how we should put first before we convert the model to zip?

[wise cedar]

It doesn't matter what files you download first. Just make sure you have both the weight (.pth) and index (.index) files of your voice model downloaded and zipped together before uploading to Huggingface.

[wispy glacier]

Oh thanks

[wise cedar]

You're welcome.

[wispy glacier]

is MB of the pth file will be smaller if we trained the model with less epoch like 20 and below?

[wise cedar]

No. The file size (or "MB (megabyte(s))") of the voice models' weights (or the ".pth" files) will not be smaller if you were to train with less epochs. Each weight that you save will have the same file size no matter what frequency epoch (or "checkpoints") you've chosen.

[wispy glacier]

Oh okay

[wispy glacier]

Hopefully and praying that MB for pth files will be smaller tho because some people including me want to to save storage of our phone,tablet or laptop/computer

[wispy glacier]

Unfortunately easyGUI seems already disappeared and I don't want to use RVC v2 Disconnected because it's hard and doesn't work to me so easyGUI was only Google Colab that I probably have a chance to train models but waiting for Astra Labs' website release for alternative to any Google Colabs since we can download the zip file of the models there since it's not connected here on Discord unlike the bot version one or i will work hard to gain a lot of Streaks on Weights although it's exhausting haha since we can download and use the zip or model from Weights although some of the models or zip from there doesn't work on Astra Labs' register command

[wise cedar]

I understand. I'd recommend using the Mainline RVC Colab notebook. Supposedly, since you use easyGUI, I'd recommend Ilaria's Mainline RVC Colab notebook because the GUI is simple and easy to use. I have the guide here for you to follow so you can able to train off of this notebook: #🤖│bots message

[wispy glacier]

Okay

[wispy glacier]

that train on RVC mainline is easy to train models too like on easyGUI?

[wispy glacier]

But pls do screenrecord of tutorial how to train models there to understand more, but pls do that on DM and DM me because probably AI Hub doesn't allow us to post screenrecords in the public chat ✌️

[wise cedar]

I don't think I can help you any further but I'll ping @nocturne elk to assist you.

[wispy glacier]

Okay Thanks

[nocturne elk]

I think we've been through this before. No PC = no guide. Pth file size and index is a non-issue since you just upload it on huggingface as a open-rail model

Would take some guide to record a Kaggle mainline guide anyways

[wispy glacier]

Yeah but I still need a screenrecord or at least screenshot tutorial to understand more at least do this on DM ✌️

[nocturne elk]

??? You do realize screenshots are the same as the screen recording right? https://rentry.co/RVC-Mainline-Kaggle . The video will be me speedrunning through it

[wispy glacier]

Okay

[wispy glacier]

I will probably rarely to train models because most of my favorite groups/artists are popular or semi-popular I mean because most of people are only focusing on popular artists and never do models for underrated artists so I rarely do models for some underrated artists but I probably still need to train models on Google Colabs although the pth file is too large to download until Astra Labs website release or the Streaks or free training on Weights won't be so limited anymore

[prisma barn]

Ayo? @wispy glacier level 14 !!!

[nocturne elk]

Actually there'll be no video because something always messes up with ngrok. Please read the guide and if you think the pth files are too large, clean your pc once in a while

you're literally the only person complaining about this

[wispy glacier]

I'm a mobile user and I don't have a laptop yet

[wispy glacier]

Oh okay sorry tho hehe

[nocturne elk]

Kaggle is meant for pc. So unfortunately you'll have to rely on Rvc disconnected with no UI

[wispy glacier]

Yeah but I already tried that Colab and it's hard and doesn't work to me sorry

[nocturne elk]

easygui is discontinued because people leave the server all the time. Nothing we can do about it

#📰│dev-updates message

[wispy glacier]

Oh okay yeah thanks, it's sad that easyGUI discontinued because that Colab is the easiest ☹️

[wispy glacier]

But I will probably work hard to gain Streaks on Weights and waiting for Astra Labs' website (but also on bot version of Astra Labs although we can't download the model or zip from there due the restriction of Discord) will come to train models of some of my underrated groups/bands/artists because I don't have a chance to train models on any Google Colabs anymore because easyGUI is already discontinued and that RVC v2 Disconnected is sucks to me because it's hard to train models there and the models that I trained from there aren't working to me ✌️ but Applio isn't work to me as well but maybe that's only works on PC

[wispy glacier]

But there's a colab(s) that alternative for both easyGui and RVC v2 Disconnected at least not too easy nor easy as easyGUI hehe

[nocturne elk]

Then don't make models. Not until you get something sustainable. Alright that'll be all I can say

-close

[wispy glacier]

Yeah I agree, I think I should give up about Google Colabs especially that RVC v2 Disconnected is so hard and doesn't work sometimes because Weights and Astra Labs are the best alternative and the easiest to me although that free training or Streaks on Weights are so limited and we can't download the model or zip on Astra Labs bot and because most of my favorite artists are popular/semi popular or from big companies so a lot of people already do models for them and I don't need to train models for them