Formerly known as MO-MODEL-Fused-V0.6-LLaMa-70B, This model is part of my ongoing experiments with merging specialized curated models. For this one, I started experimenting with gradients, to give myself more finetuned control of how the models influence the final merge.
Recommended sampler settings:
Temp 1.0
Min P 0.02
Because of the nature of this sort of 'Hyper Multi Model Merge', my recommendation is not to run this on anything lower than a Q5 quant.
If you enjoy my work, please consider supporting me, It helps me make more models like this! Support on KO-FI <3
MERGE2
This is a merge of pre-trained language models created using mergekit.
Merge Details
Merge Method
This model was merged using the DARE TIES merge method using TareksLab/MO-MODEL6-V0.1-LLaMa-70B as a base.
Models Merged
The following models were included in the merge:
- TareksLab/MO-MODEL3-V0.2-LLaMa-70B
- TareksLab/MO-MODEL5-V0.3-LLaMa-70B
- TareksLab/MO-MODEL2-V0.2-LLaMa-70B
- TareksLab/MO-MODEL1-V1-LLaMa-70B
- TareksLab/MO-MODEL4-V0.1-LLaMa-70B
Configuration
The following YAML configuration was used to produce this model:
models:
- model: TareksLab/MO-MODEL6-V0.1-LLaMa-70B
parameters:
weight: [0.1, 0.1, 0.1, 0.2, 0.5]
density: 0.5
- model: TareksLab/MO-MODEL4-V0.1-LLaMa-70B
parameters:
weight: [0.1, 0.1, 0.2, 0.4, 0.2]
density: 0.5
- model: TareksLab/MO-MODEL5-V0.3-LLaMa-70B
parameters:
weight: [0.1, 0.2, 0.4, 0.2, 0.1]
density: 0.5
- model: TareksLab/MO-MODEL3-V0.2-LLaMa-70B
parameters:
weight: [0.2, 0.4, 0.2, 0.1, 0.1]
density: 0.5
- model: TareksLab/MO-MODEL2-V0.2-LLaMa-70B
parameters:
weight: [0.5, 0.2, 0.1, 0.1, 0.1]
density: 0.5
- model: TareksLab/MO-MODEL1-V1-LLaMa-70B
parameters:
weight: 0.10
density: 0.5
merge_method: dare_ties
base_model: TareksLab/MO-MODEL6-V0.1-LLaMa-70B
parameters:
normalize: false
int8_mask: true
dtype: float32
out_dtype: bfloat16
chat_template: llama3
tokenizer:
source: base
- Downloads last month
- 109