Wav2vec2 Large Xlsr Dutch

🚀 Wav2Vec2-Large-XLSR-53-荷兰语

本项目基于 Common Voice 数据集，对荷兰语进行微调 facebook/wav2vec2-large-xlsr-53 模型。

使用此模型时，请确保语音输入的采样率为 16kHz。

🚀 快速开始

本模型在荷兰语语音识别任务上进行了微调，可用于将荷兰语语音转换为文本。使用时需注意输入语音的采样率要求。

✨ 主要特性

• 微调模型：基于 facebook/wav2vec2-large-xlsr-53 在荷兰语数据集上进行微调。
• 特定采样率要求：语音输入需采样率为 16kHz。

📦 安装指南

文档未提供安装步骤，暂不展示。

💻 使用示例

基础用法

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

test_dataset = load_dataset("common_voice", "nl", split="test[:2%]")

processor = Wav2Vec2Processor.from_pretrained("simonsr/wav2vec2-large-xlsr-dutch")
model = Wav2Vec2ForCTC.from_pretrained("simonsr/wav2vec2-large-xlsr-dutch")

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

# Preprocessing the datasets.
# We need to read the audio 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])

高级用法

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

test_dataset = load_dataset("common_voice", "nl", split="test")
wer = load_metric("wer")

processor = Wav2Vec2Processor.from_pretrained("{model_id}") #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("{model_id}") #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.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):
    batch["sentence"] = unidecode.unidecode(batch["sentence"])
    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=8)

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

📚 详细文档

评估

本模型可在 Common Voice 的荷兰语测试数据上进行如下评估：

上述高级用法中的代码即为评估代码，测试结果显示字错率（WER）为 38.74%。

训练

训练使用了 Common Voice 的 train、validation 等数据集。训练脚本可在 此处 找到（需补充训练脚本链接）。

🔧 技术细节

文档未提供足够技术细节，暂不展示。

📄 许可证

本模型采用 apache-2.0 许可证。

模型信息表格