[Bug] kik-train-00001.parquet appears corrupt (1.23 kB)
Hi,
I'm trying to load the Kikuyu TTS dataset but encountering an error with kik-train-00001.parquet.
Issue:
The file is only 1.23 kB, which seems too small to contain valid data. Loading fails with:ArrowInvalid: BatchSize must be greater than 0, got 0
Expected:
Similar files (e.g., kik-train-00000.parquet) are ~597 MB.
Steps to reproduce:
from datasets import load_dataset
kik_tts = load_dataset("google/WaxalNLP", "kik_tts")
# Fails on kik-train-00001.parquet
Workaround: Loading parquet files directly works, skipping the corrupt file.
Could you check if this file was uploaded correctly? Is there missing Kikuyu TTS data?
Thanks!
Actually, the file kik-train-00001.parquet contains 0 samples:
df = pd.read_parquet('kik-train-00001.parquet')
print(len(df)) # Output: 0
This causes the datasets library to fail when loading kik_tts.
Is this file a placeholder, or is there missing Kikuyu TTS data that should be there?
Thanks!
Hi Mark. Thanks for pointing out this issue! The empty parquet files should no longer be present. Please try again and let us know if you hit any problems.