Wav2vec2 Large Xlsr Bengali

🚀 Wav2Vec2-Large-XLSR-孟加拉語

本項目是基於 facebook/wav2vec2-large-xlsr-53 模型，使用 包含約19.6萬個語音樣本的孟加拉語自動語音識別訓練數據集 進行微調得到的。在使用該模型時，請確保輸入的語音採樣率為16kHz。

🚀 快速開始

使用此模型前，需從 該網站 下載數據集並進行相應的預處理。例如，選擇了1250個測試樣本。

💻 使用示例

基礎用法

import pandas as pd
test_dataset = pd.read_csv('utt_spk_text.tsv', sep='\t', header=None)[60000:61250]
test_dataset.columns = ["audio_path", "__", "label"]
test_dataset = test_data.drop("__", axis=1)
def add_file_path(text):
  path = "data/" + text[:2] + "/" + text + '.flac'
  return path
test_dataset['audio_path'] = test_dataset['audio_path'].map(lambda x: add_file_path(x))

高級用法

模型可以直接使用（無需語言模型），示例代碼如下：

import torch
import torchaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
processor = Wav2Vec2Processor.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
model = Wav2Vec2ForCTC.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
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["audio_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["label"][:2])

🔧 評估

可以使用以下代碼在OpenSLR的孟加拉語測試數據上評估該模型：

import torch
import torchaudio
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
model = Wav2Vec2ForCTC.from_pretrained("tanmoyio/wav2vec2-large-xlsr-bengali")
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["label"]).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=8)
print("WER: {:2f}".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))

測試結果：88.58 %

📚 訓練

訓練使用的腳本可參考 孟加拉語自動語音識別微調Wav2Vec2

📄 許可證

本項目採用CC BY-SA 4.0許可證。

📦 模型信息