{
  "best_metric": null,
  "best_model_checkpoint": null,
  "best_supernet_model_checkpoint": null,
  "epoch": 12.0,
  "global_step": 3216,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 1.0,
      "eval_loss": 0.7594008445739746,
      "eval_matthews_correlation": 0.22693312638521376,
      "eval_runtime": 0.6384,
      "eval_samples_per_second": 1633.689,
      "eval_steps_per_second": 14.097,
      "step": 268
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 1.0,
      "eval_loss": 0.7153037786483765,
      "eval_matthews_correlation": 0.33478791021206633,
      "eval_runtime": 0.7177,
      "eval_samples_per_second": 1453.27,
      "eval_steps_per_second": 12.54,
      "step": 268
    },
    {
      "compression_loss": 0.0,
      "epoch": 1.87,
      "learning_rate": 1.884896652217917e-05,
      "loss": 1.1021,
      "step": 500
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 2.0,
      "eval_loss": 0.6954232454299927,
      "eval_matthews_correlation": 0.4354344418941058,
      "eval_runtime": 0.7494,
      "eval_samples_per_second": 1391.709,
      "eval_steps_per_second": 12.009,
      "step": 536
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 2.0,
      "eval_loss": 0.6847538352012634,
      "eval_matthews_correlation": 0.4723897975346641,
      "eval_runtime": 0.8051,
      "eval_samples_per_second": 1295.486,
      "eval_steps_per_second": 11.179,
      "step": 536
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 3.0,
      "eval_loss": 0.7816486954689026,
      "eval_matthews_correlation": 0.4156507632951837,
      "eval_runtime": 0.6601,
      "eval_samples_per_second": 1580.016,
      "eval_steps_per_second": 13.634,
      "step": 804
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 3.0,
      "eval_loss": 0.7744055986404419,
      "eval_matthews_correlation": 0.43824800688521975,
      "eval_runtime": 0.74,
      "eval_samples_per_second": 1409.462,
      "eval_steps_per_second": 12.162,
      "step": 804
    },
    {
      "compression_loss": 0.0,
      "epoch": 3.73,
      "learning_rate": 1.5644725411751336e-05,
      "loss": 0.6376,
      "step": 1000
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 4.0,
      "eval_loss": 0.8254319429397583,
      "eval_matthews_correlation": 0.42409562062729317,
      "eval_runtime": 0.6033,
      "eval_samples_per_second": 1728.723,
      "eval_steps_per_second": 14.917,
      "step": 1072
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 4.0,
      "eval_loss": 0.8103329539299011,
      "eval_matthews_correlation": 0.45596899298288374,
      "eval_runtime": 0.6386,
      "eval_samples_per_second": 1633.35,
      "eval_steps_per_second": 14.094,
      "step": 1072
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 5.0,
      "eval_loss": 0.8831449747085571,
      "eval_matthews_correlation": 0.4133635813145441,
      "eval_runtime": 0.6198,
      "eval_samples_per_second": 1682.88,
      "eval_steps_per_second": 14.521,
      "step": 1340
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 5.0,
      "eval_loss": 0.8358224034309387,
      "eval_matthews_correlation": 0.4637527852072172,
      "eval_runtime": 0.63,
      "eval_samples_per_second": 1655.659,
      "eval_steps_per_second": 14.287,
      "step": 1340
    },
    {
      "compression_loss": 0.0,
      "epoch": 5.6,
      "learning_rate": 1.1130738016122495e-05,
      "loss": 0.3925,
      "step": 1500
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 6.0,
      "eval_loss": 0.8798311948776245,
      "eval_matthews_correlation": 0.43765045412808823,
      "eval_runtime": 0.6072,
      "eval_samples_per_second": 1717.727,
      "eval_steps_per_second": 14.822,
      "step": 1608
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 6.0,
      "eval_loss": 0.87115079164505,
      "eval_matthews_correlation": 0.4572856727029901,
      "eval_runtime": 0.6525,
      "eval_samples_per_second": 1598.36,
      "eval_steps_per_second": 13.792,
      "step": 1608
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 7.0,
      "eval_loss": 0.9016041159629822,
      "eval_matthews_correlation": 0.43794163748101905,
      "eval_runtime": 0.5925,
      "eval_samples_per_second": 1760.294,
      "eval_steps_per_second": 15.19,
      "step": 1876
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 7.0,
      "eval_loss": 0.9116769433021545,
