ValueError: a must be non-empty?

Question

ValueError: a must be non-empty?

LINFF1023 opened this issue 2 years ago · comments

LINBOBO commented 2 years ago

There was an error running step 6.
why?
Looking forward to the author's reply!!!

Xingtong Liu · Answer 1 · Tue Sep 13 2022 01:04:05 GMT+0800 (China Standard Time)

Could you provide more details on this issue?

LINBOBO · Answer 2 · Tue Sep 13 2022 13:32:01 GMT+0800 (China Standard Time)

Could you provide more details on this issue?

when i python3 training.py --config_path "configs/training.json", this happened.

Xingtong Liu · Answer 3 · Tue Sep 13 2022 15:56:22 GMT+0800 (China Standard Time)

This means the dataset class couldn’t find hdf5 data. Please make sure you download hdf5 data to the correct path

self.sequence_hdf5_path_list = sorted( list(self.data_root.rglob(hdf5_pattern))

LINBOBO · Answer 4 · Tue Sep 13 2022 16:05:18 GMT+0800 (China Standard Time)

This means the dataset class couldn’t find hdf5 data. Please make sure you download hdf5 data to the correct path

self.sequence_hdf5_path_list = sorted( list(self.data_root.rglob(hdf5_pattern))

i think so too...thanks ！！

LINBOBO · Answer 5 · Wed Sep 14 2022 13:55:30 GMT+0800 (China Standard Time)

Could you provide more details on this issue?

I'm sorry to bother you again.I'm a newbie but I really want to run your code successfully.I really don't understand this problem but my classmate said it might be the version problem(torch).

Xingtong Liu · Answer 6 · Wed Sep 14 2022 15:32:26 GMT+0800 (China Standard Time)

Could you try updating the Dockerfile with the latest one and try the instructions starting from building docker image again? It could be that the PyTorch version issue. I didn’t test thoroughly before updating the PyTorch base image a couple of days ago.