Having error while running train_extractive_reader.py

[tkabir1]

When I am running
python3 train_extractive_reader.py encoder.sequence_length=350 train_files="./prediction_copy.json" dev_files="./prediction.json" gold_passages_src="./gold_passages_info/nq_dev.json" gold_passages_src_dev="./gold_passages_info/nq_dev_dev.json" output_dir="./test_reader/"
I am having the error:

input_ids = torch.stack([t for t in positives_selected + negatives_selected], dim=0)
RuntimeError: stack expects a non-empty TensorList

Can anyone please help me with that?

[gaishun]

Hi tkabir1 , I have the same problem.
By the way, are you using the same train_files and dev_files ? ( Actually, I do this.)