-
Hi, If I've trained a ckpt on SD20 760, in order to re-train on top of it for a second iteration (I don't mean to resume the session, I mean to use it as the "base" ckpt instead of base SD20) do I have to tick the v2_model tick box in the model load section, or select 2.0 768 from the dropdown list? Or do I simply provide the path to my ckpt without any other selections? |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 12 replies
-
There is no script to convert the V2 ckpt for now, use the session cell to load it. |
Beta Was this translation helpful? Give feedback.
-
I tried skipping the "load model" section and reloading the session to resume training, and I get this
|
Beta Was this translation helpful? Give feedback.
-
I apologise but I do not understand what I should do different (hopefully not having to re-start the training) What I did: Step 1. Trained a model on SD20 768 by selecting "V2-768px" and ticking the "V2_model", did not fill any other fields. The training worked fine and produced a usable CKPT The resulting folder structure is: Step 2. I want to resume training the model. According to instructions, I have not touched the "model download", went straight to create session, specified the same instance name, and it returns the error above What should I do differently? Thx |
Beta Was this translation helpful? Give feedback.
-
Now I'm getting a new error when trying to resume a SD20 768 session (sorry for not copy-pasting the error but for some reason it won't copy the text) |
Beta Was this translation helpful? Give feedback.
I apologise but I do not understand what I should do different (hopefully not having to re-start the training)
What I did:
Step 1. Trained a model on SD20 768 by selecting "V2-768px" and ticking the "V2_model", did not fill any other fields. The training worked fine and produced a usable CKPT
The resulting folder structure is:
Step 2. I want to resume training the model. According to instructions, I have not touched the "model download", went straight to create session, specified the same instance name, and it returns the error above
What should I do differently?
Thx