[Bug]: new Shallow diffusion bug?

[iiallgaii]

系统平台版本号

win7

GPU 型号

3060

Python版本

3.8.9

PyTorch版本

1.13.1

sovits分支

4.0(默认)

数据集来源（用于判断数据集质量）

UVR

出现问题的环节或执行的命令

train Shallow diffusion

情况描述

May i know that is it same content between train.txt and val.txt ? the below files train.txt & val.txt are auto gen by Sovits, why i have 10 wav but only 2 wav show in val.txt? is it mean Sovits will only train
./dataset/44k/Ai_mary/ocalsV2_8.wav and ./dataset/44k/Ai_mary/ocalsV2_10.wav ? thanks !

train.txt

./dataset/44k/Ai_mary/ocalsV2_1.wav ./dataset/44k/Ai_mary/ocalsV2_2.wav ./dataset/44k/Ai_mary/ocalsV2_3.wav ./dataset/44k/Ai_mary/ocalsV2_4.wav ./dataset/44k/Ai_mary/ocalsV2_5.wav ./dataset/44k/Ai_mary/ocalsV2_6.wav ./dataset/44k/Ai_mary/ocalsV2_7.wav ./dataset/44k/Ai_mary/ocalsV2_8.wav ./dataset/44k/Ai_mary/ocalsV2_9.wav ./dataset/44k/Ai_mary/ocalsV2_10.wav

val.txt

./dataset/44k/Ai_mary/ocalsV2_8.wav ./dataset/44k/Ai_mary/ocalsV2_10.wav

日志

-

补充说明

No response

[bfloat16]

On the contrary, the audio in train.txt is the training set, while the audio in val.txt is the validation set. According to your description, those 10 audio clips will be used for training.

[iiallgaii]

On the contrary, the audio in train.txt is the training set, while the audio in val.txt is the validation set. According to your description, those 10 audio clips will be used for training.

i see , thanks!

[HuanLinOTO]

close as completed