From 83566ee414d4b8a2d27c77732a774020839cc50f Mon Sep 17 00:00:00 2001 From: zefie Date: Sun, 1 Nov 2026 18:58:34 -0500 Subject: [PATCH] compile llama for the docker --- Dockerfile | 17 ++++++++++++++++- 1 file changed, 16 insertions(+), 1 deletion(-) diff --git a/Dockerfile b/Dockerfile index 7e8ac45..b58db7e 100644 --- a/Dockerfile +++ b/Dockerfile @@ -2,7 +2,22 @@ FROM runpod/pytorch:1.5.0-cu1320-torch2130-ubuntu2404 WORKDIR /root RUN mkdir llama.cpp -COPY llama.cpp ./llama.cpp/ +COPY --exclude=build/ llama.cpp ./llama.cpp/ +WORKDIR /root/llama.cpp +ENV LD_LIBRARY_PATH=/usr/local/cuda-13.2/compat:${LD_LIBRARY_PATH} +ENV LIBRARY_PATH=/usr/local/cuda-13.2/compat:${LIBRARY_PATH} + +RUN cmake -B build -S . \ + -DCMAKE_BUILD_TYPE=Release \ + -DGGML_CUDA=ON \ + -DCMAKE_CUDA_ARCHITECTURES=120 \ + -DLLAMA_BUILD_TESTS=OFF \ + -DLLAMA_BUILD_EXAMPLES=ON \ + && cmake --build build -j$(nproc) \ + && cmake --build build --target install \ + && rm -rf build + +WORKDIR /root RUN git clone https://github.com/silveroxides/convert_to_quant && cd convert_to_quant && pip install -r requirements.txt && pip install . RUN mkdir nvfp4_gguf_quant