newsInfoQ·26 Oct 20263.9
Google and Meta open-sourced TorchTPU, a native PyTorch backend for TPUs, and Meta plans to move larger LLM workloads onto TPUs; AWS released TorchNeuron for Trainium the same week.
newsInfoQ·26 Oct 20263.7
Google open-sourced ML Drift under Apache 2.0, a cross-platform GPU compute engine for on-device inference over OpenGL ES, OpenCL, Metal and WebGPU.
blogHugging Face Blog·20 Oct 20263.5
TorchTPU is a native PyTorch backend for TPUs, so Hugging Face models run on TPU without torch_xla or JAX.
paperIEEE Xplore·1 Nov 20263.5
Hydra-MCS splits maximum common subgraph search across CPU and GPU for speedups.
newsHacker News·24 Oct 20263.3
Linux kernel module unlocks full FP and BF16 instruction issue rates on Nvidia RTX 3000 cards.