darsnack / fluxmodels.jl

Standard (pre-trained) ML models written in Flux.jl

FluxModels

fluxmodels.jl's Issues

Add standard vision models

Add standard vision models that replicates: https://pytorch.org/docs/stable/torchvision/models.html

We want Flux implementations of the architectures. These can be untrained models for now, and we can add another issue for providing pre-trained weights. The models should be input size agnostic so that they can work with CIFAR or ImageNet. PyTorch does this with an Adaptive2DPool layer between the feature extractors and the fully-connected layers. Supporting similar functionality might require a PR to Flux.jl.

List of models:

• VGG
  • VGG11 (w/ and w/o batch norm)
  • VGG13 (w/ and w/o batch norm)
  • VGG16 (w/ and w/o batch norm)
  • VGG19 (w/ and w/o batch norm)
• ResNet
  • ResNet18
  • ResNet34
  • ResNet50
  • ResNet101
  • ResNet152
• AlexNet
• InceptionNet
  • GoogLeNet
  • Inception v3
  • Inception v4

Add some simple tests

It is hard to test this package, because we don't want to run training on the CI. But, a simple test can be passing through a random tensor and ensuring the output size is the expected size. Here is an example:

m = alexnet()
@test size(m(rand(256, 256, 3, 50))) == (1000, 50)

We would add such tests for every model configuration. This ensures that the models can at least do a forward pass with no errors.