✨│ai-help

[proper shale]

But check #1159949278270193734

Hm... I've seen this error before but only when people are training

I'm not sure how to solve it to be honest

[plain sparrow]

Ok. Damn. But thanks again

[brittle wing]

Sorry but can you give me the user manual link, I'm a little bad at English

[proper shale]

#1159513888199540817

[plain sparrow]

This might help a lil. I don't understand much of what it says lol

[sand iris]

You are using a GPU with 2GB of VRAM which is really low for RVC stuff

[sage matrix]

is it better to stay on mangio or switch to applio? (local)

[crimson basalt]

I switched to applio and found it to run abit smoother and that it gives better results. Don't delete mangio first until you try applio just incase you don't like it. Also applio gets regular updates mangio don't and you get more features with applio like tts,

[sage matrix]

understood

[sand iris]

Mainline RVC seems to work fine as well

[proper wing]

Anyone know how I can stop any loss?

[steel isle]

what is S.R.

[proud elbow]

audio sample rate

[sour prawn]

do i need lots of gpu or ram to run rvc

[brittle wing]

which one do i download for intel gpu

[pearl cove]

how can i make ai voice

[marble timber]

bros how to get model master role
jointeam don't work

[brittle wing]

it says this for me when trying to run easy gui

[proper shale]

You need a good GPU.

[round cairn]

-rvc

[azure marsh]

RVC Guides (how to make ai cover)

🍏 Applio-RVC-Fork: Guide (EN/ES)

https://docs.google.com/document/d/12YmIQFLFzth1YsZJMGrA0aeTN0Go8DNZVw8LFSf-Hug

Tip: You can also watch this video tutorial made by iroaK if you need further help.

RVC v2 Tutorial for Paperspace (By LollenApe)

https://docs.google.com/document/d/1lIAK4Y0ylash_1M2UTTL_tfA3_mEzP0D2kjX2A3rfSY/edit?usp=sharing

RVC Guides in Other Languages

• Russian FAQ for RVC/Voice changer
• Italian Guides or -it_g
• RVC Guides in PT-BR (-rvc_br)

[round cairn]

-colab

[azure marsh]

Google Colabs

RVC

AI-Cover-Gen-No-UI

https://colab.research.google.com/github/ardha27/AICoverGen-NoUI-Colab/blob/main/CoverGen_No_UI.ipynb

Note: This colab is for inference only.

RVC v2 Disconnected

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS

Note: This colab is for training only.

RVC colabs translated to other languages

• PT-BR: -colabs_br

Audio Separation/Isolation

MDX Net Colab

https://colab.research.google.com/github/NaJeongMo/Colab-for-MDX_B/blob/main/MDX-Net_Colab.ipynb

UVR arch colab 5.0.2/c1.04

https://colab.research.google.com/drive/16Q44VBJiIrXOgTINztVDVeb0XKhLKHwl

KaraFan colab

https://colab.research.google.com/github/Captain-FLAM/KaraFan/blob/master/KaraFan.ipynb

[sour prawn]

is 32 gbs good

[rare goblet]

@sour prawn has reached level 1.

[sour prawn]

and 8 gb of ram

[proper shale]

... that's just RAM

what's your GPU

[still crystal]

how do i make ai covers with mangio local?

[pliant ingot]

-help

[azure marsh]

Command has expired

[sour prawn]

32 gbs

nvidia

[proper shale]

...give me a name

[sour prawn]

where do i check

[proper shale]

Task Manager

[pliant ingot]

Ok so i used this RCV Model- https://drive.google.com/file/d/1PdFVcq1uuL-98YM_w8h5plX9-fkaRYpi/view?usp=drive_link
But them when i use it, it does not work

[rare goblet]

@pliant ingot has reached level 1.

[pearl cove]

how to add the model to mangio rvc

and how to hear voice in real time

[proper shale]

That model's private - can't do anything about it...

[pliant ingot]

i trained using the RCV v2 Disconnect

[proper shale]

If you checked the guides you'd know

[proper shale]

oh

change sharing options

on the zip file

[pearl cove]

and that

[rare goblet]

@pearl cove has reached level 1.

[proper shale]

click on Share then alter the link so that anyone with the link can see

this'll surely work

[proper shale]

Download W-Okada

[pearl cove]

alr

[proper shale]

And check the guides for it

[pliant ingot]

https://drive.google.com/file/d/1PdFVcq1uuL-98YM_w8h5plX9-fkaRYpi/view?usp=sharing here

[pearl cove]

thx bro

[proper shale]

Works now

[proper shale]

Np

[pliant ingot]

thx i used the public link. it works now

[proper shale]

Yeah thats why

Always remember to do that

[copper pebble]

someone from the staff who can view my request in #1159514067187277865 ???

