Resolving Interference When Merging Models
Paper
β’
2306.01708
β’
Published
β’
15
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using mistralai/Mistral-7B-v0.1 as a base.
The following models were included in the merge:
| Pubmedqa | Medmcqa | USMLE | arc_challenge | Hellaswag |
|---|---|---|---|---|
| 76.8 | 47.21 | 47.62 | 55.89 | 63.43 |
The following YAML configuration was used to produce this model:
models:
- model: mistralai/Mistral-7B-v0.1
- model: HuggingFaceH4/zephyr-7b-beta
parameters:
density: 0.5
weight: 0.5
- model: BioMistral/BioMistral-7B
parameters:
density: 0.5
weight: 0.5
merge_method: ties
base_model: mistralai/Mistral-7B-v0.1
parameters:
normalize: false
int8_mask: true
dtype: float16