Update Dockerfile

Dockerfile

@@ -70,8 +70,12 @@ RUN pip list
 # -----------------------------
 # Start vLLM
 # -----------------------------
-RUN python3 -u -m vllm.entrypoints.openai.api_server --model "unsloth/Llama-3.2-1B-bnb-4bit" --host 0.0.0.0 --port 7860 --tensor-parallel-size 1 --gpu-memory-utilization 0.0 2>/dev/null
+COPY start_server.sh /workspace
+WORKDIR /workspace
+# ENTRYPOINT ["./start_server.sh"]
+CMD ["./start_server.sh"]
 
+# RUN python3 -u -m vllm.entrypoints.openai.api_server --model "unsloth/Llama-3.2-1B-bnb-4bit" --host 0.0.0.0 --port 7860 --tensor-parallel-size 1 --gpu-memory-utilization 0.0 2>/dev/null
 # CMD ["python3", "-m", "vllm.entrypoints.openai.api_server", \
 #      "--model", "unsloth/Llama-3.2-1B-bnb-4bit", \
 #      "--host", "0.0.0.0", \