redpj7B-lora-cnn-dailymail-results_10/checkpoint-200/trainer_state.json

{
  "best_metric": 1.8871846199035645,
  "best_model_checkpoint": "/domino/edv/afs-mrmc-data-store-rw/innovation/hf/redpj7B-lora-cnn-dailymail-results_10/checkpoint-200",
  "epoch": 1.3888888888888888,
  "global_step": 200,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.14,
      "learning_rate": 0.0,
      "loss": 1.8996,
      "step": 20
    },
    {
      "epoch": 0.28,
      "learning_rate": 0.0,
      "loss": 1.8773,
      "step": 40
    },
    {
      "epoch": 0.42,
      "learning_rate": 0.0,
      "loss": 1.8595,
      "step": 60
    },
    {
      "epoch": 0.56,
      "learning_rate": 0.0,
      "loss": 1.8621,
      "step": 80
    },
    {
      "epoch": 0.69,
      "learning_rate": 0.0,
      "loss": 1.8206,
      "step": 100
    },
    {
      "epoch": 0.83,
      "learning_rate": 0.0,
      "loss": 1.8477,
      "step": 120
    },
    {
      "epoch": 0.97,
      "learning_rate": 0.0,
      "loss": 1.9184,
      "step": 140
    },
    {
      "epoch": 1.11,
      "learning_rate": 0.0,
      "loss": 1.891,
      "step": 160
    },
    {
      "epoch": 1.25,
      "learning_rate": 0.0,
      "loss": 1.8885,
      "step": 180
    },
    {
      "epoch": 1.39,
      "learning_rate": 0.0,
      "loss": 1.8674,
      "step": 200
    },
    {
      "epoch": 1.39,
      "eval_loss": 1.8871846199035645,
      "eval_runtime": 12064.3322,
      "eval_samples_per_second": 1.108,
      "eval_steps_per_second": 0.139,
      "step": 200
    }
  ],
  "max_steps": 432,
  "num_train_epochs": 3,
  "total_flos": 1.2442809896681472e+16,
  "trial_name": null,
  "trial_params": null
}