ASSN2 SimPO Model

This model was trained with SimPO preference optimization.

Training setup

  • Initialization model: hihihikei/assn2-sft
  • Local initialization path: /home/junhee/a2/checkpoints/assn2-sft-merged-hf
  • SimPO beta: 2.0
  • SimPO gamma/beta ratio: 0.5
  • SimPO learning rate: 5e-7

Notes

This repository contains the locally merged SimPO model exported from the assignment notebook.

Downloads last month
1
Safetensors
Model size
1B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for hihihikei/assn2-simpo

Finetuned
(1)
this model