Tugsbayasgalan Manlaibaatar bb31e3f57e Add original forward names to schema so that prettify pass works (#136887) ... When we run_decomp, we retrace if it is training IR. As a result, we do need to reliably store the oroiginal forward names when we run decomp. Differential Revision: [D63064453](https://our.internmc.facebook.com/intern/diff/D63064453/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/136887 Approved by: https://github.com/angelayi		2024-10-08 04:21:02 +00:00
..
db
pass_infra
passes	[export] fix lifted constants order for 0-input graphs (#136658)	2024-09-26 17:44:24 +00:00
serde	Add original forward names to schema so that prettify pass works (#136887)	2024-10-08 04:21:02 +00:00
__init__.py	Make PT2E work with both IR simultaneously (#135769)	2024-10-02 21:05:22 +00:00
converter.py	[ts_converter] Fix prim::If buffer names (#136648)	2024-10-02 00:07:47 +00:00
error.py
non_strict_utils.py	[export] clean up dynamic markers from tensors (#137230)	2024-10-04 06:50:45 +00:00
pass_base.py
tools.py
utils.py	[export] Update swap's forward function (#137102)	2024-10-06 04:25:36 +00:00
verifier.py	Allow autocast in training ir export (#137287)	2024-10-04 17:38:51 +00:00
wrappers.py