Safetensors joins PyTorch Foundation under Linux Foundation governance
hugging face transferred safetensors to the pytorch foundation, moving governance to the linux foundation. maintainers luc and daniel stay on the technical steering committee. the roadmap includes torch core integration, device-aware loading for cuda/roc, and native support for tensor parallelism and fp8 quantization.
Hugging Face moved the Safetensors project to the PyTorch Foundation to establish vendor-neutral governance under the Linux Foundation. Core maintainers remain in charge, but the project now accepts community contributions to formal governance documents. The roadmap details integration into PyTorch core, device-aware loading to skip CPU staging, and APIs for Tensor and Pipeline parallel loading. Future updates will formally support FP8 and block-quantized formats like GPTQ and AWQ.