tomofusa/exp022-dpo-ipo-merged

exp022-dpo-ipo-merged

SFT + DPO merged model. Full 16-bit weights, no adapter loading required.

Training Pipeline

1. 1.SFT: tomofusa/exp021b-blend-h-lora
2. 2.DPO: u-10bei/dpo-dataset-qwen-cot (1 epoch, lr=5e-07, beta=0.1)

DPO Configuration

• —Learning rate: 5e-07
• —Beta: 0.1
• —Loss type: ipo
• —LoRA: r=64, alpha=128
• —Max length: 1024