# machine-learning
- a
Awesome Face (GTX 1660 Ti)
09/12/2021, 11:44 AMObviously recommended 100 or more - c
cengi
09/12/2021, 12:24 PMDo I stop the begin learning process manually or does it stop automatically ? - j
JacobLenstar
09/12/2021, 1:15 PMgetting real sick of the CUDA out of memory error - j
JacobLenstar
09/12/2021, 1:16 PMbeing able to train a talknet model with hifigan is the rare exception - j
JacobLenstar
09/12/2021, 1:17 PMI have the batch size set to five and even then it's still too much - m
MoarMosspato
09/12/2021, 1:41 PMWorking on the FCCD Squidward - k
kumori💖
09/12/2021, 6:01 PMim currently splitting up the voice clips for My Melody from Supercute Adventures - t
The Wiggles and High School D×D
09/12/2021, 6:15 PMOut of curiosity, who's testing my notebook? I know someone took a look at it and altered it for a process, but did you successfully run it? - t
The Wiggles and High School D×D
09/12/2021, 6:43 PMI don't want to keep waiting on wondering why I can't synthesise my model - c
cengi
09/12/2021, 6:54 PMhow many iterations before I stop it training? - u
user
09/12/2021, 7:33 PMwtf just happened - u
user
09/12/2021, 7:38 PMkinda better? - u
user
09/12/2021, 7:38 PMI have no fucking clue what happened - u
user
09/12/2021, 7:38 PMI have 142 voice clips, all properly transcribed - u
user
09/12/2021, 7:47 PMRetrained it with slightly different model settings Hopefully it'll only be up from here but I will check on it from time to time - u
user
09/12/2021, 7:52 PMSo It sounds pretty decent so far, but still has that robotic aftertaste so I'm likely going to change the voice clip amount to 142 to 169 - h
hecko
09/12/2021, 7:56 PMthe roboticity is likely because you're using the default hifi-gan model (which does the final polishing of the audio), once you're done training you can try finetuning it (see #841437191073955920) - h
haru0l
09/12/2021, 8:13 PMis it only for some types of models? - h
hecko
09/12/2021, 8:20 PMthere are only 2 types of models that are currently submittable tacotron2 (the 📣 section) can have hifi-gan finetuned with the fine-tuning notebook talknet (🎵) does it as part of its training - u
user
09/12/2021, 8:38 PMCan all the things required to run the fine tuner be found in the training colab or my drive? - h
hecko
09/12/2021, 8:39 PMsee #841437191073955920, second one from the bottom - d
Dyno
09/12/2021, 8:39 PMRead #answers, @hecko. - u
user
09/12/2021, 8:39 PMk - h
hecko
09/12/2021, 8:39 PMshut up dyno - h
hecko
09/12/2021, 8:39 PMyou'll need the finished tacotron model first - u
user
09/12/2021, 8:43 PMOk, thank you! - u
(Dawn) Will Draw Fictional Women
09/12/2021, 8:43 PMwhat exactly does the hifi-gan finetuner do - u
user
09/12/2021, 8:43 PMOOH, THAT'S POG - u
(Dawn) Will Draw Fictional Women
09/12/2021, 8:43 PMi wanna see if it can make any tangible improvements to my patricia wagon model