Infinitely-Kunodiculous-9B-GGUF-IQ-Imatrix

GGUF-IQ-Imatrix quants for ABX-AI/Infinitely-Kunodiculous-9B.

I'd like to extend my gratitude to @Lewdiculous for the inspiration to start trying to learn about merging and Imatrix quantization and help with answering my questions, and to @ Nitral-AI for the help with questions about merging as well.

Why Importance Matrix?

Importance Matrix, at least based on my testing, has shown to improve the output and performance of "IQ"-type quantizations, where the compression becomes quite heavy. The Imatrix performs a calibration, using a provided dataset. Testing has shown that semi-randomized data can help perserve more important segments as the compression is applied.

Related discussions in Github: [1] [2]

The imatrix.txt file that I used can be found here. The data of the imatrix is general, semi-random.

Description

This model is intended for role-playing and storywriting purposes.

This is the very firt merge I have ever tried. It seems to be working, or by the minimum does not appear to be broken :) The prime idea of this is to be an experiment and help me learn how to do merges and quants.

GGUF/IQ/Imatrix: https://huggingface.co/ABX-AI/Infinitely-Kunodiculous-9B-GGUF-IQ-Imatrix

Infinitely-Kunodiculous-9B

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

Merge Details

Merge Method

This model was merged using the passthrough merge method.

Models Merged

The following models were included in the merge:

• SanjiWatsuki/Kunoichi-DPO-v2-7B
• Nitral-AI/Infinitely-Laydiculous-9B

Configuration

The following YAML configuration was used to produce this model:

slices:
  - sources:
      - model: Nitral-AI/Infinitely-Laydiculous-9B
        layer_range: [0, 20]
  - sources:
      - model: SanjiWatsuki/Kunoichi-DPO-v2-7B
        layer_range: [12, 32]
merge_method: passthrough
dtype: float16