hyperparameters.json

{
  "model_name": "Qwen/Qwen3.5-27B",
  "model_org_name": "ab_self_promotion",
  "lora_r": 128,
  "lora_scaling_factor": 1.0,
  "lora_alpha": 128.0,
  "use_rslora": true,
  "epochs": 1,
  "max_steps": -1,
  "batch_size": 2,
  "gradient_accumulation": 8,
  "learning_rate": 0.0002,
  "weight_decay": 0.001,
  "triviaqa_pct": 0.1,
  "alpaca_pct": 0.0,
  "seed": 42,
  "max_seq_length": 4096
}