Resolving Interference When Merging Models
Paper
•
2306.01708
•
Published
•
15
More Robust with all the Darkness added.
Models Merged:
1. ReadyArt/Omega-Darker_The-Final-Directive-12B
2. pot99rta/MagcarpMell-ThinkandReasoner-12B
Preset:
Use ChatML or Mistral
ChatML works better for reasoning due to Magicap and MagMell being ChatML for their base Models.
Just realized I've been spelling Magcap with 'Magcarp' this WHOLE time..
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using ReadyArt/Omega-Darker_The-Final-Directive-12B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: ReadyArt/Omega-Darker_The-Final-Directive-12B
#no parameters necessary for base model
- model: ReadyArt/Omega-Darker_The-Final-Directive-12B
parameters:
density: 0.5
weight: 0.5
- model: pot99rta/MagcarpMell-ThinkandReasoner-12B
parameters:
density: 0.5
weight: 0.5
merge_method: ties
base_model: ReadyArt/Omega-Darker_The-Final-Directive-12B
parameters:
normalize: false
int8_mask: true
dtype: float16