Rate the model
To rate this model, either use /rate or click the buttons below.
#KLM v7s Batch II (PRE-TRAINED MODEL)
clever nestBOT
clever nestBOT
toxic talon
tiny owl
<:Baffled:1159397286833557566>
red robin
<:nails:1159569314848972891>
Is this another version of the pretrain you posted before?
tiny owl
Is this another version of the pretrain you posted before?
Yes, batch 2 has more data than batch 1
oh and um.. 40k model is still unstable. I forgot to notice.
red robin
oh and um.. 40k model is still unstable. I forgot to notice.
But the 32 and 48k ones?
And what do you mean by "unstable"?
tiny owl
But the 32 and 48k ones?
It seems that the 40k model is overtrained. It is currently being retrained
tiny owl
But the 32 and 48k ones?
If any issues arise, letting me know would be a great help! thank leo
red robin
If any issues arise, letting me know would be a great help! thank leo
You're welcome buddy.
Also.
How much hours of dataset does your Batch 2 pretrain have?
If someday i manage to gather a dataset for making a studio session-based pretrain, you'll be able to train it for me?
It's some sort of pretrain i have planned.
tiny owl
It's some sort of pretrain i have planned.
I haven't done the calculations, but it seems to have increased about double compared to batch 1. considering it's 30 minutes per actor for 22 people... that's quite a lot
red robin
I haven't done the calculations, but it seems to have increased about double com...
Probably my pretrain wont get past 4-7 hours.
tiny owl
Probably my pretrain wont get past 4-7 hours.
I don't think cramming in too much data leads to noticeable changes. In that sense, 5 to 7 hours might be more appropriate. imao
tiny owl
If someday i manage to gather a dataset for making a studio session-based pretra...
sure why not
toxic talon
oh and um.. 40k model is still unstable. I forgot to notice.
i got error on 40k
tiny owl
i got error on 40k
Thanks! I'll check it out.
tiny owl
i got error on 40k
are you using applio? it seems fine to me
toxic talon
are you using applio? it seems fine to me
hmm when i try original pretrain it gives no error, let me try again
using rvc mainline
tiny owl
hmm when i try original pretrain it gives no error, let me try again
can you try with applio?
tiny owl
hmm when i try original pretrain it gives no error, let me try again
what about the others? 32k and 48k?
tiny owl
Sorry! Please Use Applio until the issue is resolved 
grim turtle
can I ask how many hours that your pretrain dataset was trained on?
tiny owl
can I ask how many hours that your pretrain dataset was trained on?
oh hi elsa, around 2 days? or 1.5 days?
tiny owl
can I ask how many hours that your pretrain dataset was trained on?
it was less than 48 hours for sure.
grim turtle
it was less than 48 hours for sure.
I see, just korean language alone?
tiny owl
I see, just korean language alone?
Yes~ while it does include some english documents, and pop music but the majority is overwhelmingly in korean
grim turtle
alrighty thanks