[obtuse stream]

rmvpe
rmvpe_gpu

wich is better?

whats the different

who trains with what number of epochs?

[proper shale]

those numbers vary

[pliant ingot]

Approx. how much time will it take to make an AI cover on Huggin Face

[proper shale]

takes like 200-300 secs

depends on song size really

[pliant ingot]

thx

[obtuse stream]

i use 222 voice sample what is the optimal epoch number?

[rare goblet]

@obtuse stream has reached level 1.

[proper shale]

Again, it's impossible to determine

You have to analyze the graphs

[obtuse stream]

how?

what is it to you?

400 is ok?

[potent root]

i suppose so

but consult the graph

i do 750 and just rely on the graph's lowest point

[obtuse stream]

i use this https://www.youtube.com/watch?v=rnis2Hst9ec&ab_channel=TJKCreative

i dont know where the graphs

[cyan spear]

I keep trying to make an RVC voice model, but every single time I try to, it always has an error, can someone help me out?

[rare goblet]

@cyan spear has reached level 1.

[proper shale]

What does CMD say?

[cyan spear]

CMD?

What’s CMD?

Hello?

I’m using easy collab

Gradio

I’ll try it again

[lime bough]

still happening & wdym with audio file if im not getting it to start

[cyan spear]

Do I run local URL?

Or public URL?

[pearl cove]

So mangio is for audio and okada for real time

[cyan spear]

I give up with gradio

[proper shale]

Yes

[cyan spear]

Ima see if there’s any other way to make an RVC model

[proper shale]

Oh

[pearl cove]

Gotchu

[proper shale]

Use the one in #1159949278270193734

[cyan spear]

Which one

[proper shale]

RVC Disconnected

[cyan spear]

Ok

[proper shale]

Just format your dataset in the right way and it should be straightforward

[cyan spear]

How do I do that

[rare goblet]

@cyan spear has reached level 2.

[proper shale]

Put all your audios in a folder and then zip

[cyan spear]

Ok

[proper shale]

Just that really

[cyan spear]

Ok I put my audios in a file

Now what

[proper shale]

Zip, then add to a folder in your Google Drive named "dataset"

[cyan spear]

How do I zip the file?

I figured it out

Ok how do I get to google drive dataset thingy

[proper shale]

Create a folder there

[cyan spear]

how do i get to it

[proper shale]

...you have to create a folder named dataset

[cyan spear]

oh

a new one?

or the rename the audio zip file i js made

[proper shale]

yeah s

[cyan spear]

ok

[proper shale]

and put that zip on the dataset folder

[cyan spear]

okj

i did it

now what

[proper shale]

now ur gonna go on rvcDisconnected and do the steps

[cyan spear]

which are

it looks super complicated

[proper shale]

its simple

[mortal turret]

guys, im downloading 🍏 (Appilo) and its been stuck on this screen for like 20 mins, is it done or is it a bug or WHAT

[rare goblet]

@mortal turret has reached level 1.

[cyan spear]

which do i start with

and what order do i go in

[proper shale]

from first to last, skipping the ones that say (for retraining)

[cyan spear]

ok

so do i run the one that says

run me first

[proper shale]

...yeah

[cyan spear]

ok

How do I know it’s finished running

do i js need to wait until it stops?

[proper shale]

until it shows the green tick on all of em

[cyan spear]

ok

ok they all have green ticks

now what

[proper shale]

now you do the other steps, change the settings u want

should be eas

[proper shale]

sadly i dont have enough time to tell u everything so

[cyan spear]

Is it in set training variables?

[proper shale]

ig ur gonna have to deal with that stuff on your own, sorry

yes

[rare goblet]

@cyan spear has reached level 3.

[cyan spear]

dawg what do i do here

before you go

[proper shale]

the zip name

[cyan spear]

just copy the zip name?

[proper shale]

yeah

[cyan spear]

i did

and this happened

[proper shale]

should be .zip

[rare goblet]

@proper shale has reached level 18.

[cyan spear]

oh

[proper shale]

oh nvm wait

u have to put the dataset on rvcDisconnected

forgot lmfa

[cyan spear]

how

[proper shale]

Look I'm sorry but I can't keep helping you here, I know in the beginning it's confusing, but I can't keep spoonfeeding you on every step. Sorry.

