ledits_pp model/pipeline review

# `ledits_pp` model/pipeline review

Commit tested: `0f1abc4ae8b0eb2a3b40e82a310507281144c423`

Review performed against the repository review rules.

Coverage checked: fast and slow test classes exist for both SD and SDXL LEdits++; fast tests passed locally with `.venv` (`3 passed, 1 skipped` for each file). Slow tests exist but were not run. Duplicate search was run with `gh search issues/prs`; no duplicates found for the issues below. Existing open issue https://github.com/huggingface/diffusers/issues/8826 covers a separate known LEdits++ empty attention-store crash, so I am not presenting that one as new.

## Issue 1: `LEditsPPInversionPipelineOutput` is missing from lazy exports

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/__init__.py#L24-L27

Problem:
The lazy import structure exports `LEditsPPDiffusionPipelineOutput` twice and omits `LEditsPPInversionPipelineOutput`. The eager `TYPE_CHECKING` path imports both outputs, so behavior differs between slow/eager imports and normal lazy imports. `diffusers.pipelines.__init__` also imports these outputs in the eager path but does not expose them in the lazy import structure.

Impact:
Public imports documented under `pipelines.ledits_pp` fail at runtime for the inversion output class.

Reproduction:
```python
from diffusers.pipelines.ledits_pp import LEditsPPInversionPipelineOutput
# AttributeError: module diffusers.pipelines.ledits_pp has no attribute LEditsPPInversionPipelineOutput
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_output.py#L27-L42

Suggested fix:
```python
_import_structure["pipeline_output"] = [
"LEditsPPDiffusionPipelineOutput",
"LEditsPPInversionPipelineOutput",
]
```

Also either add both output classes to `src/diffusers/pipelines/__init__.py` lazy exports or remove the eager-only imports there.

## Issue 2: SDXL editing after batched inversion has inconsistent embedding batch sizes

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L538-L558
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1112-L1121

Problem:
`LEditsPPPipelineStableDiffusionXL.encode_prompt()` encodes edit concepts once per concept, not once per `(concept, image)` pair. After batched inversion, latents are repeated by `1 + enabled_editing_prompts`, but edit prompt embeddings and pooled embeddings are not expanded the same way.

Impact:
SDXL LEdits++ cannot edit a batch of inverted images with multiple edit prompts. The failure is currently untested because fast tests cover batched inversion only, not batched editing.

Reproduction:
```python
from diffusers import LEditsPPPipelineStableDiffusionXL
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion_xl import LEditsPPPipelineStableDiffusionXLFastTests

case = LEditsPPPipelineStableDiffusionXLFastTests()
pipe = LEditsPPPipelineStableDiffusionXL(**case.get_dummy_components())
pipe.set_progress_bar_config(disable=True)

inputs = case.get_dummy_inversion_inputs("cpu")
inputs["num_inversion_steps"] = 2
inputs["skip"] = 0.0
pipe.invert(**inputs)

pipe(editing_prompt=["wearing glasses", "sunshine"], output_type="latent")
# RuntimeError: mat1 and mat2 shapes cannot be multiplied ...
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L632-L645

Suggested fix:
Expand edit concept embeddings per image in `__call__` or `encode_prompt`, while keeping `num_edit_tokens` indexed per concept. The implementation should preserve the inversion path, where `editing_prompt` is reused as the source prompt and already represents one prompt per image.

## Issue 3: SDXL non-square image sizes are swapped for micro-conditioning

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1574-L1576
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1103-L1110

Problem:
`invert()` computes `width` from latent height and `height` from latent width, then stores `self.size = (height, width)`. For non-square images, SDXL added time IDs receive swapped original/target sizes. The `target_size` argument accepted by `__call__` is also ignored; the call always uses `self.size`.

Impact:
Non-square SDXL edits are conditioned on the wrong dimensions, which can degrade output and makes the documented `target_size` parameter ineffective.

