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 tdot604/Hazy-Llama-70b as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
# config.yaml
models:
- model: tdot604/Hazy-Llama-70b
parameters:
weight: 0.4
- model: Sao10K/L3-70B-Euryale-v2.1
parameters:
weight: 0.25
- model: crestf411/L3-70B-daybreak-abliterated-v0.4
parameters:
weight: 0.2
- model: TheDrummer/Moist-Miqu-70B-v1.1
parameters:
weight: 0.15
merge_method: ties # This should be a string, not a dictionary
base_model: tdot604/Hazy-Llama-70b
parameters:
density: 0.7
normalize: true
tokenizer_source: base
dtype: float16