DELLA-Merging: Reducing Interference in Model Merging through Magnitude-Based Sampling
Paper
•
2406.11617
•
Published
•
8
This is a merge of pre-trained language models created using mergekit.
This model was merged using the DELLA merge method using TroyDoesAI/BlackSheep-24B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: cognitivecomputations/Dolphin3.0-Mistral-24B
parameters:
density: 0.5
weight: 0.5
- model: huihui-ai/Mistral-Small-24B-Instruct-2501-abliterated
parameters:
density: 0.5
weight: 0.5
- model: TheDrummer/Cydonia-24B-v2
parameters:
density: 0.5
weight: 0.5
- model: huihui-ai/Arcee-Blitz-abliterated
parameters:
density: 0.5
weight: 0.5
- model: TroyDoesAI/BlackSheep-24B
parameters:
density: 0.5
weight: 0.5
merge_method: della
base_model: TroyDoesAI/BlackSheep-24B
parameters:
normalize: false
int8_mask: true
dtype: float16