernorm1): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","        (layernorm2): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","        (dropout1): Dropout(p=0.2, inplace=False)\n","        (dropout2): Dropout(p=0.2, inplace=False)\n","      )\n","    )\n","    (dropout): Dropout(p=0.2, inplace=False)\n","    (layernorm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","  )\n","  (decoder): Decoder(\n","    (dec_layers): ModuleList(\n","      (0-1): 2 x DecoderLayer(\n","        (mha1): MultiHeadAttention(\n","          (w_qs): Linear(in_features=512, out_features=512, bias=True)\n","          (w_ks): Linear(in_features=512, out_features=512, bias=True)\n","          (w_vs): Linear(in_features=512, out_features=512, bias=True)\n","          (attention): ScaledDotProductAttention(\n","            (dropout): Dropout(p=0.2, inplace=False)\n","            (softmax): Softmax(dim=2)\n","          )\n","          (fc): Linear(in_features=512, out_features=512, bias=True)\n","          (dropout): Dropout(p=0.2, inplace=False)\n","        )\n","        (mha2): MultiHeadAttention(\n","          (w_qs): Linear(in_features=512, out_features=512, bias=True)\n","          (w_ks): Linear(in_features=512, out_features=512, bias=True)\n","          (w_vs): Linear(in_features=512, out_features=512, bias=True)\n","          (attention): ScaledDotProductAttention(\n","            (dropout): Dropout(p=0.2, inplace=False)\n","            (softmax): Softmax(dim=2)\n","          )\n","          (fc): Linear(in_features=512, out_features=512, bias=True)\n","          (dropout): Dropout(p=0.2, inplace=False)\n","        )\n","        (ffn): FeedForward(\n","          (linear_1): Linear(in_features=512, out_features=2048, bias=True)\n","          (dropout): Dropout(p=0.2, inplace=False)\n","          (linear_2): Linear(in_features=2048, out_features=512, bias=True)\n","        )\n","        (layernorm1): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","        (layernorm2): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","        (layernorm3): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","        (dropout1): Dropout(p=0.2, inplace=False)\n","        (dropout2): Dropout(p=0.2, inplace=False)\n","        (dropout3): Dropout(p=0.2, inplace=False)\n","      )\n","    )\n","    (target_embedding): Embedding(31, 512, padding_idx=0)\n","    (positional_encoding): PositionalEncoding()\n","    (final_linear): Sequential(\n","      (0): Linear(in_features=512, out_features=31, bias=True)\n","    )\n","    (dropout): Dropout(p=0.2, inplace=False)\n","    (layernorm): LayerNorm((512,), eps=1e-05, elementwise_affine=True)\n","  )\n",")\n"]}]},{"cell_type":"code","source":["### Load your best P1 model checkpoint\n","model.load_state_dict(torch.load('/content/checkpoints-basic-transformer/checkpoint-best-loss-model.pth')['model_state_dict'])"],"metadata":{"id":"ETooy0d59__J","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713395560595,"user_tz":240,"elapsed":436,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"c0e0b176-adab-481d-c0a7-a0fe2b3534c9"},"execution_count":null,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<All keys matched successfully>"]},"metadata":{},"execution_count":276}]},{"cell_type":"markdown","source":["Freezing model weights except for transformer encoder layers. This is like an initialization strategy so that the encoder learns the range of values required by the decoder. After about 3 epochs when this initialzation is done, unfreeze weights of the entire model and let it train.\n","\n","\n"],"metadata":{"id":"UUg3DDMA_Bap"}},{"cell_type":"code","source":["### Copy weights from P1 model to full model\n","### Freeze the weights of full transformer input embedding, linear projection, and decoder\n","\n","full_model.encoder.embedding.load_state_dict(model.encoder.state_dict())\n","for param in full_model.encoder.embedding.parameters():\n","    param.requires_grad = False # TODO make it non-trainable\n","\n","full_model.encoder.projection.load_state_dict(model.proj.state_dict())\n","for param in full_model.encoder.projection.parameters():\n","    param.requires_grad = False # TODO make it non-trainable\n","\n","full_model.decoder.load_state_dict(model.decoder.state_dict())\n","for param in full_model.decoder.parameters():\n","    param.requires_grad = False # TODO make it non-trainable"],"metadata":{"id":"JjAH-p6B-Y2M"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Loss, Optimizer, and Scheduler Definitions"],"metadata":{"id":"PDLdbCwdghEB"}},{"cell_type":"code","source":["loss_func   = nn.CrossEntropyLoss(ignore_index = PAD_TOKEN)\n","scaler      = torch.cuda.amp.GradScaler()"],"metadata":{"id":"ymlPz5lfghl3","executionInfo":{"status":"ok","timestamp":1713554059044,"user_tz":240,"elapsed":154,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":38,"outputs":[]},{"cell_type":"code","source":["''' defining optimizer '''\n","\n","if config[\"optimizer\"] == \"SGD\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.SGD(full_model.parameters(),\n","                              lr=config[\"learning_rate\"],\n","                              momentum=config[\"momentum\"],\n","                              weight_decay=1E-4,\n","                              nesterov=config[\"nesterov\"])\n","\n","elif config[\"optimizer\"] == \"Adam\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.Adam(full_model.parameters(),\n","                               lr=float(config[\"learning_rate\"]),\n","                               weight_decay=1e-4)\n","\n","elif config[\"optimizer\"] == \"AdamW\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.AdamW(full_model.parameters(), lr=float(config[\"learning_rate\"]),\n","                                betas=(0.9, 0.999),\n","                                eps=1e-9,\n","                                weight_decay=0.01)\n","\n","''' defining scheduler '''\n","\n","if config[\"scheduler\"] == \"ReduceLR\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,\n","                factor=config[\"factor\"], patience=config[\"patience\"], min_lr=1E-8, verbose=True)\n","\n","elif config[\"scheduler\"] == \"CosineAnnealing\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer,\n","                T_max = config[\"epochs\"], eta_min=1E-8)"],"metadata":{"id":"YGR9Rk2JgkQD","executionInfo":{"status":"ok","timestamp":1713554061148,"user_tz":240,"elapsed":1392,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":39,"outputs":[]},{"cell_type":"markdown","source":["## WandB"],"metadata":{"id":"7C4QYts02vt7"}},{"cell_type":"code","source":["# using WandB? resume training?\n","\n","USE_WANDB = True\n","RESUME_LOGGING = False\n","\n","\n","# creating your WandB run\n","run_name = \"{}_Transformer_ENC-{}/{}_DEC-{}/{}_{}_{}_{}_{}_Lr:{}\".format(\n","    config[\"Name\"],\n","    config[\"enc_num_layers\"],       # only used in Part II with the Transformer Encoder\n","    config[\"enc_num_heads\"],        # only used in Part II with the Transformer Encoder\n","    config[\"dec_num_layers\"],\n","    config[\"dec_num_heads\"],\n","    config[\"d_model\"],\n","    config[\"d_ff\"],\n","    config[\"optimizer\"],\n","    config[\"scheduler\"],\n","    config['learning_rate'])\n","\n","if USE_WANDB:\n","\n","    wandb.login(key=\"b9fec08c02d03c41bdf5fe1f5268589277270fbe\", relogin=True) # TODO enter your key here\n","\n","    if RESUME_LOGGING:\n","        run_id = \"f0bqyz1j\"\n","        run = wandb.init(\n","            id     = run_id,        ### Insert specific run id here if you want to resume a previous run\n","            resume = True,          ### You need this to resume previous runs, but comment out reinit=True when using this\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","        )\n","\n","    else:\n","        run = wandb.init(\n","            name    = run_name,     ### Wandb creates random run names if you skip this field, we recommend you give useful names\n","            reinit  = True,         ### Allows reinitalizing runs when you re-run this cell\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","            config  = config        ### Wandb Config for your run\n","        )\n","\n","        ### Save your model architecture as a string with str(model)\n","        model_arch  = str(full_model)\n","\n","        ### Save it in a txt file\n","        arch_file   = open(\"model_arch.txt\", \"w\")\n","        file_write  = arch_file.write(model_arch)\n","        arch_file.close()\n","\n","        ### Log it in your wandb run with wandb.save()\n","        wandb.save(\"model_arch.txt\")"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":410,"referenced_widgets":["14f661da6a404f1ab8ee6efc229f3725","890d26224b8f4ac789af33a6d1093a40","6cbc75a0297941059000c1d5d12c4f04","9239d24f6f2b43f4b9df6bf3de9e8937","eb54eb599dc44978a47edf7e70b441c3","f76afe29621f4130a0efd2b537bd5d56","66e559fa64ed4f828ec907258a295a38","2b599b08df004fa09cd2d493f9e706cf"]},"id":"O2N1ZA5323bR","executionInfo":{"status":"ok","timestamp":1713400239754,"user_tz":240,"elapsed":6899,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"834ff7c5-9b18-4ebf-fe35-61bd01daeae8"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[33mWARNING\u001b[0m Calling wandb.login() after wandb.init() has no effect.\n"]},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Finishing last run (ID:3ijrthc2) before initializing another..."]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["VBox(children=(Label(value='0.006 MB of 0.006 MB uploaded\\r'), FloatProgress(value=1.0, max=1.0)))"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"14f661da6a404f1ab8ee6efc229f3725"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["<style>\n","    table.wandb td:nth-child(1) { padding: 0 10px; text-align: left ; width: auto;} td:nth-child(2) {text-align: left ; width: 100%}\n","    .wandb-row { display: flex; flex-direction: row; flex-wrap: wrap; justify-content: flex-start; width: 100% }\n","    .wandb-col { display: flex; flex-direction: column; flex-basis: 100%; flex: 1; padding: 10px; }\n","    </style>\n","<div class=\"wandb-row\"><div class=\"wandb-col\"><h3>Run history:</h3><br/><table class=\"wandb\"><tr><td>learning_rate</td><td>█▇▅▁</td></tr><tr><td>train_loss</td><td>█▅▃▁</td></tr><tr><td>train_perplexity</td><td>█▅▃▁</td></tr><tr><td>val_distance</td><td>█▁</td></tr></table><br/></div><div class=\"wandb-col\"><h3>Run summary:</h3><br/><table class=\"wandb\"><tr><td>learning_rate</td><td>0.0001</td></tr><tr><td>train_loss</td><td>0.66079</td></tr><tr><td>train_perplexity</td><td>1.93691</td></tr><tr><td>val_distance</td><td>33.725</td></tr></table><br/></div></div>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run <strong style=\"color:#cdcd00\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_CosineAnnealing_Lr:1E-4</strong> at: <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/3ijrthc2/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/3ijrthc2/workspace</a><br/>Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 1 other file(s)"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Find logs at: <code>./wandb/run-20240417_235037-3ijrthc2/logs</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Successfully finished last run (ID:3ijrthc2). Initializing new run:<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["wandb version 0.16.6 is available!  To upgrade, please run:\n"," $ pip install wandb --upgrade"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Tracking run with wandb version 0.16.5"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Run data is saved locally in <code>/content/wandb/run-20240418_003032-4yfyi1gn</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Syncing run <strong><a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/4yfyi1gn/workspace' target=\"_blank\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_CosineAnnealing_Lr:1E-4</a></strong> to <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View project at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24</a>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/4yfyi1gn/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/4yfyi1gn/workspace</a>"]},"metadata":{}}]},{"cell_type":"markdown","source":["## Experiments"],"metadata":{"id":"aoaXWl9IqUfi"}},{"cell_type":"code","source":["###  Now let's train the encoder to master the encoder input ranges\n","e                   = 0\n","best_loss           = 10.0\n","\n","checkpoint_root = os.path.join(os.getcwd(), 'checkpointsfull')\n","os.makedirs(checkpoint_root, exist_ok=True)\n","wandb.watch(full_model, log=\"all\")\n","\n","checkpoint_best_loss_model_filename     = 'checkpoint-best-loss-modelfull.pth'\n","checkpoint_last_epoch_filename          = 'checkpoint-epochfull-'\n","best_loss_model_path                    = os.path.join(checkpoint_root, checkpoint_best_loss_model_filename)\n","\n","if RESUME_LOGGING:\n","    # change if you want to load best test model accordingly\n","    checkpoint = torch.load(wandb.restore(checkpoint_best_loss_model_filename, run_path=\"\"+run_id).name)\n","\n","    full_model.load_state_dict(checkpoint['model_state_dict'])\n","    optimizer.load_state_dict(checkpoint['optimizer_state_dict'])\n","    e = checkpoint['epoch']\n","\n","    print(\"Resuming from epoch {}\".format(e+1))\n","    print(\"Epochs left: \", config['epochs']-e)\n","    print(\"Optimizer: \\n\", optimizer)\n","\n","torch.cuda.empty_cache()\n","gc.collect()\n","\n","epochs = config[\"epochs\"]\n","for epoch in range(e, epochs):\n","\n","    print(\"\\nEpoch {}/{}\".format(epoch+1, config[\"epochs\"]))\n","\n","    curr_lr = float(optimizer.param_groups[0][\"lr\"])\n","\n","    train_loss, train_perplexity, attention_weights = train_model(full_model, train_loader, optimizer)\n","\n","    print(\"\\nEpoch {}/{}: \\nTrain Loss {:.04f}\\t Train Perplexity {:.04f}\\t Learning Rate {:.04f}\".format(\n","        epoch + 1, config[\"epochs\"], train_loss, train_perplexity, curr_lr))\n","\n","    if (epoch % 2 == 0):    # validate every 2 epochs to speed up training\n","        levenshtein_distance = validate_fast(full_model, val_loader)\n","        print(\"Levenshtein Distance {:.04f}\".format(levenshtein_distance))\n","\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr,\n","                \"val_distance\"      : levenshtein_distance})\n","\n","    else:\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr})\n","\n","    # plotting the encoder-nearest and decoder-nearest attention weights\n","    attention_keys = list(attention_weights.keys())\n","\n","    # plot_attention_weights((attention_weights[attention_keys[0]][0]).cpu().detach().numpy())\n","    # plot_attention_weights(attention_weights[attention_keys[-1]][0].cpu().detach().numpy())\n","\n","    attention_weights_decoder_self       = attention_weights[attention_keys[0]][0].cpu().detach().numpy()\n","    attention_weights_decoder_cross      = attention_weights[attention_keys[-1]][0].cpu().detach().numpy()\n","    save_attention_plot(attention_weights_decoder_cross, epoch)\n","\n","    if config[\"scheduler\"] == \"ReduceLR\":\n","        scheduler.step(levenshtein_distance)\n","    else:\n","        scheduler.step()\n","\n","    ### Highly Recommended: Save checkpoint in drive and/or wandb if accuracy is better than your current best\n","    epoch_model_path = os.path.join(checkpoint_root, (checkpoint_last_epoch_filename + str(epoch) + '.pth'))\n","    save_model(model, optimizer, scheduler, ['train_loss', train_loss], epoch, epoch_model_path)\n","    ## wandb.save(epoch_model_path) ## Can't save on wandb for all epochs, may blow up storage\n","\n","    print(\"Saved epoch model\")\n","\n","    # the first time we add the transformer encoder block, we freeze all other weights\n","    #   except the transformer encoder weights and train this partially-frozen model\n","    #   for about 3 epochs to weight initialize the transformer encoder weights.\n","    #   This speeds up convergence of the entire network.