Resolving Interference When Merging Models
Paper β’ 2306.01708 β’ Published β’ 18
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using TheDrummer/Rocinante-12B-v1 as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: Trappu/Nemo-Picaro-12B
parameters:
density: 0.7
weight: 0.4
- model: anthracite-org/magnum-v4-12b
parameters:
density: 0.3
weight: 0.25
- model: PocketDoc/Dans-PersonalityEngine-V1.1.0-12b
parameters:
density: 0.5
weight: 0.3
merge_method: ties
base_model: TheDrummer/Rocinante-12B-v1
parameters:
normalize: true
int8_mask: true
dtype: bfloat16
chat_template: "chatml"
tokenizer:
source: union