Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
Paper
•
2203.05482
•
Published
•
7
This is a merge of pre-trained language models created using mergekit.
This model was merged using the linear merge method.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
merge_method: linear
dtype: bfloat16
models: # just your average dolphin
- model: cognitivecomputations/dolphin-2.7-mixtral-8x7b
parameters:
weight: 0.5
- model: cognitivecomputations/dolphin-2.6-mixtral-8x7b
parameters:
weight: 0.3
- model: cognitivecomputations/dolphin-2.5-mixtral-8x7b
parameters:
weight: 0.2
parameters:
normalize: true