[request] nn_flatten

[EmKarr]

Hello,

thank you for the brilliant torch package for R, I really like it :-)

Nevertheless, I am running into problems setting up a model via nn_sequential(), for me it seems like the R counterpart of PyTorch's nn.Flatten() is missing. If so, would it be possible to include such a function? Sequential models currently are a prerequisite for using the "innsight" package, which I would like to do, so using nn_module() is not a solution.

All the best,
stay well and safe,
Mike

[cregouby]

Hello @EmKarr,
Could you elaborate a bit what is missing to cover your need that is not available between
torch::torch_flatten() and tensor$flatten() ?
Thanks

[EmKarr]

Hi cregouby,
sorry, seems like I went to the wrong package to bring up the issue, should have been "torch", not "torchvision".

In torch, this issue already had beed addressed: mlverse/torch#716 (comment)
Thanks,
best wishes,
Mike