suzii/vi-whisper-large-v3-turbo
Automatic Speech Recognition • 0.8B • Updated • 671 • 14
Datasets:
Dataset Viewer
The dataset viewer is not available for this dataset.
Unexpected token '<', "<html>
<h"... is not valid JSON
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
unofficial mirror of FPT Open Speech Dataset (FOSD)
released publicly in 2018 by FPT Corporation
100h, 25.9k samples
official link (dead): https://fpt.ai/fpt-open-speech-data/
mirror: https://data.mendeley.com/datasets/k9sxg2twv4/4
DOI: 10.17632/k9sxg2twv4.4
pre-process:
- remove non-sense strings:
-N\r\n - remove 4 files because missing transcription:
Set001_V0.1_008210.mp3Set001_V0.1_010753.mp3Set001_V0.1_011477.mp3Set001_V0.1_011841.mp3
need to do: check misspelling
usage with HuggingFace:
# pip install -q "datasets[audio]"
from datasets import load_dataset
from torch.utils.data import DataLoader
dataset = load_dataset("doof-ferb/fpt_fosd", split="train", streaming=True)
dataset.set_format(type="torch", columns=["audio", "transcription"])
dataloader = DataLoader(dataset, batch_size=4)
- Downloads last month
- 340
