Pytorch studying and practicing for Deep Learning
Datasets Description:
In this project, the necessary datasets are MNIST, IMDB and CELEBA.
The MNIST and IMDB could be directly downloaded from the torchvision and torchtext.
The CELEBA could be downloaded from the Google official: https://drive.google.com/drive/folders/0B7EVK8r0v71pbWNEUjJKdDQ3dGc
For other datasets, here is the baidu cloud's link: https://pan.baidu.com/s/1-nmxzd4Wm7ijRpUMzRnQQA, code: rbo5
The above link only has the datasets for Recurrent Neural Network project.
The remaining datasets of Convolutional Neural Network would not be offered. (Neural Style Transfer and Resnet Transfer Learning)
Library Description:
python -m pip install --user --upgrade pip --upgrade setuptools -r requirements.txt
Hint: Some words split tools should be extra downloaded or setted from the internet.
Project Description:
All the projects in Deep folders have the Config.py file, which contains all the hyperparameters and directories settings.
Directly running the Trainer.py file, all the hyperparameters and directories would be applied as the default value.
There are two ways to changing the settings, one is directly modifying the default value in Config.py file.
Another one is giving the new value by running the instructions like below:
python .../Trainer.py -gpu [GPUID] -bs [BatchSize] - lr [LearningRate] -modelDir [Checkpoints Saving Dir] ...