      "eval_matthews_correlation": 0.4523199346278705,
      "eval_runtime": 0.622,
      "eval_samples_per_second": 1676.901,
      "eval_steps_per_second": 14.47,
      "step": 1876
    },
    {
      "compression_loss": 0.0,
      "epoch": 7.46,
      "learning_rate": 6.354385348824488e-06,
      "loss": 0.2697,
      "step": 2000
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 8.0,
      "eval_loss": 0.9721390008926392,
      "eval_matthews_correlation": 0.43577816963420296,
      "eval_runtime": 0.6725,
      "eval_samples_per_second": 1550.837,
      "eval_steps_per_second": 13.382,
      "step": 2144
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 8.0,
      "eval_loss": 1.0106229782104492,
      "eval_matthews_correlation": 0.43907744261562187,
      "eval_runtime": 0.7223,
      "eval_samples_per_second": 1443.926,
      "eval_steps_per_second": 12.46,
      "step": 2144
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 9.0,
      "eval_loss": 0.9457906484603882,
      "eval_matthews_correlation": 0.430812413056651,
      "eval_runtime": 0.7253,
      "eval_samples_per_second": 1438.048,
      "eval_steps_per_second": 12.409,
      "step": 2412
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 9.0,
      "eval_loss": 0.969419538974762,
      "eval_matthews_correlation": 0.43758782623801895,
      "eval_runtime": 0.706,
      "eval_samples_per_second": 1477.441,
      "eval_steps_per_second": 12.749,
      "step": 2412
    },
    {
      "compression_loss": 0.0,
      "epoch": 9.33,
      "learning_rate": 2.4239250674893345e-06,
      "loss": 0.2133,
      "step": 2500
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 10.0,
      "eval_loss": 0.9448842406272888,
      "eval_matthews_correlation": 0.44105796467098995,
      "eval_runtime": 0.611,
      "eval_samples_per_second": 1706.975,
      "eval_steps_per_second": 14.729,
      "step": 2680
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 10.0,
      "eval_loss": 0.9690240025520325,
      "eval_matthews_correlation": 0.4315719153344676,
      "eval_runtime": 0.617,
      "eval_samples_per_second": 1690.365,
      "eval_steps_per_second": 14.586,
      "step": 2680
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 11.0,
      "eval_loss": 0.9587270021438599,
      "eval_matthews_correlation": 0.4401936795964168,
      "eval_runtime": 0.6843,
      "eval_samples_per_second": 1524.176,
      "eval_steps_per_second": 13.152,
      "step": 2948
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 11.0,
      "eval_loss": 0.9697301983833313,
      "eval_matthews_correlation": 0.4243567637161194,
      "eval_runtime": 0.6973,
      "eval_samples_per_second": 1495.775,
      "eval_steps_per_second": 12.907,
      "step": 2948
    },
    {
      "compression_loss": 0.0,
      "epoch": 11.19,
      "learning_rate": 2.5352335693478615e-07,
      "loss": 0.1818,
      "step": 3000
    },
    {
      "Minimum SubNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 320, 1: 192, 2: 192, 3: 320, 4: 448, 5: 320, 6: 256, 7: 320, 8: 904, 9: 878, 10: 885, 11: 910, 12: 944, 13: 943, 14: 896, 15: 524})])",
      "epoch": 12.0,
      "eval_loss": 0.9606204628944397,
      "eval_matthews_correlation": 0.4401936795964168,
      "eval_runtime": 0.7028,
      "eval_samples_per_second": 1484.092,
      "eval_steps_per_second": 12.806,
      "step": 3216
    },
    {
      "SuperNet": "OrderedDict([(<ElasticityDim.WIDTH: 'width'>, {0: 512, 1: 512, 2: 512, 3: 512, 4: 512, 5: 512, 6: 512, 7: 512, 8: 2048, 9: 2048, 10: 2048, 11: 2048, 12: 2048, 13: 2048, 14: 2048, 15: 2048})])",
      "epoch": 12.0,
      "eval_loss": 0.969758152961731,
      "eval_matthews_correlation": 0.4243567637161194,
      "eval_runtime": 0.7371,
      "eval_samples_per_second": 1415.06,
      "eval_steps_per_second": 12.21,
      "step": 3216
    },
    {
      "epoch": 12.0,
      "step": 3216,
      "total_flos": 2008275016808448.0,
      "train_loss": 0.4476036005352267,
      "train_runtime": 3049.2312,
      "train_samples_per_second": 33.652,
      "train_steps_per_second": 1.055
    }
  ],
  "max_steps": 3216,
  "min_subnet_acc": null,
  "min_subnet_best_acc": null,
  "num_train_epochs": 12,
  "supernet_acc": null,
  "supernet_best_acc": null,
  "total_flos": 2008275016808448.0,
  "trial_name": null,
  "trial_params": null
}