{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 51.02048,
  "eval_steps": 500,
  "global_step": 10000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "balancing_entropy": 1.0919,
      "balancing_loss": 1.2522,
      "epoch": 0.0512,
      "grad_norm": 0.7890625,
      "learning_rate": 6e-05,
      "loss": 12.4605,
      "router_z_loss": 1.5055,
      "step": 10
    },
    {
      "balancing_entropy": 1.0861,
      "balancing_loss": 1.2521,
      "epoch": 0.1024,
      "grad_norm": 0.58203125,
      "learning_rate": 0.00012,
      "loss": 12.0881,
      "router_z_loss": 1.4914,
      "step": 20
    },
    {
      "balancing_entropy": 1.0744,
      "balancing_loss": 1.2507,
      "epoch": 0.1536,
      "grad_norm": 0.55859375,
      "learning_rate": 0.00017999999999999998,
      "loss": 11.2974,
      "router_z_loss": 1.4148,
      "step": 30
    },
    {
      "balancing_entropy": 1.0317,
      "balancing_loss": 1.254,
      "epoch": 0.2048,
      "grad_norm": 0.50390625,
      "learning_rate": 0.00024,
      "loss": 10.2003,
      "router_z_loss": 1.2813,
      "step": 40
    },
    {
      "balancing_entropy": 1.0752,
      "balancing_loss": 1.2505,
      "epoch": 0.256,
      "grad_norm": 0.54296875,
      "learning_rate": 0.00030000000000000003,
      "loss": 9.3136,
      "router_z_loss": 1.191,
      "step": 50
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.2494,
      "epoch": 0.3072,
      "grad_norm": 0.70703125,
      "learning_rate": 0.00035999999999999997,
      "loss": 8.8813,
      "router_z_loss": 1.0029,
      "step": 60
    },
    {
      "balancing_entropy": 1.0038,
      "balancing_loss": 1.2603,
      "epoch": 0.3584,
      "grad_norm": 2.328125,
      "learning_rate": 0.00042000000000000007,
      "loss": 8.4725,
      "router_z_loss": 0.7408,
      "step": 70
    },
    {
      "balancing_entropy": 1.0507,
      "balancing_loss": 1.2492,
      "epoch": 0.4096,
      "grad_norm": 1.15625,
      "learning_rate": 0.00048,
      "loss": 7.933,
      "router_z_loss": 0.6414,
      "step": 80
    },
    {
      "balancing_entropy": 1.0537,
      "balancing_loss": 1.2466,
      "epoch": 0.4608,
      "grad_norm": 0.431640625,
      "learning_rate": 0.00054,
      "loss": 7.3236,
      "router_z_loss": 0.3564,
      "step": 90
    },
    {
      "balancing_entropy": 1.0876,
      "balancing_loss": 1.2455,
      "epoch": 0.512,
      "grad_norm": 1.234375,
      "learning_rate": 0.0006000000000000001,
      "loss": 6.8889,
      "router_z_loss": 0.2195,
      "step": 100
    },
    {
      "balancing_entropy": 1.0897,
      "balancing_loss": 1.2363,
      "epoch": 0.5632,
      "grad_norm": 0.427734375,
      "learning_rate": 0.00066,
      "loss": 6.593,
      "router_z_loss": 0.1751,
      "step": 110
    },
    {
      "balancing_entropy": 1.0886,
      "balancing_loss": 1.2427,
      "epoch": 0.6144,
      "grad_norm": 1.0703125,
      "learning_rate": 0.0007199999999999999,
      "loss": 6.3197,
      "router_z_loss": 0.1719,
      "step": 120
    },
    {
      "balancing_entropy": 1.082,
      "balancing_loss": 1.2294,
      "epoch": 0.6656,
      "grad_norm": 0.99609375,
      "learning_rate": 0.0007800000000000001,
      "loss": 6.0763,
      "router_z_loss": 0.1539,
      "step": 130
    },
    {
      "balancing_entropy": 1.0747,
      "balancing_loss": 1.2314,
      "epoch": 0.7168,
      "grad_norm": 1.3828125,
      "learning_rate": 0.0008400000000000001,
      "loss": 5.8621,
      "router_z_loss": 0.1356,
      "step": 140
    },
    {
