training_config.json

{
  "base_model": "/root/affine-cortex/models",
  "output_dir": "./trained_model",
  "lora_r": 32,
  "lora_alpha": 64,
  "lora_dropout": 0.05,
  "stage1_epochs": 1,
  "stage1_lr": 0.0002,
  "stage1_batch_size": 2,
  "stage1_grad_accum": 8,
  "stage2_epochs": 1,
  "stage2_lr": 0.0001,
  "stage2_batch_size": 2,
  "stage2_grad_accum": 8,
  "max_length": 8192,
  "warmup_ratio": 0.03,
  "weight_decay": 0.01,
  "max_grad_norm": 1.0,
  "save_steps": 500,
  "logging_steps": 10
}