Generate: end-to-end compilation #30788
gante
merged
31 commits into
huggingface:main
from
gante:end_to_end_mvp
ydshieh
changed the title
gante
changed the title
gante
commented
on 2024-05-25
ArthurZucker
commented
on 2024-07-09
mvp
a9841793
added test (a few models need fixes)
f0336ed4
fix a few test cases
06983d86
test nits
17580dc1
harder test 😈
d06e9991
revert changes in stablelm
903b81f7
test with improved condition
08cc2c0a
add todo
683f3e7e
tmp commit
e063fc82
merged with main
0ebc4c76
nits
86c7170c
add todo
b2b60013
final corrections
2fcc2074
add docs for generation compilation
e84aedb0
docs nits
d2b45a46
add tip
64ce18b8
PR suggestions
ef4d4192
add more details to the compilation docs
d5e920d0
fix cache positions
40482d3d
cache is now init in generate; update docs
e3d9c04d
tag test as flaky
139e212b
docs
bc4ad7d8
post rebase make fixup and other nits
54c9eefa
remove unintended changes
3186b14d
whisper (encoder-decoder) not supported
484d9221
move token default updates to ; add tests for token defaults
bf9ef8ab
push changes
f2e28338
manual rebase
16f92f43
chameleon doesn't support this
838ba6a9
fix test_static_cache_mha_mqa_gqa (broken in another PR)
795d0580
docs: dynamic is better with end-to-end compilation
d2e423bb
Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub