-
Notifications
You must be signed in to change notification settings - Fork 12
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
loss值无法下降 #30
Comments
|
另外,如果是对应多说话人的训练,建议使用SOT (Serialized Output Training) 的方法构建数据集进行训练。 |
十分感谢大佬细心解答!我是改了max_length,所有的数据基本都是30秒的。看下来应该是我数据集构造的过于简单了,导致一直降不下来。我再试试您说的SOT的方式。十分感谢!!! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
问题
我准备了新的数据集,每条数据是两个角色的通话记录,用来训练说话者识别(SR)的任务。adapter和lora都是从零开始训练。数据集基本都是按照30s切割。
在训练过程中无论采用什么超参数,loss值只是来回震荡,并且降的不多。基本情况是从2.5降到1.8左右就来回震荡,训练时间再长一些loss值能够在1.8-1.2之间震荡,再长一些学习率基本降到0,loss值仍然在1.8-1.2之间震荡。无论用哪个节点的lora和adapter,输出结果都十分糟糕。
提问
详细过程
epoch
,lr
,encoder(paraformer,whisper)
,gradient_accumulation_steps
,cvmn是否做标准化
。但是loss变化基本不大,如下图所示PS
为什么你们在用cmvn做标准化是加上平均数,乘以标准差。为什么不是减和除?
The text was updated successfully, but these errors were encountered: