diffusers

diffusers copied to clipboard

addresses https://github.com/huggingface/diffusers/issues/12776 # What does this PR do? This PR keeps the tuples, but moves the splitting from tensors into tuples of tensors to the transformer blocks, to avoid issues...

dxqb

How about forcing the first and last block on device when groupoffloading is used?

13

**Is your feature request related to a problem? Please describe.** When group offloading is enabled, the offload and onload cannot be streamed between steps and this is really a big...

seed93

Fix prompt isolation tests for QwenImage edit and edit plus

1

# What does this PR do? Passing prompt embeds to the pipeline input class is an important use case and the corresponding tests were disabled for QwenImage Edit and QwenImage...

sayakpaul

_flash_attention_3 in dispatch_attention_fn is not compatible with the latest flash-atten interface.

6

### Describe the bug [FA3] Don't return lse: https://github.com/Dao-AILab/flash-attention/commit/ed209409acedbb2379f870bbd03abce31a7a51b7 but in the current diffuser version, it is not updated. https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_dispatch.py#L608 when use fa3 backend, diffuser will raise error. ### Reproduction...

hmzjwhmzjw

Qwen Image prompt encoding is not padding to max seq len

19

### Describe the bug The pipeline method for QwenImagePipeline.encode_prompts is not padding correctly; it's padding by the longest sequence length in the batch, which leaves very very short embeds that...

bghira

Fix qwen image prompt padding #12075

3

# Fix QwenImage Prompt Embedding Padding for Deterministic Outputs ## What was the issue? Issue #12075 reported that QwenImage pipelines were producing non-deterministic outputs when using the same prompt across...

sambhavnoobcoder

I see convert_wan_to_diffusers.py in diffusers repo, but no convert_diffusers_to_wan.py. Do you have plan to upload a convert scripts?

wikiwen

### Describe the bug Hi, I might be misunderstanding something here — I’m not very experienced with submitting issues to Diffusers yet, so please feel free to correct me if...

scp3125

i think there is something wrong with new/latest scripts. RuntimeError: mat1 and mat2 shapes cannot be multiplied (2x1536 and 768x3072)

2

### Describe the bug i got "RuntimeError: mat1 and mat2 shapes cannot be multiplied (2x1536 and 768x3072)" when i ran "train_dreambooth_lora_flux_advanced.py" from the latest version of diffusers or v0.35.1 but...

gerylavin

fix torchao memory check

6

The test `pytest -rA tests/quantization/torchao/test_torchao.py::TorchAoTest::test_model_memory_usage` failed with ``` > assert unquantized_model_memory / quantized_model_memory >= expected_memory_saving_ratio E assert (1416704 / 1382912) >= 2.0 tests/quantization/torchao/test_torchao.py:512: AssertionError ``` on A100. I guess it...

jiqing-feng