-
Notifications
You must be signed in to change notification settings - Fork 1.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Converter from bin to ckpt works!!!!! heres link with code from Josh to incorporate to colab !!! #53
Comments
Yes just saw it, I'll add it to the colab |
if you have colab pro, let me know if the V100 and the A100 work without issues |
I'll add to the dreambooth the prior preservation to avoid messing up all the classes and get better results. |
Awesome! im super excited , i do have colab pro but i rarely get v100 and just once per month a100, mostly p100 not even T4!!! |
``You think You can prune the ckpt so its 2gb ?
|
ok pruning is in the new code now by default ! Amazing |
The CKPT produced by this colab is 3.97 Gb, no need to prune |
Would also be grateful for attribution, @TheLastBen - if you could just put a link in the converttosd.py file that goes to my github that would be excellent. :) |
Sure, I was planning on adding that on the main Readme, thanks for the script, amazing job ! |
Developer named Josh did converter code in 2 days, works great i can use ckpt files in webui, does 4.3 gb file in like half minute time
You can paste the code into a cell and then paste your paths to model and ckpt location you want it in , its on bottom of the code.
https://gist.github.com/jachiam/ef621e765c2d88f9dc090ee48f6ced89
Amazing day for SD!!!
It would be nice to automate this in dreambooth notebook after training is done.
My best training so far - 10 images ( 4 body shots , 6 head shots) 1000 steps, no init word, just use class male or female, works best with likeness so far, yeh wrecks other stuff and all guys look like person trained but you can just swap models in automatics webui
The text was updated successfully, but these errors were encountered: