checkpoint-2500/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.249375312343828,
  "eval_steps": 500,
  "global_step": 2500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.25,
      "grad_norm": 1.5337328910827637,
      "learning_rate": 1.900049975012494e-05,
      "loss": 0.4281,
      "step": 500
    },
    {
      "epoch": 0.5,
      "grad_norm": 1.992353081703186,
      "learning_rate": 1.8000999500249876e-05,
      "loss": 0.2656,
      "step": 1000
    },
    {
      "epoch": 0.75,
      "grad_norm": 1.4741365909576416,
      "learning_rate": 1.7001499250374815e-05,
      "loss": 0.263,
      "step": 1500
    },
    {
      "epoch": 1.0,
      "grad_norm": 0.5607998967170715,
      "learning_rate": 1.600199900049975e-05,
      "loss": 0.2228,
      "step": 2000
    },
    {
      "epoch": 1.25,
      "grad_norm": 56.73757553100586,
      "learning_rate": 1.500249875062469e-05,
      "loss": 0.1552,
      "step": 2500
    }
  ],
  "logging_steps": 500,
  "max_steps": 10005,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 5,
  "save_steps": 500,
  "total_flos": 4658025936754176.0,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}