[SD3 LoRA] Fix list index out of range #8584
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. |
asomoza
changed the title
|
@sayakpaul is this test case good enough? if it is, I can upload the loras to the internal testing repo. I ran the test locally and it passes with the PR and fails without it. |
tests/lora/test_lora_layers_sd3.py
Outdated
| def test_sd3_lora(self): | ||
| components = self.get_dummy_components() | ||
|
|
||
| pipe = self.pipeline_class(**components) | ||
| pipe = pipe.to(torch_device) | ||
| pipe.set_progress_bar_config(disable=None) | ||
|
|
||
| lora_model_id = "OzzyGT/lora_test" | ||
| lora_filename = "lora_diffusers_format.safetensors" | ||
| pipe.load_lora_weights(lora_model_id, weight_name=lora_filename) | ||
| pipe.unload_lora_weights() | ||
|
|
||
| lora_model_id = "OzzyGT/lora_test" | ||
| lora_filename = "lora_peft_format.safetensors" | ||
| pipe.load_lora_weights(lora_model_id, weight_name=lora_filename) |
There was a problem hiding this comment.
Thanks for adding the tests.
We should perhaps make this a SLOW one or require this test to have a GPU (@require_torch_gpu). And also add a comment to explain why we had to add this test. I think adding a mention of this PR is just fine.
|
any reason why this was not merged and included in diffusers==0.29.1? |
|
It left some crucial checks and tests. |
|
@sayakpaul I don't have write permissions in Just in case, these are tiny loras (2kb) that I made using the |
| "vae": vae, | ||
| } | ||
|
|
||
| def test_sd3_lora(self): |
There was a problem hiding this comment.
@asomoza I think just this test could be marked with "require_torch_gpu" and be made a part of this class:
I think that will be good enough.
Once this is done, feel free to merge the PR :)
There was a problem hiding this comment.
Just a single comment: https://github.com/huggingface/diffusers/pull/8584/files#r1649020147
* fix * add check * key present is checked before * test case draft * aply suggestions * changed testing repo, back to old class * forgot docstring --------- Co-authored-by: YiYi Xu <[email protected]> Co-authored-by: Sayak Paul <[email protected]>
What does this PR do?
For some reason, the LoRA state_dict gets saved with both formats, PEFT and Diffusers, this PR add the
convert_unet_state_dict_to_peftto the loading function so it works with both.Fixes #8579 #8565
Who can review?
Anyone in the community is free to review