yunshu

PyPI page
Home page
Author: YuhuanStudio
Summary: Fast local single-node LLM/VLM inference engine for Apple Silicon — speculative decoding, prefix reuse, OpenAI/Anthropic-compatible
Latest version: 0.1.5
Required dependencies: argon2-cffi | fastapi | httpx | huggingface-hub | jsonschema | langdetect | llguidance | mlx | mlx-lm | numpy | openai | opentelemetry-api | opentelemetry-instrumentation-fastapi | opentelemetry-sdk | packaging | prometheus-client | psutil | pycountry | pydantic | pyjwt | pypdf | pypdfium2 | python-multipart | regex | rich | safetensors | scipy | sse-starlette | structlog | tokenizers | tqdm | trafilatura | transformers | typer | uvicorn | websockets
Optional dependencies: aiortc | datasets | diffusers | lz4 | matplotlib | mlx-audio | mlx-embeddings | mlx-vlm | mypy | pillow | playwright | pre-commit | pytest | pytest-asyncio | pytest-httpx | pytest-timeout | ruff | soundfile | torch | torchvision | zstandard

Downloads last day: 89
Downloads last week: 105
Downloads last month: 320
7-day smoothing