Tag: PyTorch
Why NVIDIA’s Moat Is Software, and What Would Break It
Rivals have shipped comparable silicon for years without shifting the market. The moat is the software above the chips and the rack around them.
Transformers v5: PyTorch-First Library Update
Hugging Face releases Transformers v5 with PyTorch-only backend, quantization as first-class feature, and enhanced interoperability across the AI ecosystem.
OpenEnv: Standard Agent Training Environments
OpenEnv launches a standard for agentic execution environments, enabling AI agents to operate securely across cloud and local platforms with unified protocols.
PyTorch Monarch: Distributed Programming
PyTorch announces Project Monarch, a new compiler backend that significantly improves training efficiency for Mixture-of-Experts (MoE) models on NVIDIA GPUs.