transformers
Fix CohereASR training-loss double-shift (same as Moonshine fix #46784)
#46895
Merged

Fix CohereASR training-loss double-shift (same as Moonshine fix #46784) #46895

sharmax-vikas
sharmax-vikas Fix CohereASR training-loss double-shift (same as Moonshine fix #46784)
a67e0d76
sharmax-vikas Merge branch 'main' into fix/cohere-asr-double-shift-loss
eb0822dc
vasqu
vasqu commented on 2026-06-26
sharmax-vikas Use shift_labels parameter instead of CrossEntropyLoss (per reviewer …
9aa71883
sharmax-vikas Merge branch 'fix/cohere-asr-double-shift-loss' of https://github.com…
f87d2535
vasqu
vasqu commented on 2026-06-26
sharmax-vikas Use do_shift_labels flag: only skip shift when shift_tokens_right() w…
4853a7a5
sharmax-vikas Add test for training loss no double shift
ee79d1d2
sharmax-vikas Retrigger CI
580eca0f
sharmax-vikas Merge branch 'main' into fix/cohere-asr-double-shift-loss
6037b939
sharmax-vikas Merge branch 'main' into fix/cohere-asr-double-shift-loss
ccc61821
sharmax-vikas Fix test: use ParakeetEncoderModelOutput with attention_mask
b59ec312
sharmax-vikas Merge branch 'main' into fix/cohere-asr-double-shift-loss
790cc90b
sharmax-vikas
SunMarc
SunMarc commented on 2026-06-30
sharmax-vikas Simplify: always pass shift_labels=labels (per reviewer feedback)
bd001cd2
eustlb
eustlb commented on 2026-07-06
sharmax-vikas Merge branch 'main' into fix/cohere-asr-double-shift-loss
34a15656
github-actions
github-actions
SunMarc
SunMarc approved these changes on 2026-07-10
HuggingFaceDocBuilderDev
SunMarc SunMarc merged e6347996 into main 124 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone