Kyle Daruwalla

I was thinking that all the datasets in MLDatasets (including the vision ones) should have pre-processing helpers. If I have a custom model that I plan training with CIFAR-100, then...

I'm okay with that, as well as merging the pipelines into Metalhead if that's what folks want. Is the concern here about MLDatasets deps? It already provides the common vision...

I think that is an easy first step we can do until we split off the vision datasets like Carlo suggested.

I don't think they belong in Metalhead.jl as they are not specific to vision models. The advantage of NNlib is that packages unrelated to Flux can use it. `topk` is...

cc @theabhirath please add the models you are planning on implementing to this list

A starting point for EfficientNet can be found at https://github.com/pxl-th/EfficientNet.jl (see #106).

> Hello, Can I work on inception v4 and efficient net?(Also I have already coded ESRGANs) As mentioned, there is already a PR for EfficientNet, but InceptionNet v4 would be...

> I could provide the pretrained weights for VGG and ResNet converted from PyTorch (once some minor changes to the Metalhead models are merged so that they are equivalent to...

The Flux [contribution guide](https://github.com/FluxML/Flux.jl/blob/master/CONTRIBUTING.md) has some info as well as links for how to get started with making your first contribution. I'll briefly summarize the process for this repo (apologies...

I think we can support image size, input channels, and output classes as consistent arguments to the highest level constructors (i.e. the ones that have the `pretrain` keyword). I think...