README.md

metadata

language:
  - nl
license: apache-2.0
tags:
  - automatic-speech-recognition
  - mozilla-foundation/common_voice_8_0
  - mozilla-foundation/common_voice_7_0
  - nl
  - robust-speech-event
  - model_for_talk
  - hf-asr-leaderboard
datasets:
  - mozilla-foundation/common_voice_8_0
model-index:
  - name: XLS-R-300M - Dutch
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice 8 NL
          type: mozilla-foundation/common_voice_8_0
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 46.94
          - name: Test CER
            type: cer
            value: 21.65
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Robust Speech Event - Dev Data
          type: speech-recognition-community-v2/dev_data
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: '???'
          - name: Test CER
            type: cer
            value: '???'
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Robust Speech Event - Test Data
          type: speech-recognition-community-v2/eval_data
          args: nl
        metrics:
          - name: Test WER
            type: wer
            value: 42.56

xlsr300m_cv_8.0_nl

Evaluation Commands

1. To evaluate on mozilla-foundation/common_voice_8_0 with split test

python eval.py --model_id Iskaj/xlsr300m_cv_8.0_nl --dataset mozilla-foundation/common_voice_8_0 --config nl --split test

2. To evaluate on speech-recognition-community-v2/dev_data

python eval.py --model_id Iskaj/xlsr300m_cv_8.0_nl --dataset speech-recognition-community-v2/dev_data --config nl --split validation --chunk_length_s 5.0 --stride_length_s 1.0

Inference

import torch
from datasets import load_dataset
from transformers import AutoModelForCTC, AutoProcessor
import torchaudio.functional as F

model_id = "Iskaj/xlsr300m_cv_8.0_nl"

sample_iter = iter(load_dataset("mozilla-foundation/common_voice_8_0", "nl", split="test", streaming=True, use_auth_token=True))

sample = next(sample_iter)
resampled_audio = F.resample(torch.tensor(sample["audio"]["array"]), 48_000, 16_000).numpy()

model = AutoModelForCTC.from_pretrained(model_id)
processor = AutoProcessor.from_pretrained(model_id)

inputs = processor(resampled_audio, sampling_rate=16_000, return_tensors="pt")
with torch.no_grad():
  logits = model(**inputs).logits
  predicted_ids = torch.argmax(logits, dim=-1)
  transcription = processor.batch_decode(predicted_ids)

transcription[0].lower()
#'het kontine schip lag aangemeert in de aven'