logits_to_keep bool mask for all ForXXX classes #48825
feat(gemma4): bool mask logits_to_keep, expose last_hidden_state
e9016e02
feat(gemma4): extend logits_to_keep mask support to all ForXXX classe…
7cc36012
jp1924
force pushed
from
1da2b43e
to
7cc36012
62 days ago
jp1924
changed the title [Gemma4] Support bool mask for `logits_to_keep`, expose `last_hidden_state` logits_to_keep bool mask for all ForXXX classes, expose last_hidden_state, fix negative output_hidden_states indices 62 days ago
feat: extend logits_to_keep bool-mask support to all ForCausalLM/ForC…
2c987eae
remove last_hidden_state field, rely on output_hidden_states indices …
8fa79042
Merge branch 'main' into feat/gemma-4-mask-last
f442d49a
jp1924
marked this pull request as ready for review 61 days ago
revert CSM to original logits_to_keep handling, drop stray gemma4-onl…
2ecaad5b
Merge branch 'main' into feat/gemma-4-mask-last
f04e87b8
vasqu
commented
on 2026-09-18
revert output_capturing.py
57f5cd1a
Merge branch 'main' into feat/gemma-4-mask-last
f71a9d2b
Add: slice_logits_to_keep to modeling_utils.py
413a3f17
apply to `bit`, `falcon`, `hyperclovax`, `hyperclovax_vision_v2`, `ne…
b0e0e102
Add: 2d idx process
b0d70b78
Merge branch 'main' into feat/gemma-4-mask-last
23f3e3bf
vasqu
commented
on 2026-10-19
Address review feedback: https://github.com/huggingface/transformers/…
dc4efdbc
Merge branch 'main' into feat/gemma-4-mask-last
597d3719
vasqu
approved these changes
on 2026-10-19
Address feedback: https://github.com/huggingface/transformers/pull/48…
9b41da20
Merge branch 'main' into feat/gemma-4-mask-last
953f38a6
Merge branch 'main' into feat/gemma-4-mask-last
db521a1c
vasqu
commented
on 2026-10-20
Address feedback: https://github.com/huggingface/transformers/pull/48…
a45225a8
Merge branch 'main' into feat/gemma-4-mask-last
1534479a
jp1924
changed the title logits_to_keep bool mask for all ForXXX classes, expose last_hidden_state, fix negative output_hidden_states indices logits_to_keep bool mask for all ForXXX classes 29 days ago
Update src/transformers/modeling_utils.py
f12a916e
Merge branch 'main' into feat/gemma-4-mask-last
cdc10868
Update src/transformers/modeling_utils.py
6d6dee54
Apply slice_logits_to_keep to the models that were missed
9b365d9f
jp1924
commented
on 2026-10-21
dont like the newline that was added
7f854d99
more newline
d7060de1
more
e789354d
remove more newline and nonsense assignment
c1a9367b
unrelated
63bdf9c2
Merge branch 'main' into feat/gemma-4-mask-last
87981517
more based on latest main
63d67d98
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub