Resolving Interference When Merging Models
Paper
•
2306.01708
•
Published
•
15
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using huihui-ai/QwQ-32B-abliterated as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: DataSoul/Sky-T1-32B-Flash-abliterated-merge1
parameters:
weight: 0.382
density: 1
- model: zetasepic/Qwen2.5-32B-Instruct-abliterated-v2
parameters:
weight: 0.618
density: 1
- model: huihui-ai/QwQ-32B-abliterated
parameters:
weight: 1
density: 1
merge_method: ties
base_model: huihui-ai/QwQ-32B-abliterated
parameters:
normalize: true
int8_mask: true
tokenizer:
source: "base"
pad_to_multiple_of: 152064
dtype: bfloat16