Unable to Access Intel-Optimized PyTorch Packages (403 Errors and Timeout Issues)

[martingagne]

Describe the issue

Description:
I’m attempting to install Intel-optimized PyTorch packages for testing GPU/NPU workloads, but I’ve encountered persistent issues. I am following the steps at https://intel.github.io/intel-extension-for-pytorch/index.html#introduction. Here are the details:

Command Executed:

python -m pip install torch==2.3.1.post0+cxx11.abi torchvision==0.18.1.post0+cxx11.abi torchaudio==2.3.1.post0+cxx11.abi intel-extension-for-pytorch==2.3.110.post0+xpu --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/lnl/us/

Issues:

1. US Blob:
   • The US blob returns a 403 Forbidden error:

     ERROR: HTTP error 403 while getting https://intel-extension-for-pytorch.s3.amazonaws.com/ipex_stable/lnl/torch-2.3.1.post0%2Bcxx11.abi-cp310-cp310-win_amd64.whl
     

2. CN Blob:
   • While the CN blob downloads packages, it frequently times out during the process:

     ReadTimeoutError: HTTPSConnectionPool(host='intel-optimized-pytorch.s3.cn-north-1.amazonaws.com.cn', port=443): Read timed out.
     

Steps to Reproduce:

1. Run the above command for installing the PyTorch packages using the provided --extra-index-url.
2. Attempt downloads from both US and CN blob URLs.

Expected Behavior:
Packages should download and install seamlessly without 403 errors or timeouts.

Actual Behavior:

• US blob is inaccessible due to 403 Forbidden.
• CN blob downloads timeout consistently, making installations unreliable.

Impact:
I’m unable to proceed with testing Intel GPU/NPU optimizations, which is a key part of my development workflow.

Request:
Please ensure that the US blob is accessible and address the timeout issues with the CN blob. Stable package availability is critical for developers working with Intel’s tools.

[feng-intel]

What about this command:

python -m pip install torch==2.3.1+cxx11.abi torchvision==0.18.1+cxx11.abi torchaudio==2.3.1+cxx11.abi intel-extension-for-pytorch==2.3.110+xpu oneccl_bind_pt==2.3.100+xpu --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/cn/

From https://pytorch-extension.intel.com/installation?platform=gpu&version=v2.3.110%2Bxpu&os=linux%2Fwsl2&package=pip

[Disty0]

Same with the XPU URL from US blobs on my end. US throws access denied errors.

CN is generally slow (5 MB/s~) compared to US and sometimes can be very slow (50 KB/s) even with a 1000 Mbps connection.

CN:

And this is what i get from the official PyTorch servers:

[notsyncing]

Same here, any updates? CN repo is OK, US repo reports 403.

[feng-intel]

Something went wrong to the US index link. We are working on fixing it.
Pls use the cn index URL at this moment.

[huichuno]

I encountered the same issue. China url keep getting "network timeout" error, whereas, US url is getting "403 Forbidden" error.