Intel announced that it has joined the PyTorch Foundation as a Premier member .
Intel has been actively involved in PyTorch development for five years, helping to optimize PyTorch for CPU inference performance, as well as various other optimizations to boost AI performance on Intel processors.
For example, in PyTorch 2.0, Intel has optimized TorchInductor CPU FP32 inference, improved inference/training performance of GNN in PyG, optimized Int8 inference for x86 CPU platform, and accelerated inference on CPU with oneDNN Graph API.
Intel currently has four PyTorch maintainers (three active, one retired) who maintain the CPU performance module and compiler front end. They are proactive in triaging issues and reviewing PRs from the community, having merged hundreds of PRs for PyTorch upstream.
Other jobs for maintainers include:
- Categorize issues in the PyTorch GitHub repository
- Enhanced PyTorch documentation
- Meetups and workshops to share Intel's latest applications integrating PyTorch
- Publish technical blogs, articles and white papers
- Make technical videos about PyTorch