Whats the actual difference?

#2
by ShallowDesire - opened

just found out that it has the same merges as Maginum Cydoms. I tried them both and I feel like Slimaki is better at following the character card more. Is there an actual difference? I am just going off by feeling and it does feel better at following instructions compared to Maginum Cydoms

just found out that it has the same merges as Maginum Cydoms. I tried them both and I feel like Slimaki is better at following the character card more. Is there an actual difference? I am just going off by feeling and it does feel better at following instructions compared to Maginum Cydoms

Yes, the difference is that Slimaki merges everything at once using Della.

With Maginum Cydoms, it is split into 3 parts:

  • Ties merge of Magidonia, Precog, and PaintedFantasy v3
  • Della merge of Cydonia, 420 Broken Tutu, and PaintedFantasy v2
  • A SLERP of the Ties and DELLA

With Slimaki, the main advantage of merging them all at once seems to be a reduction of refusals. I have also ran some audits you can see here to examine the actual influence of each donor model on the final merge. Running a SLERP seems to enhance refusals somewhat, and I'm not sure about regular TIES (haven't tested it much yet except OddTheGreat's models).


Note for mergers: The audit scripts I made for Model_Stock, Karcher, and Della have been released now. You can use a coding LLM to adapt this feature to other merge_methods as needed.

Download them here: https://huggingface.co/spaces/Naphula/model_tools

Slimaki_Live_Audit

Slimaki_Audit

--- DELLA AUDIT V2 START ---
Loading config: config.yaml
Base Model: B:\24B\!models--anthracite-core--Mistral-Small-3.2-24B-Instruct-2506-Text-Only
Donors: 7

Extracting BASE MODEL fingerprint...

Extracting DONOR fingerprints...

Computing Task Vector geometry...

================================================================================
ID    | Model Name
--------------------------------------------------------------------------------
#1    | anthracite-core--Mistral-Small-3.2-24B-Instruct-2506-Text-Only
#2    | TheDrummer--Cydonia-24B-v4.3
#3    | ReadyArt--4.2.0-Broken-Tutu-24b
#4    | zerofata--MS3.2-PaintedFantasy-v2-24B
#5    | TheDrummer--Magidonia-24B-v4.3
#6    | TheDrummer--Precog-24B-v1
#7    | zerofata--MS3.2-PaintedFantasy-v3-24B
================================================================================

--- MAGNITUDE ANALYSIS & DATA POINTS ---
ID    | Status     | Delta Norm   | Orig Size    | Model Name
----------------------------------------------------------------------------------------------------
#1    | OK         | 0.0000       | 83886080     | anthracite-core--Mistral-Small-3.2-24B-Instruct-2506-Text-Only
#2    | OK         | 1.2955       | 83886080     | TheDrummer--Cydonia-24B-v4.3
#3    | HIGH MAG   | 46.6745      | 83886080     | ReadyArt--4.2.0-Broken-Tutu-24b
#4    | OK         | 0.0505       | 83886080     | zerofata--MS3.2-PaintedFantasy-v2-24B
#5    | OK         | 4.5662       | 83886080     | TheDrummer--Magidonia-24B-v4.3
#6    | OK         | 4.0883       | 83886080     | TheDrummer--Precog-24B-v1
#7    | OK         | 4.8187       | 83886080     | zerofata--MS3.2-PaintedFantasy-v3-24B

Log saved to: della_scan.log
Displaying charts...

Sign up or log in to comment