Spaces:
Running
on
A100
Running
on
A100
cuda
Browse files- Dockerfile +10 -2
Dockerfile
CHANGED
|
@@ -19,7 +19,15 @@ RUN apt-get install -y --no-install-recommends \
|
|
| 19 |
libglib2.0-0 \
|
| 20 |
libgl1-mesa-glx \
|
| 21 |
libpq-dev \
|
| 22 |
-
make
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 23 |
# insall latest vllm
|
| 24 |
|
| 25 |
RUN pip install uv
|
|
@@ -34,6 +42,6 @@ RUN uv pip install vllm==0.7.2
|
|
| 34 |
RUN uv pip install --upgrade git+https://github.com/huggingface/transformers.git@336dc69d63d56f232a183a3e7f52790429b871ef
|
| 35 |
|
| 36 |
RUN uv pip install ray[adag]
|
| 37 |
-
RUN uv pip install cupy
|
| 38 |
|
| 39 |
ENTRYPOINT ["/bin/bash", "-c", ". .venv/bin/activate && export VLLM_USE_V1=1 && python -m vllm.entrypoints.openai.api_server --model convergence-ai/proxy-lite --trust-remote-code --tokenizer-pool-size 20 --limit-mm-per-prompt image=1 --enable-auto-tool-choice --tool-call-parser hermes --port 7860"]
|
|
|
|
| 19 |
libglib2.0-0 \
|
| 20 |
libgl1-mesa-glx \
|
| 21 |
libpq-dev \
|
| 22 |
+
make \
|
| 23 |
+
cuda-toolkit-12-* \
|
| 24 |
+
nvidia-cuda-toolkit
|
| 25 |
+
|
| 26 |
+
# Set CUDA environment variables
|
| 27 |
+
ENV CUDA_HOME=/usr/local/cuda
|
| 28 |
+
ENV PATH=${CUDA_HOME}/bin:${PATH}
|
| 29 |
+
ENV LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${LD_LIBRARY_PATH}
|
| 30 |
+
|
| 31 |
# insall latest vllm
|
| 32 |
|
| 33 |
RUN pip install uv
|
|
|
|
| 42 |
RUN uv pip install --upgrade git+https://github.com/huggingface/transformers.git@336dc69d63d56f232a183a3e7f52790429b871ef
|
| 43 |
|
| 44 |
RUN uv pip install ray[adag]
|
| 45 |
+
RUN uv pip install cupy-cuda12x
|
| 46 |
|
| 47 |
ENTRYPOINT ["/bin/bash", "-c", ". .venv/bin/activate && export VLLM_USE_V1=1 && python -m vllm.entrypoints.openai.api_server --model convergence-ai/proxy-lite --trust-remote-code --tokenizer-pool-size 20 --limit-mm-per-prompt image=1 --enable-auto-tool-choice --tool-call-parser hermes --port 7860"]
|