Wav2vec2 Large Xlsr 53 Polish

🚀 語音自動識別模型評估項目

本項目主要針對基於facebook/wav2vec2-large-xlsr-53-polish模型在Common Voice波蘭語測試集上進行評估，通過一系列處理流程計算字錯率（WER），為語音自動識別相關研究和應用提供數據參考。

🚀 快速開始

環境準備

確保你已經安裝了以下必要的庫：

• torchaudio
• datasets
• transformers
• torch

代碼運行

以下是在Common Voice波蘭語測試集上進行評估的代碼示例：

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

model_name = "facebook/wav2vec2-large-xlsr-53-polish"
device = "cuda"
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"]'  # noqa: W605

model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)

ds = load_dataset("common_voice", "pl", split="test", data_dir="./cv-corpus-6.1-2020-12-11")

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

def map_to_array(batch):
    speech, _ = torchaudio.load(batch["path"])
    batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
    batch["sampling_rate"] = resampler.new_freq
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
    return batch

ds = ds.map(map_to_array)


def map_to_pred(batch):
    features = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0], padding=True, return_tensors="pt")
    input_values = features.input_values.to(device)
    attention_mask = features.attention_mask.to(device)
    with torch.no_grad():
        logits = model(input_values, attention_mask=attention_mask).logits
    pred_ids = torch.argmax(logits, dim=-1)
    batch["predicted"] = processor.batch_decode(pred_ids)
    batch["target"] = batch["sentence"]
    return batch

result = ds.map(map_to_pred, batched=True, batch_size=16, remove_columns=list(ds.features.keys()))

wer = load_metric("wer")
print(wer.compute(predictions=result["predicted"], references=result["target"]))

運行結果

運行上述代碼後，得到的字錯率（WER）結果為：24.6 %

💻 使用示例

基礎用法

# 此代碼用於在Common Voice波蘭語測試集上評估模型的字錯率（WER）
import torchaudio
from datasets import load_dataset, load_metric
from transformers import (
    Wav2Vec2ForCTC,
    Wav2Vec2Processor,
)
import torch
import re
import sys

model_name = "facebook/wav2vec2-large-xlsr-53-polish"
device = "cuda"
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"]'  # noqa: W605

model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)

ds = load_dataset("common_voice", "pl", split="test", data_dir="./cv-corpus-6.1-2020-12-11")

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

def map_to_array(batch):
    speech, _ = torchaudio.load(batch["path"])
    batch["speech"] = resampler.forward(speech.squeeze(0)).numpy()
    batch["sampling_rate"] = resampler.new_freq
    batch["sentence"] = re.sub(chars_to_ignore_regex, '', batch["sentence"]).lower().replace("’", "'")
    return batch

ds = ds.map(map_to_array)


def map_to_pred(batch):
    features = processor(batch["speech"], sampling_rate=batch["sampling_rate"][0], padding=True, return_tensors="pt")
    input_values = features.input_values.to(device)
    attention_mask = features.attention_mask.to(device)
    with torch.no_grad():
        logits = model(input_values, attention_mask=attention_mask).logits
    pred_ids = torch.argmax(logits, dim=-1)
    batch["predicted"] = processor.batch_decode(pred_ids)
    batch["target"] = batch["sentence"]
    return batch

result = ds.map(map_to_pred, batched=True, batch_size=16, remove_columns=list(ds.features.keys()))

wer = load_metric("wer")
print(wer.compute(predictions=result["predicted"], references=result["target"]))

高級用法

此代碼目前主要是基礎的評估流程，若要進行高級拓展，例如更換不同的模型、調整參數等，可以參考以下示例：

# 更換模型名稱以使用不同的預訓練模型
model_name = "new_model_name"
model = Wav2Vec2ForCTC.from_pretrained(model_name).to(device)
processor = Wav2Vec2Processor.from_pretrained(model_name)

# 調整批量大小
result = ds.map(map_to_pred, batched=True, batch_size=32, remove_columns=list(ds.features.keys()))

📄 許可證

本項目使用的許可證為Apache 2.0許可證。

📋 信息表格