[Python] huggingface transformer models: KeyError: 'input_ids' message at beginning of BERT...

Using the Huggingface transformer library, I am encountering a bug in the final step when I go to fine tune the BERT language model for masked language modelling task. I am looking to fine tune it for an in domain finance corpus that model has not been trained on yet. However, I get the following error message when I call the model to train: KeyError: 'input_ids'. Provided below are the steps and code I took. Any insights are appreciated!

First, I created a dataset object from a pandas dataframe that was in turn created from a csv file with 1 column of many rows of text:

unlabelled_dataset = Dataset.from_pandas(unlabelled)


Second, I then tokenized the dataset with the following code:

tokenizerBERT = BertTokenizerFast.from_pretrained('bert-base-uncased') #BERT model tokenization & check
tokenizerBERT(unlabelled_dataset['paragraphs'], padding=True, truncation=True)
tokenizerBERT.save_pretrained('tokenizers/pytorch/labelled/BERT/')


Third, I created a data collator as instructed:

data_collator_BERT = DataCollatorForLanguageModeling(tokenizer=tokenizerBERT, mlm=True, mlm_probability=0.15)


Next, I select my model from_pretrained to get the benefits of transfer learning:

model_BERT = BertForMaskedLM.from_pretrained("bert-base-uncased")


Next, I passed my training args to the transformer trainer and initialize:

training_args_BERT = TrainingArguments(
output_dir="./BERT",
num_train_epochs=10,
evaluation_strategy='steps',
warmup_steps=10000,
weight_decay=0.01,
per_gpu_train_batch_size=64,
)

trainer_BERT = Trainer(
model=model_BERT,
args=training_args_BERT,
data_collator=data_collator_BERT,
train_dataset=unlabelled_dataset,
)


Last, I call the model to train and get the error KeyError: 'input_ids'

trainer_BERT.train()


Any insights on how to debug this approach to training the model?

Provided below is the exact error message received:

---------------------------------------------------------------------------
KeyError Traceback (most recent call last)
<ipython-input-9-83b7063dea0b> in <module>
----> 1 trainer_BERT.train()
2 trainer.save_model("./models/royalBERT")

~/anaconda3/lib/python3.7/site-packages/transformers/trainer.py in train(self, model_path, trial)
755 self.control = self.callback_handler.on_epoch_begin(self.args, self.state, self.control)
756
--> 757 for step, inputs in enumerate(epoch_iterator):
758
759 # Skip past any already trained steps if resuming training

~/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py in __next__(self)
361
362 def __next__(self):
--> 363 data = self._next_data()
364 self._num_yielded += 1
365 if self._dataset_kind == _DatasetKind.Iterable and \

~/anaconda3/lib/python3.7/site-packages/torch/utils/data/dataloader.py in _next_data(self)
401 def _next_data(self):
402 index = self._next_index() # may raise StopIteration
--> 403 data = self._dataset_fetcher.fetch(index) # may raise StopIteration
404 if self._pin_memory:
405 data = _utils.pin_memory.pin_memory(data)

~/anaconda3/lib/python3.7/site-packages/torch/utils/data/_utils/fetch.py in fetch(self, possibly_batched_index)
45 else:
46 data = self.dataset[possibly_batched_index]
---> 47 return self.collate_fn(data)

~/anaconda3/lib/python3.7/site-packages/transformers/data/data_collator.py in __call__(self, examples)
193 ) -> Dict[str, torch.Tensor]:
194 if isinstance(examples[0], (dict, BatchEncoding)):
--> 195 examples = [e["input_ids"] for e in examples]
196 batch = self._tensorize_batch(examples)
197 if self.mlm:

~/anaconda3/lib/python3.7/site-packages/transformers/data/data_collator.py in <listcomp>(.0)
193 ) -> Dict[str, torch.Tensor]:
194 if isinstance(examples[0], (dict, BatchEncoding)):
--> 195 examples = [e["input_ids"] for e in examples]
196 batch = self._tensorize_batch(examples)
197 if self.mlm:

KeyError: 'input_ids'

Continue reading...