Issues with new pytorch version for JP 6.1

On the downloads page, a new version (2.5.0…) of pytorch appears for jetpack 6.1.

However:

  1. The installation (as per the guide) requires installation of libcusparse, and the install script is not supported by the CUDA version shipped with Jetpack 6.1, as the script only supports up to 12.4. The script also has issues with root install.
  2. The numpy version it was compiled with seems to be 1.2(6?), meaning that simply installing numpy which gives version 2.1.x will result in a broken pytorch version.

It would be good to see a pytorch install that supports numpy 2.0 (pytorch 2.3 already supports this), as well as updated install documentation and libcusparse scripts - the current link appearing on the page to the docs is broken.

1 Like

To install libcusparselt, I attempted to modify install_cusparselt.sh script file to bypass the version check:

-if [[ ${CUDA_VERSION:0:4} =~ ^12\.[1-4]$ ]]; then
+if [[ ${CUDA_VERSION:0:4} =~ ^12\.[1-6]$ ]]; then

However, I am not sure if there is any impact from this.

Hi,

You can find the package for JetPack 6.1 in the below link:
https://developer.download.nvidia.cn/compute/redist/jp/v61/

Thanks.

That doesn’t resolve the issue. The issue is not the pytorch version (the one you linked to is identical to the one mentioned in the original post), but rather the fact that the libcusparselt script and the documentation isn’t updated.

Hi,

Are you able to get it work?

You can find the libcusparselt package below:

Thank.s

As brian.wu mentioned above, the cuSPARSE issue is solvable by hacking the install script, though the correct way is likely via the cuSPARSE link you provided, and should probably should be updated in the docs as well.

The pytorch version issue in JP 6.1 still remains - it’s a release candidate (2.5.0rcX) compiled against a old version of python (3.10) as well as an old version of numpy (1.26). It would be good to get the build instructions so that users can try building their own combinations.

1 Like

Hi,

You can find the build from the source instructions in the below topic if custom Python version is needed:

Thanks.

Those build form source instructions haven’t been updated for quite a while - the latest patch is for pytorch 1.11 (released March 2022). Are there updated patches or instructions available? or are none needed in modern versions?

Are there any differences between a compiled version and the prepacked wheel in terms of optimization?

Hi,

For recent PyTorch (2.x), it can be built without a custom patch.
Orin GPU architecture is already added in the building config so you don’t do that manually.

Thanks.

1 Like

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.