\n","    # if epoch == 3:\n","    #     break\n","\n","    if best_loss >= train_loss:\n","        best_loss = train_loss\n","        save_model(full_model, optimizer, scheduler, ['train_loss', train_loss], epoch, best_loss_model_path)\n","        # wandb.save(best_loss_model_path)\n","        print(\"Saved best training model\")\n","\n","### Finish your wandb run\n","# run.finish()"],"metadata":{"id":"qHEgNSYU_0tH","colab":{"base_uri":"https://localhost:8080/","height":1000},"executionInfo":{"status":"ok","timestamp":1713429593228,"user_tz":240,"elapsed":29348413,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"5cba6825-7fd0-40fc-c417-6dc24d9e7f0a"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["\n","Epoch 1/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 1/50: \n","Train Loss 0.6309\t Train Perplexity 1.8806\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:25<59:48, 85.44s/it, running_distance=37.6875]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRING THAT IN ABLE SOLUND IN A TRANCE TO HOW HIS BODY STIFF AND UNSUMPORTANTED EXCEPTED TO POINTS THE HEND AND HEALS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:40<30:15, 44.29s/it, running_distance=28.6172]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HAD GENTLY AND GRATITUDE WHERE TO YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:11<25:22, 38.06s/it, running_distance=29.8646]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAREL EXPOSED HIM TO THE PENLED HIS OF THE JULY AND LAW BOTHS IN A FEW OBER MENTS AND A SIPPERSISIONS AGE HE WAS THE CURE OF AND EVEN OF PRACE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:42<22:51, 35.17s/it, running_distance=31.0938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AND THE SCAPE MORE IT MORE FROM INTENCITY PORT SUNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:05<33:11, 52.40s/it, running_distance=34.5781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  WITH THE EDDER INTO HIS DUPEN WITH SUFFICIENT WANTER TO COVERT LET ITS TWO GENTLY TO QUITE THEN WHEN TAKE IT OUT TO COAK\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 34.5781\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 2/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 2/50: \n","Train Loss 0.5918\t Train Perplexity 1.8077\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 3/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 3/50: \n","Train Loss 0.5572\t Train Perplexity 1.7463\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:25<17:30, 25.00s/it, running_distance=31.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT IN ABLE SOLUND IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTANTED EXCEPTED TO POINTS THE HEND AND HEALS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:43<14:32, 21.28s/it, running_distance=24.4297]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WHERE TOO HAVE BE THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:19<18:29, 27.74s/it, running_distance=25.3854]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAREL EXPOSED HIM TO THE PENLED HIS OF THE JULY IN LAW BUT IN A FEEBLE GAVER MEN SAME A SIPPORSITIOUS AGE HE WAS THE KIROVER INCUNITY AND EVEN OF PRAYS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:41<32:04, 49.36s/it, running_distance=28.4219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  FLEE RANGE WHITER LAST LONGER AT THE SCEEP MORE AT MORE FROM INTENSITY PORTS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:04<38:58, 61.54s/it, running_distance=29.9469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  WITH THE UTTER INTO HIS DO PEN WITH SUFFICIENT WANDER TO COVERT LET IT STUDENLY TO QUITE THEN WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 29.9469\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 4/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 4/50: \n","Train Loss 0.5276\t Train Perplexity 1.6953\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 5/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 5/50: \n","Train Loss 0.5013\t Train Perplexity 1.6513\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:29<20:49, 29.75s/it, running_distance=24.9844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT IN ABLE SOLEND IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUMPORTANTED EXCIPTED TRUE POINTS THE HEND AND HAILS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:44<14:29, 21.20s/it, running_distance=20.1562]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WHERE TO YOU HAVE BEEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:18<17:54, 26.87s/it, running_distance=21.1510]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAREL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEEBLE GOVER MEN SAME A SUPPERSISTIOUS AGE HE WAS THE CURE OF AND KINNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:51<19:05, 29.36s/it, running_distance=23.0977]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  FOR WE RANGE WHITER BLAST LONGER AT THE SCAPE MORE AT MORE FROM INTENCITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [03:14<30:53, 48.78s/it, running_distance=24.1781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  WITH THE UTTER INTO HIS DUPEN WITH SUFFICIENT WANDER TO COVERED LET ITS TOO GENTLY TO QUITE THEN WHEN TAKE IT OUT TO COOK\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 24.1781\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 6/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 6/50: \n","Train Loss 0.4783\t Train Perplexity 1.6137\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 7/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 7/50: \n","Train Loss 0.4574\t Train Perplexity 1.5802\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:25<59:38, 85.21s/it, running_distance=28.6094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT IN ABLE SOLEN IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCIPTED TRUE POINTS THE HEND AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:42<31:04, 45.48s/it, running_distance=21.4922]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WHERE TO YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:18<27:13, 40.83s/it, running_distance=21.6615]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAREL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEEBLE GOVER MENTS AND A SUPPERSICIOUS AGE HE WAS THE KIR OF INCUNITY AND EVEN OF PRAYS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:40<37:15, 57.32s/it, running_distance=24.1016]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  FLEE RANGE WHITER BLAST LONGER AND THE SCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [05:04<42:13, 66.66s/it, running_distance=24.7344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  FOR THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TWO GENTLY TO QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 24.7344\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 8/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 8/50: \n","Train Loss 0.4404\t Train Perplexity 1.5537\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 9/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 9/50: \n","Train Loss 0.4231\t Train Perplexity 1.5270\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:26<18:51, 26.93s/it, running_distance=21.7031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT IN ABLE SOLEN IN A TRANCE TOOKLED AS BODY STIFF AND UNSUMPORTED EXCIPTED TWO POINTS THE HEND AND HAILS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:45<14:52, 21.77s/it, running_distance=17.8672]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WHERE TO YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<18:34, 27.87s/it, running_distance=18.9479]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CARROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVER MENTS AND A SUPPERSISIS AGE HE WAS THE CURE OF AND KINNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:43<32:13, 49.57s/it, running_distance=21.4219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:06<39:04, 61.71s/it, running_distance=22.1313]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  WITH THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LED ITS TO GENTLY TO QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 22.1313\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 10/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 10/50: \n","Train Loss 0.4075\t Train Perplexity 1.5033\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 11/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 11/50: \n","Train Loss 0.3957\t Train Perplexity 1.4857\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:25<18:05, 25.84s/it, running_distance=20.9219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT IN ABLE SO LONG IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCIPTED TO POINTS THE HEAD AND HEALS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:43<14:30, 21.23s/it, running_distance=16.9922]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:16<17:38, 26.45s/it, running_distance=17.4844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVER MEN AND A SUPERSISTIOUS AGE HE WAS THE CURE OF AND KINNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:39<31:41, 48.74s/it, running_distance=20.6719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:02<38:36, 60.97s/it, running_distance=21.2375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  FOR THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS DUGENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 21.2375\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 12/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 12/50: \n","Train Loss 0.3818\t Train Perplexity 1.4651\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 13/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 13/50: \n","Train Loss 0.3695\t Train Perplexity 1.4473\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:33<23:45, 33.94s/it, running_distance=21.3125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT THE NABLE SOLUNG IN A TRANSED TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TRUE POINTS THE HEND AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:51<16:37, 24.33s/it, running_distance=17.3594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD TO YOU HAVE BE THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<18:06, 27.17s/it, running_distance=17.5156]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVER MENTS AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:44<31:48, 48.94s/it, running_distance=19.8320]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:07<38:46, 61.23s/it, running_distance=20.4281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  BUT THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TWO GENTLY TOO QUITE DONE WHEN TAKE IT OUT TO COOR\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 20.4281\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 14/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 14/50: \n","Train Loss 0.3595\t Train Perplexity 1.4328\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 15/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 15/50: \n","Train Loss 0.3496\t Train Perplexity 1.4188\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:32<22:39, 32.37s/it, running_distance=20.7969]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT IN NABLE SO LONG IN A TRANST TOOPLE AS BODY STIFF AND UNSUMPORTED EXCEPTED TRUE POINTS THE HEND AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:48<15:24, 22.55s/it, running_distance=16.6797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:03<31:02, 46.56s/it, running_distance=17.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIAL GOVER MENTS AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:36<26:56, 41.44s/it, running_distance=18.9922]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SKEEP MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [03:59<35:40, 56.33s/it, running_distance=19.1594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 19.1594\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 16/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 16/50: \n","Train Loss 0.3400\t Train Perplexity 1.4052\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 17/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 17/50: \n","Train Loss 0.3312\t Train Perplexity 1.3929\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:28<19:46, 28.26s/it, running_distance=19.4688]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TRUE POINTS THE HEAD AND HEALS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:44<14:18, 20.95s/it, running_distance=16.0703]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:58<30:22, 45.57s/it, running_distance=17.0990]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVER MENT AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:21<39:11, 60.31s/it, running_distance=19.2539]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:44<43:16, 68.34s/it, running_distance=19.2281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  FOR THE UTTER INTO A STUPEN WITH SUFFICIENT WANDER TO COVER IT LED ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 19.2281\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 18/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 18/50: \n","Train Loss 0.3232\t Train Perplexity 1.3817\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 19/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 19/50: \n","Train Loss 0.3154\t Train Perplexity 1.3711\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:34<23:49, 34.04s/it, running_distance=19.0156]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TWO POINTS THE HEND AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:51<16:45, 24.53s/it, running_distance=15.0469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:28<19:55, 29.88s/it, running_distance=15.3750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEEBLE GOVER MENT AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:50<32:59, 50.76s/it, running_distance=17.5352]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:13<39:29, 62.36s/it, running_distance=18.6125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 18.6125\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 20/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 20/50: \n","Train Loss 0.3081\t Train Perplexity 1.3610\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 21/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 21/50: \n","Train Loss 0.3013\t Train Perplexity 1.3517\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:28<20:14, 28.92s/it, running_distance=18.5312]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THE NABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEND AND HILLS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:44<14:29, 21.21s/it, running_distance=14.9609]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<18:34, 27.86s/it, running_distance=14.9844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVERNMENTS AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:43<32:09, 49.49s/it, running_distance=17.2539]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:05<38:55, 61.47s/it, running_distance=17.8281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 17.8281\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 22/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 22/50: \n","Train Loss 0.2958\t Train Perplexity 1.3444\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 23/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 23/50: \n","Train Loss 0.2890\t Train Perplexity 1.3353\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:32<22:29, 32.13s/it, running_distance=18.2656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCIPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:47<15:23, 22.54s/it, running_distance=14.8203]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:17<17:10, 25.76s/it, running_distance=14.5156]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW BUT IN A FEVIL GOVER MEN AND A SUPER SISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:40<31:18, 48.17s/it, running_distance=16.7656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER BLAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:02<38:17, 60.46s/it, running_distance=17.5375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEW PEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 17.5375\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 24/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 24/50: \n","Train Loss 0.2846\t Train Perplexity 1.3294\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 25/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 25/50: \n","Train Loss 0.2787\t Train Perplexity 1.3215\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:30<21:09, 30.22s/it, running_distance=18.4844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:47<15:32, 22.74s/it, running_distance=14.6172]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<18:52, 28.32s/it, running_distance=14.3750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MEN AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:45<32:22, 49.81s/it, running_distance=16.8711]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WATER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:08<39:05, 61.72s/it, running_distance=17.0688]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 17.0688\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 26/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 26/50: \n","Train Loss 0.2740\t Train Perplexity 1.3153\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 27/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 27/50: \n","Train Loss 0.2697\t Train Perplexity 1.3097\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:30<21:01, 30.04s/it, running_distance=17.5000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLE SO ONE IN A TRANS TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TWO POINTS THE HEAD AND HILLS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:48<15:44, 