Dimenet_model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "51caaf12",
   "metadata": {},
   "source": [
    "# Utils"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "2a9cd674",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\tqdm\\auto.py:22: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "from torch import autograd\n",
    "from torch_geometric.data import Data\n",
    "from torch_geometric.nn import DimeNet\n",
    "# from torch_geometric.nn import DimeNetPlusPlus\n",
    "from tqdm import tqdm\n",
    "import torch_geometric\n",
    "from torch_geometric.loader import DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "3634b671",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.1.0\n"
     ]
    }
   ],
   "source": [
    "print(torch_geometric.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "8a4f9ddd",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import random\n",
    "import math\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import MinMaxScaler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "19738ff1",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch_geometric.datasets import QM9"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "980361a4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.11.0\n"
     ]
    }
   ],
   "source": [
    "print(torch.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "f72a65b4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[0., 1., 0., 0., 0., 6., 0., 0., 0., 0., 4.],\n",
      "        [1., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
      "        [1., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
      "        [1., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
      "        [1., 0., 0., 0., 0., 1., 0., 0., 0., 0., 0.]])\n",
      "tensor([[0, 0, 0, 0, 1, 2, 3, 4],\n",
      "        [1, 2, 3, 4, 0, 0, 0, 0]])\n",
      "tensor([[1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.],\n",
      "        [1., 0., 0., 0.]])\n",
      "tensor([[    0.0000,    13.2100,   -10.5499,     3.1865,    13.7363,    35.3641,\n",
      "             1.2177, -1101.4878, -1101.4098, -1101.3840, -1102.0229,     6.4690,\n",
      "           -17.1722,   -17.2868,   -17.3897,   -16.1519,   157.7118,   157.7100,\n",
      "           157.7070]])\n",
      "tensor([[-1.2700e-02,  1.0858e+00,  8.0000e-03],\n",
      "        [ 2.2000e-03, -6.0000e-03,  2.0000e-03],\n",
      "        [ 1.0117e+00,  1.4638e+00,  3.0000e-04],\n",
      "        [-5.4080e-01,  1.4475e+00, -8.7660e-01],\n",
      "        [-5.2380e-01,  1.4379e+00,  9.0640e-01]])\n",
      "tensor([0])\n",
      "tensor([6, 1, 1, 1, 1])\n"
     ]
    }
   ],
   "source": [
    "# sample dataset QM9\n",
    "\n",
    "path = \"open catalyst\"\n",
    "dataset_qm9 = QM9(root=path)\n",
    "target = 9\n",
    "\n",
    "for data in dataset_qm9:\n",
    "    print(data.x)\n",
    "    print(data.edge_index)\n",
    "    print(data.edge_attr)\n",
    "    print(data.y)\n",
    "    print(data.pos)\n",
    "    print(data.idx)\n",
    "    print(data.z)\n",
    "    break"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2bc78bfc",
   "metadata": {},
   "source": [
    "# Datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "ffa5f1ad",
   "metadata": {},
   "outputs": [
    {
     "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>O_x</th>\n",
       "      <th>O_y</th>\n",
       "      <th>O_z</th>\n",
       "      <th>H1_x</th>\n",
       "      <th>H1_y</th>\n",
       "      <th>H1_z</th>\n",
       "      <th>H2_x</th>\n",
       "      <th>H2_y</th>\n",
       "      <th>H2_z</th>\n",
       "      <th>energy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.008938</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.435510</td>\n",
       "      <td>-0.797301</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.597660</td>\n",
       "      <td>0.655452</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.487586</td>\n",
       "      <td>-76.273156</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.007150</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.435510</td>\n",
       "      <td>-0.783116</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.586653</td>\n",
