I am not able to set global attention mask. I have although given two sep tokens between question and context

outputs = model(**inputs)
File "C:\Users\Rudraksh\Anaconda3\envs\bert_pre\lib\site-packages\torch\nn\modules\module.py", line 547, in __call__
result = self.forward(*input, **kwargs)
File "d:\transformers\src\transformers\modeling_longformer.py", line 1345, in forward
global_attention_mask = _compute_global_attention_mask(input_ids, self.config.sep_token_id)
File "d:\transformers\src\transformers\modeling_longformer.py", line 84, in _compute_global_attention_mask
question_end_index = _get_question_end_index(input_ids, sep_token_id)
File "d:\transformers\src\transformers\modeling_longformer.py", line 73, in _get_question_end_index
), f"There should be exactly three separator tokens: {sep_token_id} in every sample for questions answering. You might also consider to set `global_attention_mask` manually in the forward function to avoid this error."
AssertionError: There should be exactly three separator tokens: 2 in every sample for questions answering. You might also consider to set `global_attention_mask` manually in the forward function to avoid this error.