Successor of pytorch-pretrained-cnns minus pytorch lightning
Scripts:
train.py
: Training with various parameterseval.py
: Runs evaluation of a given checkpoint with validation dataseteval_adv_rob.py
: Runs evaluation of a given checkpoint with validation dataset under adversarial attacks of AutoAttackplot_convolutions.py
: Plots the 3x3 conv filters of a given checkpointplot_feature_maps.py
: Plots the feature maps of selected layers of a given checkpoints for a given input sampleinit_data.py
: Prepares the data independent of training (usefull before sweeps)ckpt_info.py
: Dumps info stored in a checkpointto_onnx.py
: Exports checkpoints to ONNX
Code partially taken from https://github.com/huyvnphan/PyTorch_CIFAR10, https://github.com/1M50RRY/resnet18-preact, https://github.com/pytorch/vision/, https://github.com/andreasveit/densenet-pytorch/.