Cache: new Cache format in decoder-only models #31421
zucchini-nlp
merged
44 commits into
huggingface:main
from
zucchini-nlp:dynamic_cache_decoder_only
draft bart with new cache
183cd66a
add cache for decoder-only models
4578bca4
revert utils
9505ca4c
modify docstring
2ab28f34
revert bart
5fe4e9e3
minor fixes
09413c30
fix copies (not related)
3c276043
zucchini-nlp
commented
on 2024-06-14
zucchini-nlp
commented
on 2024-06-14
zucchini-nlp
commented
on 2024-06-14
revert tests
350acc5a
gante
commented
on 2024-06-14
remove enc-dec related code
c0adf10d
remove bloom
c18b1775
remove opt (enc-dec)
582f289d
Merge remote-tracking branch 'upstream/main' into dynamic_cache_decod…
3141a715
update docstring
33d54b49
git, codegen, gpt_neo, gpt_neox, gpj
dd05e6bf
clean up
cb878d57
copied from statements
0588791e
revert
a27b47c6
tmp
1abcf305
update warning msg
00ed88c4
forgot git
6c3b3aa7
gante
commented
on 2024-06-20
add more flags
fd5eeabb
run-slow git,codegen,gpt_neo,gpt_neox,gpj
e233f296
add cache flag to VLMs
356d578d
remove files
c9066701
Merge branch 'main' into dynamic_cache_decoder_only
08d9e6f0
style
56c05b2b
video LLMs also need a flag
85108109
style
cebb55d4
llava will go in another PR
8fd9dd1f
Merge branch 'main' into dynamic_cache_decoder_only
4b9ced10
style
aea219ba
[run-slow] codegen, falcon, git, gpt_neo, gpt_neox, gptj, idefics
49918633
Update src/transformers/models/gpt_neo/modeling_gpt_neo.py
ec306a26
copy from
cf793b7d
deprecate until v4.45 and warn if not training
c92409c7
nit
c2b97e41
fix test
35b60de9
test static cache
d2fca9a1
Merge branch 'main' into dynamic_cache_decoder_only
0933350e
add more tests and fix models
42349d49
fix copies
45c3a1bd
return sliding window mask
5f226161
run slow tests & fix + codestyle
f5af6a28
one more falcon fix for alibi
21b45c5d
Login to write a write a comment.
Login via GitHub