LibriSpeech 数据集训练

Question

Closed this issue 2 years ago · 5 comments

你好，我使用LibriSpeech 数据集混合了20000个语音进行训练，但下降并不是很明显，且在70多epoch的时候就early stop了损失只下降到-7.几，请问可能有哪些原因呢

Answer 1 · 2021-06-10T10:59:14.000Z

你用的是哪个文件夹的代码？LibriSpeech的混合应该结果不会很差的。

Answer 2 · 2021-06-10T12:37:28.000Z

pytorch那个

Answer 3 · 2021-06-10T14:09:19.000Z

不应该吧，你也可以试试lightning的版本。我在wsj的结果应该是没问题的。或者你可以检查一下yml的配置文件

Answer 4 · 2023-03-31T06:22:09.000Z

This issue is stale because it has been open for 14 days with no activity.

Answer 5 · 2023-04-08T01:21:35.000Z

This issue was closed because it has been inactive for 7 days since being marked as stale.