knifeayumu's picture
Update README.md
400a247 verified
|
raw
history blame
2.17 kB
metadata
base_model:
  - TheDrummer/Cydonia-22B-v1.1
  - mistralai/Mistral-Small-Instruct-2409
library_name: transformers
tags:
  - mergekit
  - merge
license: other

The Drummer turns into a Joshi Youchien

This is a merge of pre-trained language models created using mergekit.

GGUF quants : knifeayumu/Lite-Cydonia-22B-v1.1-Test-GGUF

Inspiration

I thought BeaverAI/Cydonia-22B-v1f-GGUF and BeaverAI/Cydonia-22B-v1e-GGUF versions being a bit too evil. The sense of morality is screwed up too much and it was a bit deterministic (swipes don't give much variety) versus the base model. Then an idea propped into my mind — why not merge it back again to the base? Give it a sense of "good" back, at least a little. Maybe that should fix some of deterministic generations too.

Quick testing shows... it works? Zero-shot evil Q&A no longer works but which a bit of persuasion, it did answer. I've also tried with both weights at 0.5 but it was too moral for my liking. Hence, I uploaded this version.

Credits to TheDrummer and BeaverAI who makes such finetunes. "Lightly decensored" is a heavy understatement in this case.

Merge Details

Merge Method

This model was merged using the task arithmetic merge method using TheDrummer/Cydonia-22B-v1.1 as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: TheDrummer/Cydonia-22B-v1.1
    parameters:
      weight: 0.75
  - model: mistralai/Mistral-Small-Instruct-2409
    parameters:
      weight: 0.25
merge_method: task_arithmetic
base_model: TheDrummer/Cydonia-22B-v1.1
dtype: float16