{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 3.998058880621158,
  "eval_steps": 1000,
  "global_step": 12356,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.0003237293622531564,
      "grad_norm": 5.294984340667725,
      "learning_rate": 0.0,
      "loss": 2.5131,
      "step": 1
    },
    {
      "epoch": 0.32372936225315635,
      "grad_norm": 3.888833522796631,
      "learning_rate": 2.424757281553398e-05,
      "loss": 1.8415,
      "step": 1000
    },
    {
      "epoch": 0.32372936225315635,
      "eval_cosine_accuracy": 0.9511937499046326,
      "eval_loss": 1.082350730895996,
      "eval_runtime": 25.045,
      "eval_samples_per_second": 377.959,
      "eval_steps_per_second": 1.477,
      "step": 1000
    },
    {
      "epoch": 0.6474587245063127,
      "grad_norm": 3.7192389965057373,
      "learning_rate": 2.7941546762589928e-05,
      "loss": 1.3696,
      "step": 2000
    },
    {
      "epoch": 0.6474587245063127,
      "eval_cosine_accuracy": 0.9616522192955017,
      "eval_loss": 0.9928585290908813,
      "eval_runtime": 24.1997,
      "eval_samples_per_second": 391.162,
      "eval_steps_per_second": 1.529,
      "step": 2000
    },
    {
      "epoch": 0.9711880867594691,
      "grad_norm": 3.8493621349334717,
      "learning_rate": 2.5243705035971223e-05,
      "loss": 1.4502,
      "step": 3000
    },
    {
      "epoch": 0.9711880867594691,
      "eval_cosine_accuracy": 0.9655609726905823,
      "eval_loss": 0.94865882396698,
      "eval_runtime": 24.3555,
      "eval_samples_per_second": 388.66,
      "eval_steps_per_second": 1.519,
      "step": 3000
    },
    {
      "epoch": 1.2947266256874799,
      "grad_norm": 3.0650620460510254,
      "learning_rate": 2.254586330935252e-05,
      "loss": 1.3141,
      "step": 4000
    },
    {
      "epoch": 1.2947266256874799,
      "eval_cosine_accuracy": 0.9704204797744751,
      "eval_loss": 0.8925007581710815,
      "eval_runtime": 24.3977,
      "eval_samples_per_second": 387.987,
      "eval_steps_per_second": 1.517,
      "step": 4000
    },
    {
      "epoch": 1.6182465221611129,
      "grad_norm": 2.806612968444824,
      "learning_rate": 1.985071942446043e-05,
      "loss": 1.1692,
      "step": 5000
    },
    {
      "epoch": 1.6182465221611129,
      "eval_cosine_accuracy": 0.9709486365318298,
      "eval_loss": 0.8780702352523804,
      "eval_runtime": 24.3201,
      "eval_samples_per_second": 389.226,
      "eval_steps_per_second": 1.521,
      "step": 5000
    },
    {
      "epoch": 1.941766418634746,
      "grad_norm": 3.0037014484405518,
      "learning_rate": 1.7152877697841724e-05,
      "loss": 1.1209,
      "step": 6000
    },
    {
      "epoch": 1.941766418634746,
      "eval_cosine_accuracy": 0.9717937707901001,
      "eval_loss": 0.8578657507896423,
      "eval_runtime": 24.5063,
      "eval_samples_per_second": 386.268,
      "eval_steps_per_second": 1.51,
      "step": 6000
    },
    {
      "epoch": 2.265286315108379,
      "grad_norm": 3.225210666656494,
      "learning_rate": 1.445773381294964e-05,
      "loss": 1.0609,
      "step": 7000
    },
    {
      "epoch": 2.265286315108379,
      "eval_cosine_accuracy": 0.9738009572029114,
      "eval_loss": 0.8649092316627502,
      "eval_runtime": 24.4052,
      "eval_samples_per_second": 387.869,
      "eval_steps_per_second": 1.516,
      "step": 7000
    },
    {
      "epoch": 2.5888062115820123,
      "grad_norm": 2.8941006660461426,
      "learning_rate": 1.1759892086330935e-05,
      "loss": 1.0507,
      "step": 8000
    },
    {
      "epoch": 2.5888062115820123,
      "eval_cosine_accuracy": 0.9725332856178284,
      "eval_loss": 0.8568609356880188,
      "eval_runtime": 28.1064,
      "eval_samples_per_second": 336.792,
      "eval_steps_per_second": 1.316,
      "step": 8000
    },
    {
      "epoch": 2.9123261080556455,
      "grad_norm": 6.115662574768066,
      "learning_rate": 9.064748201438849e-06,
      "loss": 1.0079,
      "step": 9000
    },
    {
      "epoch": 2.9123261080556455,
      "eval_cosine_accuracy": 0.9735897183418274,
      "eval_loss": 0.8492642641067505,
      "eval_runtime": 24.4414,
      "eval_samples_per_second": 387.293,
      "eval_steps_per_second": 1.514,
      "step": 9000
    },
    {
      "epoch": 3.2358460045292787,
      "grad_norm": 2.8798933029174805,
      "learning_rate": 6.366906474820144e-06,
      "loss": 1.0006,
      "step": 10000
    },
    {
      "epoch": 3.2358460045292787,
      "eval_cosine_accuracy": 0.9734840393066406,
      "eval_loss": 0.8392395973205566,
      "eval_runtime": 24.2481,
      "eval_samples_per_second": 390.381,
      "eval_steps_per_second": 1.526,
      "step": 10000
    },
    {
      "epoch": 3.559365901002912,
      "grad_norm": 3.430743932723999,
      "learning_rate": 3.6717625899280574e-06,
      "loss": 0.9947,
      "step": 11000
    },
    {
      "epoch": 3.559365901002912,
      "eval_cosine_accuracy": 0.9750686883926392,
      "eval_loss": 0.8390318155288696,
      "eval_runtime": 24.3341,
      "eval_samples_per_second": 389.001,
      "eval_steps_per_second": 1.52,
      "step": 11000
    },
    {
      "epoch": 3.8828857974765447,
      "grad_norm": 2.899552583694458,
      "learning_rate": 9.739208633093526e-07,
      "loss": 0.9774,
      "step": 12000
    },
    {
      "epoch": 3.8828857974765447,
      "eval_cosine_accuracy": 0.9748573899269104,
      "eval_loss": 0.840337872505188,
      "eval_runtime": 24.3905,
      "eval_samples_per_second": 388.101,
      "eval_steps_per_second": 1.517,
      "step": 12000
    }
  ],
  "logging_steps": 1000,
  "max_steps": 12356,
  "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
}