BuilderConfig questions

[xiaos16]

hello, When I test tedlium and fleurs-zh， I get these errors,

ERROR - load args is {'path': 'TwinkStart/tedlium', 'name': 'release1'}load dataset error: BuilderConfig 'release1' not found. Available: ['default']

ERROR - load args is {'path': 'google/fleurs', 'name': 'cmn_hans_cn', 'split': 'test'}load dataset error: BuilderConfig 'cmn_hans_cn' not found. Available: ['default']

how to solve it ? thanks ！

[UltraEval]

Unable to Reproduce the Bug

To verify the issue, please run the following code and share your results:

from datasets import get_dataset_config_names

configs = get_dataset_config_names("TwinkStart/tedlium")
print(configs)

from datasets import get_dataset_config_names

configs = get_dataset_config_names("google/fleurs")
print(configs)

[xiaos16]

from datasets import get_dataset_config_names
configs = get_dataset_config_names("TwinkStart/tedlium")
print(configs)
['default']
from datasets import get_dataset_config_names
configs = get_dataset_config_names("google/fleurs")
print(configs)
['default']

and I use comands with
'''
python audio_evals/main.py --dataset fleurs-zh --prompt mini-cpm-omni-asr-zh --model MiniCPMo2_6-audio
python audio_evals/main.py --dataset tedlium-release1 --prompt mini-cpm-omni-asr-en --model MiniCPMo2_6-audio
'''

in /UltraEval-Audio-main/registry/dataset/fleurs.yaml, it is this:

fleurs-zh:
class: audio_evals.dataset.huggingface.Huggingface
args:
subset: cmn_hans_cn
default_task: asr-zh
name: google/fleurs
ref_col: raw_transcription
split: test

[UltraEval]

from datasets import get_dataset_config_names
configs = get_dataset_config_names("TwinkStart/tedlium")
print(configs)
['default']
from datasets import get_dataset_config_names
configs = get_dataset_config_names("google/fleurs")
print(configs)
['default']

and I use comands with ''' python audio_evals/main.py --dataset fleurs-zh --prompt mini-cpm-omni-asr-zh --model MiniCPMo2_6-audio python audio_evals/main.py --dataset tedlium-release1 --prompt mini-cpm-omni-asr-en --model MiniCPMo2_6-audio '''

in /UltraEval-Audio-main/registry/dataset/fleurs.yaml, it is this:

fleurs-zh: class: audio_evals.dataset.huggingface.Huggingface args: subset: cmn_hans_cn default_task: asr-zh name: google/fleurs ref_col: raw_transcription split: test

I know this config, you should run the following code with your python shell:

from datasets import get_dataset_config_names

configs = get_dataset_config_names("TwinkStart/tedlium")
print(configs)

from datasets import get_dataset_config_names

configs = get_dataset_config_names("google/fleurs")
print(configs)

and share your results

[xiaos16]

yes, the results are both ['default']

from datasets import get_dataset_config_names
configs = get_dataset_config_names("TwinkStart/tedlium")
print(configs)
['default']
from datasets import get_dataset_config_names
configs = get_dataset_config_names("google/fleurs")
print(configs)
['default']

[xiaos16]

[UltraEval]

yes, the results are both ['default']

from datasets import get_dataset_config_names configs = get_dataset_config_names("TwinkStart/tedlium") print(configs) ['default'] from datasets import get_dataset_config_names configs = get_dataset_config_names("google/fleurs") print(configs) ['default']

You need check:

• connect hf website: https://huggingface.co/datasets/TwinkStart/tedlium

it should be like:

• upgrade datasets package

[xiaos16]

which verison do you use? I use datasets ==3.3.2. I also try 3.4.1, but it is not ok for me.

I've downloaded the data locally.

./TwinkStart/tedlium/release1/test-00000-of-00001.parquet

./google/fleurs/data/cmn_hans_cn/audio/test.tar.gz

[UltraEval]

which verison do you use? I use datasets ==3.3.2. I also try 3.4.1, but it is not ok for me.

I've downloaded the data locally.

./TwinkStart/tedlium/release1/test-00000-of-00001.parquet

./google/fleurs/data/cmn_hans_cn/audio/test.tar.gz

you can try download hf data with following code

save_path='xx'
dataset = load_dataset('TwinkStart/tedlium', name='release1',  cache_dir=save_path)

[xiaos16]

which verison do you use? I use datasets ==3.3.2. I also try 3.4.1, but it is not ok for me.
I've downloaded the data locally.
./TwinkStart/tedlium/release1/test-00000-of-00001.parquet
./google/fleurs/data/cmn_hans_cn/audio/test.tar.gz

you can try download hf data with following code

save_path='xx'
dataset = load_dataset('TwinkStart/tedlium', name='release1', cache_dir=save_path)

I will try it, thanks !