[cyan spear]

bruh

theres no instructions

[proper shale]

The steps say them tho

[cyan spear]

bro i dont even know how to put the dataset in rvcdisconnected

ive only ever used gradio

thingy

so this stuff is new to me

[livid isle]

how do i fix this D:

[proper shale]

reinstall python and check add to path

[livid isle]

ok imma reinstall python

[cyan spear]

the main problem is i dont know how to get to google drive

im tryna figure out how to do this

[proper shale]

...you create a google drive folder named "rvcDisconnected" and move the zip file to it?

[cyan spear]

how do i make a google drive folder???

i have a folder called dataset

and my zip file is inside that

[proper shale]

you could just... rename the folder

[cyan spear]

ok

so with the " or without?

[proper shale]

without

[cyan spear]

ok

ok i renamed the folder

what now

[proper shale]

try the step again

[cyan spear]

its still doing the same thing

[uncut forge]

FileNotFoundError: [Errno 2] No such file or directory: '/content/drive/MyDrive/project-main/logs/Tommy /0_gt_wavs'

huh

[safe whale]

Ah, interesting... thanks!

[oblique rain]

Hey, I'm sorry, but this is not part of "how to learn to use RVC." Not knowing what Google Drive is and how to use it is a basic lack of knowledge on your part, and we can't provide assistance for these types of questions either.

[terse nymph]

like on the website

[rare goblet]

@terse nymph has reached level 2.

[terse nymph]

u just go to ur google drive page, right click the empty space, 'new folder'
name it what it says, then do the thingy

[formal zenith]

Are there any disconnection prevention methods for the disconnected colab

[rare goblet]

@formal zenith has reached level 1.

[glad zealot]

your zip file must have a folder inside of it, and inside the folder should be the audio files

less than a sec

[cyan spear]

Alr

[cyan spear]

Ohhhh

Okay

Thanks

Wait

I already do

Look

[glad zealot]

did you import the zipfile to rvcDisconnected folder in google drive?

[cyan spear]

I don’t know how

[glad zealot]

just make a folder in google drive called rvcDisconnected

go to https://drive.google.com

[cyan spear]

Ok

[glad zealot]

on the top left, there should a bug + icon

press it and press the new folder

set the name to rvcDisconnected

then drag your zip file there

[glad zealot]

yup

[cyan spear]

so i click the +new then?

[glad zealot]

nah just open it

since you already have it

[cyan spear]

oh

alr

[glad zealot]

then drag and drop your zip file

[rare goblet]

@cyan spear has reached level 4.

[glad zealot]

why is it .zip.zip

[cyan spear]

because i renamed the original file .zip at the end

cuz i was told to

[glad zealot]

lol

[cyan spear]

but then they stopped telling me how to do it lol

[glad zealot]

thats fine anwyas

go to colab now

[cyan spear]

alr

[ember ermine]

open your file browser in your computer

[glad zealot]

on the part where it asks for your zip file

paste the whole zip file name

[cyan spear]

my rvc v2 disconnected fuckin went red and shit

do i have to do it again

[glad zealot]

just restart it

nah

[cyan spear]

how

[glad zealot]

restart the red part

[cyan spear]

ok

[glad zealot]

press the red circle

[cyan spear]

alr

[glad zealot]

before that tho

change the name to also be .zip.zip

[cyan spear]

ok

[glad zealot]

cause thats your file has that

[glad zealot]

did you change the name before running it?

[cyan spear]

yeah

[glad zealot]

oh wait did you connect to your google drive

[cyan spear]

?

i dont believe so

[glad zealot]

ye tahts why lol

[cyan spear]

how do i do that?

[glad zealot]

wait lemme get that code

[ember ermine]

go to the top right

where it says connected

click that

delete runtime

[glad zealot]

nah

[ember ermine]

reload the page

[glad zealot]

dont

[ember ermine]

then do everything all over again

just to be sure

[cyan spear]

ok

[ember ermine]

or keep it messy and wrestle it lol gl

[glad zealot]

same thing anyways lol

[ember ermine]

true

click red circle get error and then click the red circle before it

until you get to the top

[glad zealot]

on the left side press the folder icon

[ember ermine]

then you have to reload

gl

[cyan spear]

ok i did

[glad zealot]

yup

press the 3rd icon

the white folder

it should add a new cell

[glad zealot]

yup

then run that

[cyan spear]

alr i did

[glad zealot]

then press the red circle from earlier

it should work now.. maybe

[glad zealot]

fck

at least its a new error XD

[cyan spear]

lmao

[glad zealot]

thats the new rvcDisconnected right?

[rare goblet]

@glad zealot has reached level 11.

