nodes/BobJohnson24-ComfyUI-Flux2-INT8
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
173 Commits 3 Branches 0 Tags
Debug
Commit Graph
3 Commits
Author SHA1 Message Date
jjarquin aefad3aff8 Add per-row INT8 weight quantization with fused Triton kernel 
- Add int8_forward_dynamic_per_row for per-row weight scale dequantization
- Add fused Triton GEMM kernel with per-row weight scale epilogue
- Auto-detect per-row vs per-tensor scale shape during weight loading
 2026-02-19 02:05:13 +01:00
BobJohnson24 39c3e30223 Make rounding match previous behavior 2026-01-24 01:32:21 +07:00
BobJohnson24 a37b48f199 Fused Triton Kernel 2026-01-23 22:27:53 +07:00