Automatic Weight Calc based on NearSwap #179
Closed
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
adds in a new merge method based off of NearSwap it uses Dare_Ties and adds in auto weight calculation based off of near tuned interpolation on a per parameter bases. so theoretically it is gradient weights on steroids. Am looking at changing this into in a auto density function that will feed back into the weight function based on calculating a gradient and variance-based density map, with a gradient fall off for each parameter this is the ramblings of a sleep deprived idiot
fixed batch dimension logic added weight defaults
Update generalized_task_arithmetic.py
|
Why closed? |
|
There's too much going on IRL, so I couldn't push any further. I may look into it later. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adds in a merge method based off of NearSwap
It uses Dare / Ties and adds in auto weight calculation based off of near tuned interpolation on a per parameter bases, so theoretically it is gradient weights on steroids.
Am looking at changing this into in a auto density function that will feed back into the weight function based on calculating a gradient / variance-based density map, with a gradient fall off for each parameter.
this is the ramblings of a sleep deprived dude.
would love to hear from others on this as it was a 3am random thought and im sure the implementation could be changed.
I still need to change a few other files still
This is based on the work of:
https://huggingface.co/lodrick-the-lafted
&
https://huggingface.co/LilyWinter
who developed the NearSwap algorithm that was used on:
https://huggingface.co/alchemonaut/BoreanGale-70B
https://huggingface.co/alchemonaut/QuartetAnemoi-70B-t0.0001