# Unable to load image using llama-index

**URL:** <https://discuss.huggingface.co/t/unable-to-load-image-using-llama-index/38304>\
**Category:** 🤗Tokenizers\
**Created:** [May 1, 2023, 1:44pm UTC](https://discuss.huggingface.co/t/unable-to-load-image-using-llama-index/38304 "2023-05-01T13:44:57Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![itsamamir](https://avatars.discourse-cdn.com/v4/letter/i/eb9ed0/32.png) [@itsamamir](https://discuss.huggingface.co/u/itsamamir)\
**Post date:** [May 1, 2023, 1:44pm UTC](https://discuss.huggingface.co/t/unable-to-load-image-using-llama-index/38304/1 "2023-05-01T13:44:57Z")

</div>

Hi everyone,  
i am trying to load an image using llama-index but got the error below. Does anyone have any clue?

```auto
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
File C:\ProgramData\anaconda3\lib\site-packages\transformers\feature_extraction_utils.py:164, in BatchFeature.convert_to_tensors(self, tensor_type)
    163 if not is_tensor(value):
--> 164 tensor = as_tensor(value)
    166 self[key] = tensor

File C:\ProgramData\anaconda3\lib\site-packages\transformers\feature_extraction_utils.py:146, in BatchFeature.convert_to_tensors.<locals>.as_tensor(value)
    145 value = np.array(value)
--> 146 return torch.tensor(value)

RuntimeError: Could not infer dtype of numpy.float32

During handling of the above exception, another exception occurred:

ValueError Traceback (most recent call last)
Cell In[5], line 43
     41 ImageReader = download_loader("ImageReader")
     42 loader = ImageReader(text_type = "key_value")
---> 43 documents = loader.load_data(file=Path(path))
     45 index = GPTSimpleVectorIndex(documents,service_context)
     47 question = "what is the total amount of invoice?"

File C:\ProgramData\anaconda3\lib\site-packages\llama_index\readers\llamahub_modules/file/image/base.py:88, in ImageReader.load_data(self, file, extra_info)
     83 task_prompt = "<s_cord-v2>"
     84 decoder_input_ids = processor.tokenizer(
     85 task_prompt, add_special_tokens=False, return_tensors="pt",padding=True
     86 ).input_ids
---> 88 pixel_values = processor(image, return_tensors="pt").pixel_values
     90 outputs = model.generate(
     91 pixel_values.to(device),
     92 decoder_input_ids=decoder_input_ids.to(device),
   (...)
    100 return_dict_in_generate=True,
    101 )
    103 sequence = processor.batch_decode(outputs.sequences)[0]

File C:\ProgramData\anaconda3\lib\site-packages\transformers\models\donut\processing_donut.py:69, in DonutProcessor. __call__ (self, *args, **kwargs)
     66 raise ValueError("You need to specify either an `images` or `text` input to process.")
     68 if images is not None:
---> 69 inputs = self.feature_extractor(images, *args, **kwargs)
     70 if text is not None:
     71 encodings = self.tokenizer(text, **kwargs)

File C:\ProgramData\anaconda3\lib\site-packages\transformers\models\donut\feature_extraction_donut.py:209, in DonutFeatureExtractor. __call__ (self, images, return_tensors, random_padding, **kwargs)
    207 # return as BatchFeature
    208 data = {"pixel_values": images}
--> 209 encoded_inputs = BatchFeature(data=data, tensor_type=return_tensors)
    211 return encoded_inputs

File C:\ProgramData\anaconda3\lib\site-packages\transformers\feature_extraction_utils.py:75, in BatchFeature. __init__ (self, data, tensor_type)
     73 def __init__ (self, data: Optional[Dict[str, Any]] = None, tensor_type: Union[None, str, TensorType] = None):
     74 super(). __init__ (data)
---> 75 self.convert_to_tensors(tensor_type=tensor_type)

File C:\ProgramData\anaconda3\lib\site-packages\transformers\feature_extraction_utils.py:170, in BatchFeature.convert_to_tensors(self, tensor_type)
    168 if key == "overflowing_values":
    169 raise ValueError("Unable to create tensor returning overflowing values of different lengths. ")
--> 170 raise ValueError(
    171 "Unable to create tensor, you should probably activate padding "
    172 "with 'padding=True' to have batched tensors with the same length."
    173 )
    175 return self

ValueError: Unable to create tensor, you should probably activate padding with 'padding=True' to have batched tensors with the same length.

```