      "balancing_entropy": 1.0826,
      "balancing_loss": 1.2038,
      "epoch": 0.768,
      "grad_norm": 0.392578125,
      "learning_rate": 0.0009,
      "loss": 5.6384,
      "router_z_loss": 0.1192,
      "step": 150
    },
    {
      "balancing_entropy": 1.0737,
      "balancing_loss": 1.2131,
      "epoch": 0.8192,
      "grad_norm": 0.66015625,
      "learning_rate": 0.00096,
      "loss": 5.4706,
      "router_z_loss": 0.108,
      "step": 160
    },
    {
      "balancing_entropy": 1.0764,
      "balancing_loss": 1.1915,
      "epoch": 0.8704,
      "grad_norm": 0.86328125,
      "learning_rate": 0.00102,
      "loss": 5.2654,
      "router_z_loss": 0.1157,
      "step": 170
    },
    {
      "balancing_entropy": 1.0707,
      "balancing_loss": 1.1882,
      "epoch": 0.9216,
      "grad_norm": 0.75390625,
      "learning_rate": 0.00108,
      "loss": 5.1254,
      "router_z_loss": 0.1258,
      "step": 180
    },
    {
      "balancing_entropy": 1.0706,
      "balancing_loss": 1.1888,
      "epoch": 0.9728,
      "grad_norm": 1.265625,
      "learning_rate": 0.00114,
      "loss": 4.9843,
      "router_z_loss": 0.1302,
      "step": 190
    },
    {
      "balancing_entropy": 1.032,
      "balancing_loss": 1.1853,
      "epoch": 1.02048,
      "grad_norm": 0.609375,
      "learning_rate": 0.0012000000000000001,
      "loss": 4.8635,
      "router_z_loss": 0.1217,
      "step": 200
    },
    {
      "balancing_entropy": 1.0663,
      "balancing_loss": 1.182,
      "epoch": 1.07168,
      "grad_norm": 1.046875,
      "learning_rate": 0.00126,
      "loss": 4.727,
      "router_z_loss": 0.112,
      "step": 210
    },
    {
      "balancing_entropy": 1.0641,
      "balancing_loss": 1.1717,
      "epoch": 1.12288,
      "grad_norm": 0.89453125,
      "learning_rate": 0.00132,
      "loss": 4.6284,
      "router_z_loss": 0.1126,
      "step": 220
    },
    {
      "balancing_entropy": 1.0675,
      "balancing_loss": 1.1728,
      "epoch": 1.17408,
      "grad_norm": 0.8125,
      "learning_rate": 0.0013800000000000002,
      "loss": 4.5432,
      "router_z_loss": 0.1004,
      "step": 230
    },
    {
      "balancing_entropy": 1.0645,
      "balancing_loss": 1.1699,
      "epoch": 1.22528,
      "grad_norm": 0.73046875,
      "learning_rate": 0.0014399999999999999,
      "loss": 4.4039,
      "router_z_loss": 0.106,
      "step": 240
    },
    {
      "balancing_entropy": 1.0562,
      "balancing_loss": 1.174,
      "epoch": 1.27648,
      "grad_norm": 0.6640625,
      "learning_rate": 0.0015,
      "loss": 4.345,
      "router_z_loss": 0.0916,
      "step": 250
    },
    {
      "balancing_entropy": 1.0651,
      "balancing_loss": 1.1617,
      "epoch": 1.32768,
      "grad_norm": 0.69140625,
      "learning_rate": 0.0015600000000000002,
      "loss": 4.2596,
      "router_z_loss": 0.0946,
      "step": 260
    },
    {
      "balancing_entropy": 1.057,
      "balancing_loss": 1.1628,
      "epoch": 1.37888,
      "grad_norm": 0.5859375,
      "learning_rate": 0.0016200000000000001,
      "loss": 4.1665,
      "router_z_loss": 0.0921,
      "step": 270
    },
    {
      "balancing_entropy": 1.0561,
      "balancing_loss": 1.1645,
      "epoch": 1.43008,
      "grad_norm": 0.77734375,
      "learning_rate": 0.0016800000000000003,
      "loss": 4.1265,
      "router_z_loss": 0.1304,
      "step": 280
    },
    {
      "balancing_entropy": 1.0591,
      "balancing_loss": 1.1756,
      "epoch": 1.48128,
      "grad_norm": 1.296875,
      "learning_rate": 0.00174,
      "loss": 4.0338,
