CoolFace
Modelpublic

canada-quant/hy3-w4a16-mtp

sourceHugging Faceapache-2.0updated 14h agoView on Hugging Face
5likes72downloads
12 commits on main
a8db22814h ago

card: fix recipe link (recipe.yaml is at the repo root)

pastapaul
49228b92mo ago

Fix calibration dataset metadata: reflect actual code+math+chat blend + link published calibration set

pastapaul
78ab8b02mo ago

deploy guide: measured 4xH100 256K validation

pastapaul
4aa9f302mo ago

docs: Hopper deploy guide

pastapaul
25400132mo ago

docs: findings referenced by the model card (fixes 404 links)

pastapaul
5e3fada2mo ago

card: fix strikethrough rendering, headline comparison table

pastapaul
6fdc36b2mo ago

release: Hy3 W4A16-MTP weights (verified artifact, 172GB)

pastapaul
ee3e5282mo ago

release: final matrix

pastapaul
c11b15d2mo ago

reproducibility: GPTQ recipe

pastapaul
e31f8632mo ago

model card: quality matrix (4 quants + FP8), 12-config throughput, deploy guide refs

pastapaul
71b02212mo ago

Upload config.json with huggingface_hub

pastapaul
207669d2mo ago

initial commit

pastapaul