       "      <td>0.669637</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.498593</td>\n",
       "      <td>-76.273893</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.023238</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.438126</td>\n",
       "      <td>-0.572714</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.420285</td>\n",
       "      <td>0.941522</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.706480</td>\n",
       "      <td>-76.250688</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.028601</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.435510</td>\n",
       "      <td>-0.953335</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.718743</td>\n",
       "      <td>0.499417</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.366503</td>\n",
       "      <td>-76.245592</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.014300</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.436818</td>\n",
       "      <td>-0.855226</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.641063</td>\n",
       "      <td>0.628268</td>\n",
       "      <td>0.0</td>\n",
       "      <td>-0.464943</td>\n",
       "      <td>-76.264198</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        O_x  O_y       O_z      H1_x  H1_y      H1_z      H2_x  H2_y  \\\n",
       "0  0.008938  0.0  0.435510 -0.797301   0.0 -0.597660  0.655452   0.0   \n",
       "1  0.007150  0.0  0.435510 -0.783116   0.0 -0.586653  0.669637   0.0   \n",
       "2 -0.023238  0.0  0.438126 -0.572714   0.0 -0.420285  0.941522   0.0   \n",
       "3  0.028601  0.0  0.435510 -0.953335   0.0 -0.718743  0.499417   0.0   \n",
       "4  0.014300  0.0  0.436818 -0.855226   0.0 -0.641063  0.628268   0.0   \n",
       "\n",
       "       H2_z     energy  \n",
       "0 -0.487586 -76.273156  \n",
       "1 -0.498593 -76.273893  \n",
       "2 -0.706480 -76.250688  \n",
       "3 -0.366503 -76.245592  \n",
       "4 -0.464943 -76.264198  "
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df1 = pd.read_csv(\"water_coordinates\")\n",
    "df2 = pd.read_csv(\"water_zmatrix\")\n",
    "df1.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "293c4806",
   "metadata": {},
   "outputs": [],
   "source": [
    "def x2e(x):\n",
    "    \"\"\"convert xyz coordinates to inverse pairwise distance\"\"\"\n",
    "    r2 = np.sum((x - x[:, np.newaxis, :]) ** 2, axis=-1)\n",
    "    e = np.where(r2 != 0, 1 / r2, 0.0)\n",
    "    return e"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "de600c4b",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "36003\n"
     ]
    }
   ],
   "source": [
    "angle = df2[\"θ(at2,at0,at1)\"]\n",
    "# angle = torch.tensor(angle, dtype=torch.float)\n",
    "# edge_index = df2.iloc[:,:2]\n",
    "# edge_index.head()\n",
    "angle = angle.to_numpy()\n",
    "size = angle.size\n",
    "print(size)\n",
    "# zeros = np.zeros(size, dtype=float)\n",
    "# edge_arr = np.vstack((zeros,angle,zeros))\n",
    "# edge_lst = []\n",
    "# edge_atr = edge_atr.reshape(-1,3)\n",
    "# for i in range(size):\n",
    "#     x = edge_arr[:,i]\n",
    "#     edge_lst.append(x)\n",
    "    \n",
    "# for i in range(len(angle)):\n",
    "#     edge[0][0] = 1.\n",
    "#     edge[0][1] = angle[i]\n",
    "#     edge[1][0] = 1.\n",
    "#     edge[1][1] = angle[i]\n",
    "#     edge_lst.append(edge)\n",
    "# edge_atr = np.asarray(edge_lst)\n",
    "# edge_atr = torch.tensor(edge_lst)\n",
    "# edge_atr.size()\n",
    "\n",
    "edge_atr = torch.tensor([8,1,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "1d8cbed5",
   "metadata": {},
   "outputs": [],
   "source": [
    "# normalisation of data\n",
    "\n",
    "# scaler1 = MinMaxScaler()\n",
    "# scaler2 = MinMaxScaler()\n",
    "# nodes = df1.drop(columns=[\"energy\"])\n",
    "# nodes = scaler1.fit_transform(nodes)\n",
    "# nodes = torch.tensor(nodes, dtype=torch.float)\n",
    "# nodes = torch.reshape(nodes, (-1,3,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "e8198406",
   "metadata": {},
   "outputs": [],
   "source": [
    "# non-noralised data\n",
    "nodes = df1.drop(columns=[\"energy\"])\n",
    "nodes = nodes.to_numpy()\n",
    "nodes = torch.tensor(nodes, dtype=torch.float)\n",
