AlphaUncommon
Mininglamp-AI/cider
W8A8/W4A8 inference + optimized SDPA on Apple Silicon — unlocking unused INT8 TensorOps in M5 for 1.2–1.9× faster LLM prefill, plus FlashInfer-inspired GQA decode attention for up to 1.6× SDPA speedup, built as MLX custom primitives.
- Owner
- Mininglamp-AI (organization)
- Stars
- Stars: 361
- Forks
- Forks: 17
- Language
- Python
- License
- MIT
- Created
- 2026
- Topics
- apple-siliconmetalmlxquantizationw4a8w8a8
- Tier
- Alpha (250 to 699 ★) What this tier means
More repos by Mininglamp-AI
All 3 repos by Mininglamp-AI- Mininglamp-AI/Mano-PStars: 2,840RC
- Mininglamp-AI/OctoASRStars: 148PythonPrototype
Nearby in Python
All Python repos- shuyicc/MathLensStars: 361PythonAlpha
- eliautobot/my-virtual-officeStars: 361PythonAlpha
- JFGAtlas/aethergate-proStars: 361PythonAlpha
- pyelasticsearch/pyelasticsearchStars: 361PythonAlpha
- github/pycon2011Stars: 361PythonAlpha
- inconshreveable/sqltapStars: 361PythonAlpha