I have a list of BatchEncoding, I need to convert it into a format that Huggingface models accepts it
context_input_tensor =  passage_ids =  while len(passage_ids) < batch_size: idx = random.randint(0, num_questions) question_tensor, passage_id = questions[idx] if passage_id in passage_ids: continue passage_ids.append(passage_id) context_input_tensor.append(passages_dict[passage_id])
context_input_tensor is a list which contains
BatchEncoding type objects. What should I do
If I pass the
context_input_tensor as it is I get a
TypeError: forward() missing 2 required positional arguments: 'query_input_ids' and 'query_attention_mask' I was able to solve the issue by creating two different tensors and stacking
'query_attention_mask'. Is there a better way to do this ?