fix gemma4 num attention head bugs #7975
mingxiang1006
wants to merge
14 commits into
deepspeedai:master
from
mingxiang1006:master
fix gemma4 num attention head bugs
62ac4c6b
revise logic of fallback for text config
96d503df
Merge branch 'master' into master
6a5f998d
fix gemma4 num attention head bugs
c7aa332c
revise logic of fallback for text config
1f2af84d
fix(fp_quantizer): fix UB and negative shift warnings in fp_quantize_…
d7c1ca6c
mingxiang1006
commented
on 2026-04-21
fix(op_builder): avoid duplicate/wrong -gencode flags (#7974)
c604c942
Rename dequantization template parameters (#7976)
175869f1
Avoid CUDA reinit error in CI tests (#7977)
0ea3c5e9
Merge branch 'master' of https://github.com/mingxiang1006/DeepSpeed
fc2e3429
delock
commented
on 2026-04-22
remove fix here comment
e385c29a
update text fallback
e5b921f8
Merge branch 'master' into master
dd34cd99
DCO remediation sign-off
22b31c0e
Login to write a write a comment.
Login via GitHub