23.04s/it, running_distance=13.8125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:02<31:04, 46.61s/it, running_distance=14.6094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MENT AND A SUPPER SICIOUS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:25<39:36, 60.93s/it, running_distance=16.8125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT THE SKEEP MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:48<43:37, 68.87s/it, running_distance=16.9906]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.9906\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 28/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 28/50: \n","Train Loss 0.2650\t Train Perplexity 1.3036\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 29/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 29/50: \n","Train Loss 0.2609\t Train Perplexity 1.2982\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:32<22:31, 32.18s/it, running_distance=17.2812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANST TOOK OLD HIS BODY STIFF AND UNSUMPORTED EXCIPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:47<15:16, 22.37s/it, running_distance=13.5625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:19<17:40, 26.52s/it, running_distance=13.4323]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THUS IN A FEEBLE GOVER MEN AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:42<31:42, 48.79s/it, running_distance=15.8516]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:05<38:43, 61.16s/it, running_distance=16.0219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.0219\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 30/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 30/50: \n","Train Loss 0.2573\t Train Perplexity 1.2936\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 31/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 31/50: \n","Train Loss 0.2540\t Train Perplexity 1.2893\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:24<59:29, 84.99s/it, running_distance=19.5938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRINK THAT ENABLE SO LONG IN A TRANSE TOOK HOLD HIS BODY STIFF AND UNSUMPORTED EXCIPITED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:42<31:02, 45.42s/it, running_distance=14.7422]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:12<25:28, 38.22s/it, running_distance=14.0208]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSITE HIM TO THE PENDLED HE'S OF THE JULIAN LAW BUT IN A FEEBLE GOVER MEN TO A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:34<36:11, 55.67s/it, running_distance=16.1719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:57<41:28, 65.48s/it, running_distance=16.3719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERENT LET ITS DO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.3719\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 32/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 32/50: \n","Train Loss 0.2507\t Train Perplexity 1.2850\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 33/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 33/50: \n","Train Loss 0.2476\t Train Perplexity 1.2810\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:34<24:23, 34.85s/it, running_distance=16.8438]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:52<17:05, 25.01s/it, running_distance=13.2188]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY AND GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:25<18:48, 28.22s/it, running_distance=12.9740]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSITE HIM TO THE PENDLE HE'S OF THE JULIAN LAW THUS IN A FEEBLE GOVER MEN TO AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:51<17:58, 27.64s/it, running_distance=14.3789]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [03:14<30:04, 47.49s/it, running_distance=14.8000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 14.8000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 34/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 34/50: \n","Train Loss 0.2453\t Train Perplexity 1.2782\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 35/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 35/50: \n","Train Loss 0.2423\t Train Perplexity 1.2743\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:52<36:54, 52.73s/it, running_distance=17.4062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:08<21:08, 30.93s/it, running_distance=13.7656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:38<20:17, 30.44s/it, running_distance=13.1615]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MEN TO AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:00<33:06, 50.93s/it, running_distance=15.5977]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:23<39:33, 62.47s/it, running_distance=15.7344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS DO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 15.7344\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 36/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 36/50: \n","Train Loss 0.2401\t Train Perplexity 1.2715\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 37/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 37/50: \n","Train Loss 0.2374\t Train Perplexity 1.2681\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:26<18:46, 26.83s/it, running_distance=17.2188]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:42<13:51, 20.27s/it, running_distance=13.5000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:57<30:04, 45.12s/it, running_distance=14.2031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MENT AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:19<38:57, 59.94s/it, running_distance=16.1445]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INENENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:42<43:08, 68.11s/it, running_distance=16.0125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.0125\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 38/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 38/50: \n","Train Loss 0.2359\t Train Perplexity 1.2662\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 39/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 39/50: \n","Train Loss 0.2341\t Train Perplexity 1.2638\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:52<37:04, 52.96s/it, running_distance=17.5938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANS TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:08<21:11, 31.02s/it, running_distance=13.5625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORD YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:23<33:58, 50.95s/it, running_distance=14.4167]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSITE HIM TO THE PENDLED HE'S OF THE JULIAN LAW THAT IN A FEEBLE GOVER MENT AND A SUPERSISHES AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:45<41:11, 63.37s/it, running_distance=16.3086]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [05:08<44:34, 70.39s/it, running_distance=16.2094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.2094\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 40/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 40/50: \n","Train Loss 0.2325\t Train Perplexity 1.2619\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 41/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 41/50: \n","Train Loss 0.2309\t Train Perplexity 1.2599\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:26<18:51, 26.95s/it, running_distance=16.3594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO ONE IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:44<14:45, 21.59s/it, running_distance=12.8281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<18:44, 28.11s/it, running_distance=12.8854]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MEN TO AND A SUPER SICIOUS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:43<32:13, 49.57s/it, running_distance=15.1641]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INENENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:05<38:57, 61.51s/it, running_distance=15.1750]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 15.1750\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 42/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 42/50: \n","Train Loss 0.2300\t Train Perplexity 1.2587\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 43/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 43/50: \n","Train Loss 0.2292\t Train Perplexity 1.2577\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:33<23:44, 33.92s/it, running_distance=17.1250]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:49<15:52, 23.23s/it, running_distance=13.4062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:04<31:07, 46.69s/it, running_distance=14.2448]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVER MEN TO AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:26<39:34, 60.88s/it, running_distance=16.1914]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:49<43:36, 68.84s/it, running_distance=15.8469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 15.8469\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 44/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 44/50: \n","Train Loss 0.2280\t Train Perplexity 1.2562\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 45/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 45/50: \n","Train Loss 0.2279\t Train Perplexity 1.2561\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:33<23:29, 33.56s/it, running_distance=16.7812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:49<15:45, 23.06s/it, running_distance=13.1797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:04<31:11, 46.79s/it, running_distance=13.9531]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVERNMENT AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:26<39:31, 60.82s/it, running_distance=16.1641]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [04:49<43:34, 68.80s/it, running_distance=15.8594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 15.8594\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 46/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 46/50: \n","Train Loss 0.2267\t Train Perplexity 1.2546\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 47/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 47/50: \n","Train Loss 0.2265\t Train Perplexity 1.2543\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:51<36:00, 51.45s/it, running_distance=17.1250]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:07<20:45, 30.38s/it, running_distance=13.4219]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:21<33:44, 50.62s/it, running_distance=14.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVERNMENT AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [03:44<41:07, 63.27s/it, running_distance=16.1797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [05:07<44:35, 70.42s/it, running_distance=15.8344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 15.8344\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 48/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 48/50: \n","Train Loss 0.2266\t Train Perplexity 1.2544\t Learning Rate 0.0000\n","Saved epoch model\n","\n","Epoch 49/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 49/50: \n","Train Loss 0.2258\t Train Perplexity 1.2534\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [01:24<59:27, 84.95s/it, running_distance=19.1562]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH THAT ENABLE SO LONG IN A TRANSE TO HOLD HIS BODY STIFF AND UNSUMPORTAGE EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:40<30:11, 44.19s/it, running_distance=14.3672]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAD GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [02:55<38:46, 58.16s/it, running_distance=14.7448]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENDLED HIS OF THE JULIAN LAW THAT IN A FEEBLE GOVERNMENT AND A SUPER SICH HIS AGE HE WAS THE CURE OF INCUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [04:18<44:07, 67.90s/it, running_distance=16.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGE WHITER LAST LONGER AT ESCAPE MORE AT MORE AT MORE FROM INENENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["Val:  12%|█▏        | 5/43 [05:41<46:26, 73.34s/it, running_distance=16.2375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE UTTER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS TOO GENTLY TOOK QUITE DONE WHEN TAKE IT OUT TO COO\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["Levenshtein Distance 16.2375\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 50/50\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 50/50: \n","Train Loss 0.2262\t Train Perplexity 1.2539\t Learning Rate 0.0000\n","Saved epoch model\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 640x480 with 2 Axes>"],"image/png":"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\n"},"metadata":{}}]},{"cell_type":"code","source":["validate_full(full_model, val_loader)"],"metadata":{"id":"EV72WOmGARVm"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["### TODO Unfreeze the weights of full transformer input embedding and decoder\n","# @NOTE: after unfreezing, continue training the P2 model for more epochs to achieve best results!\n","\n","for param in full_model.encoder.embedding.parameters():\n","    param.requires_grad = True # TODO make it trainable\n","\n","for param in full_model.encoder.projection.parameters():\n","    param.requires_grad = True # TODO make it trainable\n","\n","for param in full_model.decoder.parameters():\n","    param.requires_grad = True # TODO make it trainable"],"metadata":{"id":"o3qqn4nbHCdj"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Testing"],"metadata":{"id":"p2h4hspIreu7"}},{"cell_type":"code","source":["full_model.eval()\n","\n","# progress bar\n","batch_bar = tqdm(total=len(test_loader), dynamic_ncols=True, leave=False, position=0, desc=\"Test\", ncols=5)\n","\n","all_predictions = []\n","\n","for i, data in enumerate(test_loader):\n","    inputs, inputs_lengths = data\n","    inputs                 = inputs.to(device)\n","\n","    with torch.inference_mode():\n","        greedy_predictions = full_model.recognize(inputs, inputs_lengths)\n","\n","    greedy_predictions = greedy_predictions.detach().cpu().numpy()\n","\n","    for batch_idx in range(greedy_predictions.shape[0]):\n","        pred_sliced = indices_to_chars(greedy_predictions[batch_idx], vocab= VOCAB)\n","        pred_string = ''.join(pred_sliced)\n","        all_predictions.append(pred_string)\n","\n","    batch_bar.update()"],"metadata":{"id":"Tf-xgN8rBAWT","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713431827048,"user_tz":240,"elapsed":2232963,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"25a91d6d-4c41-44f9-e63a-6d592de966d4"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stderr","text":["Test: 100%|██████████| 41/41 [37:12<00:00, 57.40s/it]"]}]},{"cell_type":"code","source":["%cd /content/\n","df = pd.DataFrame({\n","    \"index\" : list(range(len(all_predictions))),\n","    \"label\" : all_predictions\n","})\n","\n","df.to_csv(\"improved.csv\", index=False)"],"metadata":{"id":"Nwz_XRqnBTbr","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713440952383,"user_tz":240,"elapsed":161,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"2d8d6517-5849-4e18-a91f-0dbe4982eb5a"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["/content\n"]}]},{"cell_type":"code","source":["!kaggle competitions submit -c hw4p2-sp24 -f improved.csv -m \"It's been a ride :)\""],"metadata":{"id":"vJD6b3AgBlT-","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713365209397,"user_tz":240,"elapsed":993,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"f79eb47c-a131-447d-887f-844a83e1ce84"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Traceback (most recent call last):\n","  File \"/usr/local/bin/kaggle\", line 8, in <module>\n","    sys.exit(main())\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/cli.py\", line 67, in main\n","    out = args.func(**command_args)\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 561, in competition_submit_cli\n","    submit_result = self.competition_submit(file_name, message,\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 532, in competition_submit\n","    url_result['createUrl'], quiet)\n","KeyError: 'createUrl'\n"]}]},{"cell_type":"markdown","source":["# Part #3 (Improving your Model)\n","* Think of the input embedding, can you have something better than just an CNN-LSTM? Hint: HW3P2 might perform better.\n","* Think about the transformer encoder parameter, what about number of heads and number of layers, can this help?\n","* What about the decoder parameters what can  we improve there?\n","* We trained using an incremental training technique, can we use this technique to add more layers?\n","\n","**Good Luck!!