huggingface / candle

Minimalist ML framework for Rust
Apache License 2.0
15.91k stars 963 forks source link

Import the ggml_cuda_dp4a function. #2628

Closed LaurentMazare closed 1 week ago

LaurentMazare commented 1 week ago

Use the same hack to provide __dp4a to older architectures as in llama.cpp, see common.cuh.