search

AInnervate / TinyNets.jl

A Julia library for neural network pruning.
MIT License
1 stars 0 forks source link

Add sparse inference to the tests of Flux.jl #4

Open ArthurWalraven opened 2 years ago

ArthurWalraven commented 2 years ago

A good way of ensuring that the tools we use from Flux.jl won't break in the future is to add checks for their integrity directly to their repository. In this way no release will kill our project (like in #1).

On the other hand, this might make it more clear to them how painful supporting sparse networks can be and motivate them to drop it. Anyway, if this is going to happen, it's better that it does as early as possible.

natema commented 2 years ago

A good way of ensuring that the tools we use from Flux.jl won't break in the future is to add checks for their integrity directly to their repository. In this way no release will kill our project (like in #1).

You mean contributing to Flux.jl by adding some tests?

ArthurWalraven commented 2 years ago

A good way of ensuring that the tools we use from Flux.jl won't break in the future is to add checks for their integrity directly to their repository. In this way no release will kill our project (like in #1).

You mean contributing to Flux.jl by adding some tests?

Yes. Specifically for the things that we need but that are not currently covered by their tests.