{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.9988358556461003,
  "eval_steps": 1000,
  "global_step": 20608,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00019409937888198756,
      "grad_norm": 6.586876392364502,
      "learning_rate": 0.0,
      "loss": 4.1943,
      "step": 1
    },
    {
      "epoch": 0.19409937888198758,
      "grad_norm": 4.414008617401123,
      "learning_rate": 7.270742358078603e-06,
      "loss": 3.407,
      "step": 1000
    },
    {
      "epoch": 0.19409937888198758,
      "eval_cosine_accuracy": 0.9413691163063049,
      "eval_loss": 1.0037776231765747,
      "eval_runtime": 38.8856,
      "eval_samples_per_second": 243.432,
      "eval_steps_per_second": 0.952,
      "step": 1000
    },
    {
      "epoch": 0.38819875776397517,
      "grad_norm": 4.412827491760254,
      "learning_rate": 1.4548762736535663e-05,
      "loss": 2.7704,
      "step": 2000
    },
    {
      "epoch": 0.38819875776397517,
      "eval_cosine_accuracy": 0.9514050483703613,
      "eval_loss": 0.9424285888671875,
      "eval_runtime": 32.1879,
      "eval_samples_per_second": 294.086,
      "eval_steps_per_second": 1.15,
      "step": 2000
    },
    {
      "epoch": 0.5822981366459627,
      "grad_norm": 4.752381324768066,
      "learning_rate": 2.182678311499272e-05,
      "loss": 2.7878,
      "step": 3000
    },
    {
      "epoch": 0.5822981366459627,
      "eval_cosine_accuracy": 0.9509824514389038,
      "eval_loss": 0.9267508387565613,
      "eval_runtime": 32.1133,
      "eval_samples_per_second": 294.769,
      "eval_steps_per_second": 1.152,
      "step": 3000
    },
    {
      "epoch": 0.7763975155279503,
      "grad_norm": 3.283628463745117,
      "learning_rate": 2.910480349344978e-05,
      "loss": 3.1982,
      "step": 4000
    },
    {
      "epoch": 0.7763975155279503,
      "eval_cosine_accuracy": 0.946017324924469,
      "eval_loss": 0.8683244585990906,
      "eval_runtime": 32.1375,
      "eval_samples_per_second": 294.546,
      "eval_steps_per_second": 1.151,
      "step": 4000
    },
    {
      "epoch": 0.9704968944099379,
      "grad_norm": 5.951777458190918,
      "learning_rate": 2.840592017469368e-05,
      "loss": 3.4883,
      "step": 5000
    },
    {
      "epoch": 0.9704968944099379,
      "eval_cosine_accuracy": 0.9291146993637085,
      "eval_loss": 0.9623793959617615,
      "eval_runtime": 32.2807,
      "eval_samples_per_second": 293.24,
      "eval_steps_per_second": 1.146,
      "step": 5000
    },
    {
      "epoch": 1.1645324020178502,
      "grad_norm": 3.765852689743042,
      "learning_rate": 2.658619434671843e-05,
      "loss": 2.7146,
      "step": 6000
    },
    {
      "epoch": 1.1645324020178502,
      "eval_cosine_accuracy": 0.9578491449356079,
      "eval_loss": 0.8477861285209656,
      "eval_runtime": 32.0996,
      "eval_samples_per_second": 294.894,
      "eval_steps_per_second": 1.153,
      "step": 6000
    },
    {
      "epoch": 1.358556461001164,
      "grad_norm": 3.609743356704712,
      "learning_rate": 2.477010797039913e-05,
      "loss": 2.5822,
      "step": 7000
    },
    {
      "epoch": 1.358556461001164,
      "eval_cosine_accuracy": 0.9603845477104187,
      "eval_loss": 0.8279656767845154,
      "eval_runtime": 31.9719,
      "eval_samples_per_second": 296.072,
      "eval_steps_per_second": 1.157,
      "step": 7000
    },
    {
      "epoch": 1.552580519984478,
      "grad_norm": 3.763313055038452,
      "learning_rate": 2.2950382142423874e-05,
      "loss": 2.5244,
      "step": 8000
    },
    {
      "epoch": 1.552580519984478,
      "eval_cosine_accuracy": 0.9629199504852295,
      "eval_loss": 0.805507481098175,
      "eval_runtime": 31.6662,
      "eval_samples_per_second": 298.931,
      "eval_steps_per_second": 1.168,
      "step": 8000
    },
    {
      "epoch": 1.746604578967792,
      "grad_norm": 3.4670729637145996,
      "learning_rate": 2.11324760402766e-05,
      "loss": 2.4791,
      "step": 9000
    },
    {
      "epoch": 1.746604578967792,
      "eval_cosine_accuracy": 0.9632368683815002,
      "eval_loss": 0.8120037317276001,
      "eval_runtime": 32.3325,
      "eval_samples_per_second": 292.77,
      "eval_steps_per_second": 1.144,
      "step": 9000
    },
    {
      "epoch": 1.940628637951106,
      "grad_norm": 3.514266014099121,
      "learning_rate": 1.9312750212301347e-05,
      "loss": 2.4854,
      "step": 10000
    },
    {
      "epoch": 1.940628637951106,
      "eval_cosine_accuracy": 0.9645045399665833,
      "eval_loss": 0.801057755947113,
      "eval_runtime": 32.1401,