[cyan spear]

yeah

[ember ermine]

much better than reloading tbh

[glad zealot]

ah found it

[glad zealot]

scroll up to the cell before that

this one

[brittle wing]

is 250 epochs good on a 6 minute dataset using the rmvpe or whatever its called

[glad zealot]

hmmmmm wtf

[cyan spear]

is it because it doesnt have the .zip.zip

[glad zealot]

nah it doesnt need that

[cyan spear]

what do i do now

[glad zealot]

ngl i have no idea

can you run that cell again

then run the next one again

[cyan spear]

sure

lemme try

oh

it worked

[glad zealot]

lol

ig you forgot to run it?

[cyan spear]

maybe

[glad zealot]

yup

btw why are you using harvest?

[cyan spear]

idk its just what i use on rvc

[violet heron]

i use harvest

[glad zealot]

meh own preference ig

[violet heron]

it works when you know how to train with it

[glad zealot]

i dont :D

im new at this

XD

[violet heron]

my mysticat model sounds like a mangio model

[glad zealot]

yes

[brittle wing]

a mysticat model? can you link please?

[rare goblet]

@brittle wing has reached level 2.

[glad zealot]

thats what i use ngl

[cyan spear]

so what do i do now

[violet heron]

here https://discord.com/channels/1159260121998827560/1159290382023147650

[glad zealot]

just run the preprocessing cell and feature extraction

[cyan spear]

ok

[brittle wing]

tysm i love her squeaky voice

[glad zealot]

then just set the settings for this and run it

[glad zealot]

fck

WHYYYYYYY

go to the first cells

that has red

run those

[ember ermine]

dependencies wasnt run preperly

[mossy pasture]

maybe your bitrate is wrong? try get 40k instead of 48k

[ember ermine]

the gpu check

[compact wraith]

restart runtime

[rare goblet]

@compact wraith has reached level 7.

[compact wraith]

and try again

[glad zealot]

not that

[ember ermine]

nah he will never restart

[ember ermine]

he will waste 5 more hours listening to shit advice lol

ping me later if u need help

[cyan spear]

atleast theyre helping me though

[mossy pasture]

click x

[glad zealot]

lol

[brittle wing]

heard people say to use a alt when training

[ember ermine]

u need to reload

[brittle wing]

did you try that?

[ember ermine]

100 percent

[compact wraith]

i just got here, so i know these type of errros

[glad zealot]

ye i think so too

[ember ermine]

anyone telling you te press random red circles is wasting your time

[violet heron]

Use a alt accoumt

[glad zealot]

QwQ

[violet heron]

dont know whats causing that

[cyan spear]

a different gmail?

[violet heron]

yep

[ember ermine]

firstly

[mossy pasture]

i feel like your zip is wrong, you just need to name it cybermanaudio, not cybermanaudio.zip

[rare goblet]

@mossy pasture has reached level 2.

[ember ermine]

open your file browser

[cyan spear]

wont i have to do everything again

[ember ermine]

on windows

[glad zealot]

ye feels like your forgot to run some cells ngl

[cyan spear]

which ones??

[brittle wing]

Yay! My model is training!

[cyan spear]

i ran all of em

[glad zealot]

¯_(ツ)_/¯

[compact wraith]

Maybe it could be your pretrained path, probably somthing wrong is there

[cyan spear]

everything has green ticks aside from the newer shit

[ember ermine]

dude im trying to help and keep getting ignored xD

ill just try 1 more time

[glad zealot]

imma go make another model tomorrow so imma check if i get taht error

[cyan spear]

im listening im just looking at what everyone is saying

[ember ermine]

ok

open your file browser

click view

at the top

[glad zealot]

dang it i dont have the older version of it.. i was using it earlier no error

[cyan spear]

wheres view at

[proper shale]

apparently smth went wrong with the installation

[ember ermine]

u use windows right?

[cyan spear]

yeah

[glad zealot]

@proper shaleu use colab?

[cyan spear]

how do i solve that

[ember ermine]

click on file explorer

[glad zealot]

to train

[ember ermine]

at the top bar there will be file, edit and on the right view

[proper shale]

yes

[ember ermine]

click it

[glad zealot]

tried kaggle?

[proper shale]

for everything

yeah back in the applio days

when uh

applio banned

[brittle wing]

should I try applio?

[ember ermine]

not right

or left*

at the top

[glad zealot]

is there a way to prevent kaggle from dying of inactivity?

[ember ermine]

right under the wordss file explorer

[ember ermine]

above taht

[brittle wing]

i think you can use a autoclicker idk

[ember ermine]

that*

[proper shale]

im not sure

[glad zealot]

its 2 am AAAAAAAAAAAAA

[glad zealot]

i want to run it in the background

on colab you can just run a script that will press the conenct button every 30 sec or something

[brittle wing]

use the thing on a different device?/

[glad zealot]

dont have :D

[quaint nacelle]

files

[glad zealot]

i only have my potato laptop

that can somehow run vrchat

[mossy pasture]

i mean, it will still disconnect you after like 4-5 hours no matter what

[glad zealot]

kaggle disconencts after 15 mins i think

[rare goblet]

@ember ermine has reached level 3.

[ember ermine]

click view

[mossy pasture]

and i still have no idea why sdafsa is trying to ask britt to open file explorer when britt is using colab

[ember ermine]

then click the box next to file name extensions so that it is ON

[proper shale]

... I'm sorry here but...

that is a folder that should be in your Google Drive

[cyan spear]

ight

[ember ermine]

done?

[proper shale]

Colab does not read the files on your machine

[proper shale]

😭

[ember ermine]

ok good

now you'll never do .zip.zip again

[glad zealot]

ye they are doing that, just need to shoe file extentions cause it showing as .zip.zip lol

[cyan spear]

yes sir

[ember ermine]

now go to drive.google.com

[rare goblet]

@cyan spear has reached level 5.

[proper shale]

oh ok

[ember ermine]

once you are on drive.google.com go to my drive

name the folder rvcDisconnected

[cyan spear]

ok

[ember ermine]

and put the zip folder in it

[cyan spear]

i alr got it

[ember ermine]

if you have already uploaded it

[ember ermine]

then rename it so it doesnt say .zip.zip

just one .zip

[cyan spear]

ok

[ember ermine]

rename it

to cybermanaudio.zip

[glad zealot]

@proper shaleril?

[ember ermine]

well done!

now go to the colab page

[cyan spear]

yay

[glad zealot]

fck why is my alt acc on google on light mode

[ember ermine]

that u use for training

[ember ermine]

ok

top right

there will be a connect tab

[proper shale]

reel

[ember ermine]

there will be a down awrrow next to it

good

[glad zealot]

restart that whole got damn thing

[ember ermine]

the arrow next to RAM and DISK

click it

[cyan spear]

ok

[ember ermine]

now click disconnect and delete runtime

[cyan spear]

ok

i did

[ember ermine]

once you have done that refresh the page

ok

now

[cyan spear]

alr done

[ember ermine]

go where it says dependenceis

and click the first play button

don't need to expand the tab

[cyan spear]

k

[ember ermine]

for dependencies

[cyan spear]

im doing that rn

[ember ermine]

good good

[glad zealot]

@proper shale last question before i sleep.. does kaggle let you use it continously for 30 hours or it will auto disconnect after some time even when actively using it

[ember ermine]

just the button underneath where it says "Run me first"

and now simply wait

it will eventually prompt you to connect to drive

[cyan spear]

yeah

[proper shale]

ehhhhh i

dont know

[ember ermine]

tell me when that happens

[cyan spear]

ok

ok its

prompted me

now what

[glad zealot]

@proper shaleanother last dum question.. how do you connect google drive to kaggle.. if its even possible... i wanna train on kaggle and auto save the pth files to my drive so when it disconencts its safe

[ember ermine]

did u connect?

click connect to drive

[cyan spear]

not yet]

ok

[ember ermine]

it will pop up

[proper shale]

i dont think u can

[ember ermine]

and let you select your google account

click it

and click authorize

[cyan spear]

im connected now

[ember ermine]

ok

wait until dependencies is all done

[cyan spear]

ight

[ember ermine]

has it finished?

look at the last cell within the dependencies tab

[cyan spear]

yeah

[ember ermine]

it is called Setup CSVDB

[ember ermine]

ok good

now

choose a name for your experiment

[cyan spear]

ok

[ember ermine]

something like cyberman

to keep it simple

so where it says experiment name put in cyberman

[ember ermine]

good

now click play

until it shows a green checkmark

[ember ermine]

ok

now expand the preprocessing tab

you will see load dataset

[ember ermine]

yep

expand it

dont play it yet

[ember ermine]

good

replace zipfile.zip with cybermanaudio.zip

is that what it is called in your google drive?

[cyan spear]

yeah

[ember ermine]

ok good

wen u do that click play

[ember ermine]

ok

seems good

play

[ember ermine]

ok

now to the next cel

cell*

preprocessing

play that one

[cyan spear]

k

[ember ermine]

once that's done click feature extraction underneath it

[cyan spear]

it said end preprocess

[ember ermine]

green check mark?

