innnky / so-vits-svc

基于vits与softvc的歌声音色转换模型
GNU Affero General Public License v3.0
3.57k stars 1 forks source link

The size of tensor a (2) must match the size of tensor b (4) at non-singleton dimension 0 #171

Open hugplz opened 1 year ago

hugplz commented 1 year ago

是不是数据集的问题?

hugplz commented 1 year ago

测试了一下,发现重新用预模型训练就可以,用之前的模型就会报如上错误,请问这是什么原因导致的?

NaruseMioShirakana commented 1 year ago

请发送完整报错的上下文

hugplz commented 1 year ago

QQ截图20230303160618 而且我还发现colab好像不能保存训练后的模型?

hugplz commented 1 year ago

还有一个特点就是模型加载不上,加载G模型的时候没有下文就直接跳到图片第一行了

NaruseMioShirakana commented 1 year ago

在截图一遍完整的配置文件

hugplz commented 1 year ago

QQ截图20230303202647

hugplz commented 1 year ago

QQ截图20230303202658

hugplz commented 1 year ago

QQ截图20230303202708

NaruseMioShirakana commented 1 year ago

你是否在训练sovits4.0的模型,如果是sovits4.0,此配置文件存在非常大的问题。 sovits3.0和sovits4.0的模型、配置文件均不通用。 如果你在训练sovits4.0模型,请确保使用了sovits4.0仓库中的配置文件进行训练

hugplz commented 1 year ago

没有,我就是在训练sovits3.0的模型