Claudia Comito
Claudia Comito
I can reproduce the `torch.broadcast_shapes` error on CPU (osx-arm64). More debugging needed.
Let's get a couple lines of instructions from Sven, so next time we can update the recipe ourselves right after a release, what do you think @mrfh92 ?
Ongoing here including CUDA and ROCm build https://github.com/ClaudiaComito/spack/tree/py-heat_1.5.0 Testing in progress.
Created a PR with updated spack recipe: https://github.com/spack/spack/pull/47710 Added options ("variants" in spack speak) to build with CUDA or ROCm. I have problems with this recipe on JUWELS.
I've updated the recipe https://github.com/spack/spack/pull/47710 . Would be great if you tried the installation out on your various clusters @JuanPedroGHM @mrfh92 . You can clone my fork https://github.com/ClaudiaComito/spack/tree/py-heat_1.5.0 branch `py-heat_1.5.0`....
Will review this feature request when #749 is finished. --- Reviewed within #1109
This is still ongoing. Needs #749 which needs #914
See above. Actively working on #914 now, will finish #749 afterwards and get back to ASSET.
I've modified the conda-forge recipe so it pulls `pytorch-gpu` if a CUDA installation is available. Needs testing.
closed by #1846