star7code / minimax-h3-chunk-star7
git clone https://github.com/star7code/minimax-h3-chunk-star7.git
VRAM-efficient MiniMax H3 for ComfyUI: QKV/RoPE/MLP activation chunking, automatic OOM chunk reduction, and CK/SLA/Sol/Hybrid attention for SM75 and SM80+ GPUs.
+0
stars, 24h · #28,702
+~-16
stars, 7 days · #61,864
+~-70
stars, 30 days · #62,192
92
★ total · -15% this week
2
forks · 3 open issues
Star history (as recorded here)
LanguagePython
LicenseMIT
Created2026-08-14 (2mo ago)
Last push5d ago
OwnerUser