@abhi-glitchhg thanks for the PR, seems like torch XLA is still on python 3.7

@abhi-glitchhg thanks for the PR, seems like torch XLA is still on python 3.7

yeah; what should we do; Should i make an issue for this on XLA repo?

It may be possible that we are not correctly installing it for python 3.8. In colab they seem to pick python 3.8 : https://colab.research.google.com/github/pytorch/xla/blob/master/contrib/colab/pytorch-xla-profiling-colab.ipynb

https://storage.googleapis.com/tpu-pytorch/wheels/colab/torch_xla-1.13-cp38-cp38-linux_x86_64.whl

Hmm,

I went through the error

WARNING: Skipping torch as it is not installed.
WARNING: Skipping torchvision as it is not installed.
CommandException: No URLs matched: gs://tpu-pytorch/wheels/colab/torch-nightly-cp38-cp38m-linux_x86_64.whl
CommandException: No URLs matched: gs://tpu-pytorch/wheels/colab/torch_xla-nightly-cp38-cp38m-linux_x86_64.whl
CommandException: No URLs matched: gs://tpu-pytorch/wheels/colab/torchvision-nightly-cp38-cp38m-linux_x86_64.whl
WARNING: Requirement 'torch-nightly-cp38-cp38m-linux_x86_64.whl' looks like a filename, but the file does not exist
ERROR: torch-nightly-cp38-cp38m-linux_x86_64.whl is not a supported wheel on this platform.

We are running this set up script to download the wheels. From the error we can see that 3.8 python version is being used by the script but URL is not matching(found?).

@vfdev-5, could it be possible that nightly versions for 3.8 are not up there? or can we get the list of all the available xla wheels for 3.8 version?

Thanks

@abhi-glitchhg I do not know but I just tried in colab without any accelerators to run the following:

pip install cloud-tpu-client==0.10 --pre torch https://storage.googleapis.com/tpu-pytorch/wheels/colab/torch_xla-nightly-cp38-cp38-linux_x86_64.whl

and it seems to work...

You can try either on your fork to enable GH actions and figure out how to install torch xla or either you can do things in docker container...