tch-rs
tch-rs copied to clipboard
LaurentMazare
CUDA 11.3
The URL https://download.pytorch.org/libtorch/cu111/libtorch-shared-with-deps-1.11.0%2Bcu111.zip is AccessDenied now.
An upgrade to cu113 is necessary.
Thanks for reporting this issue, indeed cu113 seems to work fine so the build script has been updated to reflect this.
With CUDA 11.6, setting the TORCH_CUDA_VERSION=cu116:
Unexpected response code 403 for https://download.pytorch.org/libtorch/cu116/libtorch-cxx11-abi-shared-with-deps-1.12.0.zip'
I had to go to https://download.pytorch.org/libtorch/cu116/libtorch-cxx11-abi-shared-with-deps-1.12.1%2Bcu116.zip and install manually the pre-build files.
@Mec-iS that's odd. I just tried with the latest github tip and TORCH_CUDA_VERSION=cu116 and it worked well. Is it possible that you were using an older version of this crate which did not support cu116? This would have resulted in the error message you saw, anyway I just changed the build script to make understanding the root cause a bit easier.
I was installing version 0.8.0 using crate. I don't think the version is the problem as the link I posted above indeed return "Access Denied" if you try to visit it. The link that worked for me is https://download.pytorch.org/libtorch/cu116/libtorch-cxx11-abi-shared-with-deps-1.12.1%2Bcu116.zip
See the name of the file at the end is different.
Support for cu116 was added after the crate 0.8.0 release so this explains the error you saw, the next version will have a better error message for this.
ok thanks. the latest crate in crates.io is 0.8.0, where I can find the latest ones?
The latest version is 0.8.0, we usually make crate releases only on new versions of libtorch so the next one would be when PyTorch 1.13 is released. Maybe you can point your Cargo.toml at some recent github revision for the time being?
Closing this as the 1.13 version has been released for quite some time now.