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 using Undi95/PsyMedRP-v1-20B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
merge_method: linear
dtype: bfloat16
base_model: Undi95/PsyMedRP-v1-20B
models:
- model: Undi95/PsyMedRP-v1-20B
parameters:
weight: 0.8
layer_range: [0, 20]
- model: Undi95/MXLewd-L2-20B
parameters:
weight: 0.5
layer_range: [21, 40]
- model: Undi95/PsyMedRP-v1-20B
parameters:
weight: 0.3
layer_range: [41, 62]