Couldn't start the model training process

[stark zealot]

The dataset is 366mb in size and 50m in length. Not sure if it helps but I'm trying to specify everything that I could. My issue is that I can't start training the model. Info is in the screenshot. My PC's running on an i5-12400F and an RTX 3060.

[spice ingot]

Ayo? @stark zealot level 1 !!!

[steady light]

in preprocessing, switch to v1 then back to v2, select 32k again, and redo the preprocessing

[tidal finch]

I believe the problem may have been the number of CPU processes. Try using 2 and see if it works.

[graceful saddle]

show me the training page

it might be batch size issue

[graceful saddle]

save frequency 5 is insane

change it to 20 or 25

[stark zealot]

I'm just going with whatever the guide says

sorry if that was bleach to your eyes lol

[graceful saddle]

untick 'save a small final model to the weights' folder

[stark zealot]

k

[tidal finch]

actually keep that ticked

cuz if you wanna test a previous model

you can do that

[graceful saddle]

you can generate that from D and G anyways

[steady light]

nah it's fine if there's enough storage space

[tidal finch]

• every .pth is like 50 mb so not bas

[tidal finch]

It can kill the nand on the SSD tho

Constantly writing and overwriting a gigabyte of data...

[stark zealot]

I'd agree that it'd kill the SSD

[graceful saddle]

D and G is like 1gb combined

[tidal finch]

yea exactly

[stark zealot]

🤔

[stark zealot]

2 of what sir

batch size?

[steady light]

500 epochs/5 = 100 pth files
each is ~55 mb, then total 5500 mb

[tidal finch]

CPU processes, as i mentioned earlier

[tidal finch]

exactly

that's why they should set save frequency to 10 or 25

[stark zealot]

oh great it's 12

how did I not notice that

[graceful saddle]

i have 32 cores and i uses 22, works fine

[tidal finch]

can give errors tho

[graceful saddle]

probs im lucky

[steady light]

10 is still fine in order to test and find the best model snapshot

[tidal finch]

yeah exactl

[steady light]

then store the RVC and training files in HDD despite being slow

[stark zealot]

what am I even looking at

[tidal finch]

og my god

the name

has a space

and that was the problem

😭

Remove the space from the experiment name and the dataset folder

It should work

[graceful saddle]

bruh

[stark zealot]

modern problems require modern solutions

[graceful saddle]

i didnt spot that

but ik i cant have spaces in my data or exp name

[stark zealot]

I got so caught up with programming in Python that I forgot having a space in the name is an issue

[spice ingot]

Ayo? @stark zealot level 2 !!!

[graceful saddle]

sadge

[stark zealot]

issue resolved, thank you all for the help and suggestions.

[tidal finch]

WOOO

Glad it worked :) you're welcome

[stark zealot]

ngl that icon looks like gem berry from stardew valley

[tidal finch]

That's the FL Studio logo

[stark zealot]

ik

[tidal finch]

Carrot 👍

[stark zealot]