- root cause: the reference Q4_K GGUF stores the 2 noise + 2 context refiner
blocks as Q8_0; on a memory-bound CPU GEMM the same M=1024 w13 matmul is
89.4 ms (Q8_0) vs 42.9 ms (Q4_K), ~-1.5 s per 8-step run
- tools/requant_refiner.py + tools/q4k.c rewrite only the 16 refiner
attention/FFN matrices as Q4_K (adaLN kept at Q8_0); build.sh now produces
z_image_turbo-Q4_K-ffn13-rq.gguf (opt-in, pixel diff mean 6.4)
- flash-attn: hoist V packing to once per head and raise tiles to 128x128;
FLASH_ATTN_EXT 4663 -> 4451 ms (-4.5%), pixel-identical
- add stages/stage_p2_refiner notes, SmartTest rq row, README findings
3 days ago
Add OpenVINO INT4 (8 steps) to benchmark; update results and model card