      "router_z_loss": 0.1135,
      "step": 290
    },
    {
      "balancing_entropy": 1.0557,
      "balancing_loss": 1.1678,
      "epoch": 1.53248,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0018,
      "loss": 3.9945,
      "router_z_loss": 0.1018,
      "step": 300
    },
    {
      "balancing_entropy": 1.0535,
      "balancing_loss": 1.1618,
      "epoch": 1.58368,
      "grad_norm": 0.5078125,
      "learning_rate": 0.00186,
      "loss": 3.9413,
      "router_z_loss": 0.1065,
      "step": 310
    },
    {
      "balancing_entropy": 1.0671,
      "balancing_loss": 1.1601,
      "epoch": 1.6348799999999999,
      "grad_norm": 0.765625,
      "learning_rate": 0.00192,
      "loss": 3.8926,
      "router_z_loss": 0.0878,
      "step": 320
    },
    {
      "balancing_entropy": 1.0489,
      "balancing_loss": 1.172,
      "epoch": 1.68608,
      "grad_norm": 0.494140625,
      "learning_rate": 0.00198,
      "loss": 3.8661,
      "router_z_loss": 0.1065,
      "step": 330
    },
    {
      "balancing_entropy": 1.0666,
      "balancing_loss": 1.163,
      "epoch": 1.7372800000000002,
      "grad_norm": 0.59765625,
      "learning_rate": 0.00204,
      "loss": 3.7964,
      "router_z_loss": 0.1075,
      "step": 340
    },
    {
      "balancing_entropy": 1.0468,
      "balancing_loss": 1.1842,
      "epoch": 1.7884799999999998,
      "grad_norm": 0.4296875,
      "learning_rate": 0.0021,
      "loss": 3.7704,
      "router_z_loss": 0.1338,
      "step": 350
    },
    {
      "balancing_entropy": 1.0524,
      "balancing_loss": 1.1797,
      "epoch": 1.83968,
      "grad_norm": 0.51953125,
      "learning_rate": 0.00216,
      "loss": 3.7421,
      "router_z_loss": 0.1235,
      "step": 360
    },
    {
      "balancing_entropy": 1.0557,
      "balancing_loss": 1.1556,
      "epoch": 1.8908800000000001,
      "grad_norm": 0.40234375,
      "learning_rate": 0.00222,
      "loss": 3.7284,
      "router_z_loss": 0.0839,
      "step": 370
    },
    {
      "balancing_entropy": 1.0553,
      "balancing_loss": 1.1576,
      "epoch": 1.94208,
      "grad_norm": 0.69921875,
      "learning_rate": 0.00228,
      "loss": 3.693,
      "router_z_loss": 0.0978,
      "step": 380
    },
    {
      "balancing_entropy": 1.0582,
      "balancing_loss": 1.1686,
      "epoch": 1.99328,
      "grad_norm": 0.375,
      "learning_rate": 0.00234,
      "loss": 3.6605,
      "router_z_loss": 0.099,
      "step": 390
    },
    {
      "balancing_entropy": 1.0168,
      "balancing_loss": 1.1761,
      "epoch": 2.04096,
      "grad_norm": 0.59765625,
      "learning_rate": 0.0024000000000000002,
      "loss": 3.6227,
      "router_z_loss": 0.1624,
      "step": 400
    },
    {
      "balancing_entropy": 1.0592,
      "balancing_loss": 1.1807,
      "epoch": 2.09216,
      "grad_norm": 0.578125,
      "learning_rate": 0.00246,
      "loss": 3.6069,
      "router_z_loss": 0.0839,
      "step": 410
    },
    {
      "balancing_entropy": 1.0602,
      "balancing_loss": 1.1598,
      "epoch": 2.14336,
      "grad_norm": 0.6953125,
      "learning_rate": 0.00252,
      "loss": 3.6083,
      "router_z_loss": 0.0849,
      "step": 420
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.1967,
      "epoch": 2.19456,
      "grad_norm": 0.466796875,
      "learning_rate": 0.00258,
      "loss": 3.56,
      "router_z_loss": 0.1057,
      "step": 430
    },
    {
      "balancing_entropy": 1.063,
      "balancing_loss": 1.1847,
      "epoch": 2.24576,
      "grad_norm": 0.5390625,
      "learning_rate": 0.00264,