**"],"metadata":{"id":"vacdiL_WCvgW"}},{"cell_type":"code","source":["full_model = FullTransformer(\n","input_dim                   = 27,\n","enc_num_layers              = config['enc_num_layers'],\n","dec_num_layers              = config['dec_num_layers'],\n","enc_num_heads               = config['enc_num_heads'],\n","dec_num_heads               = config['dec_num_heads'],\n","\n","d_model                     = config['d_model'],\n","d_ff                        = config['d_ff'],\n","\n","target_vocab_size           = len(VOCAB),\n","eos_token                   = EOS_TOKEN,\n","sos_token                   = SOS_TOKEN,\n","pad_token                   = PAD_TOKEN,\n","\n","dropout                     = config['enc_dropout'],\n","\n","# decrease to a small number if you are just trying to implement the network\n","# we've computed the max length of the transcripts for you\n","trans_max_seq_length        = 550,\n","mfcc_max_seq_length         = dataset_max_len).to(device)\n","\n","full_model.load_state_dict(torch.load('/content/checkpointsfull/checkpoint-best-loss-modelfull.pth')['model_state_dict'])\n","\n","full_model_new = FullTransformer(\n","input_dim                   = 27,\n","enc_num_layers              = 3,\n","dec_num_layers              = config['dec_num_layers'],\n","enc_num_heads               = 2,\n","dec_num_heads               = config['dec_num_heads'],\n","\n","d_model                     = config['d_model'],\n","d_ff                        = config['d_ff'],\n","\n","target_vocab_size           = len(VOCAB),\n","eos_token                   = EOS_TOKEN,\n","sos_token                   = SOS_TOKEN,\n","pad_token                   = PAD_TOKEN,\n","\n","dropout                     = config['enc_dropout'],\n","\n","# decrease to a small number if you are just trying to implement the network\n","# we've computed the max length of the transcripts for you\n","trans_max_seq_length        = 550,\n","mfcc_max_seq_length         = dataset_max_len).to(device)"],"metadata":{"id":"V0IEKgyjyXG4","executionInfo":{"status":"ok","timestamp":1713622964827,"user_tz":240,"elapsed":949,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":84,"outputs":[]},{"cell_type":"code","source":["def load_and_freeze_parameters(old_model, old_layers_number, new_model):\n","    #\n","    for old_layer, new_layer in zip(old_model.encoder.enc_layers, new_model.encoder.enc_layers):\n","        new_layer.load_state_dict(old_layer.state_dict())\n","\n","    #\n","    for layer in new_model.encoder.enc_layers[0:old_layers_number]:\n","        for param in layer.parameters():\n","            param.requires_grad = False\n","\n","    #\n","    for layer in new_model.encoder.enc_layers[old_layers_number:]:\n","        for param in layer.parameters():\n","            param.requires_grad = True\n","\n"],"metadata":{"id":"koI0T8OTzbCj"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["load_and_freeze_parameters(full_model, 2, full_model_new)"],"metadata":{"id":"zKjPZw381_-Z"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["# using WandB? resume training?\n","\n","USE_WANDB = True\n","RESUME_LOGGING = True\n","\n","\n","# creating your WandB run\n","run_name = \"{}_Transformer_ENC-{}/{}_DEC-{}/{}_{}_{}_{}_{}_Lr:{}\".format(\n","    config[\"Name\"],\n","    2,       # only used in Part II with the Transformer Encoder\n","    2,        # only used in Part II with the Transformer Encoder\n","    config[\"dec_num_layers\"],\n","    config[\"dec_num_heads\"],\n","    config[\"d_model\"],\n","    config[\"d_ff\"],\n","    config[\"optimizer\"],\n","    config[\"scheduler\"],\n","    config['learning_rate'])\n","\n","if USE_WANDB:\n","\n","    wandb.login(key=\"b9fec08c02d03c41bdf5fe1f5268589277270fbe\", relogin=True) # TODO enter your key here\n","\n","    if RESUME_LOGGING:\n","        run_id = \"c4cdof4t\"\n","        run = wandb.init(\n","            id     = run_id,        ### Insert specific run id here if you want to resume a previous run\n","            resume = True,          ### You need this to resume previous runs, but comment out reinit=True when using this\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","        )\n","\n","    else:\n","        run = wandb.init(\n","            name    = run_name,     ### Wandb creates random run names if you skip this field, we recommend you give useful names\n","            reinit  = True,         ### Allows reinitalizing runs when you re-run this cell\n","            project = \"HW4P2-S24\",  ### Project should be created in your wandb account\n","            config  = config        ### Wandb Config for your run\n","        )\n","\n","        ### Save your model architecture as a string with str(model)\n","        model_arch  = str(full_model_new)\n","\n","        ### Save it in a txt file\n","        arch_file   = open(\"model_arch.txt\", \"w\")\n","        file_write  = arch_file.write(model_arch)\n","        arch_file.close()\n","\n","        ### Log it in your wandb run with wandb.save()\n","        wandb.save(\"model_arch.txt\")"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":410,"referenced_widgets":["f8a088954bb64422a9c15d57c46dbc5f","c252db17298d44e194878b1534a67525","cd265a1376734407ab68d3c6461deadc","8c8a5aec6f5b4669bdf21383eece0b10","60433fd3f3954044b71a888b45b93b62","c57b11408cc34c2096423684db29578f","17de0d628c1141cf819838f3dbe5c792","9ecf9ea2ceb24c59842b27a9d7c0f463"]},"id":"kJ2OIIh33Smi","executionInfo":{"status":"ok","timestamp":1713636093194,"user_tz":240,"elapsed":9235,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"09a785d5-d414-435f-ce61-bfdcb921619f"},"execution_count":94,"outputs":[{"output_type":"stream","name":"stderr","text":["\u001b[34m\u001b[1mwandb\u001b[0m: \u001b[33mWARNING\u001b[0m Calling wandb.login() after wandb.init() has no effect.\n"]},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Finishing last run (ID:c4cdof4t) before initializing another..."]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["VBox(children=(Label(value='0.006 MB of 0.006 MB uploaded\\r'), FloatProgress(value=1.0, max=1.0)))"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"f8a088954bb64422a9c15d57c46dbc5f"}},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["<style>\n","    table.wandb td:nth-child(1) { padding: 0 10px; text-align: left ; width: auto;} td:nth-child(2) {text-align: left ; width: 100%}\n","    .wandb-row { display: flex; flex-direction: row; flex-wrap: wrap; justify-content: flex-start; width: 100% }\n","    .wandb-col { display: flex; flex-direction: column; flex-basis: 100%; flex: 1; padding: 10px; }\n","    </style>\n","<div class=\"wandb-row\"><div class=\"wandb-col\"><h3>Run history:</h3><br/><table class=\"wandb\"><tr><td>learning_rate</td><td>████████████████▁▁▁▁▁▁</td></tr><tr><td>train_loss</td><td>▇██▇▇▇▆▇▆▆▅▄▄▄▄▄▃▃▁▁▁▁</td></tr><tr><td>train_perplexity</td><td>▇██▇▇▇▆▇▆▆▅▄▄▄▄▄▃▃▁▁▁▁</td></tr><tr><td>val_distance</td><td>▃▃▁▃▁█▄▃▃▄▄</td></tr></table><br/></div><div class=\"wandb-col\"><h3>Run summary:</h3><br/><table class=\"wandb\"><tr><td>learning_rate</td><td>9e-05</td></tr><tr><td>train_loss</td><td>0.07664</td></tr><tr><td>train_perplexity</td><td>1.07968</td></tr><tr><td>val_distance</td><td>14.725</td></tr></table><br/></div></div>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run <strong style=\"color:#cdcd00\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_ReduceLR_Lr:1E-4</strong> at: <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/c4cdof4t/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/c4cdof4t/workspace</a><br/>Synced 5 W&B file(s), 0 media file(s), 0 artifact file(s) and 1 other file(s)"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Find logs at: <code>./wandb/run-20240420_142300-c4cdof4t/logs</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Successfully finished last run (ID:c4cdof4t). Initializing new run:<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["wandb version 0.16.6 is available!  To upgrade, please run:\n"," $ pip install wandb --upgrade"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Tracking run with wandb version 0.16.5"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Run data is saved locally in <code>/content/wandb/run-20240420_180123-c4cdof4t</code>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":["Resuming run <strong><a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/c4cdof4t/workspace' target=\"_blank\">_Transformer_ENC-2/2_DEC-2/2_512_2048_AdamW_ReduceLR_Lr:1E-4</a></strong> to <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">Weights & Biases</a> (<a href='https://wandb.me/run' target=\"_blank\">docs</a>)<br/>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View project at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24</a>"]},"metadata":{}},{"output_type":"display_data","data":{"text/plain":["<IPython.core.display.HTML object>"],"text/html":[" View run at <a href='https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/c4cdof4t/workspace' target=\"_blank\">https://wandb.ai/11785-hw3p2/HW4P2-S24/runs/c4cdof4t/workspace</a>"]},"metadata":{}}]},{"cell_type":"markdown","metadata":{"id":"GNaYGhToyfjv"},"source":["## Loss, Optimizer, and Scheduler Definitions"]},{"cell_type":"code","source":["loss_func   = nn.CrossEntropyLoss(ignore_index = PAD_TOKEN)\n","scaler      = torch.cuda.amp.GradScaler()"],"metadata":{"executionInfo":{"status":"ok","timestamp":1713636098586,"user_tz":240,"elapsed":140,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"id":"b1F9MXFuyfjv"},"execution_count":95,"outputs":[]},{"cell_type":"code","source":["''' defining optimizer '''\n","\n","if config[\"optimizer\"] == \"SGD\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.SGD(full_model.parameters(),\n","                              lr=config[\"learning_rate\"],\n","                              momentum=config[\"momentum\"],\n","                              weight_decay=1E-4,\n","                              nesterov=config[\"nesterov\"])\n","\n","elif config[\"optimizer\"] == \"Adam\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.Adam(full_model.parameters(),\n","                               lr=float(config[\"learning_rate\"]),\n","                               weight_decay=1e-4)\n","\n","elif config[\"optimizer\"] == \"AdamW\":\n","  # feel free to change any of the initializations you like to fit your needs\n","  optimizer = torch.optim.AdamW(full_model.parameters(),\n","                                lr=float(config[\"learning_rate\"]),\n","                                weight_decay=0.01)\n","\n","''' defining scheduler '''\n","\n","if config[\"scheduler\"] == \"ReduceLR\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,\n","                factor=config[\"factor\"], patience=config[\"patience\"], min_lr=1E-8, verbose=True)\n","\n","elif config[\"scheduler\"] == \"CosineAnnealing\":\n","  #Feel Free to change any of the initializations you like to fit your needs\n","  scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer,\n","                T_max = config[\"epochs\"], eta_min=1E-8)"],"metadata":{"executionInfo":{"status":"ok","timestamp":1713636103327,"user_tz":240,"elapsed":158,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"id":"FA8iunN-yfjw"},"execution_count":96,"outputs":[]},{"cell_type":"code","source":["###  Now let's train the encoder to master the encoder input ranges\n","e                   = 0\n","best_loss           = 10.0\n","\n","checkpoint_root = os.path.join(os.getcwd(), 'checkpointsfull')\n","os.makedirs(checkpoint_root, exist_ok=True)\n","wandb.watch(full_model, log=\"all\")\n","\n","checkpoint_best_loss_model_filename     = 'checkpoint-best-loss-modelfull.pth'\n","checkpoint_last_epoch_filename          = 'checkpoint-epochfull-'\n","best_loss_model_path                    = os.path.join(checkpoint_root, checkpoint_best_loss_model_filename)\n","\n","if RESUME_LOGGING:\n","    # change if you want to load best test model accordingly\n","    # checkpoint = torch.load(wandb.restore(checkpoint_best_loss_model_filename, run_path=\"\"+run_id).name)\n","    full_model, optimizer, scheduler, epoch, metric = load_model(best_loss_model_path, full_model, metric= \"train_loss\", optimizer= optimizer, scheduler= scheduler)\n","    # model.load_state_dict(checkpoint['model_state_dict'])\n","    # optimizer.load_state_dict(checkpoint['optimizer_state_dict'])\n","    e = epoch\n","\n","    print(\"Resuming from epoch {}\".format(epoch+1))\n","    print(\"Epochs left: \", config['epochs']-epoch)\n","    print(\"Optimizer: \\n\", optimizer)\n","    print(\"Scheduler: \\n\", scheduler.state_dict())\n","    print(\"metric: \\n\", metric)\n","    scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer,\n","                factor=config[\"factor\"], patience=config[\"patience\"], min_lr=1E-8, verbose=True)\n","    print(\"Scheduler: \\n\", scheduler.state_dict())\n","\n","torch.cuda.empty_cache()\n","gc.collect()\n","\n","epochs = config[\"epochs\"]\n","for epoch in range(e, epochs):\n","\n","    print(\"\\nEpoch {}/{}\".format(epoch+1, config[\"epochs\"]))\n","\n","    curr_lr = float(optimizer.param_groups[0][\"lr\"])\n","\n","    train_loss, train_perplexity, attention_weights = train_model(full_model, train_loader, optimizer)\n","\n","    print(\"\\nEpoch {}/{}: \\nTrain Loss {:.04f}\\t Train Perplexity {:.04f}\\t Learning Rate {:.04f}\".format(\n","        epoch + 1, config[\"epochs\"], train_loss, train_perplexity, curr_lr))\n","\n","    if (epoch % 2 == 0):    # validate every 2 epochs to speed up training\n","        levenshtein_distance = validate_fast(full_model, val_loader)\n","        print(\"Levenshtein Distance {:.04f}\".format(levenshtein_distance))\n","\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr,\n","                \"val_distance\"      : levenshtein_distance})\n","\n","    else:\n","        wandb.log({\"train_loss\"     : train_loss,\n","                \"train_perplexity\"  : train_perplexity,\n","                \"learning_rate\"     : curr_lr})\n","\n","    # plotting the encoder-nearest and decoder-nearest attention weights\n","    attention_keys = list(attention_weights.keys())\n","\n","    # plot_attention_weights((attention_weights[attention_keys[0]][0]).cpu().detach().numpy())\n","    # plot_attention_weights(attention_weights[attention_keys[-1]][0].cpu().detach().numpy())\n","\n","    attention_weights_decoder_self       = attention_weights[attention_keys[0]][0].cpu().detach().numpy()\n","    attention_weights_decoder_cross      = attention_weights[attention_keys[-1]][0].cpu().detach().numpy()\n","    save_attention_plot(attention_weights_decoder_cross, epoch)\n","\n","    if config[\"scheduler\"] == \"ReduceLR\":\n","        scheduler.step(levenshtein_distance)\n","    else:\n","        scheduler.step()\n","\n","    ### Highly Recommended: Save checkpoint in drive and/or wandb if accuracy is better than your current best\n","    epoch_model_path = os.path.join(checkpoint_root, (checkpoint_last_epoch_filename + str(epoch) + '.pth'))\n","    save_model(full_model, optimizer, scheduler, ['train_loss', train_loss], epoch, epoch_model_path)\n","    ## wandb.save(epoch_model_path) ## Can't save on wandb for all epochs, may blow up storage\n","\n","    print(\"Saved epoch model\")\n","\n","    # the first time we add the transformer encoder block, we freeze all other weights\n","    #   except the transformer encoder weights and train this partially-frozen model\n","    #   for about 3 epochs to weight initialize the transformer encoder weights.\n","    #   This speeds up convergence of the entire network.