      "eval_samples_per_second": 294.523,
      "eval_steps_per_second": 1.151,
      "step": 10000
    },
    {
      "epoch": 2.13465269693442,
      "grad_norm": 3.7418572902679443,
      "learning_rate": 1.7493024384326095e-05,
      "loss": 2.4283,
      "step": 11000
    },
    {
      "epoch": 2.13465269693442,
      "eval_cosine_accuracy": 0.965983510017395,
      "eval_loss": 0.80595862865448,
      "eval_runtime": 32.1443,
      "eval_samples_per_second": 294.485,
      "eval_steps_per_second": 1.151,
      "step": 11000
    },
    {
      "epoch": 2.328676755917734,
      "grad_norm": 4.063589096069336,
      "learning_rate": 1.5673298556350846e-05,
      "loss": 2.3847,
      "step": 12000
    },
    {
      "epoch": 2.328676755917734,
      "eval_cosine_accuracy": 0.967039942741394,
      "eval_loss": 0.8105789422988892,
      "eval_runtime": 32.6458,
      "eval_samples_per_second": 289.961,
      "eval_steps_per_second": 1.133,
      "step": 12000
    },
    {
      "epoch": 2.5227008149010475,
      "grad_norm": 3.5170540809631348,
      "learning_rate": 1.3855392454203567e-05,
      "loss": 2.3488,
      "step": 13000
    },
    {
      "epoch": 2.5227008149010475,
      "eval_cosine_accuracy": 0.9672512412071228,
      "eval_loss": 0.802911639213562,
      "eval_runtime": 32.9292,
      "eval_samples_per_second": 287.465,
      "eval_steps_per_second": 1.124,
      "step": 13000
    },
    {
      "epoch": 2.7167248738843615,
      "grad_norm": 3.778048515319824,
      "learning_rate": 1.2035666626228315e-05,
      "loss": 2.3237,
      "step": 14000
    },
    {
      "epoch": 2.7167248738843615,
      "eval_cosine_accuracy": 0.9676737785339355,
      "eval_loss": 0.8115915060043335,
      "eval_runtime": 32.7903,
      "eval_samples_per_second": 288.683,
      "eval_steps_per_second": 1.128,
      "step": 14000
    },
    {
      "epoch": 2.9107489328676754,
      "grad_norm": 3.5443954467773438,
      "learning_rate": 1.0215940798253063e-05,
      "loss": 2.3515,
      "step": 15000
    },
    {
      "epoch": 2.9107489328676754,
      "eval_cosine_accuracy": 0.9682019948959351,
      "eval_loss": 0.7974018454551697,
      "eval_runtime": 32.5168,
      "eval_samples_per_second": 291.111,
      "eval_steps_per_second": 1.138,
      "step": 15000
    },
    {
      "epoch": 3.1047729918509894,
      "grad_norm": 3.9634554386138916,
      "learning_rate": 8.396214970277812e-06,
      "loss": 2.2909,
      "step": 16000
    },
    {
      "epoch": 3.1047729918509894,
      "eval_cosine_accuracy": 0.9678850769996643,
      "eval_loss": 0.7994604706764221,
      "eval_runtime": 32.2162,
      "eval_samples_per_second": 293.827,
      "eval_steps_per_second": 1.148,
      "step": 16000
    },
    {
      "epoch": 3.2987970508343034,
      "grad_norm": 3.258812427520752,
      "learning_rate": 6.580128593958511e-06,
      "loss": 2.2851,
      "step": 17000
    },
    {
      "epoch": 3.2987970508343034,
      "eval_cosine_accuracy": 0.9693640470504761,
      "eval_loss": 0.7956675291061401,
      "eval_runtime": 32.2334,
      "eval_samples_per_second": 293.671,
      "eval_steps_per_second": 1.148,
      "step": 17000
    },
    {
      "epoch": 3.4928211098176174,
      "grad_norm": 4.7701897621154785,
      "learning_rate": 4.760402765983259e-06,
      "loss": 2.2746,
      "step": 18000
    },
    {
      "epoch": 3.4928211098176174,
      "eval_cosine_accuracy": 0.968307614326477,
      "eval_loss": 0.7924105525016785,
      "eval_runtime": 32.2428,
      "eval_samples_per_second": 293.585,
      "eval_steps_per_second": 1.148,
      "step": 18000
    },
    {
      "epoch": 3.6868451688009314,
      "grad_norm": 3.8555006980895996,
      "learning_rate": 2.940676938008007e-06,
      "loss": 2.3213,
      "step": 19000
    },
    {
      "epoch": 3.6868451688009314,
      "eval_cosine_accuracy": 0.9689414501190186,
      "eval_loss": 0.7967694997787476,
      "eval_runtime": 32.1998,
      "eval_samples_per_second": 293.977,
      "eval_steps_per_second": 1.149,
      "step": 19000
    },
    {
      "epoch": 3.8808692277842454,
      "grad_norm": 3.3843157291412354,
      "learning_rate": 1.1227708358607304e-06,
      "loss": 2.296,
      "step": 20000
    },
    {
      "epoch": 3.8808692277842454,
      "eval_cosine_accuracy": 0.9692584276199341,
      "eval_loss": 0.796402633190155,
      "eval_runtime": 32.3478,
      "eval_samples_per_second": 292.632,
      "eval_steps_per_second": 1.144,
      "step": 20000
    }
  ],
  "logging_steps": 1000,
  "max_steps": 20608,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}