Tag: CUDA

Trained by reinforcement learning, ByteDance kernels outrun torch.compile

ByteDance's CUDA Agent trained a model to write GPU kernels that beat torch.compile on nearly every task.

NVIDIA Jetson Gains Agentic AI with JetPack 7.2 and NemoClaw Framework

NVIDIA's JetPack 7.2 and NemoClaw framework enable agentic AI to run directly on Jetson edge hardware for robotics,…