README.md

metadata

base_model:
  - cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser
  - Intel/neural-chat-7b-v3-3
  - meta-math/MetaMath-Mistral-7B
tags:
  - mergekit
  - merge

Ethux 7B merge

This model is a test of what is possible with mergekit, this model does not have the best performance and quality responses yet.

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

Merge Details

Merge Method

This model was merged using the TIES merge method using Intel/neural-chat-7b-v3-3 as a base.

Models Merged

The following models were included in the merge:

• cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser
• meta-math/MetaMath-Mistral-7B

Configuration

The following YAML configuration was used to produce this model:

# Main configuration for merging pre-trained models

models:
#  - model: Intel/neural-chat-7b-v3-3
#    parameters:
#      density: 0.9 # High density since generally good for most tasks
#      weight: 0.4 # Lower weight compared to openchat due to slight inferiority
  - model: meta-math/MetaMath-Mistral-7B
    parameters:
      density: 0.6 # Moderate density focusing mainly on mathematical prowess
      weight:
        - filter: mlp
          value: 0.5
        - value: 0
  - model: cognitivecomputations/dolphin-2.6-mistral-7b-dpo-laser
    parameters:
      density: 0.8 # Balanced density emphasizing coding skills over chitchat
      weight: 0.4 # Slightly higher weight acknowledging superior coding ability

base_model: Intel/neural-chat-7b-v3-3
merge_method: ties
parameters:
  normalize: true
  int8_mask: false
dtype: float16