Qualcomm AI Engine Direct - Change transpose output from NCH1 to NC1H before conv - #21707
Qualcomm AI Engine Direct - Change transpose output from NCH1 to NC1H before conv#21707chenweng-quic wants to merge 1 commit into
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/21707
Note: Links to docs will display an error until the docs builds have been completed. ❌ 4 New FailuresAs of commit afb0f01 with merge base b20f16a ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This PR needs a
|
|
cc @digantdesai |
|
@chenweng-quic Can you check the CI failures (and lint-check as well ?) Are they related ? |
|
checking, will update. |
Summary
This memory layout has better performance on the HTP backend, and lead to better performance on prefill model.
Prefill performance of third partitioned graph:
cc @cbilgin @psiddh