Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch
Paper
•
2311.03099
•
Published
•
30
This is a merge of pre-trained language models created using mergekit.
This model was merged using the DARE TIES merge method using /workspace/cache/models--BruhzWater--Eden-L3.3-70b-0.3/snapshots/723770586b6603d44ac9bb1d4527e80e86d4dc8b as a base.
The following models were included in the merge:
The following YAML configuration was used to produce this model:
models:
- model: /workspace/cache/models--BruhzWater--Liliths-Whisper-L3.3-70b-0.1/snapshots/2f5b9f0307fbb865a3f2e98eefaeb3c86955cb03
parameters:
weight: 0.25
density: 0.8
- model: /workspace/cache/models--BruhzWater--Apocrypha-L3.3-70b-0.3/snapshots/3facb4c0a7b953ff34a5caa90976830bf82a84c2
parameters:
weight: 0.25
density: 0.65
- model: /workspace/cache/models--BruhzWater--Liliths-Whisper-L3.3-70b-0.1/snapshots/2f5b9f0307fbb865a3f2e98eefaeb3c86955cb03
parameters:
weight: 0.25
density: 0.65
- model: /workspace/cache/models--BruhzWater--Eden-L3.3-70b-0.3/snapshots/723770586b6603d44ac9bb1d4527e80e86d4dc8b
parameters:
weight: 0.25
density: 0.4
merge_method: dare_ties
base_model: /workspace/cache/models--BruhzWater--Eden-L3.3-70b-0.3/snapshots/723770586b6603d44ac9bb1d4527e80e86d4dc8b
parameters:
normalize: false
lambda: 1
chat_template: llama3
pad_to_multiple_of: 8
int8_mask: true
tokenizer:
source: base
dtype: bfloat16