Reproduction:
```python
from diffusers import LEditsPPPipelineStableDiffusionXL
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion_xl import LEditsPPPipelineStableDiffusionXLFastTests

case = LEditsPPPipelineStableDiffusionXLFastTests()
pipe = LEditsPPPipelineStableDiffusionXL(**case.get_dummy_components())
pipe.set_progress_bar_config(disable=True)

inputs = case.get_dummy_inversion_inputs("cpu")
inputs["image"] = inputs["image"][0].resize((64, 32))
inputs.update({"height": 32, "width": 64, "num_inversion_steps": 2, "skip": 0.0})
pipe.invert(**inputs)

print(pipe.size) # (64, 32), expected (32, 64)
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L1030-L1031

Suggested fix:
```python
height = x0.shape[-2] * self.vae_scale_factor
width = x0.shape[-1] * self.vae_scale_factor
self.size = (height, width)
```

Then use `target_size = target_size or self.size` in `__call__`.

## Issue 4: Callback tensor allowlists include names that are not in scope

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L301-L304
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L1235-L1243
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L336-L344
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1016-L1026

Problem:
The SD pipeline allows `prompt_embeds`, but the denoising loop has no `prompt_embeds` local. The SDXL pipeline allows `negative_add_time_ids`, but that local is commented out and never defined; SDXL also comments out its `check_inputs()` call, so invalid callback tensor names are not rejected early.

Impact:
Documented callback customization crashes with `KeyError` instead of either passing the requested tensor or raising the normal validation error.

Reproduction:
```python
from diffusers import LEditsPPPipelineStableDiffusion
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion import LEditsPPPipelineStableDiffusionFastTests

case = LEditsPPPipelineStableDiffusionFastTests()
pipe = LEditsPPPipelineStableDiffusion(**case.get_dummy_components())
pipe.set_progress_bar_config(disable=True)

inputs = case.get_dummy_inversion_inputs("cpu")
inputs.update({"image": inputs["image"][0], "num_inversion_steps": 1, "skip": 0.0})
pipe.invert(**inputs)

def cb(pipe, step, timestep, kwargs):
return kwargs

pipe(
editing_prompt="cat",
output_type="latent",
use_intersect_mask=False,
callback_on_step_end=cb,
callback_on_step_end_tensor_inputs=["prompt_embeds"],
)
# KeyError: 'prompt_embeds'
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L232-L237
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L636-L640

Suggested fix:
Make `_callback_tensor_inputs` match actual denoising-loop locals, call `check_inputs()` in SDXL, and add focused fast tests for every allowed callback tensor.

## Issue 5: SDXL IP-Adapter path calls the VAE image encoder instead of IP-Adapter encoding

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1123-L1128
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1440-L1473

Problem:
`__call__()` accepts `ip_adapter_image`, but calls `self.encode_image(ip_adapter_image, device, num_images_per_prompt)`. In this pipeline, `encode_image()` is the VAE inversion helper, not the CLIP/IP-Adapter image encoder from SDXL. The positional arguments are interpreted as `dtype` and `height`.

Impact:
The advertised IP-Adapter path is unusable and fails before preparing image embeddings.

Reproduction:
```python
from diffusers import LEditsPPPipelineStableDiffusionXL
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion_xl import LEditsPPPipelineStableDiffusionXLFastTests

case = LEditsPPPipelineStableDiffusionXLFastTests()
pipe = LEditsPPPipelineStableDiffusionXL(**case.get_dummy_components())
pipe.set_progress_bar_config(disable=True)

inputs = case.get_dummy_inversion_inputs("cpu")
inputs.update({"image": inputs["image"][0], "num_inversion_steps": 1, "skip": 0.0})
pipe.invert(**inputs)

pipe(editing_prompt="cat", ip_adapter_image=inputs["image"], output_type="latent")
# ValueError: height and width must be > 0
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L522-L590
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion_xl/pipeline_stable_diffusion_xl.py#L1153-L1160

Suggested fix:
Rename the VAE helper to something like `encode_vae_image()`, restore/copy SDXL’s `encode_image()` and `prepare_ip_adapter_image_embeds()`, add `ip_adapter_image_embeds`, and include `image_encoder` in the offload sequence.

## Issue 6: `invert()` and cross-attention masking permanently replace user attention processors

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L497-L519
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L1343-L1344
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L811-L834
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1543-L1544

Problem:
`invert()` unconditionally calls `self.unet.set_attn_processor(AttnProcessor())`. `prepare_unet()` also replaces all processors with either `LEDITSCrossAttnProcessor` or plain `AttnProcessor()`. Neither path restores the original processors.

Impact:
Any configured attention backend or custom processor, including `AttnProcessor2_0`, xFormers-style processors, LoRA/IP-Adapter processors, or user-supplied processors, is silently discarded after inversion or masked editing.

Reproduction:
```python
from diffusers import LEditsPPPipelineStableDiffusion
from diffusers.models.attention_processor import AttnProcessor2_0
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion import LEditsPPPipelineStableDiffusionFastTests

