Run 35B MoE models in ~2.5 GB of RAM. Weights stream from SSD; a trained router predicts loads a step ahead. Mac, iPhone, Android, Windows supported
Python
Commits on main
README: banner image, Hugging Face badges and checkpoint links
dd2cec8 linyubupa committed 2 months ago
README: drop numerical-guard bullet from public feature list
5f70676 linyubupa committed 2 months ago
docs: translate docs to English; README prerouter A/B numbers and HF checkpoint quick start
b4e45ab linyubupa committed 2 months ago
README: report short-context peak memory (35b 2.9 GiB, 10b 1.0 GiB)
d426e65 linyubupa committed 2 months ago
README: official checkpoint names (Edge0/Edge0-35b-a3b, Edge0-10b-a1b)
ec6a7e3 linyubupa committed 2 months ago
Public-release pass: neutral naming, no internal references
9333ea1 linyubupa committed 2 months ago
README: add prefill timing (cold/warm) to the benchmark table
918d61c linyubupa committed 2 months ago
Think-off default, temp 0.7, benchmark section, prewarm (opt-in)
fdbd068 linyubupa committed 2 months ago
Open-source readiness: EN README, zero hardcoded paths, demo-profile parity
2de963b linyubupa committed 2 months ago
README rewrite + architecture doc: platform-neutral framing
5669d6a linyubupa committed 2 months ago
README/pyproject: drop Apple-Silicon-only framing; backend-isolated, multi-platform roadmap
4d78aeb linyubupa committed 2 months ago
round9 (35b) + round6 (10b) adapters; co-located layout; strip vision tower
b73af00 linyubupa committed 2 months ago
edge0 initial commit: MLX streaming MoE inference (prerouter + LoRA + SSD offload)
17021c5 linyubupa committed 2 months ago