[UNet_Spatio_Temporal_Condition] fix default num_attention_heads in unet_spatio_temporal_condition #7205
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Outdated suggestions cannot be applied.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.
What does this PR do?
Since
UNet_Spatio_Temporal_Condition
file was designed for Stable Video Diffusion, the default parameternum_attention_heads
should be[5, 10, 20, 20]
as the SVD config shows. If people initialize theUNetSpatioTemporalConditionModel
class byself.__init__()
rather than loading byload_pretrained(config)
using pretrained config file, they can not realize that they have initialized the differentAttention Head
, and also there is no warning or error when load pretrained weights into differentAttention Head
, causing the wrong attention output.We should avoid the situation that will cause the wrong attention output. This PR does that.
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.