ViperEkura 20041d7aa9 perf: extend MMA decode to arbitrary GQA ratio, add launch bounds, vectorize combine
- Multi-pass MMA: encode pass in grid blockIdx.x, compute q_head0/G in-kernel
- Fixes crash for G>32 (previously block(32,G) exceeded 1024 threads)
- Fixes alloc_split_partials using uninitialized num_splits (MAX_SPLITS=32)
- __launch_bounds__ on all MMA and prefill kernels for better register allocation
- 4x vectorized combine kernel (4 head_dim per thread)
- uint4 vectorized K loads in scalar decode kernels
- cp.async .L2::128B cache hint for K/V tile streaming
- Extract warp_reduce_sum, bf16, MAX_SPLITS to attn_warp_utils.cuh
2026-07-27 00:35:34 +08:00
2026-04-10 11:25:00 +08:00

Logo

A lightweight Transformer training & inference framework

python license release stars forks


📖 Table of Contents


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.
  • 🔬 ResearchFriendly: 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.

Getting Started

End-to-end walkthrough in 5 steps:

1. Install

git clone https://github.com/ViperEkura/AstrAI.git
cd AstrAI
pip install -e .                                          # pure PyTorch (no CUDA kernels)
# CSRC_KERNELS=true pip install -e . --no-build-isolation  # optional: fused CUDA kernels
# pip install -e ".[dev]"                                  # dev dependencies (pytest, ruff)

2. Download model

python scripts/demo/download.py    # downloads 1B checkpoint to params/

3. Preprocess data

Create pretrain.json (preprocessing config for seq strategy):

{
    "version": 1,
    "input": {"sections": [{"field": "text", "action": "train"}]},
    "preprocessing": {"max_seq_len": 2048},
    "output": {"storage_format": "bin"}
}
python scripts/tools/preprocess.py data/*.jsonl -o output/ -c pretrain.json

4. Train

export CUDA_VISIBLE_DEVICES=0,1,2,3

nohup python scripts/tools/train.py \
    --nprocs=4 \
    --parallel_mode=ddp \
    --train_type=seq \
    --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 \
    --weight_decay=0.1 \
    --window_size=2048 \
    --ckpt_interval=10000 \
    --ckpt_dir=./checkpoint \
    --random_seed=3407 \
    --label_smoothing=0.05 \
    > out.log 2> err.log &

5. Serve & query

# Terminal 1: start server
python scripts/tools/server.py --param_path ./params --device cuda

# Terminal 2: query
curl http://localhost:8000/v1/chat/completions \
  -H "Content-Type: application/json" \
  -d '{"messages":[{"role":"user","content":"Hello"}],"max_tokens":512}'

Demo

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

# Download model weights (required before running demos)
python scripts/demo/download.py                      # model → params/

# Interactive streaming chat (multi-turn, maintains history)
python scripts/demo/stream_chat.py
# Type your message after >>, type !exit to quit

# Batch generation (5 hardcoded prompts, non-streaming)
python scripts/demo/generate_batch.py

# Single-prompt autoregressive streaming
python scripts/demo/generate_ar.py

All generation demos use temperature=0.8, top_p=0.95, top_k=50, max_tokens=2048 by default and require params/ to contain model weights (run download.py first).

Watch a video walkthrough on bilibili.


See Documentation for full references beyond the examples above.

Text Generation

Batch generation from a JSONL file:

python scripts/tools/generate.py \
    --param_path ./params \
    --input_json_file input.jsonl \
    --output_json_file output.jsonl

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 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.

HTTP API Examples

Additional request examples beyond the Getting Started flow:

# 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

See Inference Guide for SSE streaming format, error codes, and stats endpoint.

Documentation

Document Description
CLI Reference Parameters for all CLI tools (train, server, generate, preprocess)
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
Preprocessing Declarative JSON-driven data preprocessing

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

License

This project is licensed under the GPL-3.0 License.


A lightweight Transformer framework designed for both high performance and ease of use.
S
Description
No description provided
Readme Apache-2.0
4.4 MiB
Languages
Python 77.9%
Cuda 19%
Shell 2%
C++ 0.7%
CMake 0.2%
Other 0.2%