[Reland] horizontal concat fusion #111437
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/111437
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (4 Unrelated Failures)As of commit f94a6e5 with merge base 2dc1726 ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
BROKEN TRUNK - The following jobs failed but was present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
mlazos
added
ciflow/inductor
ciflow/trunk
|
@mlazos has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@mlazos has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
@pytorchbot merge |
Merge startedYour change will be merged once all checks pass (ETA 0-4 Hours). Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
| if ( | ||
| input_unwrapped.is_input_buffer() | ||
| and inputs[i].get_device().type == "cuda" | ||
| and not is_dynamic(input_buffer) | ||
| ): | ||
| buffer_names.append(input_buffer.get_name()) | ||
|
|
||
| if len(buffer_names) > 1: | ||
| V.graph.register_list(buffer_names) |
There was a problem hiding this comment.
Hi @mlazos, with this code we'll only launch a single concat kernel instead of [num-operand] concat kernels, correct?
5 participants
Reland #108115
The main fix is to disallow nop nodes to be included in foreach scheduler nodes
cc @voznesenskym @penguinwu @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @aakhundov @ColinPeppler