metadata
dataset_info:
features:
- name: audio
dtype:
audio:
sampling_rate: 24000
- name: clip_id
dtype: string
- name: failure_mode
dtype: string
- name: failure_mechanism
dtype: string
- name: source
dtype: string
- name: ground_truth
dtype: string
- name: kyutai_transcript
dtype: string
- name: whisper_transcript
dtype: string
- name: slm_transcript
dtype: string
- name: wer_whisper
dtype: float32
- name: wer_slm
dtype: float32
- name: expected_kyutai
dtype: string
- name: expected_whisper_multilingual
dtype: string
- name: note
dtype: string
splits:
- name: train
num_bytes: 7866062
num_examples: 7
download_size: 7878030
dataset_size: 7866062
configs:
- config_name: default
data_files:
- split: train
path: data/train-*