{
  "train_batch_size": 24,
  "train_micro_batch_size_per_gpu": 3,
  "steps_per_print": 1,
  "optimizer": {
    "type": "Adam",
    "params": {
      "lr": 3e-5,
      "weight_decay": 0.0,
      "bias_correction": false
    }
  },
  "gradient_clipping": 1.0,
  "fp16": {
    "enabled": true
  },
  "tensorboard": {
    "enabled": true,
    "output_path": "/tmp/eigenvalue_quantize_output",
    "job_name": "eigenvalue_quantize"
  },
  "eigenvalue": {
      "enabled": true,
      "verbose": true,
      "max_iter": 50,
      "tol": 1e-2,
      "stability": 0,
      "gas_boundary_resolution": 1,
      "model_name": "bert-large"
  },
  "quantize_training": {
    "quantize_bits": {
      "start_bits": 12,
      "target_bits": 4
    },
    "quantize_type": "symmetric",
    "quantize_schedule": {
      "quantize_period": 400,
      "schedule_offset": 400
    },
    "quantize_groups": 16,
    "fp16_mixed_quantize": {
      "enabled": true,
      "quantize_change_ratio": 0.001
    },
    "quantize_verbose": true,
    "quantize_eigenvalue": true
  }
}