trainer_state.json

{
  "best_global_step": 30,
  "best_metric": 0.0006636497564613819,
  "best_model_checkpoint": "outputs/microsoft/Phi-3.5-mini-instruct/countdown_n50_mcl_256/checkpoint-30",
  "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": 0.0,
      "learning_rate": 8.888888888888888e-07,
      "loss": 0.0021,
      "step": 5
    },
    {
      "epoch": 0.4,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 249.95,
      "eval_kl": 4.139542579650879e-05,
      "eval_loss": 0.013958251103758812,
      "eval_num_tokens": 31702.0,
      "eval_reward": 0.0375,
      "eval_reward_std": 0.075,
      "eval_rewards/equation_reward_func": 0.025,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 282.7565,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 5
    },
    {
      "epoch": 0.8,
      "grad_norm": 0.02207428403198719,
      "learning_rate": 7.5e-07,
      "loss": 0.007,
      "step": 10
    },
    {
      "epoch": 0.8,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.1625,
      "eval_kl": 4.143416881561279e-05,
      "eval_loss": 0.0009803941939026117,
      "eval_num_tokens": 63209.0,
      "eval_reward": 0.0375,
      "eval_reward_std": 0.075,
      "eval_rewards/equation_reward_func": 0.0375,
      "eval_rewards/format_reward_func": 0.0,
      "eval_runtime": 282.4506,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 10
    },
    {
      "epoch": 1.24,
      "grad_norm": 0.023838160559535027,
      "learning_rate": 6.111111111111112e-07,
      "loss": 0.004,
      "step": 15
    },
    {
      "epoch": 1.24,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 247.625,
      "eval_kl": 4.603266716003418e-05,
      "eval_loss": 0.0016482158098369837,
      "eval_num_tokens": 94609.0,
      "eval_reward": 0.0375,
      "eval_reward_std": 0.075,
      "eval_rewards/equation_reward_func": 0.025,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 282.5452,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 15
    },
    {
      "epoch": 1.6400000000000001,
      "grad_norm": 5.511666358870571e-07,
      "learning_rate": 4.722222222222222e-07,
      "loss": 0.0006,
      "step": 20
    },
    {
      "epoch": 1.6400000000000001,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.5625,
      "eval_kl": 4.3252110481262206e-05,
      "eval_loss": 0.0033958435524255037,
      "eval_num_tokens": 126068.0,
      "eval_reward": 0.025,
      "eval_reward_std": 0.05,
      "eval_rewards/equation_reward_func": 0.025,
      "eval_rewards/format_reward_func": 0.0,
      "eval_runtime": 282.4816,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 20
    },
    {
      "epoch": 2.08,
      "grad_norm": 0.020114393904805183,
      "learning_rate": 3.333333333333333e-07,
      "loss": 0.0308,
      "step": 25
    },
    {
      "epoch": 2.08,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.9375,
      "eval_kl": 3.747940063476562e-05,
      "eval_loss": 0.007938861846923828,
      "eval_num_tokens": 157153.0,
      "eval_reward": 0.05,
      "eval_reward_std": 0.07287135720252991,
      "eval_rewards/equation_reward_func": 0.025,
      "eval_rewards/format_reward_func": 0.025,
      "eval_runtime": 282.3454,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 25
    },
    {
      "epoch": 2.48,
      "grad_norm": 7.424079626616731e-07,
      "learning_rate": 1.9444444444444445e-07,
      "loss": 0.0,
      "step": 30
    },
    {
      "epoch": 2.48,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.325,
      "eval_kl": 3.4242868423461914e-05,
      "eval_loss": 0.0006636497564613819,
      "eval_num_tokens": 188560.0,
      "eval_reward": 0.0625,
      "eval_reward_std": 0.10386751294136047,
      "eval_rewards/equation_reward_func": 0.05,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 282.4117,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "step": 30
    },
    {
      "epoch": 2.88,
      "grad_norm": 4.1665236949484097e-07,
      "learning_rate": 5.555555555555555e-08,
      "loss": 0.006,
      "step": 35
    },
    {
      "epoch": 2.88,
      "eval_clip_ratio": 0.0,
      "eval_completion_length": 251.825,
      "eval_kl": 4.251301288604736e-05,
      "eval_loss": 0.000741269439458847,
      "eval_num_tokens": 220089.0,
      "eval_reward": 0.0375,
      "eval_reward_std": 0.075,
      "eval_rewards/equation_reward_func": 0.025,
      "eval_rewards/format_reward_func": 0.0125,
      "eval_runtime": 282.813,
      "eval_samples_per_second": 0.071,
      "eval_steps_per_second": 0.018,
      "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
}