last-checkpoint/trainer_state.json

{
  "best_global_step": 200,
  "best_metric": 0.6018396615982056,
  "best_model_checkpoint": "./t5gemma-math-corrector/checkpoint-200",
  "epoch": 2.5645549318364074,
  "eval_steps": 200,
  "global_step": 800,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.16038492381716118,
      "grad_norm": 2.080294370651245,
      "learning_rate": 5.212765957446809e-05,
      "loss": 1.8373,
      "step": 50
    },
    {
      "epoch": 0.32076984763432237,
      "grad_norm": 1.3640859127044678,
      "learning_rate": 9.94061757719715e-05,
      "loss": 0.028,
      "step": 100
    },
    {
      "epoch": 0.48115477145148355,
      "grad_norm": 0.47297751903533936,
      "learning_rate": 9.346793349168646e-05,
      "loss": 0.0138,
      "step": 150
    },
    {
      "epoch": 0.6415396952686447,
      "grad_norm": 0.6050183176994324,
      "learning_rate": 8.752969121140144e-05,
      "loss": 0.0085,
      "step": 200
    },
    {
      "epoch": 0.6415396952686447,
      "eval_loss": 0.6018396615982056,
      "eval_runtime": 5.1929,
      "eval_samples_per_second": 50.454,
      "eval_steps_per_second": 12.71,
      "step": 200
    },
    {
      "epoch": 0.8019246190858059,
      "grad_norm": 0.2022608071565628,
      "learning_rate": 8.15914489311164e-05,
      "loss": 0.0064,
      "step": 250
    },
    {
      "epoch": 0.9623095429029671,
      "grad_norm": 0.8748292922973633,
      "learning_rate": 7.565320665083135e-05,
      "loss": 0.0047,
      "step": 300
    },
    {
      "epoch": 1.1218925421010426,
      "grad_norm": 0.28713762760162354,
      "learning_rate": 6.971496437054633e-05,
      "loss": 0.0025,
      "step": 350
    },
    {
      "epoch": 1.2822774659182037,
      "grad_norm": 0.16400396823883057,
      "learning_rate": 6.377672209026129e-05,
      "loss": 0.0029,
      "step": 400
    },
    {
      "epoch": 1.2822774659182037,
      "eval_loss": 1.9704385995864868,
      "eval_runtime": 5.1665,
      "eval_samples_per_second": 50.711,
      "eval_steps_per_second": 12.775,
      "step": 400
    },
    {
      "epoch": 1.4426623897353648,
      "grad_norm": 0.35608166456222534,
      "learning_rate": 5.783847980997625e-05,
      "loss": 0.0015,
      "step": 450
    },
    {
      "epoch": 1.603047313552526,
      "grad_norm": 0.33180880546569824,
      "learning_rate": 5.190023752969121e-05,
      "loss": 0.0015,
      "step": 500
    },
    {
      "epoch": 1.7634322373696873,
      "grad_norm": 0.16627037525177002,
      "learning_rate": 4.596199524940617e-05,
      "loss": 0.0009,
      "step": 550
    },
    {
      "epoch": 1.9238171611868484,
      "grad_norm": 0.04953346028923988,
      "learning_rate": 4.002375296912114e-05,
      "loss": 0.001,
      "step": 600
    },
    {
      "epoch": 1.9238171611868484,
      "eval_loss": 3.8351235389709473,
      "eval_runtime": 5.1639,
      "eval_samples_per_second": 50.737,
      "eval_steps_per_second": 12.781,
      "step": 600
    },
    {
      "epoch": 2.0834001603849237,
      "grad_norm": 0.13095279037952423,
      "learning_rate": 3.408551068883611e-05,
      "loss": 0.0009,
      "step": 650
    },
    {
      "epoch": 2.2437850842020852,
      "grad_norm": 0.1417347490787506,
      "learning_rate": 2.8147268408551068e-05,
      "loss": 0.0005,
      "step": 700
    },
    {
      "epoch": 2.4041700080192463,
      "grad_norm": 0.07039827853441238,
      "learning_rate": 2.2209026128266035e-05,
      "loss": 0.0005,
      "step": 750
    },
    {
      "epoch": 2.5645549318364074,
      "grad_norm": 0.2078932374715805,
      "learning_rate": 1.6270783847980998e-05,
      "loss": 0.0005,
      "step": 800
    },
    {
      "epoch": 2.5645549318364074,
      "eval_loss": 3.8249130249023438,
      "eval_runtime": 5.2131,
      "eval_samples_per_second": 50.258,
      "eval_steps_per_second": 12.66,
      "step": 800
    }
  ],
  "logging_steps": 50,
  "max_steps": 936,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 200,
  "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": 3090020422385664.0,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}