Cleaned using RX 10 MVSEP Audacity
Dataset : 2 Seconds
Batch File: 16
Trained & Tested On Applio(Kaggle)
https://huggingface.co/notiscBalbalEko/myrvcmodel/resolve/main/BalbalLaugh_2500e_5000s.zip?download=true
hope u like my laugh >:)
#Balbal Laugh(RVC V2)(RMVPE)(Original)(40k)(2500 Epochs - 5000 Steps)
harsh stone
worn rivetBOT
worn rivetBOT
hoary agate
You could have looped the dataset to 1 minute and train to 800 epochs instead, but good job xD
harsh stone
You could have looped the dataset to 1 minute and train to 800 epochs instead, b...
it was training for four hours
and stopped at like 7pm
hoary agate
it was training for four hours
I made a toolgun model with looped dataset to 1 min and 800 epochs (also batch file was set to 8) and it took way less
Don't tell me you're using P100 GPU or something
But I don't say that the model is bad yk
harsh stone
Don't tell me you're using P100 GPU or something
nop the t4
hoary agate
nop the t4
Strange, a very high batch size with a 2 second dataset training for too long
harsh stone
my internet was shet
hoary agate
It's not your internet, probably the GPU being the GPU
harsh stone
It's not your internet, probably the GPU being the GPU
or kaggle being kaggle
hoary agate
lol
magic totem
You could have looped the dataset to 1 minute and train to 800 epochs instead, b...
who told u:
to loop the dataset
x amount of epochs mean better model
because both of those are wrong lmao
hoary agate
who told u:
to loop the dataset
x amount of epochs mean better model
because bo...
no fucking way now making a non voice model is hard xD
hoary agate
who told u:
to loop the dataset
x amount of epochs mean better model
because bo...
I know:
looping the dataset is necessary just in case, model makers like mrsneakydick loop their datasets for non voices and whatsoever and the models sound good
more epochs doesn't mean better model
so I'm not wrong probably
magic totem
I know:
looping the dataset is necessary just in case, model makers like mrsneak...
About the quality, that depends on the dataset, good quality datasets are gonna always sound good regardless of what u do 
(yes im charging my phone now ffs)
harsh stone
good thing i don't loop mine
hoary agate
About the quality, that depends on the dataset, good quality datasets are gonna ...
well then im not looping datasets on non voice anymore
if it's 1 second i don't give a shit and train it
magic totem
👍
harsh stone
<:trolley:1159468147133395025> 👍
hoary agate
harsh stone
okay stop messaging now cuz the huggingface link is gonna get blocked by our messages 
hoary agate
huggingface link is on the top, probably we're flooding the post 
arctic mason
or kaggle being kaggle
ew kaggle
arctic mason
whats wrong with kaggle?
It's confusing to me, I prefer Google collab
I personally don't like the website
harsh stone
It's confusing to me, I prefer Google collab
nah bruh google colab is trash it has a 4 hour usage
arctic mason
How long of a usage does kaggle have?
harsh stone
How long of a usage does kaggle have?
30 hours
arctic mason
Does it give the same quality with ai models?
harsh stone
Does it give the same quality with ai models?
yea
arctic mason
Oh
harsh stone
it uses two t4 gpus
harsh stone
bruh i just woke up
arctic mason
I didn't know that
harsh stone
ask @hoary agate
hoary agate
Could you show me how to set it up tho
ask litsa
compact crow
no fucking way now making a non voice model is hard xD
Looping same data gives 0 advantages
hoary agate
Looping same data gives 0 advantages
Yea ik
lofty finchBOT