checkpoint-2250/trainer_state.json

{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 1.3266509433962264,
  "eval_steps": 2250,
  "global_step": 2250,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.589622641509434,
      "grad_norm": 9.137022018432617,
      "learning_rate": 2.9481132075471702e-05,
      "loss": 2.8435,
      "step": 1000
    },
    {
      "epoch": 1.179245283018868,
      "grad_norm": 6.052332401275635,
      "learning_rate": 5.8962264150943403e-05,
      "loss": 1.791,
      "step": 2000
    },
    {
      "epoch": 1.3266509433962264,
      "eval_accuracy_background": 0.6658729859800179,
      "eval_accuracy_double_plant": 0.44796971861142204,
      "eval_accuracy_drydown": 0.838452272819733,
      "eval_accuracy_endrow": 0.6564099936974693,
      "eval_accuracy_nutrient_deficiency": 0.5255760511997533,
      "eval_accuracy_planter_skip": 0.7062276764551357,
      "eval_accuracy_water": 0.9873387227342924,
      "eval_accuracy_waterway": 0.5770296385642565,
      "eval_accuracy_weed_cluster": 0.4428120632960455,
      "eval_iou_background": 0.5904190410028399,
      "eval_iou_double_plant": 0.21713659970128166,
      "eval_iou_drydown": 0.5191145408680865,
      "eval_iou_endrow": 0.146705809299966,
      "eval_iou_nutrient_deficiency": 0.3433700256479391,
      "eval_iou_planter_skip": 0.19867844175237825,
      "eval_iou_water": 0.3151796347651091,
      "eval_iou_waterway": 0.17710299517028666,
      "eval_iou_weed_cluster": 0.2871479144864637,
      "eval_loss": 1.6307891607284546,
      "eval_mean_accuracy": 0.6497432359286807,
      "eval_mean_iou": 0.31053944474381673,
      "eval_overall_accuracy": 0.6703095203666921,
      "eval_runtime": 550.4413,
      "eval_samples_per_second": 15.489,
      "eval_steps_per_second": 0.485,
      "step": 2250
    }
  ],
  "logging_steps": 1000,
  "max_steps": 33920,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 20,
  "save_steps": 2250,
  "total_flos": 3.831820232429037e+19,
  "train_batch_size": 32,
  "trial_name": null,
  "trial_params": null
}