set hparams.epochs to 200

thats usually how many i do

only 200?

yes

will 300 sound worse than 200

depends on the dataset

since validation loss isnt being displayed on the terminal you need to run tensorboard alongside train.py to see how youre doing

if validation loss goes up thats when you have to stop

@Userit doesnt save properly, when it finishes it doesnt do anything, the checkpoint_0 file stays the same

* * thats weird

something is definitely broken it doesnt save when its done

should have told you the validation loss

try adding the rank command line argument

what does that do? i will do that tomorrow i have to turn my pc off now

tells train.py the rank of the gpu is 0 or somethin

g

the gpu isnt a problem

its just not saving the file when its done

idk rank just doesnt really mean anything to me and its best when its put to 0

oh ok

but i need to fix the problem where it wont save the model

Could someone please help and train this model? It's not doing very well and everytime I do it, it keeps having a stroke. Will link the necessary files in a sec

Empire Carpet Guy (Elmer Lynn Hauldren) Model (Make a copy): https://drive.google.com/file/d/1L6tl6JCjdGbXjW2TrseFg5766bFt6iZg/view?usp=sharing Transcripts/Audio files: https://drive.google.com/file/d/17ghAeziXMKpdyqi06qAjDQHIn_rHaSEA/view?usp=sharing

Er, actually I think Uberduck has an option for that

Welp, I just submitted my model. Maybe Uberduck might be able to train it some more

eminem en real

eminem w fałszywym

w