Actions: pytorch/TensorRT
Actions
1,602 workflow runs
1,602 workflow runs
FxGraphCachePickler.get_hash(new_gm)
takes up a large portion of the total compile time, which makes reusing cached engine slow
Blossom-CI
#3622:
Issue comment #3249 (comment)
created
by
zewenli98
STRIP_PLAN
+ REFIT
Blossom-CI
#3621:
Issue comment #3255 (comment)
created
by
zewenli98
dict
type module_outputs
in infer_module_output_dtypes
Blossom-CI
#3615:
Issue comment #3133 (comment)
created
by
jiwoong-choi