Dataset Viewer issue: JobManagerCrashedError
The dataset viewer is not working. I've tested only a single shard (train/recitation_0/shard_0006.parquet) worked fine but when adding more shards it fails.
I've successfully loaded the dataset in my machine.
Note: the dataset is uploaded manually not with push_to_hup method
Error details:
Error code: JobManagerCrashedError
Hi ! I refreshed the viewer and it's working now :) sorry for the delay
Btw I's suggest you to store your audio data as encoded audio (wav, mp3), which is possible with Parquet. This will make the dataset much smaller and easier to use.
You can use the datasets library to save audio data in Parquet with the Audio() type, or use pandas with https://github.com/lhoestq/pandas-audio-methods
Thank you for the reply I know you have too many issues to work with ! I appreciate that !
I indeed parsed the dataset and save it as parquet with hugginface Audio() feature.
I still have an issue with the viewer: Despite the audio column is a hugginface Audio feature, the viewer does not display it as an Audio.
Why I did not save the dataset as wav or mp3 ? As you know when you convert mp3 and wav to arrays because it is a costly process (if you consider working with GPU machine this may means money too). Another thing is you may ran into errors due to difference voice codec types. I'm open to any idea or technique to load and save dataset more efficiently as I preparing a larger public audio dataset