语言: as 数据集:

• common_voice 指标:
• 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: Joydeep Bhattacharjee XLSR Wav2Vec2 Large 53 阿萨姆语 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 as 类型: common_voice 参数: as 指标:
    • 名称: 测试 WER 类型: wer 值: 69.63

Wav2Vec2-Large-XLSR-53-阿萨姆语

基于facebook/wav2vec2-large-xlsr-53对阿萨姆语进行微调，使用了Common Voice数据集。 使用此模型时，请确保您的语音输入采样率为16kHz。

使用方法

该模型可以直接使用（无需语言模型）如下：

import torch
import torchaudio
from datasets import load_dataset
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
test_dataset = load_dataset("common_voice", "as", split="test[:2%]")
processor = Wav2Vec2Processor.from_pretrained("infinitejoy/Wav2Vec2-Large-XLSR-53-Assamese")
model = Wav2Vec2ForCTC.from_pretrained("infinitejoy/Wav2Vec2-Large-XLSR-53-Assamese")
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# 预处理数据集。
# 我们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:", processor.batch_decode(predicted_ids))
print("参考:", test_dataset["sentence"][:2])

评估

该模型可以在Common Voice的阿萨姆语测试数据上进行如下评估。

import torch
import torchaudio
from datasets import load_dataset, load_metric
from transformers import Wav2Vec2ForCTC, Wav2Vec2Processor
import re
test_dataset = load_dataset("common_voice", "as", split="test")
wer = load_metric("wer")
processor = Wav2Vec2Processor.from_pretrained("infinitejoy/Wav2Vec2-Large-XLSR-53-Assamese")
model = Wav2Vec2ForCTC.from_pretrained("infinitejoy/Wav2Vec2-Large-XLSR-53-Assamese")
model.to("cuda")
chars_to_ignore_regex = '[\,\?\.\!\-\;\:\"\“\%\‘\”\�\।]'
resampler = torchaudio.transforms.Resample(48_000, 16_000)
# 预处理数据集。
# 我们需要将音频文件读取为数组
def speech_file_to_array_fn(batch):
    batch["sentence"] = re.sub('’ ',' ',batch["sentence"])
    batch["sentence"] = re.sub(' ‘',' ',batch["sentence"])
    batch["sentence"] = re.sub('’|‘','\'',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)
# 预处理数据集。
# 我们需要将音频文件读取为数组
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"])))

测试结果: 69.63%

训练

使用了Common Voice的train和validation数据集进行训练。