    "nodes = torch.reshape(nodes, (-1,3,3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "52fdbd33",
   "metadata": {},
   "outputs": [],
   "source": [
    "#normalised targets\n",
    "\n",
    "# y = df1[\"energy\"]\n",
    "# y = y.to_numpy().reshape(-1,1)\n",
    "# y = scaler2.fit_transform(y)\n",
    "# y = torch.tensor(y, dtype=torch.float).reshape((-1,1,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "a047d8a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "#non-normalised targets\n",
    "\n",
    "y = df1[\"energy\"]\n",
    "y = y.to_numpy()\n",
    "y = torch.tensor(y, dtype=torch.float).reshape((-1,1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "b0a407b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "# edge_index = torch.tensor([[1.,1.,0.],[1.,0.,1.]],dtype=torch.float)\n",
    "# edge_index = torch.LongTensor([[1,1,0],[1,0,1]])\n",
    "edge_index = torch.tensor([[0, 1, 1, 2], [1, 0, 2, 1]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "a87d0c1b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "36003"
      ]
     },
     "execution_count": 46,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dataset = []\n",
    "for i in range(size):\n",
    "    data = Data(pos=nodes[i], edge_index=edge_index, edge_attr=edge_atr, y=y[i])\n",
    "    dataset.append(data)\n",
    "\n",
    "len(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "4d072220",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Data(edge_index=[2, 4], edge_attr=[3], y=[1], pos=[3, 3])"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample = dataset[0]\n",
    "sample"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "57b251b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "random.shuffle(dataset)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "e9376ac3",
   "metadata": {},
   "outputs": [],
   "source": [
    "data_tr = dataset[:math.floor(0.7*size)]\n",
    "data_va = dataset[math.floor(0.7*size):math.floor(0.9*size)]\n",
    "data_te = dataset[math.floor(0.9*size):]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bc8921a3",
   "metadata": {},
   "source": [
    "# Data Loaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "e87d77d2",
   "metadata": {},
   "outputs": [],
   "source": [
    "loader_train = DataLoader(data_tr, batch_size=32, shuffle=True)\n",
    "loader_val = DataLoader(data_va, batch_size=32, shuffle=True)\n",
    "loader_test = DataLoader(data_te, batch_size=32, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7009343b",
   "metadata": {},
   "source": [
    "# Dimenet Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "a8abf036",
   "metadata": {},
   "outputs": [],
   "source": [
    "model = DimeNet(hidden_channels=128, out_channels=1, num_blocks=6, num_bilinear=8, num_spherical=7, num_radial=6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "5f4ffed8",
   "metadata": {},
   "outputs": [],
   "source": [
    "model.reset_parameters()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "915c5fe9",
   "metadata": {},
   "source": [
    "# Loss Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "3f88bfa2",
   "metadata": {},
   "outputs": [],
   "source": [
    "def loss(y_hat,y):\n",
    "    loss = torch.nn.MSELoss(reduction=\"mean\")\n",
    "    output = loss(y_hat, y)\n",
    "    return output"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "95ee9db3",
   "metadata": {},
   "source": [
    "# Optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "b6aa36be",
   "metadata": {},
   "outputs": [],
   "source": [
    "optim1 = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
    "optim2 = torch.optim.RAdam(model.parameters(), lr=1e-4)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c63355fe",
   "metadata": {},
   "source": [
    "# Training Loop"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "b3c0890d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_on_batch(data, model, opt, loss):\n",
    "    y_hat = model.forward(z=data.edge_attr, pos=data.pos, batch=data.batch)\n",
    "    loss_mean = loss(y_hat, data.y)\n",
    "    loss_mean.backward()\n",
    "    opt.step()  \n",
