Learning resources

• Deeplearning AI
• Pytorch book
  • transformer explained
  • NN explanation
  • CNN explainer
  • Github source code
• PyTorch for Deep Learning Bootcamp
• Pytorch popular transfomers
  • AlexNet
  • ResNet
• OCR for Latex

information

• twitter for new papers
• papers with code

Zero to Mastery Learn PyTorch for Deep Learning

Pytorch model

PyTorch 4 most import modules:  [torch.nn](https://pytorch.org/docs/stable/nn.html), [torch.optim](https://pytorch.org/docs/stable/optim.html), [torch.utils.data.Dataset](https://pytorch.org/docs/stable/data.html#torch.utils.data.Dataset) and [torch.utils.data.DataLoader](https://pytorch.org/docs/stable/data.html).

Pytorch training loops

1. Forward pass - The model goes through all of the training data once, performing its forward() function calculations (model(x_train)).
2. Calculate the loss - The model's outputs (predictions) are compared to the ground truth and evaluated to see how wrong they are (loss = loss_fn(y_pred, y_train).
3. Zero gradients - The optimizers gradients are set to zero (they are accumulated by default) so they can be recalculated for the specific training step (optimizer.zero_grad()).
4. Perform backpropagation on the loss - Computes the gradient of the loss with respect for every model parameter to be updated (each parameter with requires_grad=True). This is known as backpropagation, hence "backwards" (loss.backward()).
5. Step the optimizer (gradient descent) - Update the parameters with requires_grad=True with respect to the loss gradients in order to improve them (optimizer.step()).

Pytorch layers

- Common Layer Types
    - Linear Layers
    - Convolutional Layers
    - Recurrent Layers
    - Transformers

Pytorch datasets

build custom datasets

1. Subclass torch.utils.data.Dataset.
2. Initialize our subclass with a targ_dir parameter (the target data directory) and transform parameter (so we have the option to transform our data if needed).
3. Create several attributes for paths (the paths of our target images), transform (the transforms we might like to use, this can be None), classes and class_to_idx (from our find_classes() function).
4. Create a function to load images from file and return them, this could be using PIL or [torchvision.io](https://pytorch.org/vision/stable/io.html#image) (for input/output of vision data).
5. Overwrite the __len__ method of torch.utils.data.Dataset to return the number of samples in the Dataset, this is recommended but not required. This is so you can call len(Dataset).
6. Overwrite the __getitem__ method of torch.utils.data.Dataset to return a single sample from the Dataset, this is required.

CNN

https://poloclub.github.io/cnn-explainer/

Pytorch Modular

from torchinfo import summary

Pytorch transfer learning

Freezing the base model and changing the output layer

weights = torchvision.models.EfficientNet_B0_Weights.DEFAULT # .DEFAULT = best available weights 
model = torchvision.models.efficientnet_b0(weights=weights).to(device)
# Freeze all base layers in the "features" section of the model (the feature extractor) by setting requires_grad=False
for param in model.features.parameters():
    param.requires_grad = Falsec
# Get the length of class_names (one output unit for each class)
output_shape = len(class_names)
# Recreate the classifier layer and seed it to the target device
model.classifier = torch.nn.Sequential(
    torch.nn.Dropout(p=0.2, inplace=True), 
    torch.nn.Linear(in_features=1280, 
                    out_features=output_shape, # same number of output units as our number of classes
                    bias=True)).to(device)