Skip to content

Improve execution time and eliminate flakiness in data_generation_v3_test. - #171

Open
copybara-service[bot] wants to merge 1 commit into
mainfrom
cl/972720308
Open

Improve execution time and eliminate flakiness in data_generation_v3_test.#171
copybara-service[bot] wants to merge 1 commit into
mainfrom
cl/972720308

Conversation

@copybara-service

Copy link
Copy Markdown

Improve execution time and eliminate flakiness in data_generation_v3_test.

  • In data_generation_v3_test.py, configure TabularConfig to default to MSTConfig(pgm_iters=100) instead of 5000 iterations for unit tests.
  • In mst.py, accept pgm_iters in _select_two_way_marginal_queries from MSTConfig so initial independent model estimation does not hardcode 2500 iterations.
  • In workload regression tests, explicitly configure baseline IndependentConfig to pgm_iters=500 to match tested mechanism iterations.
  • In MaxRecordsPerUserTest, relax delta in Poisson calibrate tests to 1e-3 to reduce PLD grid size during binary search.
  • In BUILD, set shard_count = 4 on data_generation_v3_test so tests run concurrently.

…test.

- In data_generation_v3_test.py, configure TabularConfig to default to MSTConfig(pgm_iters=100) instead of 5000 iterations for unit tests.
- In mst.py, accept pgm_iters in _select_two_way_marginal_queries from MSTConfig so initial independent model estimation does not hardcode 2500 iterations.
- In workload regression tests, explicitly configure baseline IndependentConfig to pgm_iters=500 to match tested mechanism iterations.
- In MaxRecordsPerUserTest, relax delta in Poisson calibrate tests to 1e-3 to reduce PLD grid size during binary search.
- In BUILD, set shard_count = 4 on data_generation_v3_test so tests run concurrently.

PiperOrigin-RevId: 972720308
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

0 participants