DeepSpeed
[CPU] Allow deepspeed.comm.inference_all_reduce in torch.compile graph
#5604
Merged
Merged
[CPU] Allow deepspeed.comm.inference_all_reduce in torch.compile graph #5604
tohtana
merged
21 commits into
deepspeedai:master
from
delock:gma/inference_all_reduce_in_graph
remove no compile for torchbackend inference_all_reduce
1b89ebd4
add myadd op
358bc46f
myadd -> inference_all
649fb53d
call inference_all_reduce_ in op implementation
284967e4
test pytorch C++ ops
8b24c2bd
cleanup
ee704a0c
enable inference_all_reduce_ as op
ba03ac12
add fallback path of all_reduce op
ee16062b
remove unused functions
39f73289
Merge branch 'master' into gma/inference_all_reduce_in_graph
1af7c0db
fix typo
4cb6c336
fix format
9cc13e3f
fix format
16e79cc7
change 'foo' in code into 'x'
84b039b6
remove debug print code
07a15764
temp save
26db8a07
remove fallback path
620efd1b
Merge branch 'master' into gma/inference_all_reduce_in_graph
78afbfbd
Merge branch 'master' into gma/inference_all_reduce_in_graph
bf38cf26
tohtana
approved these changes
on 2024-06-21
fix format
d6627c0f
Merge branch 'master' into gma/inference_all_reduce_in_graph
3baf3e9f
Login to write a write a comment.
Login via GitHub