case = LEditsPPPipelineStableDiffusionFastTests()
pipe = LEditsPPPipelineStableDiffusion(**case.get_dummy_components())
pipe.unet.set_attn_processor(AttnProcessor2_0())

before = {type(p).__name__ for p in pipe.unet.attn_processors.values()}

inputs = case.get_dummy_inversion_inputs("cpu")
inputs.update({"image": inputs["image"][0], "num_inversion_steps": 1, "skip": 0.0})
pipe.set_progress_bar_config(disable=True)
pipe.invert(**inputs)

after = {type(p).__name__ for p in pipe.unet.attn_processors.values()}
print(before, after) # {'AttnProcessor2_0'} {'AttnProcessor'}
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/models/attention_processor.py#L2696-L2786

Suggested fix:
Store the original `unet.attn_processors` before installing LEdits processors and restore them in a `finally` block after the temporary operation. For non-intercepted attention layers, preserve the existing processor instead of replacing it with bare `AttnProcessor()`.

## Issue 7: CPU offload paths use `self.device` instead of `_execution_device`

Affected code:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L948-L995
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion.py#L1388-L1398
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1010-L1014
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1459-L1467
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/ledits_pp/pipeline_leditspp_stable_diffusion_xl.py#L1650-L1655

Problem:
Both pipelines declare `model_cpu_offload_seq`, but many runtime tensors are moved to `self.device`. Under `enable_model_cpu_offload()`, `self.device` can remain CPU while `_execution_device` is the actual accelerator. Standard pipelines use `_execution_device` at call time for this reason.

Impact:
CPU offload can put prompts, latents, masks, smoothing kernels, and inversion noise on the wrong device, causing device mismatches or unexpectedly running heavy work on CPU.

Reproduction:
```python
import torch
from diffusers import LEditsPPPipelineStableDiffusionXL
from tests.pipelines.ledits_pp.test_ledits_pp_stable_diffusion_xl import LEditsPPPipelineStableDiffusionXLFastTests

if not torch.cuda.is_available():
raise SystemExit("requires CUDA to exercise model CPU offload")

case = LEditsPPPipelineStableDiffusionXLFastTests()
pipe = LEditsPPPipelineStableDiffusionXL(**case.get_dummy_components())
pipe.enable_model_cpu_offload()
print(pipe.device, pipe._execution_device) # CPU vs CUDA under offload

inputs = case.get_dummy_inversion_inputs("cpu")
inputs.update({"image": inputs["image"][0], "num_inversion_steps": 2, "skip": 0.0})
pipe.invert(**inputs)
pipe(editing_prompt="cat", output_type="latent", use_cross_attn_mask=True)
```

Relevant precedent:
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L955-L955
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_text2img.py#L786-L786
https://github.com/huggingface/diffusers/blob/0f1abc4ae8b0eb2a3b40e82a310507281144c423/src/diffusers/pipelines/latent_consistency_models/pipeline_latent_consistency_img2img.py#L849-L849

Suggested fix:
At the start of `__call__()` and `invert()`, set `device = self._execution_device` and use that local for tensor movement/allocation. Pass `device` into VAE image encoding helpers instead of reading `self.device` inside them.

Start with the affected LEdits++ pipeline files and the fast tests in tests/pipelines/ledits_pp/test_ledits_pp_stable_diffusion.py and test_ledits_pp_stable_diffusion_xl.py. Reproduce each reported import, batching, sizing, callback, IP-Adapter, and attention-processor case separately, then add focused regression coverage and run the relevant fast test classes. Done means the reported failures are covered and the existing inversion and editing tests still pass.

Escrito por el modelo de indexación a partir del texto del issue.