load_dataset from local squad.py, raise error: TypeError: 'NoneType' object is not callable

[xixiaoyao]

version: 1.0.2

train_dataset  = datasets.load_dataset('squad') 

The above code can works. However, when I download the squad.py from your server, and saved as my_squad.py to local. I run followings raise errors.

train_dataset  = datasets.load_dataset('./my_squad.py')                                                                                                

---

TypeError Traceback (most recent call last)
in
----> 1 train_dataset = nlp.load_dataset('./my_squad.py')

/opt/conda/lib/python3.7/site-packages/datasets/load.py in load_dataset(path, name, data_dir, data_files, split, cache_dir, features, download_config, download_mode, ignore_verifications, save_infos, script_version, **config_kwargs)
602 hash=hash,
603 features=features,
--> 604 **config_kwargs,
605 )
606

TypeError: 'NoneType' object is not callable

[lhoestq]

Hi !
Thanks for reporting.
It looks like no object inherits from datasets.GeneratorBasedBuilder (or more generally from datasets.DatasetBuilder) in your script.

Could you check that there exist at least one dataset builder class ?

[lhoestq]

Hi @xixiaoyao did you manage to fix your issue ?

[lhoestq]

No activity, closing

[ailian8025]

It happened when try to change the old project which use 'nlp' to new project which use 'datasets'. You should check you old 'my_squad.py' file, change the inherit class from nlp.xxx to datasets.xxx. Otherwise datasets - load.py - import_main_class() if inspect.isclass(obj) and issubclass(obj, main_cls_type): can not find the main_cls.