torch

torch copied to clipboard

mrchypark

Distributions

1

A checklist of distributions we'd like to port from [PyTorch](https://pytorch.org/docs/stable/distributions.html): - [x] Bernoulli - [ ] Beta - [ ] Binomial - [x] Categorical - [ ] Cauchy - [x]...

krzjoa

[Feature request] Warn when overwriting modules of nn_sequential

1

I think one should get at least a warning when overwriting already present modules. ``` r library(torch) n = nn_sequential() n$add_module("a", nn_relu()) n$add_module("a", nn_linear(1, 2)) # ── Modules ───────────────────────────────────────────────────────────────────── #>...

sebffischer

[feature request] Way to load state dict for corrup model (invalid external pointers)

1

Hey, thank you very much for taking the time! I have some questions regarding the serialization of torch models. In mlr3torch we have learners (R6 classes) that have a field...

sebffischer

Rstudio crashes when fitting using luz

1

Hi, I have followed installation instructions here https://torch.mlverse.org/docs/articles/installation.html and sintall CUDA 11.3 an cuddn 8.4. I have tried to run example here https://mlverse.github.io/luz/articles/examples/dogs-vs-cats-binary-classification.html but my R studio crasheh afer I...

MislavSag

# Current behavour `$movedim()` use 0-indexed dimensions for its parameters source and destination. This is not expected, create a lot of confusion is the {torch} perfect world where everything is...

cregouby

Support later versions of cuda such as cuda 11.7 please

1

Torch only supports cuda 11.3 or earlier versions, unfortunately my cuda version is 11.7, so GPU accelerating is not available.

Newtype07

https://github.com/pytorch/pytorch/releases/tag/v1.12.0

hsbadr

R crash

10

This is probably not the finest nor precise question... but here we go. In some applications we are bootstrapping a torch-based model (nothing too fancy, mostly some variant of a...

philgoucou

the section on multi step prediction is very, very slow on mac os. Any tips or tricks to speed it up? ``` n_forecast

jzicker