trainer_state.json

{
  "best_metric": 0.8608760424983216,
  "best_model_checkpoint": "result/my-sup-simcse-bert-base-uncased",
  "epoch": 10.0,
  "global_step": 1660,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.75,
      "eval_avg_sts": 0.8315040583627913,
      "eval_sickr_spearman": 0.8021320742272611,
      "eval_stsb_spearman": 0.8608760424983216,
      "step": 125
    },
    {
      "epoch": 1.51,
      "eval_avg_sts": 0.8276682684969617,
      "eval_sickr_spearman": 0.8024215096445272,
      "eval_stsb_spearman": 0.8529150273493963,
      "step": 250
    },
    {
      "epoch": 2.26,
      "eval_avg_sts": 0.8292414115876203,
      "eval_sickr_spearman": 0.8035029298934062,
      "eval_stsb_spearman": 0.8549798932818345,
      "step": 375
    },
    {
      "epoch": 3.01,
      "learning_rate": 3.4939759036144585e-05,
      "loss": 0.1927,
      "step": 500
    },
    {
      "epoch": 3.01,
      "eval_avg_sts": 0.8242638816167347,
      "eval_sickr_spearman": 0.8022541692871167,
      "eval_stsb_spearman": 0.8462735939463528,
      "step": 500
    },
    {
      "epoch": 3.77,
      "eval_avg_sts": 0.8233283779153091,
      "eval_sickr_spearman": 0.7945779827485818,
      "eval_stsb_spearman": 0.8520787730820363,
      "step": 625
    },
    {
      "epoch": 4.52,
      "eval_avg_sts": 0.8215947619403889,
      "eval_sickr_spearman": 0.7907777620027984,
      "eval_stsb_spearman": 0.8524117618779796,
      "step": 750
    },
    {
      "epoch": 5.27,
      "eval_avg_sts": 0.8227357575888109,
      "eval_sickr_spearman": 0.7947862936355189,
      "eval_stsb_spearman": 0.8506852215421029,
      "step": 875
    },
    {
      "epoch": 6.02,
      "learning_rate": 1.9879518072289157e-05,
      "loss": 0.0995,
      "step": 1000
    },
    {
      "epoch": 6.02,
      "eval_avg_sts": 0.8234624613863508,
      "eval_sickr_spearman": 0.7961339983107709,
      "eval_stsb_spearman": 0.8507909244619307,
      "step": 1000
    },
    {
      "epoch": 6.78,
      "eval_avg_sts": 0.8199047038936794,
      "eval_sickr_spearman": 0.7917320919572722,
      "eval_stsb_spearman": 0.8480773158300866,
      "step": 1125
    },
    {
      "epoch": 7.53,
      "eval_avg_sts": 0.8194132158921612,
      "eval_sickr_spearman": 0.7884059381826829,
      "eval_stsb_spearman": 0.8504204936016395,
      "step": 1250
    },
    {
      "epoch": 8.28,
      "eval_avg_sts": 0.8188285236637569,
      "eval_sickr_spearman": 0.7876833102615466,
      "eval_stsb_spearman": 0.8499737370659672,
      "step": 1375
    },
    {
      "epoch": 9.04,
      "learning_rate": 4.819277108433735e-06,
      "loss": 0.0874,
      "step": 1500
    },
    {
      "epoch": 9.04,
      "eval_avg_sts": 0.8193720271717073,
      "eval_sickr_spearman": 0.7877823023616106,
      "eval_stsb_spearman": 0.850961751981804,
      "step": 1500
    },
    {
      "epoch": 9.79,
      "eval_avg_sts": 0.8188424049556793,
      "eval_sickr_spearman": 0.7873486775778272,
      "eval_stsb_spearman": 0.8503361323335314,
      "step": 1625
    },
    {
      "epoch": 10.0,
      "step": 1660,
      "train_runtime": 4081.7638,
      "train_samples_per_second": 0.407
    }
  ],
  "max_steps": 1660,
  "num_train_epochs": 10,
  "total_flos": 26836143897968640,
  "trial_name": null,
  "trial_params": null
}