Missing 'text'.py in load_dataset in run.py

[jefflink]

I tried running the examples given but for the bash run_xx.sh which uses run.py, it fails in the following line

datasets = load_dataset("text.py", data_files={'train': data_args.train_file,
                                                'validation': data_args.validation_file,
                                                'test': data_args.test_file})

Is there a missing text.py somewhere?

[cathyry]

I met the same trouble with you , and author told me :"Hi，it seems that the file cannot be downloaded automatically from huggingface in your device. You can manually download it, and put it in the folder." However , this file couldn't be find in huggingface .

[jefflink]

@cathyry After looking through the code, I believed the author was using an old version of datasets and hence using some custom script. It can now be replaced with just the following:

datasets = load_dataset("text", data_files={'train': data_args.train_file,
                                                'validation': data_args.validation_file,
                                                'test': data_args.test_file})

I have tested and it seems to be running ok so far.

[cathyry]

@jefflink It works , thanks a lot .