      "loss": 3.5387,
      "router_z_loss": 0.0886,
      "step": 440
    },
    {
      "balancing_entropy": 1.0744,
      "balancing_loss": 1.1659,
      "epoch": 2.29696,
      "grad_norm": 0.443359375,
      "learning_rate": 0.0027,
      "loss": 3.52,
      "router_z_loss": 0.1237,
      "step": 450
    },
    {
      "balancing_entropy": 1.0616,
      "balancing_loss": 1.1807,
      "epoch": 2.34816,
      "grad_norm": 0.349609375,
      "learning_rate": 0.0027600000000000003,
      "loss": 3.5329,
      "router_z_loss": 0.0816,
      "step": 460
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.182,
      "epoch": 2.39936,
      "grad_norm": 0.60546875,
      "learning_rate": 0.00282,
      "loss": 3.5086,
      "router_z_loss": 0.0947,
      "step": 470
    },
    {
      "balancing_entropy": 1.0762,
      "balancing_loss": 1.1809,
      "epoch": 2.45056,
      "grad_norm": 0.390625,
      "learning_rate": 0.0028799999999999997,
      "loss": 3.5194,
      "router_z_loss": 0.0582,
      "step": 480
    },
    {
      "balancing_entropy": 1.0702,
      "balancing_loss": 1.2034,
      "epoch": 2.50176,
      "grad_norm": 0.5234375,
      "learning_rate": 0.00294,
      "loss": 3.455,
      "router_z_loss": 0.0567,
      "step": 490
    },
    {
      "balancing_entropy": 1.0682,
      "balancing_loss": 1.1884,
      "epoch": 2.55296,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 3.4843,
      "router_z_loss": 0.1035,
      "step": 500
    },
    {
      "balancing_entropy": 1.0736,
      "balancing_loss": 1.1813,
      "epoch": 2.6041600000000003,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 3.4361,
      "router_z_loss": 0.0677,
      "step": 510
    },
    {
      "balancing_entropy": 1.0731,
      "balancing_loss": 1.1945,
      "epoch": 2.65536,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 3.4581,
      "router_z_loss": 0.0595,
      "step": 520
    },
    {
      "balancing_entropy": 1.0711,
      "balancing_loss": 1.1982,
      "epoch": 2.70656,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 3.4099,
      "router_z_loss": 0.1024,
      "step": 530
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.1914,
      "epoch": 2.75776,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 3.4428,
      "router_z_loss": 0.0894,
      "step": 540
    },
    {
      "balancing_entropy": 1.0679,
      "balancing_loss": 1.1902,
      "epoch": 2.80896,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 3.424,
      "router_z_loss": 0.0935,
      "step": 550
    },
    {
      "balancing_entropy": 1.0654,
      "balancing_loss": 1.2015,
      "epoch": 2.86016,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 3.3786,
      "router_z_loss": 0.0614,
      "step": 560
    },
    {
      "balancing_entropy": 1.0764,
      "balancing_loss": 1.2189,
      "epoch": 2.91136,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 3.3701,
      "router_z_loss": 0.062,
      "step": 570
    },
    {
      "balancing_entropy": 1.0509,
      "balancing_loss": 1.1867,
      "epoch": 2.96256,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 3.3533,
      "router_z_loss": 0.1797,
      "step": 580
    },
    {
      "balancing_entropy": 1.0168,
      "balancing_loss": 1.2074,
      "epoch": 3.01024,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 3.3435,
      "router_z_loss": 0.2381,
      "step": 590
    },
    {
      "balancing_entropy": 1.0741,
      "balancing_loss": 1.2004,
      "epoch": 3.06144,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 3.3283,
