语言: km
数据集:

• OpenSLR
• common_voice
  评估指标:
• wer
  标签:
• audio
• automatic-speech-recognition
• speech
• xlsr-fine-tuning-week
  许可证: apache-2.0
  模型索引:
• 名称: Gagan Bhatia训练的wav2vec2-xlsr-高棉语模型
  结果:
  • 任务:
    名称: 语音识别
    类型: automatic-speech-recognition
    数据集:
    名称: OpenSLR km
    类型: OpenSLR
    参数: km
    指标:
    • 名称: 测试WER
      类型: wer
      值: 24.96

Wav2Vec2-Large-XLSR-53-高棉语

基于facebook/wav2vec2-large-xlsr-53模型，使用Common Voice和OpenSLR Kh的高棉语数据进行微调。

使用此模型时，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），如下所示：

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

!wget https://www.openslr.org/resources/42/km_kh_male.zip  
!unzip km_kh_male.zip  
!ls km_kh_male  

colnames=['path','sentence']   
df  = pd.read_csv('/content/km_kh_male/line_index.tsv',sep='\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\t',header=None,names = colnames)  
df['path'] = '/content/km_kh_male/wavs/'+df['path'] +'.wav'  

train, test = train_test_split(df, test_size=0.1)  

test.to_csv('/content/km_kh_male/line_index_test.csv')  

test_dataset = load_dataset('csv', data_files='/content/km_kh_male/line_index_test.csv',split = 'train')  

processor = Wav2Vec2Processor.from_pretrained("gagan3012/wav2vec2-xlsr-nepali")  
model = Wav2Vec2ForCTC.from_pretrained("gagan3012/wav2vec2-xlsr-nepali")   

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

# 数据预处理  
# 需要将音频文件读取为数组  
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("预测结果:", processor.batch_decode(predicted_ids))  
print("参考文本:", test_dataset["sentence"][:2])  

结果

预测结果: ['पारानाको ब्राजिली राज्यमा रहेको राजधानी', 'देवराज जोशी त्रिभुवन विश्वविद्यालयबाट शिक्षाशास्त्रमा स्नातक हुनुहुन्छ']

参考文本: ['पारानाको ब्राजिली राज्यमा रहेको राजधानी', 'देवराज जोशी त्रिभुवन विश्वविद्यालयबाट शिक्षाशास्त्रमा स्नातक हुनुहुन्छ']

评估

该模型可在Common Voice的高棉语测试数据上评估如下：

import torch  
import torchaudio  
from datasets import load_dataset, load_metric  
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor  
import re  
from sklearn.model_selection import train_test_split  
import pandas as pd  
from datasets import load_dataset  

!wget https://www.openslr.org/resources/42/km_kh_male.zip  
!unzip km_kh_male.zip  
!ls km_kh_male  

colnames=['path','sentence']   
df  = pd.read_csv('/content/km_kh_male/line_index.tsv',sep='\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\t',header=None,names = colnames)  
df['path'] = '/content/km_kh_male/wavs/'+df['path'] +'.wav'  

train, test = train_test_split(df, test_size=0.1)  

test.to_csv('/content/km_kh_male/line_index_test.csv')  

test_dataset = load_dataset('csv', data_files='/content/km_kh_male/line_index_test.csv',split = 'train')  
wer = load_metric("wer")  
cer = load_metric("cer")  

processor = Wav2Vec2Processor.from_pretrained("gagan3012/wav2vec2-xlsr-khmer")  
model = Wav2Vec2ForCTC.from_pretrained("gagan3012/wav2vec2-xlsr-khmer")   
model.to("cuda")  

chars_to_ignore_regex = '[\\\\,\\\\?\\\\.\\\\!\\\\-\\\\;\\\\:\\\\"\\\\“]'    
resampler = torchaudio.transforms.Resample(48_000, 16_000)  

# 数据预处理  
# 需要将音频文件读取为数组  
def speech_file_to_array_fn(batch):  
    batch["text"] = re.sub(chars_to_ignore_regex, '', batch["text"]).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)  

# 评估函数  
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  

cer = load_metric("cer")  

result = test_dataset.map(evaluate, batched=True, batch_size=8)  

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

测试结果: 24.96 %

WER: 24.962519
CER: 6.950925

训练

训练脚本可在此处找到链接