Resolving Interference When Merging Models
Paper
• 2306.01708 • Published
• 17
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using E:\tabbyAPI\models\deepseek-ai_DeepSeek-V2-Lite as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: E:\tabbyAPI\models\deepseek-ai_DeepSeek-V2-Lite
parameters:
weight: 0.2
density: 0.2
- model: E:\tabbyAPI\models\deepseek-ai_DeepSeek-Coder-V2-Lite-Instruct
parameters:
weight: 0.8
density: 0.8
merge_method: ties
base_model: E:\tabbyAPI\models\deepseek-ai_DeepSeek-V2-Lite
parameters:
weight: 1
density: 1
normalize: true
int8_mask: true
tokenizer_source: E:\tabbyAPI\models\deepseek-ai_DeepSeek-Coder-V2-Lite-Instruct
dtype: bfloat16