it's been training for a few minutes and now looks like this

not good, right?

just got this tho

@User The synthesis notebook won't download my model

NameError Traceback (most recent call last) in () 23 try: ---> 24 initialized 25 except NameError: NameError: name 'initialized' is not defined During handling of the above exception, another exception occurred: Exception Traceback (most recent call last) 1 frames in get_Tactron2(MODEL_ID) 129 gdown.download(d+MODEL_ID, tacotron2_pretrained_model, quiet=False) 130 if not exists(tacotron2_pretrained_model): --> 131 raise Exception("Tacotron2 model failed to download!") 132 # Load Tacotron2 and Config 133 hparams = create_hparams() Exception: Tacotron2 model failed to download!

hm

did you set it to

Anyone with the link

oh no

sorry lol

@User every time I type into the synthesis, do I need to put in punctuation?

yes

uberduck puts in a period at the end but the notebook doesn't

For replacing the English-Russian model, I made a g2p that did this The problem comes with synthesis of the audio, as I didn't have a metadata.csv file (and for the LJ dataset I still can't find one)

Is this good attention? I would say so but I'm not sure

This is good stonks!

and it reached 0.15 loss already, in my opinion the model is pretty much done

Do you think I should stop training the model now?

yup

Damn...It produces this every time I make it say that line, would there be a way to fix this if you know or no? (within the model itself)

Would training it up to 250 epoch help it?

did you put a period at the end when synthesizing? could be that

Yeah I added a period

and there were periods on your dataset aswell i assume

Oh yeah of course

weird, send me the dataset in an hour so i can see if theres something wrong with it

My upcoming Billy West Ren update gots stonks on the top at epoch 10!

so im training some ripped voice clips for Peach from New Super Mario Bros 2 but idk what the fuck im doing

i'll send the voice to the uberduck website and see if they can fix it

im using glowtts

colab