trainer_state.json

{
  "best_global_step": 90,
  "best_metric": 0.6871705055236816,
  "best_model_checkpoint": "tam2003/SFT-Qwen3-30B-dsetV3/checkpoint-90",
  "epoch": 3.75,
  "eval_steps": 30,
  "global_step": 180,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.625,
      "grad_norm": 0.3467564582824707,
      "learning_rate": 0.00018189655172413794,
      "loss": 0.7767,
      "step": 30
    },
    {
      "epoch": 0.625,
      "eval_loss": 0.7523366808891296,
      "eval_runtime": 328.6687,
      "eval_samples_per_second": 0.359,
      "eval_steps_per_second": 0.073,
      "step": 30
    },
    {
      "epoch": 1.25,
      "grad_norm": 0.19959872961044312,
      "learning_rate": 0.0001560344827586207,
      "loss": 0.568,
      "step": 60
    },
    {
      "epoch": 1.25,
      "eval_loss": 0.7073029279708862,
      "eval_runtime": 327.8879,
      "eval_samples_per_second": 0.36,
      "eval_steps_per_second": 0.073,
      "step": 60
    },
    {
      "epoch": 1.875,
      "grad_norm": 0.18368889391422272,
      "learning_rate": 0.00013017241379310345,
      "loss": 0.5549,
      "step": 90
    },
    {
      "epoch": 1.875,
      "eval_loss": 0.6871705055236816,
      "eval_runtime": 330.1164,
      "eval_samples_per_second": 0.357,
      "eval_steps_per_second": 0.073,
      "step": 90
    },
    {
      "epoch": 2.5,
      "grad_norm": 0.5208514928817749,
      "learning_rate": 0.00010431034482758621,
      "loss": 0.3923,
      "step": 120
    },
    {
      "epoch": 2.5,
      "eval_loss": 0.6951888203620911,
      "eval_runtime": 331.8127,
      "eval_samples_per_second": 0.356,
      "eval_steps_per_second": 0.072,
      "step": 120
    },
    {
      "epoch": 3.125,
      "grad_norm": 1.5229783058166504,
      "learning_rate": 7.844827586206897e-05,
      "loss": 0.4057,
      "step": 150
    },
    {
      "epoch": 3.125,
      "eval_loss": 0.7338975667953491,
      "eval_runtime": 331.5375,
      "eval_samples_per_second": 0.356,
      "eval_steps_per_second": 0.072,
      "step": 150
    },
    {
      "epoch": 3.75,
      "grad_norm": 0.46281370520591736,
      "learning_rate": 5.2586206896551724e-05,
      "loss": 0.213,
      "step": 180
    },
    {
      "epoch": 3.75,
      "eval_loss": 0.7525293827056885,
      "eval_runtime": 332.3109,
      "eval_samples_per_second": 0.355,
      "eval_steps_per_second": 0.072,
      "step": 180
    },
    {
      "epoch": 3.75,
      "step": 180,
      "total_flos": 6.46890831893631e+17,
      "train_loss": 0.485091945860121,
      "train_runtime": 21477.3859,
      "train_samples_per_second": 0.111,
      "train_steps_per_second": 0.011
    },
    {
      "epoch": 3.75,
      "eval_loss": 0.6871705055236816,
      "eval_runtime": 332.1425,
      "eval_samples_per_second": 0.355,
      "eval_steps_per_second": 0.072,
      "step": 180
    }
  ],
  "logging_steps": 30,
  "max_steps": 240,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 30,
  "stateful_callbacks": {
    "EarlyStoppingCallback": {
      "args": {
        "early_stopping_patience": 3,
        "early_stopping_threshold": 0.0
      },
      "attributes": {
        "early_stopping_patience_counter": 3
      }
    },
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.46890831893631e+17,
  "train_batch_size": 5,
  "trial_name": null,
  "trial_params": null
}