README.md

metadata

language: de
datasets:
  - common_voice (trained on 3%)
  - wer
tags:
  - audio
  - automatic-speech-recognition
  - speech
  - xlsr-fine-tuning-week
license: apache-2.0
model-index:
  - name: XLSR Wav2Vec2 Large 53
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice de
          type: common_voice
          args: de
        metrics:
          - name: Test WER
            type: wer
            value: 29.48

Wav2Vec2-Large-XLSR-53-German

Fine-tuned facebook/wav2vec2-large-xlsr-53 on German using 3% of the Common Voice dataset. #TODO: replace {language} with your language, e.g. French and eventually add more datasets that were used and eventually remove common voice if model was not trained on common voice When using this model, make sure that your speech input is sampled at 16kHz.

Usage

The model can be used directly (without a language model) as follows:

import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor

test_dataset = load_dataset("common_voice", "de", split="test[:2%]") #TODO: replace {lang_id} in your language code here. Make sure the code is one of the *ISO codes* of [this](https://huggingface.co/languages) site.

processor = Wav2Vec2Processor.from_pretrained("marcel/wav2vec2-large-xlsr-german-demo") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`
model = Wav2Vec2ForCTC.from_pretrained("marcel/wav2vec2-large-xlsr-german-demo/") #TODO: replace {model_id} with your model id. The model id consists of {your_username}/{your_modelname}, *e.g.* `elgeish/wav2vec2-large-xlsr-53-arabic`

resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
\\treturn batch

test_dataset = test_dataset.map(speech_file_to_array_fn)
inputs = processor(test_dataset["speech"][:2], sampling_rate=16_000, return_tensors="pt", padding=True)

with torch.no_grad():
\\tlogits = model(inputs.input_values, attention_mask=inputs.attention_mask).logits

predicted_ids = torch.argmax(logits, dim=-1)

print("Prediction:", processor.batch_decode(predicted_ids))
print("Reference:", test_dataset["sentence"][:2])

Evaluation

The model can be evaluated as follows on the {language} test data of Common Voice.

import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re

test_dataset = load_dataset("common_voice", "de", split="test") [this](https://huggingface.co/languages) site.
wer = load_metric("wer")

processor = Wav2Vec2Processor.from_pretrained("de")
`elgeish/wav2vec2-large-xlsr-53-arabic`
model = Wav2Vec2ForCTC.from_pretrained("de") 
`elgeish/wav2vec2-large-xlsr-53-arabic`
model.to("cuda")

chars_to_ignore_regex = '[\\,\\?\\.\\!\\-\\;\\:\\"\\“]'  
resampler = torchaudio.transforms.Resample(48_000, 16_000)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def speech_file_to_array_fn(batch):
\\\\tbatch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
\\\\tbatch["sentence"] = re.sub('\\ß', 'ss', batch["sentence"])
\\\\tspeech_array, sampling_rate = torchaudio.load(batch["path"])
\\\\tbatch["speech"] = resampler(speech_array).squeeze().numpy()
\\\\treturn batch

test_dataset = test_dataset.map(speech_file_to_array_fn)

# Preprocessing the datasets.
# We need to read the aduio files as arrays
def evaluate(batch):
\\tinputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)

\\twith torch.no_grad():
\\t\\tlogits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits

\\tpred_ids = torch.argmax(logits, dim=-1)
\\tbatch["pred_strings"] = processor.batch_decode(pred_ids)
\\treturn batch

result = test_dataset.map(evaluate, batched=True, batch_size=8)

print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))

Test Result: 29.48 %

Training

The 3% of the Common Voice train, validation datasets were used for training.

The script used for training can be found TODO