update vllm container

.devcontainer/Dockerfile-vllm

@@ -1,6 +1,8 @@
 # syntax = edrevo/dockerfile-plus
 # ^^^ this line enables the INCLUDE+ directive
 
+FROM nvcr.io/nvidia/pytorch:23.09-py3
+
 INCLUDE+ cuda-settings.dockerfile
 INCLUDE+ common.dockerfile
 
@@ -17,3 +19,13 @@ RUN pip install git+https://github.com/NVIDIA/TransformerEngine.git@stable
 # RUN pip install typing_extensions==4.5.0
 
 RUN pip install -U flash-attn
+
+
+# RUN pip install torch==2.1.0 nvidia-cuda-runtime
+# the .so file seems to be missing
+RUN ln -s /usr/local/lib/python3.10/dist-packages/nvidia/cuda_runtime/lib/libcudart.so{.12,}
+
+# perf tool
+RUN apt-get install -y linux-tools-`uname -r`
+
+RUN source /usr/local/nvm/nvm.sh && npm install -g yarn