[WIP] Reducing eval vision tests runtime

[SalmanMohammadi]

Context

What is the purpose of this PR? Is it to

• add a new feature
• fix a bug
• update tests and/or documentation
• other (please add here)

Please link to any issues this PR addresses.

Changelog

What are the changes made in this PR?
*

Test plan

Please make sure to do each of the following if applicable to your PR. If you're unsure about any one of these just ask and we will happily help. We also have a contributing page for some guidance on contributing.

• run pre-commit hooks and linters (make sure you've first installed via pre-commit install)
• add unit tests for any new functionality
• update docstrings for any new or updated methods or classes
• run unit tests via pytest tests
• run recipe tests via pytest tests -m integration_test
• manually run any new or modified recipes with sufficient proof of correctness
• include relevant commands and any other artifacts in this summary (pastes of loss curves, eval results, etc.)

UX

If your function changed a public API, please add a dummy example of what the user experience will look like when calling it.
Here is a docstring example
and a tutorial example

• I did not change any public API
• I have added an example to docs or docstrings

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/torchtune/2022

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• [DomainsOnly] Jobs fail with GLIBC version not found

✅ No Failures

As of commit bfedad2 with merge base d5c54f3 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[joecummings]

At a glance, I'm worried this isn't a good enough signal for changes in the recipe due to the fact that the majority of MMMU is multiple choice.

I'd postulate (ha big word) that a minimum of 3 is good, but I'd like to understand the tradeoff better. What are the run times for 1 vs. 2 vs. 3 vs. 4? Also what are the questions for each subject that come up? Are the open or mc?

[SalmanMohammadi]

These are going to fail due to #1995 (comment) so partly trying to debug that here. The current runtime for 4 samples is ~5 minutes which is pretty terrible.

[SalmanMohammadi]

Okay it's not coming up anymore so I'm going slightly insane.

Also what are the questions for each subject that come up? Are the open or mc?

They're all MC questions I think.

What are the run times for 1 vs. 2 vs. 3 vs. 4?

The runtime for limit=1 was around 80s, so I'd postulate that 2 is 160s, 3 is 240s, etc.

[SalmanMohammadi]

this branch

224.22s call     tests/recipes/test_eleuther_eval.py::TestEleutherEval::test_meta_eval_vision
212.66s call     tests/recipes/test_eleuther_eval.py::TestEleutherEval::test_hf_eval_vision

vs main

277.36s call     tests/recipes/test_eleuther_eval.py::TestEleutherEval::test_hf_eval_vision
276.25s call     tests/recipes/test_eleuther_eval.py::TestEleutherEval::test_meta_eval_vision

It's timeboxing time. @joecummings I humbly offer my meagre ~2 minutes speedup.

[joecummings]

I accept your 2 minute speed up and thank you for your service to this great project.

I solemnly swear we will find a better way to handle this sometime in the future (likely when it starts to annoy us again).