[inductor] skip bmm when converting channel last #159459
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/159459
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 217eaff with merge base 1e8e9f7 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
|
130e41a to
217eaff
Compare
|
@pytorchbot label "topic: not user facing" |
| @@ -800,12 +800,17 @@ def find_nodes_prefer_channels_last(self) -> OrderedSet[Node]: | |||
| With rule 2, we makes sure all the tensors in the chain uses channels last layout. So both copies | |||
| can be saved. | |||
| """ | |||
| last_conv = None | |||
| nodes_cannot_propagate = [torch.ops.aten.bmm.default] | |||
There was a problem hiding this comment.
How does the perf compare if we do a copy inside of aten.bmm.default if the dense dimension dim 0 ?
There was a problem hiding this comment.
The current behavior in aten bmm is to add a clone to do the transpose for both cuda and xpu. I don't know where I can add something like require_stride_order to make bmm dim[-1] dense, but this seems to add another transpose.
Workaround of #159458 by remove some nodes output channel last set
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @chenyang78 @kadeng @muchulee8 @amjames @chauhang @aakhundov @coconutruben