Julia deep learning library which supports array mutation?

I am searching for a Julia deep learning library which supports array mutation so I don’t need to store the arrays in a buffer.

what library doesn’t support this?

When I use Flux, I have to use Zygote.Buffer. Some times I don’t know whether I have to use Buffer or not. This is really annoying which makes the code ugly.

Out of curiosity, what’s your use case for array mutation? Do you need it for performance reasons or it’s required by model logic?

1 Like

It is required.