[CI][Perf] Add nightly PR labels, consolidate pipeline, and switch benchmark flag to --test-config-file#2816
Conversation
|
Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits. |
|
BLOCKER scan:
OVERALL: NO BLOCKERS VERDICT: COMMENT This is a straightforward CI infrastructure update. The changes to support --config-file for performance tests are useful, and the documentation is updated accordingly. Minor suggestion: Consider keeping the explicit error checks in _resolve_baseline_value() for sweep_index validation. The old code provided clearer error messages that would help users debug configuration issues more quickly. Overall, the PR is ready to merge once the blocked checks pass. |
…models - Enhanced the nightly pipeline to include additional labels for triggering tests. - Removed the obsolete `test-nightly-diffusion.yml` file. - Updated `test-nightly.yml` to include new performance tests for Omni and TTS models. - Introduced new performance test configurations in `test_qwen_omni.json` and `test_tts.json`. - Added new benchmark scripts for Omni and diffusion models. - Updated documentation to reflect changes in performance test configurations. Signed-off-by: wangyu <410167048@qq.com> Co-authored-by: inaniloquentee <inani_@stu.xjtu.edu.cn>
…rmance testing of Omni and TTS models. - Updated the nightly pipeline configuration to reflect changes in test script names and parameters. - Introduced `run_diffusion_benchmark.py` for benchmarking diffusion models. - Adjusted documentation to align with new test script usage and configuration options. Signed-off-by: wangyu <410167048@qq.com> Co-authored-by: inaniloquentee <inani_@stu.xjtu.edu.cn>
…pr-label Signed-off-by: wangyu <410167048@qq.com>
Signed-off-by: wangyu <410167048@qq.com>
|
I have modified some performance test scripts for the following reasons, @amy-why-3459 @fhfuih PTAL:
|
Signed-off-by: wangyu <410167048@qq.com>
Signed-off-by: wangyu <410167048@qq.com>
fhfuih
left a comment
There was a problem hiding this comment.
For the diffusion part, generally looks good to me. Left some comments on documentation
| /tests/e2e/offline_inference/test_{model_name}_expansion.py<br> | ||
| <strong>Performance:</strong><br> | ||
| /tests/dfx/perf/tests/test.json<br> | ||
| /tests/dfx/perf/tests/test_qwen_omni.json (Omni) and test_tts.json (TTS)<br> |
There was a problem hiding this comment.
And there is /tests/dfx/perf/tests/test_{some diffusion models}_vllm_omni.json Maybe you would like to mention them in the doc
There was a problem hiding this comment.
It is also related to your change in docs/contributing/ci/test_examples/l4_performance_tests.inc.md and some changes below within this file
| ├── test_cache_dit.py | ||
| ├── test_teacache.py | ||
| ├── test_stable_audio_expansion.py | ||
| ├── test_stable_audio_model.py |
There was a problem hiding this comment.
Is this unintentional? There isn't a test_stable_audio_model anymore after the L4 test for stable audio is merged
Signed-off-by: wangyu <410167048@qq.com>
…configurations Signed-off-by: wangyu <410167048@qq.com>
| path: /mnt/hf-cache | ||
| type: DirectoryOrCreate | ||
|
|
||
| - label: ":full_moon: Diffusion · Qwen-Image · Accuracy Test" |
There was a problem hiding this comment.
You mistakenly deleted this nightly test. Please add it back.
…nchmark flag to --test-config-file (vllm-project#2816) Signed-off-by: wangyu <410167048@qq.com> Co-authored-by: Y. Fisher <yukexiong1@huawei.com> Co-authored-by: inaniloquentee <inani_@stu.xjtu.edu.cn>
PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.
Purpose
fix #2410
Summary
Key Changes
CI:
.buildkite/pipeline.yml
.buildkite/test-nightly.yml
Deleted .buildkite/test-nightly-diffusion.yml
Perf tests:
tests/dfx/perf/scripts/run_benchmark.py
tests/dfx/perf/scripts/run_diffusion_benchmark.py
tests/dfx/conftest.py
Deleted tests/dfx/perf/tests/test.json
Added/using tests/dfx/perf/tests/test_qwen_omni.json
Docs & tooling:
docs/contributing/ci/CI_5levels.md
docs/contributing/ci/test_guide.md
docs/contributing/ci/test_examples/l4_performance_tests.inc.md
tools/nightly/generate_nightly_perf_excel.py
Why
Test Plan
1.Run perf test in local:
pytest -s -v tests/dfx/perf/scripts/run_benchmark.py --test-config-file tests/dfx/perf/tests/test_tts.json2.CI Nightly: Validate each of the tag paths (omni/tts/diffusion) once, triggering each one individually.
Test Result
omni-test:

diffusion-x2v-test:

diffusion-x2iat-test:

nightly-test:

Essential Elements of an Effective PR Description Checklist
supported_models.mdandexamplesfor a new model. Please runmkdocs serveto sync the documentation editions to./docs.BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)