    "    return loss_mean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "id": "1d85330f",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sudka\\AppData\\Local\\Temp\\ipykernel_18132\\1701331734.py:1: UserWarning: Anomaly Detection has been enabled. This mode will increase the runtime and should only be enabled for debugging.\n",
      "  with autograd.detect_anomaly():\n",
      "  0%|                                                                                          | 0/788 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "DataBatch(edge_index=[2, 128], edge_attr=[96], y=[32], pos=[96, 3], batch=[96], ptr=[33])\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\nn\\modules\\loss.py:529: UserWarning: Using a target size (torch.Size([32])) that is different to the input size (torch.Size([1])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.\n",
      "  return F.mse_loss(input, target, reduction=self.reduction)\n",
      "C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\autograd\\__init__.py:173: UserWarning: Error detected in MseLossBackward0. Traceback of forward call that caused the error:\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\runpy.py\", line 197, in _run_module_as_main\n",
      "    return _run_code(code, main_globals, None,\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\runpy.py\", line 87, in _run_code\n",
      "    exec(code, run_globals)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel_launcher.py\", line 17, in <module>\n",
      "    app.launch_new_instance()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\traitlets\\config\\application.py\", line 846, in launch_instance\n",
      "    app.start()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\kernelapp.py\", line 712, in start\n",
      "    self.io_loop.start()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\tornado\\platform\\asyncio.py\", line 215, in start\n",
      "    self.asyncio_loop.run_forever()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\asyncio\\base_events.py\", line 601, in run_forever\n",
      "    self._run_once()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\asyncio\\base_events.py\", line 1905, in _run_once\n",
      "    handle._run()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\asyncio\\events.py\", line 80, in _run\n",
      "    self._context.run(self._callback, *self._args)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\kernelbase.py\", line 510, in dispatch_queue\n",
      "    await self.process_one()\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\kernelbase.py\", line 499, in process_one\n",
      "    await dispatch(*args)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\kernelbase.py\", line 406, in dispatch_shell\n",
      "    await result\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\kernelbase.py\", line 730, in execute_request\n",
      "    reply_content = await reply_content\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\ipkernel.py\", line 383, in do_execute\n",
      "    res = shell.run_cell(\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\ipykernel\\zmqshell.py\", line 528, in run_cell\n",
      "    return super().run_cell(*args, **kwargs)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2881, in run_cell\n",
      "    result = self._run_cell(\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 2936, in _run_cell\n",
      "    return runner(coro)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\async_helpers.py\", line 129, in _pseudo_sync_runner\n",
      "    coro.send(None)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3135, in run_cell_async\n",
      "    has_raised = await self.run_ast_nodes(code_ast.body, cell_name,\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3338, in run_ast_nodes\n",
      "    if await self.run_code(code, result, async_=asy):\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\IPython\\core\\interactiveshell.py\", line 3398, in run_code\n",
      "    exec(code_obj, self.user_global_ns, self.user_ns)\n",
      "  File \"C:\\Users\\sudka\\AppData\\Local\\Temp\\ipykernel_18132\\1701331734.py\", line 7, in <cell line: 1>\n",
      "    loss_mean = train_on_batch(data,model,optim1,loss)\n",
      "  File \"C:\\Users\\sudka\\AppData\\Local\\Temp\\ipykernel_18132\\2551471310.py\", line 15, in train_on_batch\n",
      "    loss_mean = loss(y_hat, data.y)\n",
