Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
Paper • 2203.05482 • Published • 8
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:
models:
- model: YanLabs/gemma3-27b-it-abliterated-normpreserve-v1
parameters:
weight: 1.0
- model: YanLabs/gemma3-27b-it-abliterated-normpreserve-v1+allura-forge/g27bslop-ep2
parameters:
weight: 0.1
- model: YanLabs/gemma3-27b-it-abliterated-normpreserve-v1+allura-forge/g27bslop-ep1
parameters:
weight: 0.4
merge_method: linear
parameters:
normalize: true
int8_mask: true
dtype: bfloat16