README.md

metadata

language: zh-HK
datasets:
  - common_voice
metrics:
  - cer
tags:
  - audio
  - automatic-speech-recognition
  - speech
  - xlsr-fine-tuning-week
license: apache-2.0
model-index:
  - name: wav2vec2-large-xlsr-cantonese
    results:
      - task:
          name: Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: Common Voice zh-HK
          type: common_voice
          args: zh-HK
        metrics:
          - name: Test CER
            type: cer
            value: 17.81

Wav2Vec2-Large-XLSR-53-Cantonese

Fine-tuned facebook/wav2vec2-large-xlsr-53 on Cantonese using the 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", "zh-HK", split="test[:2%]")

processor = Wav2Vec2Processor.from_pretrained("ctl/wav2vec2-large-xlsr-cantonese") 
model = Wav2Vec2ForCTC.from_pretrained("ctl/wav2vec2-large-xlsr-cantonese")

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):
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(speech_array).squeeze().numpy()
    return 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():
    logits = 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. # TODO: replace #TODO: replace language with your {language}, e.g. French

!mkdir cer
!wget -O cer/cer.py https://huggingface.co/ctl/wav2vec2-large-xlsr-cantonese/raw/main/cer.py
import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
import argparse

lang_id = "zh-HK" 
model_id = "ctl/wav2vec2-large-xlsr-cantonese" 

parser = argparse.ArgumentParser(description='hanles checkpoint loading')
parser.add_argument('--checkpoint', type=str, default=None)
args = parser.parse_args()
model_path = model_id
if args.checkpoint is not None:
    model_path += "/checkpoint-" + args.checkpoint


chars_to_ignore_regex = '[\,\?\.\!\-\;\:"\“\%\‘\”\�\．\⋯\！\－\：\–\。\》\,\）\,\？\；\～\~\…\︰\，\（\」\‧\《\﹔\、\—\／\,\「\﹖\·\']'

test_dataset = load_dataset("common_voice", f"{lang_id}", split="test") 
cer = load_metric("./cer")

processor = Wav2Vec2Processor.from_pretrained(f"{model_id}") 
model = Wav2Vec2ForCTC.from_pretrained(f"{model_path}") 
model.to("cuda")

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):
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower()
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(speech_array).squeeze().numpy()
    return 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):
    inputs = processor(batch["speech"], sampling_rate=16_000, return_tensors="pt", padding=True)
    with torch.no_grad():
        logits = model(inputs.input_values.to("cuda"), attention_mask=inputs.attention_mask.to("cuda")).logits

    pred_ids = torch.argmax(logits, dim=-1)
    batch["pred_strings"] = processor.batch_decode(pred_ids)
    return batch

result = test_dataset.map(evaluate, batched=True, batch_size=16)

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

Character Error Rate implementation

Adapting code from wer

@datasets.utils.file_utils.add_start_docstrings(_DESCRIPTION, _KWARGS_DESCRIPTION)
class CER(datasets.Metric):
    def _info(self):
        ...

    def _compute(self, predictions, references):
        preds = [char for seq in predictions for char in list(seq)]
        refs = [char for seq in references for char in list(seq)]
        return wer(refs, preds)

Test Result: 17.81 %

Training

The Common Voice train, validation were used for training.

The script used for training will be posted here