Add Docker setup and example notebook for Ray Data LLM

Author: 0-mostafa-rezaee-0

docker/.dockerignore

@@ -0,0 +1,61 @@
+# Git
+.git
+.gitignore
+
+# Python
+__pycache__/
+*.py[cod]
+*$py.class
+*.so
+.Python
+env/
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+*.egg-info/
+.installed.cfg
+*.egg
+
+# Virtual Environment
+venv/
+.env
+.venv
+env/
+ENV/
+
+# Docker
+docker/.dockerignore
+docker/Dockerfile
+docker/docker-compose.yml
+
+# IDE
+.idea/
+.vscode/
+*.swp
+*.swo
+
+# Ray
+/tmp/
+ray_bootstrap_key.pem
+ray_bootstrap_config.yaml
+
+# Data and notebooks
+data/large_files/
+*.csv
+*.parquet
+*.arrow
+*.pkl
+*.h5
+.ipynb_checkpoints/
+
+# Logs
+logs/
+*.log 

docker/Dockerfile

@@ -0,0 +1,37 @@
+FROM rayproject/ray:2.9.3-py310
+
+# Set working directory
+WORKDIR /app
+
+# Install additional Python dependencies
+RUN pip install --no-cache-dir \
+    "ray[data]>=2.9.0" \
+    vllm \
+    pandas \
+    numpy \
+    torch \
+    transformers \
+    openai \
+    tqdm \
+    jupyterlab \
+    jupyter \
+    matplotlib \
+    scikit-learn \
+    huggingface_hub \
+    datasets
+
+# Copy requirements file if needed
+COPY requirements.txt /app/
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy project files to the container
+COPY . /app/
+
+# Set environment variables
+ENV RAY_ADDRESS=auto
+
+# Expose Ray ports
+EXPOSE 6379 8265 10001 8000 8888
+
+# Default command
+CMD ["ray", "start", "--head", "--dashboard-host=0.0.0.0", "--port=6379", "--dashboard-port=8265", "--block"] 

docker/README.md

@@ -0,0 +1,59 @@
+# Docker Setup for Ray Data LLM
+
+This directory contains Docker configuration files for running the Batch LLM Inference with Ray Data LLM project.
+
+## Requirements
+
+- Docker
+- Docker Compose
+- NVIDIA Container Toolkit (for GPU support)
+
+## Quick Start
+
+1. Build and start the containers:
+
+```bash
+docker-compose up -d
+```
+
+2. Access the Jupyter Lab interface:
+   - Open your browser and navigate to `http://localhost:8888`
+   - No password is required (configured for development)
+
+3. Access the Ray Dashboard:
+   - Open your browser and navigate to `http://localhost:8265`
+
+## Configuration
+
+The Docker setup includes:
+
+- **ray-head**: The main Ray head node that coordinates all computation
+- **jupyter**: A Jupyter Lab environment connected to the Ray cluster
+
+## GPU Support
+
+The setup is configured to use GPUs if available. Make sure you have the NVIDIA Container Toolkit installed:
+
+```bash
+# Check if NVIDIA Container Toolkit is installed
+sudo docker run --gpus all nvidia/cuda:11.8.0-base-ubuntu22.04 nvidia-smi
+```
+
+## Data Persistence
+
+Data is persisted through a Docker volume:
+
+- **ray-data**: Stores Ray temporary data
+
+## Customization
+
+You can customize the setup by:
+
+1. Modifying the `Dockerfile` to add additional dependencies
+2. Adjusting the `docker-compose.yml` file to change port mappings or resource allocations
+3. Updating the `requirements.txt` file to include specific package versions
+
+## Troubleshooting
+
+- If you encounter issues with GPU access, make sure your NVIDIA drivers and NVIDIA Container Toolkit are properly installed.
+- If Ray cannot access enough resources, try adjusting the memory and CPU allocations in the `docker-compose.yml` file. 

docker/docker-compose.yml

@@ -0,0 +1,58 @@
+version: '3.8'
+
+services:
+  ray-head:
+    build:
+      context: ..
+      dockerfile: docker/Dockerfile
+    container_name: ray-head
+    ports:
+      - "6379:6379"   # Ray client port
+      - "8265:8265"   # Ray dashboard
+      - "10001:10001" # Ray head node
+      - "8000:8000"   # For serving OpenAI-compatible API endpoints
+    volumes:
+      - ../:/app
+      - ray-data:/tmp/ray
+    environment:
+      - RAY_ADDRESS=auto
+    command: >
+      ray start --head
+        --dashboard-host=0.0.0.0
+        --port=6379
+        --dashboard-port=8265
+        --block
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
+
+  jupyter:
+    build:
+      context: ..
+      dockerfile: docker/Dockerfile
+    container_name: ray-jupyter
+    ports:
+      - "8888:8888"  # Jupyter notebook
+    volumes:
+      - ../:/app
+      - ray-data:/tmp/ray
+    environment:
+      - RAY_ADDRESS=ray://ray-head:10001
+    depends_on:
+      - ray-head
+    command: jupyter lab --ip=0.0.0.0 --port=8888 --no-browser --allow-root --NotebookApp.token='' --NotebookApp.password=''
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: 1
+              capabilities: [gpu]
+
+volumes:
+  ray-data:
+    driver: local 

examples/batch_inference_example.ipynb

@@ -0,0 +1 @@
+ 

requirements.txt

@@ -0,0 +1,14 @@
+ray[data]>=2.9.0
+vllm>=0.2.0
+pandas>=2.0.0
+numpy>=1.23.0
+torch>=2.0.0
+transformers>=4.30.0
+openai>=1.0.0
+tqdm>=4.65.0
+jupyterlab>=4.0.0
+jupyter>=1.0.0
+matplotlib>=3.7.0
+scikit-learn>=1.2.0
+huggingface_hub>=0.16.0
+datasets>=2.12.0