语言: 格鲁吉亚语
数据集:

• 通用语音(Common Voice)
  评估指标:
• 词错误率(WER)
  标签:
• 音频
• 自动语音识别
• 语音
• XLSR微调周
  许可证: Apache-2.0
  模型索引:
• 名称: 针对格鲁吉亚语微调的XLSR Wav2Vec模型 结果:
  • 任务: 名称: 语音识别 类型: 自动语音识别 数据集: 名称: Common Voice ka 类型: common_voice 参数: ka 指标:
    • 名称: 测试集词错误率 类型: wer 值: 45.28

Wav2Vec2-Large-XLSR-53-格鲁吉亚语版

本模型基于facebook/wav2vec2-large-xlsr-53使用通用语音数据集对格鲁吉亚语进行微调。使用时请确保语音输入采样率为16kHz。

使用方法

该模型可直接使用（无需语言模型），示例如下：

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

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

processor = Wav2Vec2Processor.from_pretrained("xsway/wav2vec2-large-xlsr-georgian")
model = Wav2Vec2ForCTC.from_pretrained("xsway/wav2vec2-large-xlsr-georgian") 

resampler = lambda sampling_rate, y: librosa.resample(y.numpy().squeeze(), sampling_rate, 16_000)

# 数据预处理
def speech_file_to_array_fn(batch):
    speech_array, sampling_rate = torchaudio.load(batch["path"])
    batch["speech"] = resampler(sampling_rate, speech_array).squeeze()
    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("预测结果:", processor.batch_decode(predicted_ids))
print("参考文本:", test_dataset["sentence"][:2])

评估

可通过以下方式在通用语音数据集的格鲁吉亚语测试集上评估模型：

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

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

processor = Wav2Vec2Processor.from_pretrained("xsway/wav2vec2-large-xlsr-georgian") 
model = Wav2Vec2ForCTC.from_pretrained("xsway/wav2vec2-large-xlsr-georgian") 
model.to("cuda")

chars_to_ignore_regex = '[\\\\,\\?\\.\\!\\-\\;\\:\\"\\“]' 
resampler = lambda sampling_rate, y: librosa.resample(y.numpy().squeeze(), sampling_rate, 16_000)

# 数据预处理
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(sampling_rate, speech_array).squeeze()
    return batch

test_dataset = test_dataset.map(speech_file_to_array_fn)

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("词错误率: {:.2f}%".format(100 * wer.compute(predictions=result["pred_strings"], references=result["sentence"])))

测试结果: 45.28%

训练过程

训练使用了通用语音数据集的train和validation子集。训练脚本详见此处