trainer_state.json

{
  "best_metric": 2.6738457679748535,
  "best_model_checkpoint": "/home/zhengxinyong/outputs/bloom-1b7_de_continual-pretrain_100000samples_-1vocab_original/checkpoint-25000",
  "epoch": 2.7477950267893942,
  "global_step": 25000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.27,
      "learning_rate": 0.0001,
      "loss": 3.1242,
      "step": 2500
    },
    {
      "epoch": 0.55,
      "learning_rate": 0.0001,
      "loss": 2.7945,
      "step": 5000
    },
    {
      "epoch": 0.55,
      "eval_loss": 2.839371681213379,
      "eval_runtime": 188.5847,
      "eval_samples_per_second": 4.263,
      "eval_steps_per_second": 2.132,
      "step": 5000
    },
    {
      "epoch": 0.82,
      "learning_rate": 0.0001,
      "loss": 2.6979,
      "step": 7500
    },
    {
      "epoch": 1.1,
      "learning_rate": 0.0001,
      "loss": 2.5465,
      "step": 10000
    },
    {
      "epoch": 1.1,
      "eval_loss": 2.7334208488464355,
      "eval_runtime": 188.8824,
      "eval_samples_per_second": 4.257,
      "eval_steps_per_second": 2.128,
      "step": 10000
    },
    {
      "epoch": 1.37,
      "learning_rate": 0.0001,
      "loss": 2.3964,
      "step": 12500
    },
    {
      "epoch": 1.65,
      "learning_rate": 0.0001,
      "loss": 2.4114,
      "step": 15000
    },
    {
      "epoch": 1.65,
      "eval_loss": 2.676481246948242,
      "eval_runtime": 188.8898,
      "eval_samples_per_second": 4.256,
      "eval_steps_per_second": 2.128,
      "step": 15000
    },
    {
      "epoch": 1.92,
      "learning_rate": 0.0001,
      "loss": 2.402,
      "step": 17500
    },
    {
      "epoch": 2.2,
      "learning_rate": 0.0001,
      "loss": 2.1601,
      "step": 20000
    },
    {
      "epoch": 2.2,
      "eval_loss": 2.703705072402954,
      "eval_runtime": 188.8146,
      "eval_samples_per_second": 4.258,
      "eval_steps_per_second": 2.129,
      "step": 20000
    },
    {
      "epoch": 2.47,
      "learning_rate": 0.0001,
      "loss": 2.1107,
      "step": 22500
    },
    {
      "epoch": 2.75,
      "learning_rate": 0.0001,
      "loss": 2.157,
      "step": 25000
    },
    {
      "epoch": 2.75,
      "eval_loss": 2.6738457679748535,
      "eval_runtime": 188.7842,
      "eval_samples_per_second": 4.259,
      "eval_steps_per_second": 2.129,
      "step": 25000
    }
  ],
  "max_steps": 25000,
  "num_train_epochs": 3,
  "total_flos": 1.4852250484437156e+18,
  "trial_name": null,
  "trial_params": null
}