-
Notifications
You must be signed in to change notification settings - Fork 749
Arm backend: Eliminate one memory copy in executor runner #12992
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/12992
Note: Links to docs will display an error until the docs builds have been completed. ❌ 2 New Failures, 1 Unrelated FailureAs of commit 4082ceb with merge base 9f8c2f6 ( NEW FAILURES - The following jobs have failed:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
Prior to this patch, temporary tensors/buffers were used to hold the input data which was then copied over to the actual input tensors for running the inference. This patch removes this copying by instead writing the input data directly to the input tensors. Change-Id: I20565a67dea8d8070efe9593aff4796c54e5fe90 Signed-off-by: [email protected]
e689661 to
4082ceb
Compare
|
https://github.com/pytorch/executorch/actions/runs/16622225895/job/47029216900 seems the size has gone above the threshold of 109000 resulting in "Fail 109236 > 109000". I dont know where the threshold value is coming from. |
Seems the threshold was increased a bit in 69f3da0e332 by @swolchok just a few days ago. |
|
mac os fails also unrelated |
Arbitrary, this is to keep large, unintended size increases to sneak past without getting noticed. I can bump it again. |
) Prior to this patch, temporary tensors/buffers were used to hold the input data which was then copied over to the actual input tensors for running the inference. This patch removes this copying by instead writing the input data directly to the input tensors. Signed-off-by: [email protected] cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 Signed-off-by: [email protected] Co-authored-by: Martin Lindström <[email protected]>
Prior to this patch, temporary tensors/buffers were used to hold the input data which was then copied over to the actual input tensors for running the inference. This patch removes this copying by instead writing the input data directly to the input tensors.
Change-Id: I20565a67dea8d8070efe9593aff4796c54e5fe90
Signed-off-by: [email protected]
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218