adjust unit tests for test_save_load_float16
#12500
Open
+9
−9
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Liu, Kaixuan <kaixuan.liu@intel.com>
|
@a-r-r-o-w @DN6 pls help review, thx! |
|
Not sure I understand the issue here. This specific T5 module is kept in fp32 on purpose, why forcing a fp16 cast in the test? |
|
@regisss Hi, the purpose of this test case is to compare the output of pipelines using fp16 dtype( |
sayakpaul
reviewed
Oct 27, 2025
Comment on lines
+1449
to
+1451
| if hasattr(component, "half"): | ||
| # Although all components for pipe_loaded should be float16 now, some submodules still use fp32, like in https://github.com/huggingface/transformers/blob/v4.57.1/src/transformers/models/t5/modeling_t5.py#L783, so we need to do the conversion again manally to align with the datatype we use in pipe exactly | ||
| component = component.to(torch_device).half() |
There was a problem hiding this comment.
This doesn't seem right at all. torch_dtype should be able to take care of it. I just ran it on my GPU for SD and it worked fine.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When we run unit test like
pytest -rA tests/pipelines/wan/test_wan_22.py::Wan22PipelineFastTests::test_save_load_float16, we found that the pipeline runs w/ all fp16 datatype, but after save and reload, some parts of text-encoder inpipe_loadeduses fp32, although we set torch_dtype to fp16 explicitly. Deep investigation found that the root cause is here: L783. Here we made an adjustment to the test case to manually add thecomponent = component.to(torch_device).half()operation to align excatly with the behavior inpipe