23 lines
735 B
Docker
23 lines
735 B
Docker
FROM rocm/pytorch:latest
|
|
|
|
ENV DEBIAN_FRONTEND=noninteractive
|
|
# Specifica l'architettura target per la compilazione dei kernel C++/HIP
|
|
ENV PYTORCH_ROCM_ARCH="gfx1100"
|
|
ENV HSA_OVERRIDE_GFX_VERSION="11.0.0"
|
|
ENV HIP_VISIBLE_DEVICES="0"
|
|
|
|
RUN apt-get update -y && apt-get install -y git build-essential python3-dev && rm -rf /var/lib/apt/lists/*
|
|
|
|
WORKDIR /workspace
|
|
|
|
# Clona vLLM e usa un tag stabile recente che supporta GGUF (0.6.0+)
|
|
RUN git clone https://github.com/vllm-project/vllm.git . && git checkout v0.6.3
|
|
|
|
RUN pip install -U pip && pip install -r requirements-rocm.txt
|
|
|
|
# Compilazione (richiederà tempo)
|
|
RUN python3 setup.py install
|
|
|
|
EXPOSE 8000
|
|
ENTRYPOINT ["python3", "-m", "vllm.entrypoints.openai.api_server"]
|