Resolving Interference When Merging Models
Paper • 2306.01708 • Published • 17
If you get prompts with 'user' or 'assistant' at the end put the words into stopping strings. The problem should go away.
This is a merge of pre-trained language models created using mergekit.
This model was merged using the TIES merge method using pot99rta/MagcarpMell-ThinkandReasoner-12B as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: pot99rta/MagcarpMell-ThinkandReasoner-12B
#no parameters necessary for base model
- model: pot99rta/MagcarpMell-ThinkandReasoner-12B
parameters:
density: 0.5
weight: 0.5
- model: pot99rta/PatriSlush-DarkLorablated-LongStock-12B
parameters:
density: 0.5
weight: 0.5
merge_method: ties
base_model: pot99rta/MagcarpMell-ThinkandReasoner-12B
parameters:
normalize: false
int8_mask: true
dtype: float16