Daniel Falbel

Actually it seems to be a bug with using the `$data()` method. Here's an equivalent implementation: ``` r library(torch) m 0 #> [ CPUFloatType{} ] ``` Created on 2020-10-19 by...

Ok! This should be an easy fix, I'll do it as soon as possible.

Wait, that doesn't seem to error for me: ``` r library(torch) m torch_tensor #> -3.66959 #> [ CPUFloatType{} ] with_no_grad({ m$weight[m$weight < -100] torch_tensor #> -3.66959 #> [ CPUFloatType{} ]...

Ok, so IIUC the `with_no_grad` solution is working as expected?

Yes, this is still kind of expected. We still need to make performance improvements in the R side. Specially related to dataloading and in the optimizers code. However, small examples...

Yeah! We are interested in adding this to torch but have a few other priorities first in the line. It would be awesome to have your contribution and I can...

Oh I didn't know that was possible either. Sure, there's no problem in supporting only reverse-mode in a first iteration.

I like `agf_` as a prefix. Although since the `autograd_` namespace is quite small, I think we could also omit the functional references and use `autograd_` for them too. So...

Hi @bearloga! thanks very much! Interesting!! It seems that it's possible to compile LibTorch on raspberry pi (eg: https://stackoverflow.com/questions/62755739/libtorch-on-raspberry-cant-load-pt-file-but-working-on-ubuntu). We could then, in theory, create a case for it here:...

Torch for R can be built from source with https://github.com/mlverse/torch/blob/master/tools/buildlantern.R and then `devtools::install`. However we assume a pre-built LibTorch binary, which is currently not provided by the PyTorch team for...