Datasets:

fsicoli

/

common_voice_15_0

like 5

Hello Jan,

Thanks for your message and for pointing that out.
Actually, I generated the dataset with big shards so the cs/other data is entirely in one 4GB shard. I updated the n_shards.json file to reflect that.

But I will split the shards into 1GB chunks for easier handling in the next couple of days, so it has better overall handling.

Dear @fsicoli , the number of shards for cs/other given by n_shards.json is 4, but only cs/other/cs_other_0.tar exists in the data ... The only dataset I can comkpare this to is the CV12, which has only one cs/other shard.

Dear @wsogmm ,

I updated the Czech dataset, dividing it into 40.000 files shards, as it is usually done in the CV dataset.

I hope it works for you now.