Add Dockerfile for containerized deployment

[Jah-yee]

Summary

Adds a Dockerfile to enable running BitNet in containerized environments like Kubernetes.

Changes

• Added Dockerfile with:
  • Python 3.11-slim base image
  • Build dependencies (cmake, build-essential, libomp-dev)
  • Python dependencies from requirements.txt
  • Support for model volume mounting
  • Default command shows help

Motivation

Resolves issue #433 - enables deployment on powerful CPUs via infrastructure like Kubernetes.

Usage

# Build the image
docker build -t bitnet .

# Run inference
docker run --rm -v /path/to/models:/workspace/models bitnet \
  python run_inference.py -m models/BitNet-b1.58-2B-4T/ggml-model-i2_s.gguf -p "Your prompt" -cnv

Testing

The Dockerfile follows best practices:

• Multi-stage not needed since this is a Python + CMake build
• Uses slim image to minimize size
• Includes all necessary build tools for native extension compilation