# AttributeError: 'list' object has no attribute '\_\_module\_\_' when loading model from file system with from\_pretrained

**URL:** <https://discuss.huggingface.co/t/attributeerror-list-object-has-no-attribute-module-when-loading-model-from-file-system-with-from-pretrained/81260>\
**Category:** 🧨 Diffusers\
**Created:** [April 11, 2024, 1:41pm UTC](https://discuss.huggingface.co/t/attributeerror-list-object-has-no-attribute-module-when-loading-model-from-file-system-with-from-pretrained/81260 "2024-04-11T13:41:47Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![dferguson992](https://avatars.discourse-cdn.com/v4/letter/d/2bfe46/32.png) [@dferguson992](https://discuss.huggingface.co/u/dferguson992)\
**Post date:** [April 11, 2024, 1:41pm UTC](https://discuss.huggingface.co/t/attributeerror-list-object-has-no-attribute-module-when-loading-model-from-file-system-with-from-pretrained/81260/1 "2024-04-11T13:41:47Z")

</div>

Original issue posted [here](https://github.com/huggingface/diffusers/issues/7612).

Essentially I’m loading & saving the stable cascade pipeline from HF using the following code snippet:

```auto
model = StableCascadeCombinedPipeline.from_pretrained(HF_MODEL_ID, 
                                                      variant="bf16",
                                                      torch_dtype=torch.float16,
                                                      cache_dir=CACHE_DIR)
model.save_pretrained(save_directory="model/", from_pt=True)

```

This works great until I try to load that model from my file system:

```auto
model = StableCascadeCombinedPipeline.from_pretrained("model/", 
                                                      variant="bf16",
                                                      torch_dtype=torch.float16,
                                                      local_files_only=True)
Loading pipeline components...: 100%
 9/9 [00:05<00:00, 2.32it/s]

---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[12], line 1
----> 1 model = StableCascadeCombinedPipeline.from_pretrained("model/", 
      2 variant="bf16",
      3 torch_dtype=torch.float16,
      4 local_files_only=True)

File /opt/conda/lib/python3.10/site-packages/huggingface_hub/utils/_validators.py:119, in validate_hf_hub_args.<locals>._inner_fn(*args, **kwargs)
    116 if check_use_auth_token:
    117 kwargs = smoothly_deprecate_use_auth_token(fn_name=fn. __name__ , has_token=has_token, kwargs=kwargs)
--> 119 return fn(*args, **kwargs)

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/pipeline_utils.py:910, in DiffusionPipeline.from_pretrained(cls, pretrained_model_name_or_path, **kwargs)
    905 raise ValueError(
    906 f"Pipeline {pipeline_class} expected {expected_modules}, but only {passed_modules} were passed."
    907 )
    909 # 8. Instantiate the pipeline
--> 910 model = pipeline_class(**init_kwargs)
    912 # 9. Save where the model was instantiated from
    913 model.register_to_config(_name_or_path=pretrained_model_name_or_path)

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/stable_cascade/pipeline_stable_cascade_combined.py:88, in StableCascadeCombinedPipeline. __init__ (self, tokenizer, text_encoder, decoder, scheduler, vqgan, prior_prior, prior_text_encoder, prior_tokenizer, prior_scheduler, prior_feature_extractor, prior_image_encoder)
     72 def __init__ (
     73 self,
     74 tokenizer: CLIPTokenizer,
   (...)
     84 prior_image_encoder: Optional[CLIPVisionModelWithProjection] = None,
     85 ):
     86 super(). __init__ ()
---> 88 self.register_modules(
     89 text_encoder=text_encoder,
     90 tokenizer=tokenizer,
     91 decoder=decoder,
     92 scheduler=scheduler,
     93 vqgan=vqgan,
     94 prior_text_encoder=prior_text_encoder,
     95 prior_tokenizer=prior_tokenizer,
     96 prior_prior=prior_prior,
     97 prior_scheduler=prior_scheduler,
     98 prior_feature_extractor=prior_feature_extractor,
     99 prior_image_encoder=prior_image_encoder,
    100 )
    101 self.prior_pipe = StableCascadePriorPipeline(
    102 prior=prior_prior,
    103 text_encoder=prior_text_encoder,
   (...)
    107 feature_extractor=prior_feature_extractor,
    108 )
    109 self.decoder_pipe = StableCascadeDecoderPipeline(
    110 text_encoder=text_encoder,
    111 tokenizer=tokenizer,
   (...)
    114 vqgan=vqgan,
    115 )

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/pipeline_utils.py:154, in DiffusionPipeline.register_modules(self, **kwargs)
    152 register_dict = {name: (None, None)}
    153 else:
--> 154 library, class_name = _fetch_class_library_tuple(module)
    155 register_dict = {name: (library, class_name)}
    157 # save model index config

File /opt/conda/lib/python3.10/site-packages/diffusers/pipelines/pipeline_loading_utils.py:488, in _fetch_class_library_tuple(module)
    486 # register the config from the original module, not the dynamo compiled one
    487 not_compiled_module = _unwrap_model(module)
--> 488 library = not_compiled_module. __module__.split(".")[0]
    490 # check if the module is a pipeline module
    491 module_path_items = not_compiled_module. __module__.split(".")

AttributeError: 'list' object has no attribute ' __module__'

```

I’m not sure where I’m going wrong here, some suggested workarounds were to modify `src/diffusers/pipelines/pipeline_loading_utils.py` but I’d rather do this by the book.