[cyan spear]

ye]

[ember ermine]

then it succeeded good

now do feature extraction

[cyan spear]

ight

[ember ermine]

btw how big os ur dataset?

because that was kinda fast

[livid isle]

reinstalled python but still didnt work :(

[rare goblet]

@livid isle has reached level 1.

[ember ermine]

but whatever we are here to learn training

[cyan spear]

all feature done

[ember ermine]

ok

[cyan spear]

you mean the audios?

[ember ermine]

now save preprocessed data to drive

[cyan spear]

i have a really good pc

k

[ember ermine]

how big it the .zip file you uploaded?

and how long is the total audio time in it?

[cyan spear]

like

13 mins

[ember ermine]

should be enough

[cyan spear]

its the same audio renamed and replicated

[ember ermine]

if it's clean

[cyan spear]

yeah it is

[ember ermine]

nope

if it's the same one file then it might not be good

you can't scam the training like that lol

[cyan spear]

itl be fine

hopefully lma

[ember ermine]

but whatever we are here to learn training

now

expand training tab

[cyan spear]

k

[ember ermine]

since you have a small dataset set save frequency to 20

leave total epochs alone

[ember ermine]

click the arrow on the left of training

[ember ermine]

ok

now in the training tab

set save frequency to 20

leave total epochs at 500

[cyan spear]

thats a lot

itll take ages

[ember ermine]

save frequency has no effect on training speed

[rare goblet]

@ember ermine has reached level 4.

[cyan spear]

ok

[ember ermine]

it sometimes halts training to save the model

but increasing it makes it go faster n

now slower

[ember ermine]

not slower*

anyways

batch size make it 20

this will actually make it go faster

[cyan spear]

ok

[ember ermine]

now click the play button

[cyan spear]

ight its playin

thanks for helping me man

[ember ermine]

np sir

[cyan spear]

couldnt have done it without you

[ember ermine]

you still need to learn to save and resume

[cyan spear]

yeah

[ember ermine]

and when to know u actually have a model

so, now that it is playing you will get a little window inside of the training tab

called tensorboard

let me know when it shows up

[ember ermine]

very well

wait a bit longer

[cyan spear]

until trainings done?

[ember ermine]

no

for your first 20th epoch

[cyan spear]

ok

its at 4 rn

[ember ermine]

seems like it is doing 1 epoch every 33 seconds

so that should be 10 mins

[cyan spear]

ok

[ember ermine]

send another screenshot

[ember ermine]

o ok even better

24secs/epoch

so that's less than 10 mins

[cyan spear]

ok

[ember nest]

how i can find the links?

[ember ermine]

go to guides channel

[ember nest]

#1159513888199540817 ?

[ember ermine]

yes

the one being used here is rvcDisconnected on colab

https://colab.research.google.com/drive/1XIPCP9ken63S7M6b5ui1b36Cs17sP-NS#scrollTo=HDsTxpbTqHol

here it is if u can't find it

[ember nest]

not works o yes?

[ember ermine]

worked for me

[ember ermine]

good

now

the yellow strip

that says tensorboard

on the right side

there will be a reload button

click it

[ember ermine]

yesp

right of inactive

there will be a refresh button

click that

[cyan spear]

ok

[ember ermine]

good

click on scalars

right of time series

[cyan spear]

k

[ember ermine]

now where it says filter tags

search for g/total

you'll get something like loss/g/total

as a graph

[pseudo flint]

can I AFK train in huggingface?

[ember ermine]

yep that one

see those 3 blue buttons?

at the bottom?

[cyan spear]

ye

[ember ermine]

click the left most

then the right most

now go on the left side, there will be a slider called smoothing

[ember ermine]

max that out till it says 0.999

[cyan spear]

wym

[ember ermine]

left of this graph

[ember ermine]

look here

[cyan spear]

yeah

[ember ermine]

at the left

there's a slider

called smoothing

[livid isle]

somebody help me pls

[cyan spear]

ok

[rare goblet]

@cyan spear has reached level 6.

[ember ermine]

search windows for system variables

[ember ermine]

ok

now

[livid isle]

wdym?

[ember ermine]

you see the graph? that line is gonna keep going down

[cyan spear]

ok

[ember ermine]

give me like 5 minutes until i finish with britt and i will help you 🙂

[ember ermine]

ok

it will keep going down

eventually...

it will flatten

horizontally

[livid isle]

okk

[ember ermine]

once that happens

your training is done

[cyan spear]

ok

[ember ermine]

now

[cyan spear]

cant i just watch the epochs?

