Add feature-extraction AlexNet

[SurajK7]

For now, I have used lists for data-loading so the implementation is slow and not memory efficient(~19 GB footprint).
The model achieves ~80% training accuracy after 5000 iterations but validation accuracy seems rather dismal, I will try to figure what's going wrong.
Meanwhile, any suggestions or feedback would be greatly appreciated!

[SurajK7]

Everything seems to work fine now. I trained the final layer for 5 epochs and achieved an accuracy of ~70% on the test set which is good enough for a fine-grained classification dataset such as ours.

[junjihashimoto]

Awesome!

[junjihashimoto]

I feel a little strange.
Before this pr, alexnet-examples loaded pre-existing weight including top-layer.
This pr loads pre-existing weight except for top-layer, and trains the weight of top-layer to learn new features.
So previous example's use-case is different from this one.

[SurajK7]

The previous example was just to showcase the loading of pretrained parameters and could be used to run inference on images as per ImageNet categories. This example feature-extracts AlexNet for non-trivial data-sets. So as such, there was no use-case of the previous example.

[SurajK7]

I wrote a PyTorch like dataset-abstraction for better loading of data but I doubt that would be of any use now.

[junjihashimoto]

I don't care about the dataload abstraction in this example.
So it's ok.

[austinvhuang]

Thanks for this contribution @SurajK7 . Just two minor requests, would like to merge this:

1. Add the gist code snippets in the directory. Makes it easier to reproduce end-to-end and gets rid of an unneeded dependency on a gist.

2. Move the google docs artifact someplace that's more intended for artifacts, could even be a github release artifact in your own repo (or we can move it to a release here if that works better) as @junjihashimoto suggested

Optional but would be nice - is there no way to avoid having to use both C++ and Python handle the serialization stuff? Would be nice if it could just be one Python script.

[SurajK7]

I think it would be better if we could move the extracted parameters here.
Also, I don't think a single python script would suffice.

[austinvhuang]

Thanks for adding the code @SurajK7 .

Move the serialized data artifact you mean? We're trying to keep artifacts/binaries out of the repo if at all possible.

[austinvhuang]

@SurajK7 also I think if you merge master into your branch that might clear the last CI issue you have.

[SurajK7]

or we can move it to a release here if that works better

@austinvhuang, I meant we can go ahead with the quoted suggestion.

[austinvhuang]

@SurajK7 - I created a release for the artifact in this repo - https://github.com/hasktorch/hasktorch-artifacts/releases/tag/alexnet

See if it works to use this link instead https://github.com/hasktorch/hasktorch-artifacts/releases/download/alexnet/alexNet.pt

[austinvhuang]

Thanks!

[junjihashimoto]

@austinvhuang
Thanks for supporting this pr.
@SurajK7
Thanks!