metadata
language:
- ko
- en
license: mit
Model Card for free-evo-qwen72b-v0.8
Developed by : Freewheelin AI Technical Team
1st place : 2024 4th May - avg. 81.28 Open Llm Leaderboard
but this kicked away. maybe the explanation was not enough.
Method
- We were inspired by this Sakana project
Process
- you need two models with the same architecture
- choose one model and finetune the model to make a gap between the original one and fine-tuned one. it doesn't matter the evaluation score is higher or lower.
- merge two of them
- evaluate the merged model
- finetune a specific evaluation part if you need to increase score of the part of the model. (sure it's not gonna work like you think. but try it)
- merge again
- evaluate again
- keep going until evaluate avg is higher then original one
that's it. simple.
Base Architecture
- QWEN2
Base Models
- several QWEN2 based models