\n","    # if epoch == 3:\n","    #     break\n","\n","    if best_loss >= train_loss:\n","        best_loss = train_loss\n","        save_model(full_model, optimizer, scheduler, ['train_loss', train_loss], epoch, best_loss_model_path)\n","        # wandb.save(best_loss_model_path)\n","        print(\"Saved best training model\")\n","\n","### Finish your wandb run\n","# run.finish()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":1000},"id":"Ho1rR1oo3jA1","executionInfo":{"status":"error","timestamp":1713674049309,"user_tz":240,"elapsed":12122759,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"6ccc1ee5-fb2c-4993-a7d3-9613dc77a430"},"execution_count":101,"outputs":[{"output_type":"stream","name":"stdout","text":["Resuming from epoch 21\n","Epochs left:  80\n","Optimizer: \n"," AdamW (\n","Parameter Group 0\n","    amsgrad: False\n","    betas: (0.9, 0.999)\n","    capturable: False\n","    differentiable: False\n","    eps: 1e-08\n","    foreach: None\n","    fused: None\n","    lr: 9e-05\n","    maximize: False\n","    weight_decay: 0.01\n",")\n","Scheduler: \n"," {'factor': 0.9, 'min_lrs': [1e-08], 'patience': 10, 'verbose': True, 'cooldown': 0, 'cooldown_counter': 0, 'mode': 'min', 'threshold': 0.0001, 'threshold_mode': 'rel', 'best': 11.825, 'num_bad_epochs': 5, 'mode_worse': inf, 'eps': 1e-08, 'last_epoch': 21, '_last_lr': [9e-05]}\n","metric: \n"," 0.07648311088471402\n","Scheduler: \n"," {'factor': 0.9, 'min_lrs': [1e-08], 'patience': 6, 'verbose': True, 'cooldown': 0, 'cooldown_counter': 0, 'mode': 'min', 'threshold': 0.0001, 'threshold_mode': 'rel', 'best': inf, 'num_bad_epochs': 0, 'mode_worse': inf, 'eps': 1e-08, 'last_epoch': 0}\n","\n","Epoch 21/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 21/100: \n","Train Loss 0.0761\t Train Perplexity 1.0791\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:42<29:40, 42.40s/it, running_distance=21.7812]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<18:38, 27.29s/it, running_distance=15.6328]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:25<17:57, 26.94s/it, running_distance=12.9271]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:40<29:54, 46.03s/it, running_distance=15.7656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.9594\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 22/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 22/100: \n","Train Loss 0.0758\t Train Perplexity 1.0788\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 23/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 23/100: \n","Train Loss 0.0758\t Train Perplexity 1.0788\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:42<29:50, 42.62s/it, running_distance=17.5000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<18:43, 27.41s/it, running_distance=12.0234]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:26<18:03, 27.09s/it, running_distance=10.2760]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWELLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:41<30:05, 46.29s/it, running_distance=15.8867]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST BLONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 15.3562\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 24/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 24/100: \n","Train Loss 0.0756\t Train Perplexity 1.0785\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 25/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 25/100: \n","Train Loss 0.0747\t Train Perplexity 1.0776\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<26:53, 38.41s/it, running_distance=16.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:54<17:28, 25.57s/it, running_distance=11.3516]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORNS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:21<17:26, 26.17s/it, running_distance=9.7865]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:36<29:34, 45.50s/it, running_distance=12.4570]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DUE GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 12.5094\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 26/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 26/100: \n","Train Loss 0.0751\t Train Perplexity 1.0780\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 27/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 27/100: \n","Train Loss 0.0747\t Train Perplexity 1.0776\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<27:14, 38.93s/it, running_distance=22.5312]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:55<17:45, 25.98s/it, running_distance=14.6719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:32, 26.31s/it, running_distance=12.1667]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPER SISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:21<25:34, 39.34s/it, running_distance=14.3594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DUE GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.5781\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 28/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 28/100: \n","Train Loss 0.0741\t Train Perplexity 1.0770\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 29/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 29/100: \n","Train Loss 0.0741\t Train Perplexity 1.0769\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:43<30:29, 43.56s/it, running_distance=15.2031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANS TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:00<18:55, 27.70s/it, running_distance=10.8516]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:27<18:13, 27.33s/it, running_distance=9.4115]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:42<30:08, 46.37s/it, running_distance=11.6055]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DUE GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 11.6813\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 30/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 30/100: \n","Train Loss 0.0737\t Train Perplexity 1.0765\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 31/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 31/100: \n","Train Loss 0.0734\t Train Perplexity 1.0762\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<26:41, 38.13s/it, running_distance=19.9531]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:53<16:59, 24.87s/it, running_distance=13.4062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<17:09, 25.75s/it, running_distance=11.2344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISS'S AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:36<29:31, 45.42s/it, running_distance=15.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST BLONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 15.9656\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 32/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 32/100: \n","Train Loss 0.0733\t Train Perplexity 1.0761\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 33/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 33/100: \n","Train Loss 0.0728\t Train Perplexity 1.0756\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:45<31:48, 45.45s/it, running_distance=27.9688]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:02<19:35, 28.66s/it, running_distance=17.3281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:29<18:30, 27.76s/it, running_distance=13.5365]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISSUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:44<30:18, 46.62s/it, running_distance=15.9492]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 15.8031\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 34/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 34/100: \n","Train Loss 0.0726\t Train Perplexity 1.0754\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 35/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 35/100: \n","Train Loss 0.0718\t Train Perplexity 1.0745\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:41<28:54, 41.29s/it, running_distance=15.8594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:58<18:20, 26.85s/it, running_distance=11.1797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:24<17:50, 26.76s/it, running_distance=9.7604]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW THUS IN A FEEBLE GOVERNMENT AND A SUPER SISSUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:40<29:58, 46.10s/it, running_distance=13.5195]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WHITER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.8688\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 36/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 36/100: \n","Train Loss 0.0723\t Train Perplexity 1.0750\t Learning Rate 0.0001\n","Epoch 00016: reducing learning rate of group 0 to 8.1000e-05.\n","Saved epoch model\n","\n","Epoch 37/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 37/100: \n","Train Loss 0.0706\t Train Perplexity 1.0732\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:45<31:41, 45.27s/it, running_distance=15.4062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:01<19:06, 27.96s/it, running_distance=11.1641]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORNDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:28<18:19, 27.48s/it, running_distance=9.6615]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPER SISSUS AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:43<30:13, 46.50s/it, running_distance=13.3477]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.0344\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 38/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 38/100: \n","Train Loss 0.0708\t Train Perplexity 1.0734\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 39/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 39/100: \n","Train Loss 0.0700\t Train Perplexity 1.0726\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:41<28:51, 41.24s/it, running_distance=17.5625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:56<17:55, 26.23s/it, running_distance=12.3672]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORNS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:23<17:39, 26.48s/it, running_distance=10.6406]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPER SISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:39<29:49, 45.88s/it, running_distance=14.1992]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DUE GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.2000\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 40/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 40/100: \n","Train Loss 0.0703\t Train Perplexity 1.0729\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 41/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 41/100: \n","Train Loss 0.0694\t Train Perplexity 1.0719\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:34<24:19, 34.74s/it, running_distance=16.2344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:50<16:02, 23.48s/it, running_distance=11.3594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:16<16:36, 24.91s/it, running_distance=9.8125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULIE IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISS'S AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:06<22:34, 34.74s/it, running_distance=11.9883]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 11.7438\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 42/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 42/100: \n","Train Loss 0.0692\t Train Perplexity 1.0717\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 43/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 43/100: \n","Train Loss 0.0693\t Train Perplexity 1.0717\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:24<17:24, 24.87s/it, running_distance=13.1719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:40<13:17, 19.45s/it, running_distance=10.0469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:07<15:10, 22.76s/it, running_distance=8.8906]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULIE IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISSUS AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:53<20:51, 32.10s/it, running_distance=11.2344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 11.2438\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 44/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 44/100: \n","Train Loss 0.0697\t Train Perplexity 1.0723\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 45/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 45/100: \n","Train Loss 0.0682\t Train Perplexity 1.0707\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<27:08, 38.78s/it, running_distance=14.3594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:55<17:36, 25.77s/it, running_distance=10.5781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WERE NO YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:31, 26.28s/it, running_distance=9.2188]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW THAT'S IN A FEEBLE GOVERNMENT AND A SUPER SISSUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:37<29:42, 45.70s/it, running_distance=11.9023]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 11.9094\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 46/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 46/100: \n","Train Loss 0.0691\t Train Perplexity 1.0716\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 47/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 47/100: \n","Train Loss 0.0687\t Train Perplexity 1.0712\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:41<28:43, 41.05s/it, running_distance=18.8594]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:57<18:05, 26.47s/it, running_distance=12.8438]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:23<17:39, 26.48s/it, running_distance=10.9323]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:39<29:48, 45.87s/it, running_distance=13.8203]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.2781\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 48/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 48/100: \n","Train Loss 0.0681\t Train Perplexity 1.0705\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 49/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 49/100: \n","Train Loss 0.0683\t Train Perplexity 1.0707\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:39<27:57, 39.93s/it, running_distance=17.6406]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:56<17:53, 26.19s/it, running_distance=12.0625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:23<17:36, 26.42s/it, running_distance=10.2448]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:38<29:43, 45.73s/it, running_distance=12.8906]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DUE GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.2719\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 50/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 50/100: \n","Train Loss 0.0677\t Train Perplexity 1.0701\t Learning Rate 0.0001\n","Epoch 00030: reducing learning rate of group 0 to 7.2900e-05.\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 51/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 51/100: \n","Train Loss 0.0675\t Train Perplexity 1.0698\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:43<30:24, 43.44s/it, running_distance=13.1094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDER'S BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<18:50, 27.58s/it, running_distance=9.9062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:26<18:04, 27.12s/it, running_distance=8.8125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:26<26:07, 40.20s/it, running_distance=12.1680]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 12.2281\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 52/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 52/100: \n","Train Loss 0.0670\t Train Perplexity 1.0694\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 53/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 53/100: \n","Train Loss 0.0661\t Train Perplexity 1.0683\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:41<29:04, 41.52s/it, running_distance=20.7969]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:58<18:24, 26.93s/it, running_distance=13.5859]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:24<17:53, 26.83s/it, running_distance=11.2604]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPORCED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:11<22:29, 34.61s/it, running_distance=14.1719]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.9469\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 54/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 54/100: \n","Train Loss 0.0665\t Train Perplexity 1.0688\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 55/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 55/100: \n","Train Loss 0.0660\t Train Perplexity 1.0683\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:44<30:53, 44.14s/it, running_distance=15.8438]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<18:38, 27.28s/it, running_distance=11.5703]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:26<17:56, 26.92s/it, running_distance=9.8958]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTENCE AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:41<29:56, 46.07s/it, running_distance=13.3242]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.5844\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 56/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 56/100: \n","Train Loss 0.0663\t Train Perplexity 1.0686\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 57/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 57/100: \n","Train Loss 0.0658\t Train Perplexity 1.0680\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:44<31:16, 44.67s/it, running_distance=17.7031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:00<18:48, 27.53s/it, running_distance=12.1797]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:26<18:03, 27.09s/it, running_distance=10.1146]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:13<22:41, 34.91s/it, running_distance=13.0391]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 12.8469\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Epoch 00037: reducing learning rate of group 0 to 6.5610e-05.