Replies: 1 comment
-
you can use the method wih slices ... as it gives you the chance t select the layer range hence your merge may create a different B ... ie if you choose slices 0-10 then for the next model 10-20 then you model wil be 20 layers only ... hence if they were both 7b models then the model will be smaller ! .. for 7b 32 layers ... |
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
-
I have a question that may sound silly. Has anyone used mergekit to shrink model size instead of combining several models ?
For example we could take as sources the same model, consider it's 2 models, and use merge kit to merge layers so that we would go for example from 24 layers to 12 layers. That could be a good base then to distill the created student on the teacher.
Beta Was this translation helpful? Give feedback.
All reactions