torch-ccl
torch-ccl copied to clipboard
oneCCL Bindings for Pytorch*
The error occurs when I am trying to install by `` python -m pip install oneccl_bind_pt -f https://developer.intel.com/ipex-whl-stable-xpu `` with Pytorch 1.13.0+cu117, Python 3.11.3, Ubuntu 22.04.1 LTS. Would you like...
I am trying to implement a concurrent asynchronous send and receive between multiple processes. This results in deadlock. Minimum code to reproduce this is as follows: ``` import torch.nn.parallel import...
How can I build this with the latest upstream pytorch ? Latest, pytorch has XPU support by default. Additionally, XPU triton backend is also moving towards working with the upstream...
fix failing due to cmake not found - Make it clear to the user that they have to install cmake.
Fixed a typo.
changed oneCCL patch to remove unnecessary files during installation.
We've had success using torch-ccl with resnet and other AI workloads to test with libfabric over psm3 but when we try to use libmlx-fi.so, torch-ccl does not seem to see...