Metal 4 GEMM and GPU tensor runtime for Apple silicon. MPP TensorOps, on-device Qwen3.5 training, EmbeddingGemma 2, BERT, and PyTorch bindings.
rust machine-learning deep-learning gpu zero-copy transformers pytorch matrix-multiplication neural-networks mpp gpu-computing bert gemm apple-silicon bf16 llm embeddinggemma qwen35 metal-4 tensorops
-
Updated
Oct 14, 2026 - Rust