Refactor embedding input/output getter/setter #39339
molbap
merged
39 commits into
main
from
draft_VLM_apis
simplify common get/set
d107f9e6
remove some noise
98e54674
change some 5 years old modeling utils
cc8b91c4
update examples
077af3d4
fix copies
7d9b16c3
revert some changes
32c7c003
fixes, gah
33e0f06c
format
91537aa8
move to Mixin
57aac344
remove smolvlm specific require grad
39322e64
Merge branch 'main' into draft_VLM_apis
951d2698
skip
d2dd1298
ArthurZucker
commented
on 2025-07-15
force defaults
92e22962
Merge branch 'main' into draft_VLM_apis
00b676f8
remodularise some stuff
c5a3676c
remodularise more stuff
12651407
add safety for audio models
b7764814
style
d9693a31
have a correct fallback, you daft donkey
4ab8399b
remove this argh
2bee0841
change heuristic for audio models
f0fe775d
Merge branch 'main' into draft_VLM_apis
773a9ad2
fixup
a233db26
revert
d0fbe201
this works
5436239e
revert again
41221d8a
ðŸ§
8556767e
aaah ESM has two modelings aaah
21820d7b
Merge branch 'main' into draft_VLM_apis
7b614f86
Merge branch 'main' into draft_VLM_apis
429e119c
add informative but short comment
6e1fbdd9
Merge branch 'draft_VLM_apis' of github.com:huggingface/transformers …
807eaf07
add `input_embed_layer` mixin attribute
3edfa426
style
b1af42a1
Merge branch 'main' into draft_VLM_apis
8455e0f9
walrus has low precedence
153a0b05
modular fix
31404b4e
this was breaking parser
822dd448
Merge branch 'main' into draft_VLM_apis
395d3f90
Login to write a write a comment.
Login via GitHub