Update Dockerfile

Dockerfile

@@ -1,7 +1,106 @@
-FROM ghcr.io/ggml-org/llama.cpp:full
+# Dockerfile for Qwen2.5-Coder-7B-Instruct-GGUF on Hugging Face Spaces (Free Tier)
+# Optimized for: 2 vCPU, 16GB RAM, CPU-only
+# Uses llama-cpp-python for OpenAI-compatible API
 
-RUN apt update && apt install wget -y
+FROM python:3.11-slim-bookworm
 
-RUN wget "https://huggingface.co/Qwen/Qwen2.5-Coder-7B-Instruct-GGUF" -O /Qwen2.5-Coder-7B.gguf
+# Install system dependencies
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    build-essential \
+    cmake \
+    curl \
+    libopenblas-dev \
+    && rm -rf /var/lib/apt/lists/*
 
-CMD ["--server", "-m", "/Qwen2.5-Coder-7B.gguf", "--port", "7860", "--host", "0.0.0.0", "-n", "4096"]
+# Set environment variables for build optimization
+ENV CMAKE_ARGS="-DLLAMA_BLAS=ON -DLLAMA_BLAS_VENDOR=OpenBLAS"
+ENV FORCE_CMAKE=1
+
+# Install llama-cpp-python with OpenBLAS acceleration (CPU-optimized)
+RUN pip install --no-cache-dir llama-cpp-python==0.3.2 --extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
+
+# Install huggingface-hub for model download
+RUN pip install --no-cache-dir huggingface-hub gradio
+
+# Set working directory
+WORKDIR /app
+
+# Download Q4_K_M quantized model (smallest viable for coding tasks)
+# Q4_K_M: ~4.7GB, good balance for 16GB RAM limit
+RUN python3 -c "from huggingface_hub import hf_hub_download; \
+    hf_hub_download( \
+        repo_id='Qwen/Qwen2.5-Coder-7B-Instruct-GGUF', \
+        filename='qwen2.5-coder-7b-instruct-q4_k_m.gguf', \
+        local_dir='/app/models', \
+        local_dir_use_symlinks=False \
+    )"
+
+# Create Gradio app.py for Hugging Face Spaces interface
+RUN cat > /app/app.py << 'EOF'
+import gradio as gr
+from llama_cpp import Llama
+import os
+
+# Initialize model with aggressive CPU optimization
+llm = Llama(
+    model_path="/app/models/qwen2.5-coder-7b-instruct-q4_k_m.gguf",
+    n_ctx=4096,           # Reduced context for memory efficiency
+    n_batch=512,          # Smaller batch size for 2 vCPU
+    n_threads=2,          # Match Hugging Face free tier vCPUs
+    n_threads_batch=2,
+    verbose=False,
+    use_mmap=True,        # Memory mapping for efficient loading
+    use_mlock=False       # Disable memory locking for container compatibility
+)
+
+SYSTEM_PROMPT = "You are Qwen2.5-Coder, a helpful AI coding assistant. Provide concise, accurate code solutions."
+
+def generate_code(message, history):
+    # Format with ChatML template
+    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
+    
+    # Add history
+    for human, assistant in history:
+        messages.append({"role": "user", "content": human})
+        messages.append({"role": "assistant", "content": assistant})
+    
+    messages.append({"role": "user", "content": message})
+    
+    # Generate with conservative parameters for free tier
+    output = llm.create_chat_completion(
+        messages=messages,
+        max_tokens=1024,      # Limit response length
+        temperature=0.7,
+        top_p=0.9,
+        repeat_penalty=1.1,
+        stop=["<|im_end|>", "<|im_start|>"]
+    )
+    
+    return output["choices"][0]["message"]["content"]
+
+# Create Gradio interface
+demo = gr.ChatInterface(
+    fn=generate_code,
+    title="Qwen2.5-Coder-7B",
+    description="Optimized for Hugging Face Spaces Free Tier (2 vCPU, 16GB RAM)",
+    examples=[
+        "Write a Python function to reverse a string",
+        "Explain how quicksort works with code",
+        "Create a simple REST API with Flask"
+    ],
+    cache_examples=False  # Disable to save memory
+)
+
+if __name__ == "__main__":
+    demo.launch(server_name="0.0.0.0", server_port=7860)
+EOF
+
+# Expose Hugging Face Spaces port
+EXPOSE 7860
+
+# Health check
+HEALTHCHECK --interval=30s --timeout=10s --start-period=120s --retries=3 \
+    CMD curl -f http://localhost:7860/ || exit 1
+
+# Run the Gradio app
+CMD ["python3", "/app/app.py"]