#torch-compile

The latest AI and developer news about #torch-compile, with the original source and a short summary.

vLLM Introduces Hardware-Agnostic Layer Architecture to Support Diverse Accelerators

The new architecture maintains within 3.4% of native token throughput on NVIDIA H100 GPUs while enabling support for out-of-tree accelerators and older hardware.

·3·2026.09.25 19:00

TorchTPU: Running PyTorch Natively on TPUs at Google Scale

Google has unveiled TorchTPU, which brings PyTorch natively onto TPUs.

·1·2026.04.10 15:00

How PyTorch Handles Dynamic Tensor Shapes

torch.compile recompiles when shapes change and restricts execution scope using guards.

·1·2025.11.04 09:00