\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 58/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 58/100: \n","Train Loss 0.0649\t Train Perplexity 1.0670\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 59/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 59/100: \n","Train Loss 0.0644\t Train Perplexity 1.0666\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:44<31:29, 44.98s/it, running_distance=16.4375]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLDEST BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:01<19:17, 28.23s/it, running_distance=11.6562]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:28<18:23, 27.59s/it, running_distance=9.9062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:43<30:06, 46.33s/it, running_distance=12.6953]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 12.4062\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 60/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 60/100: \n","Train Loss 0.0645\t Train Perplexity 1.0666\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 61/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 61/100: \n","Train Loss 0.0640\t Train Perplexity 1.0662\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:21<14:49, 21.18s/it, running_distance=15.9844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:37<12:31, 18.34s/it, running_distance=11.3281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:04<14:44, 22.11s/it, running_distance=9.6302]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPER SISSUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [01:50<20:40, 31.82s/it, running_distance=11.6562]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 11.4187\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 62/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 62/100: \n","Train Loss 0.0639\t Train Perplexity 1.0660\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 63/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 63/100: \n","Train Loss 0.0637\t Train Perplexity 1.0658\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:39<27:40, 39.54s/it, running_distance=16.0625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:56<17:49, 26.07s/it, running_distance=11.2344]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:35, 26.38s/it, running_distance=9.5729]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:38<29:45, 45.79s/it, running_distance=12.9453]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.0781\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 64/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 64/100: \n","Train Loss 0.0636\t Train Perplexity 1.0656\t Learning Rate 0.0001\n","Epoch 00044: reducing learning rate of group 0 to 5.9049e-05.\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 65/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 65/100: \n","Train Loss 0.0630\t Train Perplexity 1.0650\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:36<25:31, 36.46s/it, running_distance=18.8906]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:51<16:27, 24.09s/it, running_distance=12.7031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY AND GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:18<16:50, 25.25s/it, running_distance=10.5938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULIE IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:33<29:15, 45.03s/it, running_distance=13.7031]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.0844\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 66/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 66/100: \n","Train Loss 0.0624\t Train Perplexity 1.0644\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 67/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 67/100: \n","Train Loss 0.0624\t Train Perplexity 1.0644\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<26:50, 38.34s/it, running_distance=15.1562]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:53<17:00, 24.90s/it, running_distance=10.9062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<17:07, 25.70s/it, running_distance=9.5000]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:36<29:29, 45.38s/it, running_distance=12.9102]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.0563\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 68/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 68/100: \n","Train Loss 0.0620\t Train Perplexity 1.0640\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 69/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 69/100: \n","Train Loss 0.0623\t Train Perplexity 1.0643\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:21<14:45, 21.08s/it, running_distance=18.5625]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:37<12:35, 18.42s/it, running_distance=12.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:04<14:49, 22.23s/it, running_distance=10.7135]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:20<28:09, 43.31s/it, running_distance=14.4570]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPE MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.5406\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 70/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 70/100: \n","Train Loss 0.0622\t Train Perplexity 1.0641\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 71/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 71/100: \n","Train Loss 0.0626\t Train Perplexity 1.0646\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<27:04, 38.67s/it, running_distance=18.2500]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:53<17:01, 24.92s/it, running_distance=12.5156]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY AND GRATITUDE WORDS YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:20<17:06, 25.66s/it, running_distance=10.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:35<29:25, 45.27s/it, running_distance=13.3867]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.0625\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Epoch 00051: reducing learning rate of group 0 to 5.3144e-05.\n","Saved epoch model\n","\n","Epoch 72/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 72/100: \n","Train Loss 0.0617\t Train Perplexity 1.0636\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 73/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 73/100: \n","Train Loss 0.0611\t Train Perplexity 1.0630\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:42<29:49, 42.61s/it, running_distance=12.4844]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:59<18:36, 27.22s/it, running_distance=9.6328]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:25<17:58, 26.96s/it, running_distance=8.5469]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULIE IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:41<29:58, 46.13s/it, running_distance=12.4688]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVERT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.2281\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 74/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 74/100: \n","Train Loss 0.0611\t Train Perplexity 1.0630\t Learning Rate 0.0001\n","Saved epoch model\n","\n","Epoch 75/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 75/100: \n","Train Loss 0.0610\t Train Perplexity 1.0629\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:39<27:24, 39.17s/it, running_distance=23.5938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:55<17:42, 25.90s/it, running_distance=14.8672]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:28, 26.22s/it, running_distance=12.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JEWLY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:13<23:30, 36.17s/it, running_distance=13.9062]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.2000\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 76/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 76/100: \n","Train Loss 0.0608\t Train Perplexity 1.0627\t Learning Rate 0.0001\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 77/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 77/100: \n","Train Loss 0.0609\t Train Perplexity 1.0628\t Learning Rate 0.0001\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:34<24:22, 34.82s/it, running_distance=20.1875]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPTED TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:51<16:29, 24.14s/it, running_distance=13.3281]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:18<16:53, 25.33s/it, running_distance=11.1510]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:33<29:22, 45.19s/it, running_distance=15.3125]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPE MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 15.4406\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 78/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 78/100: \n","Train Loss 0.0604\t Train Perplexity 1.0622\t Learning Rate 0.0001\n","Epoch 00058: reducing learning rate of group 0 to 4.7830e-05.\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 79/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 79/100: \n","Train Loss 0.0605\t Train Perplexity 1.0624\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:43<30:19, 43.31s/it, running_distance=15.5781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [01:00<18:56, 27.73s/it, running_distance=11.0781]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:26<18:11, 27.30s/it, running_distance=9.5260]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THIS DEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:42<30:09, 46.39s/it, running_distance=13.2148]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 13.8656\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 80/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 80/100: \n","Train Loss 0.0596\t Train Perplexity 1.0615\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 81/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 81/100: \n","Train Loss 0.0594\t Train Perplexity 1.0612\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:43<30:13, 43.19s/it, running_distance=13.0938]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:58<18:25, 26.97s/it, running_distance=9.9297]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:25<17:53, 26.84s/it, running_distance=8.6354]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW THAT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:40<29:53, 45.98s/it, running_distance=11.4961]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE EDDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 12.0875\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","Saved best training model\n","\n","Epoch 82/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 82/100: \n","Train Loss 0.0595\t Train Perplexity 1.0614\t Learning Rate 0.0000\n","Saved epoch model\n","\n","Epoch 83/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 83/100: \n","Train Loss 0.0594\t Train Perplexity 1.0612\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:39<27:58, 39.97s/it, running_distance=21.6094]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:56<17:53, 26.19s/it, running_distance=14.1250]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:22<17:31, 26.29s/it, running_distance=11.4635]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:38<29:42, 45.71s/it, running_distance=14.5117]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET IT'S DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 14.0531\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Saved epoch model\n","\n","Epoch 84/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 84/100: \n","Train Loss 0.0593\t Train Perplexity 1.0612\t Learning Rate 0.0000\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 85/100\n"]},{"output_type":"stream","name":"stderr","text":[]},{"output_type":"stream","name":"stdout","text":["\n","Epoch 85/100: \n","Train Loss 0.0586\t Train Perplexity 1.0604\t Learning Rate 0.0000\n"]},{"output_type":"stream","name":"stderr","text":["Val:   2%|▏         | 1/43 [00:38<26:59, 38.56s/it, running_distance=21.2656]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE STRENGTH THAT ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT AT TWO POINTS THE HEAD AND HEELS\n","Prediction   :  THE STRENGTH OF ENABLES SOMEONE IN A TRANCE TO HOLD HIS BODY STIFF AND UNSUPPORTED EXCEPT A TWO POINTS THE HEAD AND HEELS\n"]},{"output_type":"stream","name":"stderr","text":["Val:   5%|▍         | 2/43 [00:55<17:36, 25.76s/it, running_distance=13.8203]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WEREN'T YOU HAPPY THEN AT ALL\n","Prediction   :  SHE PRESSED HIS HAND GENTLY IN GRATITUDE WORN'T YOU HAPPY THEN AT ALL\n"]},{"output_type":"stream","name":"stderr","text":["Val:   7%|▋         | 3/43 [01:21<17:25, 26.14s/it, running_distance=11.2396]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  THE ZEAL OF CYRIL EXPOSED HIM TO THE PENALTIES OF THE JULIAN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSTITIOUS AGE HE WAS SECURE OF IMPUNITY AND EVEN OF PRAISE\n","Prediction   :  THE ZEAL OF CAROL EXPOSED HIM TO THE PENALTIES OF THE JULY IN LAW BUT IN A FEEBLE GOVERNMENT AND A SUPERSISTES AGE HE WAS THE CURE OF IMPUNITY AND EVEN OF PRAISE\n"]},{"output_type":"stream","name":"stderr","text":["Val:   9%|▉         | 4/43 [02:37<29:41, 45.67s/it, running_distance=14.5742]"]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  WE RANGE WIDER LAST LONGER AND ESCAPE MORE AND MORE FROM INTENSITY TOWARDS UNDERSTANDING\n","Prediction   :  WE RANGED WIDER LAST LONGER AND ESCAPED MORE AT MORE FROM INTENSITY TOWARDS UNDERSTANDING\n"]},{"output_type":"stream","name":"stderr","text":["                                                                             "]},{"output_type":"stream","name":"stdout","text":["\n","Ground Truth :  PUT THE UDDER INTO A STEWPAN WITH SUFFICIENT WATER TO COVER IT LET IT STEW GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Prediction   :  PUT THE ADDER INTO A STEWPEN WITH SUFFICIENT WANDER TO COVER IT LET ITS DO GENTLY TILL QUITE DONE WHEN TAKE IT OUT TO COOL\n","Levenshtein Distance 15.2594\n"]},{"output_type":"stream","name":"stderr","text":["\r"]},{"output_type":"stream","name":"stdout","text":["Epoch 00065: reducing learning rate of group 0 to 4.3047e-05.\n","Saved epoch model\n","Saved best training model\n","\n","Epoch 86/100\n"]},{"output_type":"stream","name":"stderr","text":["Train:  88%|████████▊ | 392/446 [06:36<00:53,  1.01it/s, loss=0.0582, perplexity=1.0600]"]},{"output_type":"error","ename":"KeyboardInterrupt","evalue":"","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)","\u001b[0;32m<ipython-input-101-e1f74306b03e>\u001b[0m in \u001b[0;36m<cell line: 34>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     38\u001b[0m     \u001b[0mcurr_lr\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfloat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moptimizer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mparam_groups\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"lr\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     39\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 40\u001b[0;31m     \u001b[0mtrain_loss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_perplexity\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mattention_weights\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtrain_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfull_model\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtrain_loader\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moptimizer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     41\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     42\u001b[0m     print(\"\\nEpoch {}/{}: \\nTrain Loss {:.04f}\\t Train Perplexity {:.04f}\\t Learning Rate {:.04f}\".format(\n","\u001b[0;32m<ipython-input-50-36f261d3ecd3>\u001b[0m in \u001b[0;36mtrain_model\u001b[0;34m(model, train_loader, optimizer)\u001b[0m\n\u001b[1;32m     27\u001b[0m             \u001b[0mloss\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msum\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0mpadding_mask\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msum\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     28\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 29\u001b[0;31m         \u001b[0mscaler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mscale\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mloss\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbackward\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m   \u001b[0;31m# This is a replacement for loss.backward()\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     30\u001b[0m         \u001b[0mscaler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstep\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0moptimizer\u001b[0m\u001b[0;34m)\u001b[0m          \u001b[0;31m# This is a replacement for optimizer.step()\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m         \u001b[0mscaler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mupdate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m                 \u001b[0;31m# This is something added just for FP16\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/_tensor.