      "  File \"C:\\Users\\sudka\\AppData\\Local\\Temp\\ipykernel_18132\\2551471310.py\", line 8, in loss\n",
      "    output = loss(y_hat, y)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\nn\\modules\\module.py\", line 1110, in _call_impl\n",
      "    return forward_call(*input, **kwargs)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\nn\\modules\\loss.py\", line 529, in forward\n",
      "    return F.mse_loss(input, target, reduction=self.reduction)\n",
      "  File \"C:\\Users\\sudka\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\nn\\functional.py\", line 3262, in mse_loss\n",
      "    return torch._C._nn.mse_loss(expanded_input, expanded_target, _Reduction.get_enum(reduction))\n",
      " (Triggered internally at  C:\\cb\\pytorch_1000000000000\\work\\torch\\csrc\\autograd\\python_anomaly_mode.cpp:104.)\n",
      "  Variable._execution_engine.run_backward(  # Calls into the C++ engine to run the backward pass\n",
      "  0%|                                                                                          | 0/788 [00:15<?, ?it/s]\n"
     ]
    },
    {
     "ename": "RuntimeError",
     "evalue": "Function 'MseLossBackward0' returned nan values in its 0th output.",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mRuntimeError\u001b[0m                              Traceback (most recent call last)",
      "Input \u001b[1;32mIn [127]\u001b[0m, in \u001b[0;36m<cell line: 1>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      5\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m data \u001b[38;5;129;01min\u001b[39;00m tqdm(loader_train):\n\u001b[0;32m      6\u001b[0m     \u001b[38;5;28mprint\u001b[39m(data)\n\u001b[1;32m----> 7\u001b[0m     loss_mean \u001b[38;5;241m=\u001b[39m \u001b[43mtrain_on_batch\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdata\u001b[49m\u001b[43m,\u001b[49m\u001b[43mmodel\u001b[49m\u001b[43m,\u001b[49m\u001b[43moptim1\u001b[49m\u001b[43m,\u001b[49m\u001b[43mloss\u001b[49m\u001b[43m)\u001b[49m\n\u001b[0;32m      8\u001b[0m \u001b[38;5;28mprint\u001b[39m(loss_mean)\n\u001b[0;32m      9\u001b[0m train_loss\u001b[38;5;241m.\u001b[39mappend(loss_mean\u001b[38;5;241m.\u001b[39mdetach()\u001b[38;5;241m.\u001b[39mnumpy())\n",
      "Input \u001b[1;32mIn [124]\u001b[0m, in \u001b[0;36mtrain_on_batch\u001b[1;34m(data, model, opt, loss)\u001b[0m\n\u001b[0;32m     14\u001b[0m y_hat \u001b[38;5;241m=\u001b[39m model\u001b[38;5;241m.\u001b[39mforward(z\u001b[38;5;241m=\u001b[39mdata\u001b[38;5;241m.\u001b[39medge_attr, pos\u001b[38;5;241m=\u001b[39mdata\u001b[38;5;241m.\u001b[39mpos)\n\u001b[0;32m     15\u001b[0m loss_mean \u001b[38;5;241m=\u001b[39m loss(y_hat, data\u001b[38;5;241m.\u001b[39my)\n\u001b[1;32m---> 16\u001b[0m \u001b[43mloss_mean\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mbackward\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[0;32m     17\u001b[0m opt\u001b[38;5;241m.\u001b[39mstep()  \n\u001b[0;32m     18\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m loss_mean\n",
      "File \u001b[1;32m~\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\_tensor.py:363\u001b[0m, in \u001b[0;36mTensor.backward\u001b[1;34m(self, gradient, retain_graph, create_graph, inputs)\u001b[0m\n\u001b[0;32m    354\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m has_torch_function_unary(\u001b[38;5;28mself\u001b[39m):\n\u001b[0;32m    355\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m handle_torch_function(\n\u001b[0;32m    356\u001b[0m         Tensor\u001b[38;5;241m.\u001b[39mbackward,\n\u001b[0;32m    357\u001b[0m         (\u001b[38;5;28mself\u001b[39m,),\n\u001b[1;32m   (...)\u001b[0m\n\u001b[0;32m    361\u001b[0m         create_graph\u001b[38;5;241m=\u001b[39mcreate_graph,\n\u001b[0;32m    362\u001b[0m         inputs\u001b[38;5;241m=\u001b[39minputs)\n\u001b[1;32m--> 363\u001b[0m \u001b[43mtorch\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mautograd\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mbackward\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mgradient\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mretain_graph\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mcreate_graph\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43minputs\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43minputs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[1;32m~\\anaconda3\\envs\\aquas\\lib\\site-packages\\torch\\autograd\\__init__.py:173\u001b[0m, in \u001b[0;36mbackward\u001b[1;34m(tensors, grad_tensors, retain_graph, create_graph, grad_variables, inputs)\u001b[0m\n\u001b[0;32m    168\u001b[0m     retain_graph \u001b[38;5;241m=\u001b[39m create_graph\n\u001b[0;32m    170\u001b[0m \u001b[38;5;66;03m# The reason we repeat same the comment below is that\u001b[39;00m\n\u001b[0;32m    171\u001b[0m \u001b[38;5;66;03m# some Python versions print out the first line of a multi-line function\u001b[39;00m\n\u001b[0;32m    172\u001b[0m \u001b[38;5;66;03m# calls in the traceback and some print out the last line\u001b[39;00m\n\u001b[1;32m--> 173\u001b[0m \u001b[43mVariable\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_execution_engine\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mrun_backward\u001b[49m\u001b[43m(\u001b[49m\u001b[43m  \u001b[49m\u001b[38;5;66;43;03m# Calls into the C++ engine to run the backward pass\u001b[39;49;00m\n\u001b[0;32m    174\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtensors\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mgrad_tensors_\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mretain_graph\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mcreate_graph\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43minputs\u001b[49m\u001b[43m,\u001b[49m\n\u001b[0;32m    175\u001b[0m \u001b[43m    \u001b[49m\u001b[43mallow_unreachable\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mTrue\u001b[39;49;00m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43maccumulate_grad\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mTrue\u001b[39;49;00m\u001b[43m)\u001b[49m\n",
      "\u001b[1;31mRuntimeError\u001b[0m: Function 'MseLossBackward0' returned nan values in its 0th output."
     ]
    }
   ],
   "source": [
    "with autograd.detect_anomaly():\n",
    "    epochs = 5\n",
    "    train_loss = []\n",
    "    for epoch in range(epochs):\n",
    "        for data in tqdm(loader_train):\n",
    "            print(data)\n",
    "            loss_mean = train_on_batch(data,model,optim1,loss)\n",
    "        print(loss_mean)\n",
    "        train_loss.append(loss_mean.detach().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c823c4c2",
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.plot(train_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cea54fc0",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fcbaa95d",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "22a4d12b",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c15acfe1",
   "metadata": {},
   "outputs": [],
   "source": [
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b649c57",
   "metadata": {},
   "source": [
    "# Sample code with 3 data points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "id": "81524db0",
   "metadata": {},
   "outputs": [],
   "source": [
    "edge_index = torch.tensor([[0, 1, 1, 2], [1, 0, 2, 1]])\n",
    "edge_atr = torch.tensor([8,1,1])\n",
    "\n",
    "# pos = np.asarray([[ 0.0089,  0.0000,  0.4355,\n",
    "#          -0.7973,  0.0000, -0.5977,\n",
    "#           0.6555,  0.0000, -0.4876],\n",
    "\n",
    "#         [ 0.0072,  0.0000,  0.4355,\n",
    "#          -0.7831,  0.0000, -0.5867,\n",
    "#           0.6696,  0.0000, -0.4986],\n",
    "\n",
    "#         [-0.0232,  0.0000,  0.4381,\n",
    "#          -0.5727,  0.0000, -0.4203,\n",
    "#           0.9415,  0.0000, -0.7065]])\n",
    "\n",
    "# target = np.asarray([-76.273156,-76.273893,-76.250688])\n",
    "\n",
    "#non-normalised inputs\n",
    "# pos = torch.tensor(pos, dtype=torch.float)\n",
    "# pos = torch.reshape(pos, (-1,3,3))\n",
    "\n",
    "# target = torch.tensor(target, dtype=torch.float).reshape((-1,1))\n",
    "\n",
    "#normalised inputs\n",
    "\n",
    "# scaler1 = MinMaxScaler()\n",
    "# scaler2 = MinMaxScaler()\n",
    "# pos = scaler1.fit_transform(pos)\n",
    "# pos = torch.tensor(pos, dtype=torch.float)\n",
    "# pos = torch.reshape(pos, (-1,3,3))\n",
    "\n",
    "# target = scaler2.fit_transform(target.reshape(-1,1))\n",
    "# target = torch.tensor(target, dtype=torch.float)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "id": "f5ee96e2",
