trainer_state.json

{
  "best_metric": 81.63811371088175,
  "best_model_checkpoint": "/root/turkic_qa/ru_kaz_models/ru_kaz_xlm_roberta_large_squad_model/checkpoint-2208",
  "epoch": 5.0,
  "eval_steps": 500,
  "global_step": 2760,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 1.0,
      "step": 552,
      "train_exact_match": 67.83216783216783,
      "train_f1": 84.61477737180347,
      "train_runtime": 23.2961,
      "train_samples_per_second": 43.612,
      "train_steps_per_second": 1.588
    },
    {
      "epoch": 1.0,
      "grad_norm": 209.68630981445312,
      "learning_rate": 1e-05,
      "loss": 1.187,
      "step": 552
    },
    {
      "epoch": 1.0,
      "eval_exact_match": 62.8125,
      "eval_f1": 80.31583567739357,
      "eval_runtime": 74.7802,
      "eval_samples_per_second": 43.875,
      "eval_steps_per_second": 1.578,
      "step": 552
    },
    {
      "epoch": 2.0,
      "step": 1104,
      "train_exact_match": 77.12287712287713,
      "train_f1": 90.22014020687811,
      "train_runtime": 23.5003,
      "train_samples_per_second": 43.744,
      "train_steps_per_second": 1.574
    },
    {
      "epoch": 2.0,
      "grad_norm": 55.167510986328125,
      "learning_rate": 7.500000000000001e-06,
      "loss": 0.8237,
      "step": 1104
    },
    {
      "epoch": 2.0,
      "eval_exact_match": 64.59375,
      "eval_f1": 81.19927962759496,
      "eval_runtime": 74.7421,
      "eval_samples_per_second": 43.898,
      "eval_steps_per_second": 1.579,
      "step": 1104
    },
    {
      "epoch": 3.0,
      "step": 1656,
      "train_exact_match": 80.01998001998003,
      "train_f1": 92.75806800650366,
      "train_runtime": 23.2732,
      "train_samples_per_second": 43.698,
      "train_steps_per_second": 1.59
    },
    {
      "epoch": 3.0,
      "grad_norm": 60.38401794433594,
      "learning_rate": 5e-06,
      "loss": 0.5947,
      "step": 1656
    },
    {
      "epoch": 3.0,
      "eval_exact_match": 65.34375,
      "eval_f1": 81.60673478065382,
      "eval_runtime": 75.1587,
      "eval_samples_per_second": 43.654,
      "eval_steps_per_second": 1.57,
      "step": 1656
    },
    {
      "epoch": 4.0,
      "step": 2208,
      "train_exact_match": 85.11488511488511,
      "train_f1": 95.27935906423745,
      "train_runtime": 23.5078,
      "train_samples_per_second": 43.135,
      "train_steps_per_second": 1.574
    },
    {
      "epoch": 4.0,
      "grad_norm": 106.58255004882812,
      "learning_rate": 2.5e-06,
      "loss": 0.4463,
      "step": 2208
    },
    {
      "epoch": 4.0,
      "eval_exact_match": 65.3125,
      "eval_f1": 81.63811371088175,
      "eval_runtime": 75.6398,
      "eval_samples_per_second": 43.377,
      "eval_steps_per_second": 1.56,
      "step": 2208
    },
    {
      "epoch": 5.0,
      "step": 2760,
      "train_exact_match": 88.21178821178822,
      "train_f1": 96.20658088193036,
      "train_runtime": 23.4676,
      "train_samples_per_second": 43.549,
      "train_steps_per_second": 1.577
    },
    {
      "epoch": 5.0,
      "grad_norm": 25.936321258544922,
      "learning_rate": 0.0,
      "loss": 0.3537,
      "step": 2760
    },
    {
      "epoch": 5.0,
      "eval_exact_match": 65.25,
      "eval_f1": 81.53757682259878,
      "eval_runtime": 75.1221,
      "eval_samples_per_second": 43.676,
      "eval_steps_per_second": 1.571,
      "step": 2760
    },
    {
      "epoch": 5.0,
      "step": 2760,
      "total_flos": 5.376168689054976e+16,
      "train_loss": 0.6810590108235677,
      "train_runtime": 4970.3143,
      "train_samples_per_second": 15.529,
      "train_steps_per_second": 0.555
    }
  ],
  "logging_steps": 500,
  "max_steps": 2760,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 5.376168689054976e+16,
  "train_batch_size": 28,
  "trial_name": null,
  "trial_params": null
}