[ember ermine]

you can

but not all datasets finish at the same epochs

[cyan spear]

ok

[ember ermine]

so its better to use the graph

but

[cyan spear]

alr

[ember ermine]

one thing to nice

colab will probably disconnect before that

[ember ermine]

since it's free and wont run forever

so. we will learn how to save and resume

now

once it says 41%

pause this training cell

[cyan spear]

ok

[ember ermine]

yep

click that

pause it

[cyan spear]

ok i did

[ember ermine]

it will stop and give u a green checkmakr

now, press the one right under it

[ember ermine]

that says export model from notebook to drive

[ember ermine]

all good

normal

[cyan spear]

where

[ember ermine]

right under this cell

[ember ermine]

there's another one

after this cell under it

there's a cell that says Export model from notebook to drive

[ember ermine]

run it

[cyan spear]

k

says all done

[ember ermine]

excellent

congratulations u have trained a model

u can now go to drive.google.com

[cyan spear]

is it done??

[ember ermine]

not at all lol

[cyan spear]

lol

[ember ermine]

technically u can use it now

but only 40 epochs will sound awful

[ember ermine]

so, let's RESUME training!

[cyan spear]

ok

[ember ermine]

it's in the folder

called cyberman_e40 or something

[cyan spear]

what is

[ember ermine]

your model

[cyan spear]

oh yeah

i see it

[ember ermine]

ok

[cyan spear]

lets resume

[ember ermine]

good

resume time

go back to colab

[cyan spear]

k

[ember ermine]

now. let's IMAGINE that you're doing this after rebooting your computer

imagine ur doing this tomorrow

you will login

run dependencies just like u did earlier

[cyan spear]

you want me to run dependencies at the top again?

[ember ermine]

then you'll go to set training variables

[ember ermine]

nononono

[cyan spear]

ok

[rare goblet]

@ember ermine has reached level 5.

@ember ermine has reached level 5.

[ember ermine]

we are just pretending ur resuming in a different session

it is what u would do if u were to do this in another time

[cyan spear]

ok

[ember ermine]

you will run dependencies

you will go to set training variables

you will set the exact same training variables as u did today

important

which in ur case

just set experiment name to cyberman

[cyan spear]

mhm

[ember ermine]

and leave everything else the same

ok

[cyan spear]

do i just click play to resume

[ember ermine]

now u will SKIP the whole processing tabs

[ember ermine]

not yet

[cyan spear]

ok

[ember ermine]

we're pretending ur resuming

so let's do it properly

once u skip preprocessing ull be in training

now

click load preprocessed dataset files

it will be the first tab in training

[cyan spear]

now?

[ember ermine]

says (for resuming)

[cyan spear]

or when its another time

[ember ermine]

yes

it wont be a problem to do it npw

[cyan spear]

ok

[ember ermine]

so might aswell

once it's done, let me know

should be quick

[ember ermine]

run it

[cyan spear]

got a checkmark

[ember ermine]

ok

now the next one

[cyan spear]

ok

[ember ermine]

Import model from drive to notebook

set stepcount to 2333333

right now it says 000

make sure it's 2333333 before u play this

[cyan spear]

k

[ember ermine]

and after this, just click training 🙂

and you're done!

[cyan spear]

do i not run import model?

[ember ermine]

hold

on

[cyan spear]

ok

[ember ermine]

once u set it to 2333333

before u start training again

like i said

make SURE your parameters are the same as last time

save frequency 20

[cyan spear]

ok

[ember ermine]

total epochs 500

and batch size 23

20*

[cyan spear]

yes

[ember ermine]

theoretically it shouldn't be a problem to change these across training sessions

[ember ermine]

but ive been told not to so lets be safe

yep

now if u click training it will resume from 40

instead of from the beginning

[cyan spear]

ok

ill click it now

[ember ermine]

yep

and that should be it

i recommend u keep it at 150 ish epochs per session on colab

before u pause and export

[ember ermine]

because if u disconnect before saving u will lose all progress since last save

[ember ermine]

good good

everything is working fine

now u know how to train and to resume

good luck!

[cyan spear]

ye

[ember ermine]

sorry it took a while

u still there?

[cyan spear]

do i go afk now

[ember ermine]

yep

[cyan spear]

k tysm

deadass

[ember ermine]

but not for too long

[cyan spear]

saved my day

[ember ermine]

no problem

just know

[cyan spear]

whast the max time

until it fucks up

[ember ermine]

sometimes colab will do those are you there prompts

with a captcha

to check if ur afk

just make sure to check on it once in a while

[cyan spear]

ok

so every 30 mins?