   "metadata": {},
   "outputs": [],
   "source": [
    "pos = torch.tensor([[[ 0.0089,  0.0000,  0.4355],\n",
    "         [-0.7973,  0.0000, -0.5977],\n",
    "         [ 0.6555,  0.0000, -0.4876]],\n",
    "\n",
    "        [[ 0.0072,  0.0000,  0.4355],\n",
    "         [-0.7831,  0.0000, -0.5867],\n",
    "         [ 0.6696,  0.0000, -0.4986]],\n",
    "\n",
    "        [[-0.0232,  0.0000,  0.4381],\n",
    "         [-0.5727,  0.0000, -0.4203],\n",
    "         [ 0.9415,  0.0000, -0.7065]]])\n",
    "\n",
    "target = torch.tensor([[-76.2732],\n",
    "        [-76.2739],\n",
    "        [-76.2507]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "id": "b7262f58",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_2 = DimeNet(hidden_channels=128, out_channels=1, num_blocks=6, num_bilinear=8, num_spherical=7, num_radial=6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "id": "d7d49945",
   "metadata": {},
   "outputs": [],
   "source": [
    "model_2.reset_parameters()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "cbe49c20",
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = []\n",
    "for i in range(3):\n",
    "    data = Data(pos=pos[i], edge_index=edge_index, edge_attr=edge_atr, y=target[i])\n",
    "    dataset.append(data)\n",
    "    \n",
    "def loss(y_hat,y):\n",
    "    loss = torch.nn.MSELoss(reduction=\"mean\")\n",
    "    output = loss(y_hat, y)\n",
    "    return output\n",
    "\n",
    "opt = torch.optim.Adam(model_2.parameters(), lr=1e-5)\n",
    "\n",
    "def train_on_batch(data, model, opt, loss):\n",
    "    y_hat = model.forward(z=data.edge_attr, pos=data.pos)\n",
    "    loss_mean = loss(y_hat, data.y)\n",
    "    loss_mean.backward()\n",
    "    opt.step()  \n",
    "    return loss_mean"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "id": "f50a4362",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\sudka\\AppData\\Local\\Temp\\ipykernel_18132\\3595110761.py:1: UserWarning: Anomaly Detection has been enabled. This mode will increase the runtime and should only be enabled for debugging.\n",
      "  with autograd.detect_anomaly():\n",
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.06s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5809.4023, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:02<00:00,  1.09it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5798.9844, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:02<00:00,  1.04it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5782.9448, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:02<00:00,  1.15it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5758.5488, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.15s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5721.7905, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.08s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5667.1230, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.01s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5586.7251, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:02<00:00,  1.08it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5469.0142, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:03<00:00,  1.02s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5298.3540, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████████| 3/3 [00:02<00:00,  1.07it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(5052.3711, grad_fn=<MseLossBackward0>)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "with autograd.detect_anomaly():\n",
    "    epochs = 10\n",
    "    train_loss = []\n",
    "    for epoch in range(epochs):\n",
    "        for data in tqdm(dataset):\n",
    "            loss_mean = train_on_batch(data,model_2,opt,loss)\n",
    "        print(loss_mean)\n",
    "        train_loss.append(loss_mean.detach().numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "id": "a26e2de5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x289e49d30a0>]"
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "90e2f4b1",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:aquas]",
   "language": "python",
   "name": "conda-env-aquas-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}