Inference not working

[peter-ch]

I installed the diffusers package as specified on the HF page, and I use this code for simple inference:

import os
import sys
import torch
from diffusers import StableCascadeDecoderPipeline, StableCascadePriorPipeline

torch.cuda.device_count()
device = 'cpu'
num_images_per_prompt = 2

prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", torch_dtype=torch.bfloat16).to(device)
decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade",  torch_dtype=torch.float16).to(device)

It's the same exact code as in the HF page. It breaks with this error:

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[5], [line 4](vscode-notebook-cell:?execution_count=5&line=4)
      [1](vscode-notebook-cell:?execution_count=5&line=1) prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", torch_dtype=torch.bfloat16, 
      [2](vscode-notebook-cell:?execution_count=5&line=2)                                                    low_cpu_mem_usage=False, ignore_mismatched_sizes=True
      [3](vscode-notebook-cell:?execution_count=5&line=3)                                                    ).to(device)
----> [4](vscode-notebook-cell:?execution_count=5&line=4) decoder = StableCascadeDecoderPipeline.from_pretrained("stabilityai/stable-cascade",  torch_dtype=torch.float16, 
      [5](vscode-notebook-cell:?execution_count=5&line=5)                                                        low_cpu_mem_usage=False, ignore_mismatched_sizes=True
      [6](vscode-notebook-cell:?execution_count=5&line=6)                                                        ).to(device)

File [c:\Python310\lib\site-packages\huggingface_hub\utils\_validators.py:118](file:///C:/Python310/lib/site-packages/huggingface_hub/utils/_validators.py:118), in validate_hf_hub_args.<locals>._inner_fn(*args, **kwargs)
    [115](file:///C:/Python310/lib/site-packages/huggingface_hub/utils/_validators.py:115) if check_use_auth_token:
    [116](file:///C:/Python310/lib/site-packages/huggingface_hub/utils/_validators.py:116)     kwargs = smoothly_deprecate_use_auth_token(fn_name=fn.__name__, has_token=has_token, kwargs=kwargs)
--> [118](file:///C:/Python310/lib/site-packages/huggingface_hub/utils/_validators.py:118) return fn(*args, **kwargs)

File [c:\Python310\lib\site-packages\diffusers\pipelines\pipeline_utils.py:1263](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1263), in DiffusionPipeline.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
   [1260](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1260)     loaded_sub_model = passed_class_obj[name]
   [1261](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1261) else:
   [1262](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1262)     # load sub model
-> [1263](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1263)     loaded_sub_model = load_sub_model(
   [1264](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1264)         library_name=library_name,
   [1265](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1265)         class_name=class_name,
   [1266](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1266)         importable_classes=importable_classes,
   [1267](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1267)         pipelines=pipelines,
   [1268](file:///C:/Python310/lib/site-packages/diffusers/pipelines/pipeline_utils.py:1268)         is_pipeline_module=is_pipeline_module,
...
    [846](file:///C:/Python310/lib/site-packages/diffusers/models/modeling_utils.py:846)     )

RuntimeError: Error(s) in loading state_dict for StableCascadeUnet:
	size mismatch for embedding.1.weight: copying a param with shape torch.Size([320, 16, 1, 1]) from checkpoint, the shape in current model is torch.Size([320, 64, 1, 1]).
	You may consider adding `ignore_mismatched_sizes=True` in the model `from_pretrained` method.

Nothing helps including the suggestions in the error message.

[pabloppp]

The diffusers branch is currently broken, will be fixed very soon, meanwhile you can install it from an older commit:

pip3 install git+https://github.com/kashif/diffusers.git@a3dc21385b7386beb3dab3a9845962ede6765887

[hos3inoo]

wait

[FurkanGozukara]

ye it is hard to make it work

here our developed app check this out : #125