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
Merge pull request #10 from Edge0-AI/fix/mlx-0.30.6-nax-a18
375582a linyubupa committed 2 months ago
fix(deps): pin mlx 0.30.6 and document the A18 NAX mis-gating
2e168c5 linyubupa committed 2 months ago
fix(deps): pin mlx 0.30.6 and document the A18 NAX mis-gating
0d52bf8 linyubupa committed 2 months ago
Merge pull request #2 from ahrazzle/fix/fetch-models-tuple-replace
fbab5f8 Alichua committed 2 months ago
fix(fetch): use tuple element for MODEL env var in done hint
bc353dc ahrazzle committed 2 months ago
README: describe Recover-LoRA (post-quantization distillation training)
ac4a256 linyubupa committed 2 months ago
rename HF repo id to Edge0/Edge0-35B-A3B-preview (capital B/A)
fc51cf4 linyubupa committed 2 months ago
rename edge0-10b tier to edge0-8b (8B-class model, Ling 3.0 tiny base); HF repo Edge0/Edge0-8b-a1b-preview
4823516 linyubupa committed 2 months ago