Refactor Deepfloyd IF tests. #6855
Conversation
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
| mask_image = floats_tensor((1, 3, 256, 256), rng=random.Random(1)).to(torch_device) | ||
|
|
||
| output = pipe_2( | ||
| output = pipe_super_res( |
|
|
||
| mem_bytes = torch.cuda.max_memory_allocated() | ||
| assert mem_bytes < 10 * 10**9 | ||
| assert mem_bytes < 13 * 10**9 |
| image = floats_tensor((1, 3, 64, 64), rng=random.Random(0)).to(torch_device) | ||
|
|
||
| output = pipe_2( | ||
| output = pipe_super_res( |
There was a problem hiding this comment.
I think we shouldn't do super-resolution here.
Instead, it could make sense to pre-serialize the inputs and pass them accordingly to the pipeline for testing. This should reduce some timing, too.
WDYT?
There was a problem hiding this comment.
Actually, I should just move them to their individual files. We have fast tests for the individual pipelines. So the Slow tests can live there as well.
sayakpaul
left a comment
sayakpaul
left a comment
There was a problem hiding this comment.
This IS the way to go! Thanks!
My only concern is about how we're doing super-resolution after every core pipeline tasks. I don't think that is necessary. I have left a comment about it.
yiyixuxu
left a comment
yiyixuxu
left a comment
There was a problem hiding this comment.
makes sense and PR looks good to me:)
I'm ok if you want to split it into the existing individual files too
* update * update * update
What does this PR do?
The Deepfloyd IF slow tests runs all pipelines in a single test. This is a bit difficult to debug when there are timeouts/failed tests. Splitting these into individual tests.
Fixes # (issue)
Before submitting
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.