pytorch/torch/_decomp
Tugsbayasgalan Manlaibaatar 73b07df042 Preserve custom ops via run_decomps (#136882)
This is re-apply of https://github.com/pytorch/pytorch/pull/136773?fbclid=IwZXh0bgNhZW0CMTEAAR3SmginkvZcILVY7G2XDa_KosnV4DPmq1l6pkjPIM255QgJLKVAR90rGAU_aem_ZWpcVdUsmAGzOGiwbjtBDg.

Note that this doesn't completely remove the _preserve_ops list from export mainly because we want to have small change to address failing executorch tests. All the complications included in this PR is deleted in the next PR.

Differential Revision: [D63553086](https://our.internmc.facebook.com/intern/diff/D63553086/)
Pull Request resolved: https://github.com/pytorch/pytorch/pull/136882
Approved by: https://github.com/bdhirsh
2024-10-01 17:38:00 +00:00
..
__init__.py Preserve custom ops via run_decomps (#136882) 2024-10-01 17:38:00 +00:00
decompositions.py Fix AOT Graph capture not propagating non_blocking copy parameter to … (#136513) 2024-10-01 00:32:47 +00:00
decompositions_for_jvp.py [BE][Easy][15/19] enforce style for empty lines in import segments in torch/_d*/ (#129767) 2024-07-31 21:18:11 +00:00
decompositions_for_rng.py Add None return type to init (#132335) 2024-08-01 15:26:45 +00:00