ggml-org/llama.cpp

View on GitHub·2 presentations

Subscribe

Maintain this repository? Sign in with GitHub

LatestPopular

3:11

Offloaded MoE experts decode up to 2.2× faster.

@ggml-org/llama-cpp · 2 days ago

3:43

WebGPU now runs Q2_0 matrix multiplication.

@ggml-org/llama-cpp · 2 days ago