PyTorch 11 days ago
#PyTorch 2.12 adds torch.accelerator.Graph, up to 100x faster batched eigendecomposition on CUDA, microscaling quantization support in torch.export.save, and expanded CUDA, ROCm, XPU, MPS, and Arm support.
Live Q&A on May 20, 10 AM PT: https://streamyard.com/watch/MC2PrCkw6UaZ
#OpenSourceAI