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[1;32m    490\u001b[0m                 \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    491\u001b[0m             )\n\u001b[0;32m--> 492\u001b[0;31m         torch.autograd.backward(\n\u001b[0m\u001b[1;32m    493\u001b[0m             \u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mgradient\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mretain_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcreate_graph\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minputs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0minputs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    494\u001b[0m         )\n","\u001b[0;32m/usr/local/lib/python3.10/dist-packages/torch/autograd/__init__.py\u001b[0m in \u001b[0;36mbackward\u001b[0;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[1;32m    249\u001b[0m     \u001b[0;31m# some Python versions print out the first line of a multi-line function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    250\u001b[0m     \u001b[0;31m# calls in the traceback and some print out the last line\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 251\u001b[0;31m     Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass\n\u001b[0m\u001b[1;32m    252\u001b[0m         \u001b[0mtensors\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    253\u001b[0m         \u001b[0mgrad_tensors_\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mKeyboardInterrupt\u001b[0m: "]},{"output_type":"display_data","data":{"text/plain":["<Figure size 640x480 with 2 Axes>"],"image/png":"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\n"},"metadata":{}}]},{"cell_type":"code","source":["def unfreeze_model(model):\n","    for param in model.parameters():\n","        param.requires_grad = True\n","\n"],"metadata":{"id":"-IvGsDGT34Qg","executionInfo":{"status":"ok","timestamp":1713623141714,"user_tz":240,"elapsed":158,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":88,"outputs":[]},{"cell_type":"code","source":["unfreeze_model(full_model)"],"metadata":{"id":"7bgjV99fC_n3","executionInfo":{"status":"ok","timestamp":1713623142762,"user_tz":240,"elapsed":160,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}}},"execution_count":89,"outputs":[]},{"cell_type":"code","source":["full_model.load_state_dict(torch.load('/content/checkpointsfull/checkpoint-epochfull-4.pth')['model_state_dict'])\n"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"gIfP5k7K76Qb","executionInfo":{"status":"ok","timestamp":1713674138674,"user_tz":240,"elapsed":1459,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"f2045b82-8eec-4198-92a0-46369bc4b6e3"},"execution_count":102,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<All keys matched successfully>"]},"metadata":{},"execution_count":102}]},{"cell_type":"markdown","source":["## Testing"],"metadata":{"id":"6PFQ849OegR_"}},{"cell_type":"code","source":["full_model.eval()\n","\n","# progress bar\n","batch_bar = tqdm(total=len(test_loader), dynamic_ncols=True, leave=False, position=0, desc=\"Test\", ncols=5)\n","\n","all_predictions = []\n","\n","for i, data in enumerate(test_loader):\n","    inputs, inputs_lengths = data\n","    inputs                 = inputs.to(device)\n","\n","    with torch.inference_mode():\n","        greedy_predictions = full_model.recognize(inputs, inputs_lengths)\n","\n","    greedy_predictions = greedy_predictions.detach().cpu().numpy()\n","\n","    for batch_idx in range(greedy_predictions.shape[0]):\n","        pred_sliced = indices_to_chars(greedy_predictions[batch_idx], vocab= VOCAB)\n","        pred_string = ''.join(pred_sliced)\n","        all_predictions.append(pred_string)\n","\n","    batch_bar.update()"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713676076544,"user_tz":240,"elapsed":1929797,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"d3a5f393-01fe-466f-f6f8-77f862813423","id":"B0jv1iQVegSA"},"execution_count":103,"outputs":[{"output_type":"stream","name":"stderr","text":["Test: 100%|██████████| 41/41 [32:09<00:00, 54.92s/it]"]}]},{"cell_type":"code","source":["%cd /content/\n","df = pd.DataFrame({\n","    \"index\" : list(range(len(all_predictions))),\n","    \"label\" : all_predictions\n","})\n","\n","df.to_csv(\"improved.csv\", index=False)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713676076544,"user_tz":240,"elapsed":11,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"adafaed9-b728-4a06-b4a8-29a0b4ba3855","id":"KuDKSXsWegSA"},"execution_count":104,"outputs":[{"output_type":"stream","name":"stdout","text":["/content\n"]}]},{"cell_type":"code","source":["!kaggle competitions submit -c hw4p2-sp24 -f improved.csv -m \"It's been a ride :)\""],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1713365209397,"user_tz":240,"elapsed":993,"user":{"displayName":"sun jiancheng","userId":"12683695838139180534"}},"outputId":"f79eb47c-a131-447d-887f-844a83e1ce84","id":"Stkhbjq0egSA"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Traceback (most recent call last):\n","  File \"/usr/local/bin/kaggle\", line 8, in <module>\n","    sys.exit(main())\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/cli.py\", line 67, in main\n","    out = args.func(**command_args)\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 561, in competition_submit_cli\n","    submit_result = self.competition_submit(file_name, message,\n","  File \"/usr/local/lib/python3.10/dist-packages/kaggle/api/kaggle_api_extended.py\", line 532, in competition_submit\n","    url_result['createUrl'], quiet)\n","KeyError: 'createUrl'\n"]}]},{"cell_type":"markdown","source":["# HW3P2 model\n"],"metadata":{"id":"zDyjswBhjpCj"}},{"cell_type":"markdown","metadata":{"id":"PB6eh3gnMUzy"},"source":["### Pyramid Bi-LSTM (pBLSTM)"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"qd4BEX_yMUzz"},"outputs":[],"source":["# Utils for network\n","torch.cuda.empty_cache()\n","# from torch.autograd import Variable\n","\n","class PermuteBlock(torch.nn.Module):\n","    def forward(self, x):\n","        return x.transpose(1, 2)\n","\n","class LockedDropout(nn.Module):\n","    \"\"\" LockedDropout applies the same dropout mask to every time step.\n","\n","    **Thank you** to Sales Force for their initial implementation of :class:`WeightDrop`. Here is\n","    their `License\n","    <https://github.com/salesforce/awd-lstm-lm/blob/master/LICENSE>`__.\n","\n","    Args:\n","        p (float): Probability of an element in the dropout mask to be zeroed.\n","    \"\"\"\n","\n","    def __init__(self, p=0.25):\n","        self.p = p\n","        super().__init__()\n","\n","    def forward(self, x):\n","        \"\"\"\n","        Args:\n","            x (:class:`torch.FloatTensor` [sequence length, batch size, rnn hidden size]): Input to\n","                apply dropout too.\n","        \"\"\"\n","        if not self.training or not self.p:\n","            return x\n","        x = x.clone()\n","        mask = x.new_empty(1, x.size(1), x.size(2), requires_grad=False).bernoulli_(1 - self.p)\n","        mask = mask.div_(1 - self.p)\n","        mask = mask.expand_as(x)\n","        return x * mask\n","\n","\n","    def __repr__(self):\n","        return self.__class__.__name__ + '(' \\\n","            + 'p=' + str(self.p) + ')'\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"OmdyXI6KMUzz"},"outputs":[],"source":["class pBLSTM(torch.nn.Module):\n","\n","    '''\n","    Pyramidal BiLSTM\n","    Read the write up/paper and understand the concepts and then write your implementation here.\n","\n","    At each step,\n","    1. Pad your input if it is packed (Unpack it)\n","    2. Reduce the input length dimension by concatenating feature dimension\n","        (Tip: Write down the shapes and understand)\n","        (i) How should  you deal with odd/even length input?\n","        (ii) How should you deal with input length array (x_lens) after truncating the input?\n","    3. Pack your input\n","    4. Pass it into LSTM layer\n","\n","    To make our implementation modular, we pass 1 layer at a time.\n","    '''\n","\n","    def __init__(self, input_size, hidden_size):\n","        super(pBLSTM, self).__init__()\n","\n","        self.blstm = torch.nn.LSTM(input_size=2*input_size, hidden_size=hidden_size, num_layers=3, bidirectional=True, dropout = 0.3)  # TODO: Initialize a single layer bidirectional LSTM with the given input_size and hidden_size\n","        #could drop out\n","    def forward(self, x_packed): # x_packed is a PackedSequence\n","\n","        # TODO: Pad Packed Sequence\n","        seq_unpacked, lens_unpacked = pad_packed_sequence(x_packed, batch_first=True)\n","        # print(f'in the plstm section after pad {seq_unpacked.shape},{lens_unpacked.shape}')\n","        # Call self.trunc_reshape() which downsamples the time steps of x and increases the feature dimensions as mentioned above\n","        x, x_lens = self.trunc_reshape(x=seq_unpacked, x_lens=lens_unpacked)\n","        # print(f'in the plstm section after truncate {x.shape},{x_lens.shape}')\n","        # self.trunc_reshape will return 2 outputs. What are they? Think about what quantites are changing.\n","        # TODO: Pack Padded Sequence. What output(s) would you get?\n","        PackedSequence = pack_padded_sequence(input=x,lengths=x_lens,batch_first=True,enforce_sorted=False)\n","        # TODO: Pass the sequence through bLSTM\n","        packed_output, (_, _) = self.blstm(PackedSequence)\n","        # What do you return?\n","\n","        return packed_output\n","\n","    def trunc_reshape(self, x, x_lens):\n","        # TODO: If you have odd number of timesteps, how can you handle it? (Hint: You can exclude them)\n","        # TODO: Reshape x. When reshaping x, you have to reduce number of timesteps by a downsampling factor while increasing number of features by the same factor\n","        # TODO: Reduce lengths by the same downsampling factor\n","        # The shape of x is [batch_size, frames, features]\n","        batch_size = x.shape[0]\n","        frames = x.shape[1]\n","        features = x.shape[2]\n","        if frames % 2 != 0:\n","            x = x[:, :-1, :]\n","        x = x.reshape((batch_size, x.shape[1]//2, features*2))\n","        x_lens = x_lens/2\n","        return x, x_lens"]},{"cell_type":"markdown","metadata":{"id":"g3ZQ75OcMUz0"},"source":["### Encoder"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"GEzw5_xmMUz0"},"outputs":[],"source":["class Encoder(torch.nn.Module):\n","    '''\n","    The Encoder takes utterances as inputs and returns latent feature representations\n","    '''\n","    def __init__(self, input_size, encoder_hidden_size):\n","        super(Encoder, self).__init__()\n","\n","        self.PermuteBlock = PermuteBlock()\n","        self.embedding = torch.nn.Sequential(torch.nn.Conv1d(in_channels=input_size,out_channels=64,kernel_size=3,padding=1,stride=1),\n","                                             torch.nn.BatchNorm1d(64),\n","                                             torch.nn.GELU(),\n","                                             torch.nn.Conv1d(in_channels=64, out_channels=128, kernel_size=3, padding=1, stride=1),\n","                                             torch.nn.BatchNorm1d(128),\n","                                             torch.nn.GELU(),\n","                                             torch.nn.Conv1d(in_channels=128, out_channels=256, kernel_size=3, padding=1, stride=1),\n","                                             torch.nn.BatchNorm1d(256),\n","                                             torch.nn.GELU()\n","                                             )\n","        self.lstm_layers = torch.nn.ModuleList()\n","        for i in range(2):\n","            input_size_pblstm = 256 if i == 0 else encoder_hidden_size * 2\n","            self.lstm_layers.append(pBLSTM(input_size=input_size_pblstm, hidden_size=encoder_hidden_size))\n","            self.lstm_layers.append(LockedDropout())\n","\n","    def forward(self, x, x_lens):\n","        # Where are x and x_lens coming from? The dataloader\n","        #TODO: Call the permute layer and embedding layer\n","        x = self.PermuteBlock(x)\n","        x = self.embedding(x)\n","        x = self.PermuteBlock(x)\n","        # # TODO: Pack Padded Sequence\n","        # packed = pack_padded_sequence(input=x,lengths=x_lens,batch_first=True,enforce_sorted=False)\n","        # # TODO: Pass Sequence through the pyramidal Bi-LSTM layer\n","        # x = self.pBLSTM1(packed)\n","        # # TODO: Pad Packed Sequence\n","        # encoder_outputs, encoder_lens= pad_packed_sequence(sequence=x,batch_first=True)\n","        # print(f'The output from plstm, {encoder_outputs.shape},{encoder_lens.shape}')\n","        for layer in self.lstm_layers:\n","            if isinstance(layer, pBLSTM):\n","                x_packed = pack_padded_sequence(x, x_lens, batch_first=True, enforce_sorted=False)\n","                x_packed = layer(x_packed)\n","                x, x_lens = pad_packed_sequence(x_packed, batch_first=True)\n","            else:\n","                x = torch.permute(x, (1, 0, 2))\n","                x = layer(x)\n","                x = torch.permute(x,(1,0,2))\n","        # Remember the number of output(s) each function returns\n","\n","        return x, x_lens\n"]},{"cell_type":"markdown","metadata":{"id":"kg82HXa3MUz1"},"source":["### Decoder"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"qmHf6pFiMUz1"},"outputs":[],"source":["class HW3P2_encoder(torch.nn.Module):\n","\n","    def __init__(self, input_size, embed_size= 192):\n","        super().__init__()\n","\n","        self.encoder        = Encoder(input_size=input_size, encoder_hidden_size=embed_size*2)\n","\n","    def forward(self, x, lengths_x):\n","\n","        encoder_out, encoder_lens   = self.encoder(x, lengths_x)\n","        return encoder_out, encoder_lens"]},{"cell_type":"markdown","metadata":{"id":"EV7DMPDoMUz2"},"source":["## Initialize ASR Network"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"oaaDsnnLMUz2","outputId":"174ec9a1-3818-4328-f911-86802cbb155b"},"outputs":[{"name":"stdout","output_type":"stream","text":["ASRModel(\n","  (augmentations): Sequential()\n","  (encoder): Encoder(\n","    (PermuteBlock): PermuteBlock()\n","    (embedding): Sequential(\n","      (0): Conv1d(27, 64, kernel_size=(3,), stride=(1,), padding=(1,))\n","      (1): BatchNorm1d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","      (2): GELU(approximate='none')\n","      (3): Conv1d(64, 128, kernel_size=(3,), stride=(1,), padding=(1,))\n","      (4): BatchNorm1d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","      (5): GELU(approximate='none')\n","      (6): Conv1d(128, 256, kernel_size=(3,), stride=(1,), padding=(1,))\n","      (7): BatchNorm1d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","      (8): GELU(approximate='none')\n","    )\n","    (lstm_layers): ModuleList(\n","      (0): pBLSTM(\n","        (blstm): LSTM(512, 384, num_layers=3, dropout=0.3, bidirectional=True)\n","      )\n","      (1): LockedDropout(p=0.25)\n","      (2): pBLSTM(\n","        (blstm): LSTM(1536, 384, num_layers=3, dropout=0.3, bidirectional=True)\n","      )\n","      (3): LockedDropout(p=0.25)\n","    )\n","  )\n","  (decoder): Decoder(\n","    (mlp): Sequential(\n","      (0): PermuteBlock()\n","      (1): BatchNorm1d(768, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n","      (2): PermuteBlock()\n","      (3): Linear(in_features=768, out_features=128, bias=True)\n","      (4): GELU(approximate='none')\n","      (5): Linear(in_features=128, out_features=64, bias=True)\n","      (6): GELU(approximate='none')\n","      (7): Linear(in_features=64, out_features=41, bias=True)\n","    )\n","    (softmax): LogSoftmax(dim=2)\n","  )\n",")\n","===============================================================================================\n","                                         Kernel Shape     Output Shape  \\\n","Layer                                                                    \n","0_augmentations                                     -   [64, 2936, 27]   \n","1_encoder.PermuteBlock_PermuteBlock                 -   [64, 27, 2936]   \n","2_encoder.embedding.Conv1d_0              [27, 64, 3]   [64, 64, 2936]   \n","3_encoder.embedding.BatchNorm1d_1                [64]   [64, 64, 2936]   \n","4_encoder.embedding.GELU_2                          -   [64, 64, 2936]   \n","5_encoder.embedding.Conv1d_3             [64, 128, 3]  [64, 128, 2936]   \n","6_encoder.embedding.BatchNorm1d_4               [128]  [64, 128, 2936]   \n","7_encoder.embedding.GELU_5                          -  [64, 128, 2936]   \n","8_encoder.embedding.Conv1d_6            [128, 256, 3]  [64, 256, 2936]   \n","9_encoder.embedding.BatchNorm1d_7               [256]  [64, 256, 2936]   \n","10_encoder.embedding.GELU_8                         -  [64, 256, 2936]   \n","11_encoder.PermuteBlock_PermuteBlock                -  [64, 2936, 256]   \n","12_encoder.lstm_layers.0.LSTM_blstm                 -     [21060, 768]   \n","13_encoder.lstm_layers.LockedDropout_1              -  [1468, 64, 768]   \n","14_encoder.lstm_layers.2.LSTM_blstm                 -     [10517, 768]   \n","15_encoder.lstm_layers.LockedDropout_3              -   [734, 64, 768]   \n","16_decoder.mlp.PermuteBlock_0                       -   [64, 768, 734]   \n","17_decoder.mlp.BatchNorm1d_1                    [768]   [64, 768, 734]   \n","18_decoder.mlp.PermuteBlock_2                       -   [64, 734, 768]   \n","19_decoder.mlp.Linear_3                    [768, 128]   [64, 734, 128]   \n","20_decoder.mlp.GELU_4                               -   [64, 734, 128]   \n","21_decoder.mlp.Linear_5                     [128, 64]    [64, 734, 64]   \n","22_decoder.mlp.GELU_6                               -    [64, 734, 64]   \n","23_decoder.mlp.Linear_7                      [64, 41]    [64, 734, 41]   \n","24_decoder.LogSoftmax_softmax                       -    [64, 734, 41]   \n","\n","                                           Params    Mult-Adds  \n","Layer                                                           \n","0_augmentations                                 -            -  \n","1_encoder.PermuteBlock_PermuteBlock             -            -  \n","2_encoder.embedding.Conv1d_0               5.248k   15.220224M  \n","3_encoder.embedding.BatchNorm1d_1           128.0         64.0  \n","4_encoder.embedding.GELU_2                      -            -  \n","5_encoder.embedding.Conv1d_3              24.704k   72.155136M  \n","6_encoder.embedding.BatchNorm1d_4           256.0        128.0  \n","7_encoder.embedding.GELU_5                      -            -  \n","8_encoder.embedding.Conv1d_6               98.56k  288.620544M  \n","9_encoder.embedding.BatchNorm1d_7           512.0        256.0  \n","10_encoder.embedding.GELU_8                     -            -  \n","11_encoder.PermuteBlock_PermuteBlock            -            -  \n","12_encoder.lstm_layers.0.LSTM_blstm     9.848832M      9.8304M  \n","13_encoder.lstm_layers.LockedDropout_1          -            -  \n","14_encoder.lstm_layers.2.LSTM_blstm     12.99456M   12.976128M  \n","15_encoder.lstm_layers.LockedDropout_3          -            -  \n","16_decoder.mlp.PermuteBlock_0                   -            -  \n","17_decoder.mlp.BatchNorm1d_1               1.536k        768.0  \n","18_decoder.mlp.PermuteBlock_2                   -            -  \n","19_decoder.mlp.Linear_3                   98.432k      98.304k  \n","20_decoder.mlp.GELU_4                           -            -  \n","21_decoder.mlp.Linear_5                    8.256k       8.192k  \n","22_decoder.mlp.GELU_6                           -            -  \n","23_decoder.mlp.Linear_7                    2.665k       2.624k  \n","24_decoder.LogSoftmax_softmax                   -            -  \n","-----------------------------------------------------------------------------------------------\n","                           Totals\n","Total params           23.083689M\n","Trainable params       23.083689M\n","Non-trainable params          0.0\n","Mult-Adds             398.912768M\n","===============================================================================================\n"]},{"data":{"text/html":["<div>\n","<style scoped>\n","    .dataframe tbody tr th:only-of-type {\n","        vertical-align: middle;\n","    }\n","\n","    .dataframe tbody tr th {\n","        vertical-align: top;\n","    }\n","\n","    .dataframe thead th {\n","        text-align: right;\n","    }\n","</style>\n","<table border=\"1\" class=\"dataframe\">\n","  <thead>\n","    <tr style=\"text-align: right;\">\n","      <th></th>\n","      <th>Kernel Shape</th>\n","      <th>Output Shape</th>\n","      <th>Params</th>\n","      <th>Mult-Adds</th>\n","    </tr>\n","    <tr>\n","      <th>Layer</th>\n","      <th></th>\n","      <th></th>\n","      <th></th>\n","      <th></th>\n","    </tr>\n","  </thead>\n","  <tbody>\n","    <tr>\n","      <th>0_augmentations</th>\n","      <td>-</td>\n","      <td>[64, 2936, 27]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>1_encoder.PermuteBlock_PermuteBlock</th>\n","      <td>-</td>\n","      <td>[64, 27, 2936]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>2_encoder.embedding.Conv1d_0</th>\n","      <td>[27, 64, 3]</td>\n","      <td>[64, 64, 2936]</td>\n","      <td>5248.0</td>\n","      <td>15220224.0</td>\n","    </tr>\n","    <tr>\n","      <th>3_encoder.embedding.BatchNorm1d_1</th>\n","      <td>[64]</td>\n","      <td>[64, 64, 2936]</td>\n","      <td>128.0</td>\n","      <td>64.0</td>\n","    </tr>\n","    <tr>\n","      <th>4_encoder.embedding.GELU_2</th>\n","      <td>-</td>\n","      <td>[64, 64, 2936]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>5_encoder.embedding.Conv1d_3</th>\n","      <td>[64, 128, 3]</td>\n","      <td>[64, 128, 2936]</td>\n","      <td>24704.0</td>\n","      <td>72155136.0</td>\n","    </tr>\n","    <tr>\n","      <th>6_encoder.embedding.BatchNorm1d_4</th>\n","      <td>[128]</td>\n","      <td>[64, 128, 2936]</td>\n","      <td>256.0</td>\n","      <td>128.0</td>\n","    </tr>\n","    <tr>\n","      <th>7_encoder.embedding.GELU_5</th>\n","      <td>-</td>\n","      <td>[64, 128, 2936]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>8_encoder.embedding.Conv1d_6</th>\n","      <td>[128, 256, 3]</td>\n","      <td>[64, 256, 2936]</td>\n","      <td>98560.0</td>\n","      <td>288620544.0</td>\n","    </tr>\n","    <tr>\n","      <th>9_encoder.embedding.BatchNorm1d_7</th>\n","      <td>[256]</td>\n","      <td>[64, 256, 2936]</td>\n","      <td>512.0</td>\n","      <td>256.0</td>\n","    </tr>\n","    <tr>\n","      <th>10_encoder.embedding.GELU_8</th>\n","      <td>-</td>\n","      <td>[64, 256, 2936]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>11_encoder.PermuteBlock_PermuteBlock</th>\n","      <td>-</td>\n","      <td>[64, 2936, 256]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>12_encoder.lstm_layers.0.LSTM_blstm</th>\n","      <td>-</td>\n","      <td>[21060, 768]</td>\n","      <td>9848832.0</td>\n","      <td>9830400.0</td>\n","    </tr>\n","    <tr>\n","      <th>13_encoder.lstm_layers.LockedDropout_1</th>\n","      <td>-</td>\n","      <td>[1468, 64, 768]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>14_encoder.lstm_layers.2.LSTM_blstm</th>\n","      <td>-</td>\n","      <td>[10517, 768]</td>\n","      <td>12994560.0</td>\n","      <td>12976128.0</td>\n","    </tr>\n","    <tr>\n","      <th>15_encoder.lstm_layers.LockedDropout_3</th>\n","      <td>-</td>\n","      <td>[734, 64, 768]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>16_decoder.mlp.PermuteBlock_0</th>\n","      <td>-</td>\n","      <td>[64, 768, 734]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>17_decoder.mlp.BatchNorm1d_1</th>\n","      <td>[768]</td>\n","      <td>[64, 768, 734]</td>\n","      <td>1536.0</td>\n","      <td>768.0</td>\n","    </tr>\n","    <tr>\n","      <th>18_decoder.mlp.PermuteBlock_2</th>\n","      <td>-</td>\n","      <td>[64, 734, 768]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>19_decoder.mlp.Linear_3</th>\n","      <td>[768, 128]</td>\n","      <td>[64, 734, 128]</td>\n","      <td>98432.0</td>\n","      <td>98304.0</td>\n","    </tr>\n","    <tr>\n","      <th>20_decoder.mlp.GELU_4</th>\n","      <td>-</td>\n","      <td>[64, 734, 128]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>21_decoder.mlp.Linear_5</th>\n","      <td>[128, 64]</td>\n","      <td>[64, 734, 64]</td>\n","      <td>8256.0</td>\n","      <td>8192.0</td>\n","    </tr>\n","    <tr>\n","      <th>22_decoder.mlp.GELU_6</th>\n","      <td>-</td>\n","      <td>[64, 734, 64]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","    <tr>\n","      <th>23_decoder.mlp.Linear_7</th>\n","      <td>[64, 41]</td>\n","      <td>[64, 734, 41]</td>\n","      <td>2665.0</td>\n","      <td>2624.0</td>\n","    </tr>\n","    <tr>\n","      <th>24_decoder.LogSoftmax_softmax</th>\n","      <td>-</td>\n","      <td>[64, 734, 41]</td>\n","      <td>NaN</td>\n","      <td>NaN</td>\n","    </tr>\n","  </tbody>\n","</table>\n","</div>"],"text/plain":["                                         Kernel Shape     Output Shape  \\\n","Layer                                                                    \n","0_augmentations                                     -   [64, 2936, 27]   \n","1_encoder.PermuteBlock_PermuteBlock                 -   [64, 27, 2936]   \n","2_encoder.embedding.Conv1d_0              [27, 64, 3]   [64, 64, 2936]   \n","3_encoder.embedding.BatchNorm1d_1                [64]   [64, 64, 2936]   \n","4_encoder.embedding.GELU_2                          -   [64, 64, 2936]   \n","5_encoder.embedding.Conv1d_3             [64, 128, 3]  [64, 128, 2936]   \n","6_encoder.embedding.BatchNorm1d_4               [128]  [64, 128, 2936]   \n","7_encoder.embedding.GELU_5                          -  [64, 128, 2936]   \n","8_encoder.embedding.Conv1d_6            [128, 256, 3]  [64, 256, 2936]   \n","9_encoder.embedding.BatchNorm1d_7               [256]  [64, 256, 2936]   \n","10_encoder.embedding.GELU_8                         -  [64, 256, 2936]   \n","11_encoder.PermuteBlock_PermuteBlock                -  [64, 2936, 256]   \n","12_encoder.lstm_layers.0.LSTM_blstm                 -     [21060, 768]   \n","13_encoder.lstm_layers.LockedDropout_1              -  [1468, 64, 768]   \n","14_encoder.lstm_layers.2.LSTM_blstm                 -     [10517, 768]   \n","15_encoder.lstm_layers.LockedDropout_3              -   [734, 64, 768]   \n","16_decoder.mlp.PermuteBlock_0                       -   [64, 768, 734]   \n","17_decoder.mlp.BatchNorm1d_1                    [768]   [64, 768, 734]   \n","18_decoder.mlp.PermuteBlock_2                       -   [64, 734, 768]   \n","19_decoder.mlp.Linear_3                    [768, 128]   [64, 734, 128]   \n","20_decoder.mlp.GELU_4                               -   [64, 734, 128]   \n","21_decoder.mlp.Linear_5                     [128, 64]    [64, 734, 64]   \n","22_decoder.mlp.GELU_6                               -    [64, 734, 64]   \n","23_decoder.mlp.Linear_7                      [64, 41]    [64, 734, 41]   \n","24_decoder.LogSoftmax_softmax                       -    [64, 734, 41]   \n","\n","                                            Params    Mult-Adds  \n","Layer                                                            \n","0_augmentations                                NaN          NaN  \n","1_encoder.PermuteBlock_PermuteBlock            NaN          NaN  \n","2_encoder.embedding.Conv1d_0                5248.0   15220224.0  \n","3_encoder.embedding.BatchNorm1d_1            128.0         64.0  \n","4_encoder.embedding.GELU_2                     NaN          NaN  \n","5_encoder.embedding.Conv1d_3               24704.0   72155136.0  \n","6_encoder.embedding.BatchNorm1d_4            256.0        128.0  \n","7_encoder.embedding.GELU_5                     NaN          NaN  \n","8_encoder.embedding.Conv1d_6               98560.0  288620544.0  \n","9_encoder.embedding.BatchNorm1d_7            512.0        256.0  \n","10_encoder.embedding.GELU_8                    NaN          NaN  \n","11_encoder.PermuteBlock_PermuteBlock           NaN          NaN  \n","12_encoder.lstm_layers.0.LSTM_blstm      9848832.0    9830400.0  \n","13_encoder.lstm_layers.LockedDropout_1         NaN          NaN  \n","14_encoder.lstm_layers.2.LSTM_blstm     12994560.0   12976128.0  \n","15_encoder.lstm_layers.LockedDropout_3         NaN          NaN  \n","16_decoder.mlp.PermuteBlock_0                  NaN          NaN  \n","17_decoder.mlp.BatchNorm1d_1                1536.0        768.0  \n","18_decoder.mlp.PermuteBlock_2                  NaN          NaN  \n","19_decoder.mlp.Linear_3                    98432.0      98304.0  \n","20_decoder.mlp.GELU_4                          NaN          NaN  \n","21_decoder.mlp.Linear_5                     8256.0       8192.0  \n","22_decoder.mlp.GELU_6                          NaN          NaN  \n","23_decoder.mlp.Linear_7                     2665.0       2624.0  \n","24_decoder.LogSoftmax_softmax                  NaN          NaN  "]},"execution_count":30,"metadata":{},"output_type":"execute_result"}],"source":["HW3P2_encoder = HW3P2_encoder(\n","    input_size  = 27,\n","    embed_size  = 192,\n",").to(device)\n","print(HW3P2_encoder)"]}],"metadata":{"colab":{"provenance":[],"collapsed_sections":["-njBvl2Opd6I","gP7QK6rD_mT9","hgFYFaBGeBqM"]},"kernelspec":{"display_name":"Python 3","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.8.0"},"widgets":{"application/vnd.jupyter.widget-state+json":{"472fa7722fc74c09bff6ad00d7cc2451":{"model_module":"@jupyter-widgets/controls","model_name":"VBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"VBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"VBoxView","box_style":"","children":["IPY_MODEL_193f09f303d547328c03c9b7f82cbbb2","IPY_MODEL_c3463b441fda4798b0b10cead972933c"],"layout":"IPY_MODEL_5c9a63aad0fa4bc28af3661b0caf4366"}},"193f09f303d547328c03c9b7f82cbbb2":{"model_module":"@jupyter-widgets/controls","model_name":"LabelModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"LabelModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"LabelView","description":"","description_tooltip":null,"layout":"IPY_MODEL_fab7f97206ad4e41b0fc89d04ee4148e","placeholder":"​","style":"IPY_MODEL_9de0938b74ee4942b39104d04f15fc08","value":"0.015 MB of 0.015 MB uploaded\r"}},"c3463b441fda4798b0b10cead972933c":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_d2bef54bcb974f68b55f273b9e5cb54c","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_6fe6c18183a04ee8b531b5f0069293ae","value":1}},"5c9a63aad0fa4bc28af3661b0caf4366":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"fab7f97206ad4e41b0fc89d04ee4148e":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"9de0938b74ee4942b39104d04f15fc08":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"d2bef54bcb974f68b55f273b9e5cb54c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"6fe6c18183a04ee8b531b5f0069293ae":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"14f661da6a404f1ab8ee6efc229f3725":{"model_module":"@jupyter-widgets/controls","model_name":"VBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"VBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"VBoxView","box_style":"","children":["IPY_MODEL_890d26224b8f4ac789af33a6d1093a40","IPY_MODEL_6cbc75a0297941059000c1d5d12c4f04"],"layout":"IPY_MODEL_9239d24f6f2b43f4b9df6bf3de9e8937"}},"890d26224b8f4ac789af33a6d1093a40":{"model_module":"@jupyter-widgets/controls","model_name":"LabelModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"LabelModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"LabelView","description":"","description_tooltip":null,"layout":"IPY_MODEL_eb54eb599dc44978a47edf7e70b441c3","placeholder":"​","style":"IPY_MODEL_f76afe29621f4130a0efd2b537bd5d56","value":"0.674 MB of 0.674 MB uploaded\r"}},"6cbc75a0297941059000c1d5d12c4f04":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_66e559fa64ed4f828ec907258a295a38","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_2b599b08df004fa09cd2d493f9e706cf","value":1}},"9239d24f6f2b43f4b9df6bf3de9e8937":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"eb54eb599dc44978a47edf7e70b441c3":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"f76afe29621f4130a0efd2b537bd5d56":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"66e559fa64ed4f828ec907258a295a38":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"2b599b08df004fa09cd2d493f9e706cf":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"f8a088954bb64422a9c15d57c46dbc5f":{"model_module":"@jupyter-widgets/controls","model_name":"VBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"VBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"VBoxView","box_style":"","children":["IPY_MODEL_c252db17298d44e194878b1534a67525","IPY_MODEL_cd265a1376734407ab68d3c6461deadc"],"layout":"IPY_MODEL_8c8a5aec6f5b4669bdf21383eece0b10"}},"c252db17298d44e194878b1534a67525":{"model_module":"@jupyter-widgets/controls","model_name":"LabelModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"LabelModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"LabelView","description":"","description_tooltip":null,"layout":"IPY_MODEL_60433fd3f3954044b71a888b45b93b62","placeholder":"​","style":"IPY_MODEL_c57b11408cc34c2096423684db29578f","value":"0.415 MB of 0.415 MB uploaded\r"}},"cd265a1376734407ab68d3c6461deadc":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"","description":"","description_tooltip":null,"layout":"IPY_MODEL_17de0d628c1141cf819838f3dbe5c792","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_9ecf9ea2ceb24c59842b27a9d7c0f463","value":1}},"8c8a5aec6f5b4669bdf21383eece0b10":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"60433fd3f3954044b71a888b45b93b62":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"c57b11408cc34c2096423684db29578f":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"17de0d628c1141cf819838f3dbe5c792":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"9ecf9ea2ceb24c59842b27a9d7c0f463":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}}}}},"nbformat":4,"nbformat_minor":0}