For the past 12 years, cuDNN has been completely closed sourced (besides the .h files), until this week! OVER 20 MoE kernels & NSA sparse attention kernels from cuDNN has been open sourced! Great work to @manicely6005 & the rest of the team on seeing that parts of NVIDIA are moving towards open kernels! open source kernels drive innovation! (1/3) 🧵
@PyTorch@AMD@RedHat We spotlighted IBM's adoption of vLLM and torch.compile to integrate emerging accelerators like the IBM Spyre AI accelerator.
Researchers are working on a Spyre backend compiler and vLLM plugin for paged attention to boost memory efficiency and scalability for LLM inference.
@PyTorch@AMD@RedHat Full IBM recap of #PyTorchCon and how we’re expanding AI model training and inference for the open-source community here ⤵️ https://t.co/Q69c1ENFcQ
Today, @IBM announced the general commercial availability of the Spyre Accelerator for IBM z17 and LinuxONE 5 systems on October 28, and for Power11 servers in early December: https://t.co/s4FklnqdYM