pytorch / pytorch/executorch

Android Benchmark App Reports Success Despite Empty Results

Open
#11,620 3 comments 0 reactions 1 assignee View on GitHub

@kirklandsign is already working on this.

Since Jun 12, 2025.

Dominant language
Python
Stars
5k
Forks
1.2k
Avg merge
2d 10h
Merged PRs (30d)
581

Description

🐛 Describe the bug

The example job here: https://github.com/pytorch/executorch/actions/runs/15597347625/job/43931376798

It shows the benchmark job running gemma-3-1b on-device is finished successfully, but when checking the results in the later step (https://github.com/pytorch/executorch/actions/runs/15597347625/job/43932265620#step:8:449), it doesn't report any results from this run. If you downloaded the artifacts from the benchmark job for gemma-3-1b and inspect the results manually, it's empty. That is, there are failures durning inference but the app just silents it and report fake success instead.

For this given model, we know it will crash the runtime due to this bug (https://github.com/pytorch/executorch/issues/11618), so similar to running the llama runner binary, the correct behavior of the app is to fail explicitly and surface the error

Versions

trunk

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.