Bug with datasets configs?

🤗Datasets
1

Hi Everyone,
I created several config, one per lang, for that dataset : EuropeanParliament/cellar_eurovoc · Datasets at Hugging Face

configs:
  - config_name: default
    data_files: cellar.jsonl.gz
  - config_name: en-GB
    data_files: cellar_en.jsonl.gz
  - config_name: fr-FR
    data_files: cellar_fr.jsonl.gz

It’s appear correctly (more or less) in the website but not in the code

from datasets import get_dataset_config_names
get_dataset_config_names("EuropeanParliament/cellar_eurovoc")

I still have one set

Any idea?

2

cc @polinaeterna

3

Hello @severo , nous avons travaillé dans la meme équipe à l’IRISA a priori, le monde est petit
++

1 Like
4

Hi! The “dataset configs in the YAML” feature requires the latest release of datasets (>=2.14.0), so updating the installation with pip install -U datasets should fix the problem.

5

Thx I updated it but I may have a cache issue now

import datasets
datasets.__version__

‘2.14.4’

and

from datasets import get_dataset_config_names
get_dataset_config_names("EuropeanParliament/cellar_eurovoc")

if you want to test, give

AttributeError: ‘NoneType’ object has no attribute ‘BUILDER_CONFIG_CLASS’

6

Try passing download_mode="force_redownload" to redownload the files or, if this doesn’t help, cache_dir="path/to/new/cache_dir" to use a different directory for caching.

7

fixed :+1:
thx