Update Dockerfile

Dockerfile

@@ -1,21 +1,25 @@
-# Start with the official openEuler vLLM CPU base image
-FROM openeuler/vllm-cpu:latest
+# Use the openeuler/vllm-cpu base (includes Python, pip, and vLLM pre-installed)
+FROM openeuler/vllm-cpu:0.8.5-oe2403lts
 
-# Set the working directory
-WORKDIR /app
+# Ensure Python uses the CPU device (vLLM expects VLLM_TARGET_DEVICE for inference)
+ENV VLLM_TARGET_DEVICE=cpu
+ENV PYTHONUNBUFFERED=1
 
-# (Optional) Install any additional dependencies your model or application might need
-# Example: 
-# RUN yum install -y some-package
+# Set working directory
+WORKDIR /workspace
 
-# Copy your model files into the container (assuming they are in a 'model' directory in your context)
-# You might need to adjust this depending on how you're providing the model
-COPY ./model /app/model
+# Upgrade pip and install CPU-only PyTorch, Transformers, Accelerate, Unsloth, etc.
+# Use the official PyTorch CPU wheel index for performance on CPU:contentReference[oaicite:5]{index=5}.
+RUN pip3 install --upgrade pip \
+ && pip3 install torch --index-url https://download.pytorch.org/whl/cpu \
+ && pip3 install transformers accelerate unsloth
 
-# Set the environment variable for Hugging Face token (not strictly needed as it's not a gated model, but good practice)
-# You can uncomment this and set it if you prefer.
-# ENV HUGGING_FACE_HUB_TOKEN="<YOUR_HUGGINGFACE_TOKEN>" 
+# (Optional) Install unsloth_zoo or other utilities if needed:
+# RUN pip3 install unsloth-zoo
 
-# Command to run the vLLM OpenAI-compatible server with your model
-# Replace "your-model-name" with the actual model ID from Hugging Face
-CMD ["python", "-m", "vllm.entrypoints.openai.api_server", "--model", "unsloth/Llama-3.2-3B-bnb-4bit", "--host", "0.0.0.0", "--port", "7860"] 
+# Copy an example inference script into the container
+# (This script should load the model and do a sample generation.)
+COPY inference.py /workspace/inference.py
+
+# Default command: run the inference script to verify setup
+CMD ["python3", "/workspace/inference.py"]