-
Notifications
You must be signed in to change notification settings - Fork 722
Upload Android benchmark results to OSS benchmark database #5808
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/5808
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit c942ab4 with merge base 2f9f94a ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
This updates the mobile job to return the list of artifacts from AWS to the caller to use as they see fit. One of the use case is for ExecuTorch to extract the benchmark results JSON from `CUSTOMER_ARTIFACT` and `TESTSPEC_OUTPUT` (2 types of artifacts from AWS). My plan here is to: * [x] Get the list of artifacts (this PR) * [ ] On ExecuTorch side, extract the benchmark JSON from the artifacts pytorch/executorch#5808 * [ ] Create a new GitHub action on test-infra to upload the JSON (This could be done on ExecuTorch side too, but I plan to make this generic and reusable by other projects, so it must be on test-infra) Minor fixes: * Also update the console output to print out the context of `TESTSPEC_OUTPUT`, which is basically the main output from the device. * Print additional information about the job name (the device name from AWS in disguise) and app type (Android, iOS) * Some lint fix here and there.
|
@huydhn has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
guangy10
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM
|
The testing looks good https://github.com/pytorch/executorch/actions/runs/11189433899/job/31111425608, I can see the data on the database. |
This adds a GHA to upload benchmark results. I'm using the existing `torchci-oss-ci-benchmark` database for now until the new generic database from https://fburl.com/gdoc/ossgtvte is ready. So, it's backed by dynamoDB instead of S3 for now. I also include a quick test with some dummy Android benchmark records. * [x] Get the list of artifacts #5727 * [x] On ExecuTorch side, extract the benchmark JSON from the artifacts pytorch/executorch#5808 * [x] Create a new GitHub action on test-infra to upload the JSON #5742 * [x] Create a new GHA role to upload the results meta-pytorch/pytorch-gha-infra#483
|
@huydhn has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
Summary: This PR adds a job to upload Android benchmark results to the benchmark database. It transforms the `benchmark_results.json` file slightly to fit into the current schema. We are going to have a better schema soon https://fburl.com/gdoc/ossgtvte, but landing this first would unblock the work on building the dashboard before the launch. Updating the schema can be done later. * The job processes what it finds, so if one model fails, the rest will still be uploaded. * I will follow up with another PR for iOS later. No need to wait for the TPS metric there, we'll upload what available first. There are still some TODO pending: * pytorch/test-infra#5742 * meta-pytorch/pytorch-gha-infra#483 But the structure of the CI job is ready to review. Pull Request resolved: #5808 Reviewed By: guangy10, kirklandsign Differential Revision: D63869876 Pulled By: huydhn
542db36 to
c942ab4
Compare
|
This pull request was exported from Phabricator. Differential Revision: D63869876 |
This PR adds a job to upload Android benchmark results to the benchmark database. It transforms the
benchmark_results.jsonfile slightly to fit into the current schema. We are going to have a better schema soon https://fburl.com/gdoc/ossgtvte, but landing this first would unblock the work on building the dashboard before the launch. Updating the schema can be done later.There are still some TODO pending:
But the structure of the CI job is ready to review.