adding missing ld_library_path for cuda stubs in Dockerfile

This commit is contained in:
Morgan Funtowicz 2024-07-22 15:16:39 +00:00
parent 3fd2bb70c3
commit d0a34a95f2
2 changed files with 10 additions and 1 deletions

View File

@ -53,7 +53,7 @@ RUN mkdir /usr/local/tgi && mkdir /usr/local/tgi/include && mkdir /usr/local/tgi
FROM nvcr.io/nvidia/pytorch:24.05-py3
WORKDIR /usr/local/tgi/bin
ENV LD_LIBRARY_PATH="/usr/local/tgi/lib:$LD_LIBRARY_PATH"
ENV LD_LIBRARY_PATH="/usr/local/tgi/lib:/usr/local/cuda/lib64/stubs:$LD_LIBRARY_PATH"
RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/libcuda.so.1 && \
ln -s /usr/local/cuda/lib64/stubs/libnvidia-ml.so /usr/local/cuda/lib64/stubs/libnvidia-ml.so.1
@ -61,3 +61,6 @@ RUN ln -s /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/cuda/lib64/stubs/lib
COPY --from=cuda-builder /usr/local/tensorrt /usr/local/tensorrt
COPY --from=cuda-builder /usr/local/tgi /usr/local/tgi
COPY --from=cuda-builder /usr/src/text-generation-inference/target/release/text-generation-backends-trtllm /usr/local/tgi/bin/text-generation-launcher
ENTRYPOINT ["text-generation-launcher"]
CMD ["--executor-worker", "/usr/local/tgi/bin/executorWorker"]

View File

@ -0,0 +1,6 @@
```mermaid
sequenceDiagram
TensorRtLlmBackend -->> TensorRtLlmBackendImpl: New thread which instantiates actual backend impl
TensorRtLlmBackendImpl -->> TensorRtLlmBackendImpl.Receiver: Awaits incoming request sent throught the queue
```