fix pipeline

#4
by not-lain - opened
Tunisia.AI org
TypeError                                 Traceback (most recent call last)
<ipython-input-10-62dd828076db> in <cell line: 1>()
----> 1 pipe("i love you")

3 frames
/usr/local/lib/python3.10/dist-packages/transformers/pipelines/text_classification.py in __call__(self, *args, **kwargs)
    154             If `top_k` is used, one such dictionary is returned per label.
    155         """
--> 156         result = super().__call__(*args, **kwargs)
    157         # TODO try and retrieve it in a nicer way from _sanitize_parameters.
    158         _legacy = "top_k" not in kwargs

/usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py in __call__(self, inputs, num_workers, batch_size, *args, **kwargs)
   1138             )
   1139         else:
-> 1140             return self.run_single(inputs, preprocess_params, forward_params, postprocess_params)
   1141 
   1142     def run_multi(self, inputs, preprocess_params, forward_params, postprocess_params):

/usr/local/lib/python3.10/dist-packages/transformers/pipelines/base.py in run_single(self, inputs, preprocess_params, forward_params, postprocess_params)
   1144 
   1145     def run_single(self, inputs, preprocess_params, forward_params, postprocess_params):
-> 1146         model_inputs = self.preprocess(inputs, **preprocess_params)
   1147         model_outputs = self.forward(model_inputs, **forward_params)
   1148         outputs = self.postprocess(model_outputs, **postprocess_params)

/usr/local/lib/python3.10/dist-packages/transformers/pipelines/text_classification.py in preprocess(self, inputs, **tokenizer_kwargs)
    178                 ' dictionary `{"text": "My text", "text_pair": "My pair"}` in order to send a text pair.'
    179             )
--> 180         return self.tokenizer(inputs, return_tensors=return_tensors, **tokenizer_kwargs)
    181 
    182     def _forward(self, model_inputs):

TypeError: 'NoneType' object is not callable

seems like the model thing has no tokenizer, if it exists, you need to add it in model_thing_whatever.tokenizer

Tunisia.AI org

i just tried that, but it didn't work, i was postponing finishing the pipline because i don't know what do the 4 categories in this model stand for. but i'll just add some place holder for that now.
also i did fix it in 5cf6da561def4284f2c6cbcd7e12c6f791443543 but i disabled it, not sure why.
i'll try to debug this further else this is another method of fixing the pipeline https://huggingface.co/docs/transformers/add_new_pipeline.

Tunisia.AI org

cross that, i'll need to check https://huggingface.co/docs/transformers/add_new_pipeline to make it work

Tunisia.AI org

@parsee-mizuhashi fixed ✅

Tunisia.AI org

false alarm forget about that

Tunisia.AI org

@parsee-mizuhashi
finally 😭, there are some stuff that i don't understand about the original model, but you can now use the model via pipeline ( •̀ ω •́ )y

Sign up or log in to comment