llama.cpp 发布: b11182

来源摘要

llama : add `llama_prec_policy` + model-driven W4A4 path (#24364) * Rebase and update based on #26675 Signed-off-by: ynankani * CI failure fix(launh_bounds overload on HIP) and cleanup Signed-off-by: ynankani * Address review comments Signed-off-by: ynankani * Use ggml tensor instead of name in act policy map Signed-off-by: ynankani * Address review comments and cleanup Signed-off-by: ynankani * Address review comments Signed-off-by: ynankani * Rename changes Signed-off-by: ynankani * Update ggml/src/ggml-cuda/mmq.cu Co-authored-by: Georgi Gerganov * MXFP4 dispatch changes for higher src prec Signed-off-by: ynankani * Refactor and address review comments Signed-off-by: ynankani * Updates based on review comments Signed-off-by: ynankani * Apply batched suggestions from code review Co-authored-by: Johannes Gäßler * Address review comments Signed-off-by: ynankani * Apply patch from review Signed-off-by: ynankani --------- Signed-off-by: ynankani Co-authored-by: Georgi Gerganov Co-authored-by: Johannes Gäßler **Website:** - **Attestations:** - **macOS/iOS:** - [macOS Apple Silicon (arm64)](https://github.com/ggml-org/llama.cpp/releases/download/b11182/llama-b11182-bin-macos-arm64.tar.gz) - macOS Apple Silicon (arm64, KleidiAI enabled) [DISABLED](https://github.com/ggml-org/llama.cpp/pull/23780) - [macOS Intel (x64)](https://github.com/ggml-org/llama.cpp/releases/download/b11182/llama-b11182-bin-macos-x64.tar.gz) - [iOS XCFramework](https://github.com/ggml-org/llama.cpp/releases/download/b11182/llama-b11182-xcframework.zip) **Linux:** - [Ubuntu x64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11182/llama-b11182-bin-ubuntu-x64.tar.gz) - [Ubuntu arm64 (CPU)](https://github.com/ggml-org/llama.cpp/releases/download/b11182/llama-b11182-bin-ubuntu-arm64.tar.gz)