s2t-medium-librispeech-asr是一个用于自动语音识别(ASR)的语音到文本转换器(S2T)模型。该S2T模型在 this paper 中提出并在 this repository 中发布。
S2T是一个端到端的序列到序列转换器模型。它使用标准的自回归交叉熵损失进行训练,并自动生成转录。
该模型可用于端到端的语音识别(ASR)。查看 model hub 以获取其他S2T检查点。
由于这是一个标准的序列到序列转换器模型,您可以使用generate方法通过向模型传递语音特征来生成转录。
注意:Speech2TextProcessor对象使用 torchaudio 来提取滤波器组特征。在运行此示例之前,请确保安装了torchaudio包。
您可以使用pip install transformers"[speech, sentencepiece]"命令安装它们作为额外的语音依赖项,或者通过使用pip install torchaudio sentencepiece命令分别安装这些软件包。
import torch from transformers import Speech2TextProcessor, Speech2TextForConditionalGeneration from datasets import load_dataset import soundfile as sf model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-medium-librispeech-asr") processor = Speech2Textprocessor.from_pretrained("facebook/s2t-medium-librispeech-asr") def map_to_array(batch): speech, _ = sf.read(batch["file"]) batch["speech"] = speech return batch ds = load_dataset( "patrickvonplaten/librispeech_asr_dummy", "clean", split="validation" ) ds = ds.map(map_to_array) input_features = processor( ds["speech"][0], sampling_rate=16_000, return_tensors="pt" ).input_features # Batch size 1 generated_ids = model.generate(input_features=input_features) transcription = processor.batch_decode(generated_ids)对LibriSpeech测试的评估
以下脚本显示了如何在 LibriSpeech 的“clean”和“other”测试数据集上评估此模型。
from datasets import load_dataset from evaluate import load from transformers import Speech2TextForConditionalGeneration, Speech2TextProcessor librispeech_eval = load_dataset("librispeech_asr", "clean", split="test") # change to "other" for other test dataset wer = load("wer") model = Speech2TextForConditionalGeneration.from_pretrained("facebook/s2t-medium-librispeech-asr").to("cuda") processor = Speech2TextProcessor.from_pretrained("facebook/s2t-medium-librispeech-asr", do_upper_case=True) def map_to_pred(batch): features = processor(batch["audio"]["array"], sampling_rate=16000, padding=True, return_tensors="pt") input_features = features.input_features.to("cuda") attention_mask = features.attention_mask.to("cuda") gen_tokens = model.generate(input_features=input_features, attention_mask=attention_mask) batch["transcription"] = processor.batch_decode(gen_tokens, skip_special_tokens=True)[0] return batch result = librispeech_eval.map(map_to_pred, remove_columns=["audio"]) print("WER:", wer.compute(predictions=result["transcription"], references=result["text"]))结果(WER):
S2T-MEDIUM-LIBRISPEECH-ASR是在 LibriSpeech ASR Corpus 上训练的,该数据集包含约1000小时的16kHz英文朗读语音。
通过PyKaldi或torchaudio从WAV/FLAC音频文件中自动提取符合Kaldi的80通道对数梅尔滤波器组特征进行预处理。对每个示例应用了句子级CMVN(倒谱均值和方差归一化)。
文本以小写形式进行标准化,并使用SentencePiece和10000个词汇大小进行标记化。
该模型使用标准的自回归交叉熵损失进行训练,并使用 SpecAugment 。编码器接收语音特征,解码器自回归生成转录。
@inproceedings{wang2020fairseqs2t, title = {fairseq S2T: Fast Speech-to-Text Modeling with fairseq}, author = {Changhan Wang and Yun Tang and Xutai Ma and Anne Wu and Dmytro Okhonko and Juan Pino}, booktitle = {Proceedings of the 2020 Conference of the Asian Chapter of the Association for Computational Linguistics (AACL): System Demonstrations}, year = {2020}, }