fix: 'PaintByExampleImageEncoder' object has no attribute 'all_tied_w…

[kaixuanliu]

@yiyixuxu and @asomoza, pls help review, thx! This PR fixes bug for following sample code:

import torch
from diffusers import PaintByExamplePipeline
from PIL import Image, ImageDraw

# Load the pipeline
pipe = PaintByExamplePipeline.from_pretrained(
    "Fantasy-Studio/Paint-by-Example",
    torch_dtype=torch.float16,
)
pipe.to("cuda")

# 1. Source image (the image to be edited)
image = Image.new("RGB", (512, 512), color="white")

# 2. Mask image (white = replace, black = keep)
mask = Image.new("RGB", (512, 512), color="black")
draw = ImageDraw.Draw(mask)
draw.rectangle([128, 128, 384, 384], fill="white")

# 3. Example image (reference content to paint in masked area)
example_image = Image.new("RGB", (256, 256), color="blue")

# Run inference
output = pipe(
    image=image,
    mask_image=mask,
    example_image=example_image,
    num_inference_steps=30,
).images[0]

output.save("output.png")

Before this PR, it will throw error:

Traceback (most recent call last):
  File "/root/upstream/diffusers/test.py", line 6, in <module>                                                                     pipe = PaintByExamplePipeline.from_pretrained(                                                                               File "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_validators.py", line 89, in _inner_fn
    return fn(*args, **kwargs)
  File "/root/upstream/diffusers/src/diffusers/pipelines/pipeline_utils.py", line 1049, in from_pretrained
    loaded_sub_model = load_sub_model(                                                                                           File "/root/upstream/diffusers/src/diffusers/pipelines/pipeline_loading_utils.py", line 885, in load_sub_model
    loaded_sub_model = load_method(os.path.join(cached_folder, name), **loading_kwargs)
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py", line 4138, in from_pretrained
    loading_info = cls._finalize_model_loading(model, load_config, loading_info)
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py", line 4279, in _finalize_model_loading
    model.mark_tied_weights_as_initialized(loading_info)
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py", line 4598, in mark_tied_weights_as_initialized
    for tied_param in self.all_tied_weights_keys.keys():
  File "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/module.py", line 1965, in __getattr__
    raise AttributeError(
AttributeError: 'PaintByExampleImageEncoder' object has no attribute 'all_tied_weights_keys'. Did you mean: '_tied_weights_keys
'?

[kaixuanliu]

And the same case w/ LDMBertModel