mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-07 12:21:27 +01:00
When the constants are used twice in the exported graph (second one is returned as output), the lifting constant pass doesn't account for the second one being the output. THis PR fixes that. Differential Revision: [D64406108](https://our.internmc.facebook.com/intern/diff/D64406108/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/137993 Approved by: https://github.com/avikchaudhuri |
||
|---|---|---|
| .. | ||
| __init__.py | ||
| _node_metadata_hook.py | ||
| add_runtime_assertions_for_constraints_pass.py | ||
| collect_tracepoints_pass.py | ||
| constant_folding.py | ||
| functionalize_side_effectful_ops_pass.py | ||
| lift_constants_pass.py | ||
| remove_runtime_assertions.py | ||
| replace_autocast_with_hop_pass.py | ||
| replace_quantized_ops_with_standard_ops_pass.py | ||
| replace_set_grad_with_hop_pass.py | ||
| replace_view_ops_with_view_copy_ops_pass.py | ||
| replace_with_hop_pass_util.py | ||