[ember ermine]

depends on the workload

i had it sometimes run for 6 hours sometimes not even 3

[ember ermine]

yea something like that

[cyan spear]

ok ty

[ember ermine]

np

hope the other guy it still around

is

[livid isle]

sorry i was watching youtube, hi!

[ember ermine]

i'll post it here too in case it might help someone one with the same error

go to windows search bar and search system variables, click on edit system environment variables
then under the Advanced tab, click Environment variables in the bottom right
in the upper section, click on Path, then edit
in the popup, click new
name it python (shouldn't matter what u name it)
once u name it, click on it and click browse
navigate to the folder where u have python installed (the one that has python.exe)
click ok for everything

do this if u get the "Program is not recognized as an internal or external" error in the cmd window when trying to run python

[rugged stirrup]

Can someone tell me whats wrong here?

I cant send messages

[rare goblet]

@rugged stirrup has reached level 1.

[rugged stirrup]

Oh nvm

There is whats wrong

[wheat copper]

how is this not working???/

[rare goblet]

@wheat copper has reached level 2.

[proper shale]

could be wrong account

[wheat copper]

its not

just checked

[lean owl]

i wish there was a guide for rvc disconnected

[ember ermine]

scroll up

[silk axle]

I have a problem that Mangio RVC doesnt see the dataset, does anybody know how to fix that?

[ember ermine]

i just has a long interaction with someone

detailed step by step walkthrough

[lean owl]

i really want someone to make a guide of rvc disconnected in a google doc so its easier to understand

[wispy burrow]

The dataset cannot be zipped.
that is one reason it can't find your Dataset.

[silk axle]

I just put a few .wav files in dataset folder and it still doesnt work

[wispy burrow]

Place those inside a new folder?

[proper shale]

ehhhhh im doing that

[rugged stirrup]

Could anyone help?

[silk axle]

I did as you said but still it just ignores the files

[rare goblet]

@silk axle has reached level 1.

[proper shale]

have you put in the folder path where it told you correctly?

[silk axle]

yes I did

[wispy burrow]

Is the path correct?

[rare goblet]

@wispy burrow has reached level 2.

[silk axle]

Yes it is

[hexed granite]

Is it possible to have too much audio in a dataset?

[proper shale]

yea

like 5 hours id say

[hexed granite]

Oops

Oh, nvm

I got about 40 minutes for one character

30 for another

[wispy burrow]

Did you extract the Mangio-RVC folder in OneDrive or something?

[proper shale]

That's very good

[hexed granite]

All perfectly divided into 10 second increments ofc

[silk axle]

No I extracted it in disk C

[wispy burrow]

Hmmm...

[proper shale]

....you didn't need to do that

[hexed granite]

oh

[rare goblet]

@hexed granite has reached level 1.

[hexed granite]

I’ve been lied to

Or maybe I misunderstood something

Yeah, that sounds more likely

[proper shale]

I mean

RVC does that automatically

[hexed granite]

I was definitely not informed of that

[wispy burrow]

There are a few people that keep saying you should "as it increases quality".
I really don't see why that would be the case though...

[proper shale]

...it makes your training times slower

[hexed granite]

As my roles imply, I’m still an rvc nub

[wispy burrow]

But RVC slices it regardless...

[proper shale]

A 30 min dataset with 1,2k files separated every 3-4 seconds is literally taking 1m 30s per epoch

[silk axle]

I tried to import dataset through that google colab thing for making AI covers, but result is the same

[hexed granite]

The training is faster if I glue all my audio together?

[proper shale]

yeah that wouldnt work either way

[proper shale]

If you separate every 30 mins you should be good

[silk axle]

I just noticed that every time I try to process dataset this Error pops up

[wispy burrow]

Ah, that could be that there are spaces or special characters in the path to the RVC folder.
It is recommended to not have any spaces or other special characters...

[silk axle]

Now I see, I had a space in the first folder name, thx for help

[wispy burrow]

So it works now?

[silk axle]

Yes it is, thx

[cyan spear]

what do i do once my training goes red

figurwed it out

[clear hound]

my rvc taking like 30 seconds before audio appears

[livid isle]

how do i use inference in rvc locally? do i have to download every model that i'll be using on my computer?

[mossy pasture]

1. you gotta have the RVC itself
2. just download the model from discord/huggingface, and put the .pth into the 'weight' folder, index on logs folder/just copy the path of index

[loud hazel]

hi all, hi One

[loud hazel]

u download what's u need

[mossy pasture]

sup dude, not really into helping anymore, didn't touch the newest updates much

[livid isle]

so i have to download all the models that i'll use? dang :(

[mossy pasture]

how much you wanna use lol