libritts missing `Gender.txt`

[eschmidbauer]

Hi-thanks for sharing this work! I'm trying to train based on libritts but it appears the Pattern_Generator.py is looking for a file that does not exist, named Gender.txt I was able to find speaker info file for libritts which contains the gender but it doesn't match the expected format of Gender.txt See attached speaker_info.txt speaker_info.txt

[eschmidbauer]

I was able to use the speaker info file if i made this change. but im curious if this is the right file since the code references a Gender.txt

--- a/Pattern_Generator.py
+++ b/Pattern_Generator.py
@@ -685,7 +685,8 @@ def Libri_Info_Load(path: str):
     emotion_dict = {path: 'Neutral' for path in paths}
     language_dict = {path: 'English' for path in paths}
     gender_dict = {
-        'Libri.{:04d}'.format(int(line.strip().split('\t')[0])): line.strip().split('\t')[1]
+        'Libri.{:04d}'.format(int(line.strip().split('|')[0])): line.strip().split('|')[1]

[CODEJIN]

Dear ecshmidbauer,

Hello, thank you for your interest in the repository. The file you mentioned 'Gender.txt' was a file I created for my convenience. Instead of sharing that file, I decided it would be better to modify Pattern_Generator.py, so that it can extract gender information from the default dataset. I have pushed the updated version of Pattern_Generator.py without that file. Please check this commit.

Best regards,

Heejo

[eschmidbauer]

thanks - would you be able to share SPEAKERS.txt ? I cannot find it in the default libritts dataset here or here

[CODEJIN]

Here is the file: SPEAKERS.txt

However, that's strange. When I obtained the dataset, this file 'SPEAKERS.TXT' was included in the LibriTTS compressed file. I'm not sure if the dataset has been changed.

[eschmidbauer]

appears to be the same speaker_info.txt i attached in the first post, thank you for the quick response Excited to see results from training this dataset!