farsi_dataset / README.md
m522t's picture
Update README.md
fb48ddc verified
metadata
dataset_info:
  features:
    - name: audio
      dtype: audio
    - name: sentence
      dtype: string
  splits:
    - name: train
      num_bytes: 3083531819.2434187
      num_examples: 249148
    - name: validation
      num_bytes: 389113937.2382908
      num_examples: 31144
    - name: test
      num_bytes: 389790011.19029075
      num_examples: 31144
  download_size: 3756964682
  dataset_size: 3862435767.6720004
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/train-*
      - split: validation
        path: data/validation-*
      - split: test
        path: data/test-*

This dataset consists of some validated rows of Common Voice Farsi 18.0.0. All rows are readily downsampled to 16kHz in order to be consistent with Whisper model.