{
  "best_global_step": 25,
  "best_metric": -0.0033106361515820026,
  "best_model_checkpoint": null,
  "epoch": 2.96,
  "eval_steps": 5,
  "global_step": 36,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.4,
      "grad_norm": 0.03278883174061775,
      "learning_rate": 8.888888888888888e-07,
      "loss": 0.0226,
      "step": 5
    },
    {
      "epoch": 0.4,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 243.8875,
      "eval_kl": 3.4737586975097654e-05,
      "eval_loss": 0.012693313881754875,
      "eval_num_tokens": 30947.0,
      "eval_reward": 0.025,
      "eval_reward_std": 0.05,
      "eval_rewards/equation_reward_func": 0.0,
      "eval_rewards/format_reward_func": 0.025,
      "eval_runtime": 648.9536,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 5
    },
    {
      "epoch": 0.8,
      "grad_norm": 7.537942110502627e-07,
      "learning_rate": 7.5e-07,
      "loss": 0.0275,
      "step": 10
    },
    {
      "epoch": 0.8,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 253.3,
      "eval_kl": 3.842711448669434e-05,
      "eval_loss": 0.005588035099208355,
      "eval_num_tokens": 61638.0,
      "eval_reward": 0.0375,
      "eval_reward_std": 0.075,
      "eval_rewards/equation_reward_func": 0.0,
      "eval_rewards/format_reward_func": 0.0375,
      "eval_runtime": 649.655,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 10
    },
    {
      "epoch": 1.24,
      "grad_norm": 0.023165149614214897,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.022,
      "step": 15
    },
    {
      "epoch": 1.24,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 243.65,
      "eval_kl": 3.121495246887207e-05,
      "eval_loss": 0.012429839000105858,
      "eval_num_tokens": 92954.0,
      "eval_reward": 0.0625,
      "eval_reward_std": 0.08273502588272094,
      "eval_rewards/equation_reward_func": 0.05,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 649.1092,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 15
    },
    {
      "epoch": 1.6400000000000001,
      "grad_norm": 0.016736147925257683,
      "learning_rate": 4.722222222222222e-07,
      "loss": 0.0022,
      "step": 20
    },
    {
      "epoch": 1.6400000000000001,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 250.075,
      "eval_kl": 3.5655498504638675e-05,
      "eval_loss": 0.01138834934681654,
      "eval_num_tokens": 123546.0,
      "eval_reward": 0.0125,
      "eval_reward_std": 0.025,
      "eval_rewards/equation_reward_func": 0.0,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 649.9636,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 20
    },
    {
      "epoch": 2.08,
      "grad_norm": 0.0003201386716682464,
      "learning_rate": 3.333333333333333e-07,
      "loss": -0.0003,
      "step": 25
    },
    {
      "epoch": 2.08,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.15,
      "eval_kl": 3.8629770278930665e-05,
      "eval_loss": -0.0033106361515820026,
      "eval_num_tokens": 155009.0,
      "eval_reward": 0.025,
      "eval_reward_std": 0.05,
      "eval_rewards/equation_reward_func": 0.0125,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 649.8591,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 25
    },
    {
      "epoch": 2.48,
      "grad_norm": 4.24052643666073e-07,
      "learning_rate": 1.9444444444444445e-07,
      "loss": 0.0146,
      "step": 30
    },
    {
      "epoch": 2.48,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 246.9375,
      "eval_kl": 3.491044044494629e-05,
      "eval_loss": 0.006897517945617437,
      "eval_num_tokens": 185888.0,
      "eval_reward": 0.025,
      "eval_reward_std": 0.05,
      "eval_rewards/equation_reward_func": 0.0125,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 650.2735,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 30
    },
    {
      "epoch": 2.88,
      "grad_norm": 1.9729814084712416e-06,
      "learning_rate": 5.555555555555555e-08,
      "loss": 0.0,
      "step": 35
    },
    {
      "epoch": 2.88,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 245.1,
      "eval_kl": 4.070103168487549e-05,
      "eval_loss": 0.01672522909939289,
      "eval_num_tokens": 216927.0,
      "eval_reward": 0.025,
      "eval_reward_std": 0.05,
      "eval_rewards/equation_reward_func": 0.0,
      "eval_rewards/format_reward_func": 0.025,
      "eval_runtime": 649.2667,
      "eval_samples_per_second": 0.031,
      "eval_steps_per_second": 0.008,
      "step": 35
    }
  ],
  "logging_steps": 5,
  "max_steps": 36,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 15,
  "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": 4,
  "trial_name": null,
  "trial_params": null
}