PR #1086 Fix MXFP/NVFP + FP8 Attn/KV

Fix MXFP/NVFP + FP8 Attn/KV #1086

yiliu30 merged 10 commits into main from fp8-kv-attn-fix

add fp8 kv

493c3845

calib for kv/attn

c0aff354

add ut

aea1c1c0

[pre-commit.ci] auto fixes from pre-commit.com hooks

e5473c34

yiliu30 closed this 25 days ago

revert

831e770b

Merge branch 'fp8-kv-attn-fix' of https://github.com/intel/auto-round…

b0dede05

yiliu30 reopened this 25 days ago

yiliu30 changed the title ~~Fix MXFP/NVFP + FP8 Attn/kv~~ Fix MXFP/NVFP + FP8 Attn/KV 25 days ago

update config

97e61b22

yiliu30 requested a review from

WeiweiZhang1 25 days ago

yiliu30 requested a review from

n1ck-guo 25 days ago

Merge branch 'main' into fp8-kv-attn-fix

72e6c85e

yiliu30 added this to the 0.9.2 milestone 25 days ago

yiliu30 enabled auto-merge (squash) 25 days ago

n1ck-guo commented on 2025-12-03

disabled auto-merge 25 days ago
Manually disabled by user

n1ck-guo approved these changes on 2025-12-03

reduce model

bf9e4494

Merge branch 'main' into fp8-kv-attn-fix

83128bb7

yiliu30 merged caa246f9 into main 25 days ago

yiliu30 deleted the fp8-kv-attn-fix branch 25 days ago

Reviewers

n1ck-guo

WeiweiZhang1

Assignees

No one assigned

Labels

None yet

Milestone

0.9.3