A lightweight Transformer training & inference framework

English • 中文 • Issue Tracker • Discussions • HuggingFace

📖 Table of Contents

• Features
• Quick Start
• Documentation
• Contributing
• Community
• License

---

English

Features

• 🚀 High Performance: Optimized for both training and inference with efficient parallelization.
• 🔧 Flexible: Support for seq/sft/dpo/grpo training, customizable model architectures.
• 💡 Easy to Use: Simple API with comprehensive examples and demos.
• 📦 Lightweight: Minimal dependencies, easy to deploy.
• 🔬 Research‑Friendly: Modular design, easy to experiment with new ideas.
• 🤗 HuggingFace-Style API: AutoModel/AutoTokenizer APIs inspired by HuggingFace for easy model and tokenizer loading.
• 🔌 Dual API Compatibility: Supports both OpenAI and Anthropic chat completion APIs out of the box.

Quick Start

Installation

git clone https://github.com/ViperEkura/AstrAI.git
cd AstrAI
pip install -e .

For development dependencies:

pip install -e ".[dev]"

Download Pre-trained Model

Download pre-trained model weights (1B bilingual checkpoint) to params/:

python scripts/demo/download.py

Or download manually from HuggingFace into params/.

Train a Model

export CUDA_VISIBLE_DEVICES=0,1,2,3

nohup python scripts/tools/train.py \
    --nprocs=4 \
    --train_type=pt \
    --data_root_path=/path/to/dataset \
    --param_path=/path/to/model \
    --batch_per_device=4 \
    --grad_accum_steps=8 \
    --warmup_ratio=0.05 \
    --max_lr=1e-4 \
    --max_grad_norm=1.0 \
    --adamw_beta1=0.95 \
    --adamw_beta2=0.99 \
    --adamw_weight_decay=0.01 \
    --window_size=2048 \
    --ckpt_interval=10000 \
    --ckpt_dir=./checkpoint \
    --random_seed=3407 \
    --label_smoothing=0.05 \
    > out.log 2> err.log &

Full reference at Parameter Guide.

Generate Text

python scripts/tools/generate.py \
    --param_path /path/to/model \
    --input_json_file /path/to/input.json \
    --output_json_file /path/to/output.json

Docker

Build and run with Docker (recommended for GPU environments):

# Build image
docker build -t astrai:latest .

# Run with GPU support
docker run --gpus all -it astrai:latest

# Run with specific GPUs
docker run --gpus '"device=0,1"' -it astrai:latest

# Run inference server
docker run --gpus all -p 8000:8000 astrai:latest \
  python -m scripts.tools.server --port 8000 --device cuda

# Run with volume mount for data
docker run --gpus all -v /path/to/data:/data -it astrai:latest

# Docker Compose (GPU, default)
docker compose up -d

# Docker Compose (CPU only)
docker compose --profile cpu up -d

Note: --gpus all is required for CUDA support. Without it, torch.cuda.is_available() will return False.

Start HTTP Server

Start the inference server with OpenAI and Anthropic-compatible HTTP API:

python -m scripts.tools.server --port 8000 --device cuda

Make requests:

# OpenAI-compatible
curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Hello"}],
    "max_tokens": 512
  }'

# OpenAI-compatible streaming
curl -X POST http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Tell a story"}],
    "stream": true,
    "max_tokens": 500
  }'

# Anthropic-compatible
curl -X POST http://localhost:8000/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "astrai",
    "system": "You are a helpful assistant.",
    "messages": [{"role": "user", "content": "Hello"}],
    "max_tokens": 512
  }'

# Anthropic-compatible streaming with stop sequences
curl -X POST http://localhost:8000/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "astrai",
    "messages": [{"role": "user", "content": "Write a story"}],
    "max_tokens": 500,
    "stream": true,
    "stop_sequences": ["The end"]
  }'

# Health check
curl http://localhost:8000/health

Demo

Check out the demos in the scripts/demo/ folder:

# Download pre‑processed data (required before running demos)
python scripts/demo/download.py

# Interactive streaming chat
python scripts/demo/stream_chat.py

# Batch generation
python scripts/demo/generate_batch.py

# Auto‑regressive generation
python scripts/demo/generate_ar.py

Watch a video walkthrough on bilibili.

Documentation

Document	Description
Parameter Guide	Training & inference parameters
Architecture	System architecture, class diagram & design patterns
Training	Training loop, strategies & formulas
Inference	KVCache, continuous batching, sampling & HTTP API
Data Flow	Data pipeline, storage backends & dataset architecture

Contributing

We welcome contributions! Please see our Contributing Guidelines for details.

1. Fork the repository.
2. Create a feature branch.
3. Commit your changes.
4. Open a Pull Request.

For major changes, please open an issue first to discuss what you would like to change.

Community

• GitHub Issues: Issue Tracker
• Discussions: GitHub Discussions
• HuggingFace: Model Hub

License

This project is licensed under the GPL-3.0 License.

---

A lightweight Transformer framework designed for both high performance and ease of use.