Pytorch tutorial with jupyter notebook
pip install jupyter
pip install tqdm
torch
andtorchvision
. Follow instructions in https://pytorch.org/get-started/locally/.
- Clone this repository to any folder.
- Open your favorite shell, and
cd
to that folder. - Run
jupyter notebook
- Shift+Enter through each notebook and cell.
- Tensors and Dynamic Graphs
- Basic Operations involving
torch.Tensor
- Introduction to the dynamic computational graphs of
torch.Autograd
- Basic Operations involving
- Modules and Custom Models
- Adding predefined modules from
torch.nn
totorch.nn.Sequential
- Creating custom models by inheriting
torch.nn.Module
- Adding predefined modules from
- Training and Evaluating a Model
- General training setup for cpu or cuda
- Managing and loading predefined/custom datasets
- Experiencing the conventions of training and evaluating (actually train a CNN on MNIST to accuracy 96%)
- Saving and Loading Model States
- Using predefined models in
torchvision.models
- Saving and loading model/optimizer parameters
- Using predefined models in