采用预训练模型提取语音特征，怎么处理长语音，直接切割或滑窗处理？

TencentGameMate / chinese_speech_pretrain

chinese speech pretrained models

997 stars 84 forks source link

Open Owen1234560 opened 1 year ago

Owen1234560 commented 1 year ago

采用hubert提取特征时，发现语音太长，占用的显存变多。

mazai-xiangji commented 1 year ago

你好，请问你最终是怎么解决这个问题的

milely commented 10 months ago

同问