      "router_z_loss": 0.1607,
      "step": 600
    },
    {
      "balancing_entropy": 1.0655,
      "balancing_loss": 1.2132,
      "epoch": 3.11264,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 3.3083,
      "router_z_loss": 0.0619,
      "step": 610
    },
    {
      "balancing_entropy": 1.0676,
      "balancing_loss": 1.2093,
      "epoch": 3.16384,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 3.3066,
      "router_z_loss": 0.1001,
      "step": 620
    },
    {
      "balancing_entropy": 1.0766,
      "balancing_loss": 1.1986,
      "epoch": 3.21504,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 3.2646,
      "router_z_loss": 0.1068,
      "step": 630
    },
    {
      "balancing_entropy": 1.0596,
      "balancing_loss": 1.1986,
      "epoch": 3.26624,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 3.2712,
      "router_z_loss": 0.0909,
      "step": 640
    },
    {
      "balancing_entropy": 1.0739,
      "balancing_loss": 1.2105,
      "epoch": 3.31744,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 3.2741,
      "router_z_loss": 0.1265,
      "step": 650
    },
    {
      "balancing_entropy": 1.0708,
      "balancing_loss": 1.2126,
      "epoch": 3.36864,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 3.2479,
      "router_z_loss": 0.0757,
      "step": 660
    },
    {
      "balancing_entropy": 1.0843,
      "balancing_loss": 1.2076,
      "epoch": 3.4198399999999998,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 3.2491,
      "router_z_loss": 0.0672,
      "step": 670
    },
    {
      "balancing_entropy": 1.0762,
      "balancing_loss": 1.202,
      "epoch": 3.47104,
      "grad_norm": 0.224609375,
      "learning_rate": 0.003,
      "loss": 3.2309,
      "router_z_loss": 0.0487,
      "step": 680
    },
    {
      "balancing_entropy": 1.0807,
      "balancing_loss": 1.22,
      "epoch": 3.52224,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 3.233,
      "router_z_loss": 0.0618,
      "step": 690
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.2142,
      "epoch": 3.5734399999999997,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 3.2017,
      "router_z_loss": 0.0687,
      "step": 700
    },
    {
      "balancing_entropy": 1.0816,
      "balancing_loss": 1.2198,
      "epoch": 3.62464,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 3.1872,
      "router_z_loss": 0.0773,
      "step": 710
    },
    {
      "balancing_entropy": 1.0775,
      "balancing_loss": 1.2156,
      "epoch": 3.67584,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 3.2035,
      "router_z_loss": 0.065,
      "step": 720
    },
    {
      "balancing_entropy": 1.0794,
      "balancing_loss": 1.2035,
      "epoch": 3.72704,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 3.189,
      "router_z_loss": 0.0684,
      "step": 730
    },
    {
      "balancing_entropy": 1.0793,
      "balancing_loss": 1.2141,
      "epoch": 3.7782400000000003,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 3.1681,
      "router_z_loss": 0.0752,
      "step": 740
    },
    {
      "balancing_entropy": 1.0715,
      "balancing_loss": 1.2139,
      "epoch": 3.82944,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 3.1804,
      "router_z_loss": 0.0867,
      "step": 750
    },
    {
      "balancing_entropy": 1.0559,
      "balancing_loss": 1.2236,
      "epoch": 3.88064,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 3.1613,
      "router_z_loss": 0.0844,
      "step": 760
    },
    {
      "balancing_entropy": 1.0654,
      "balancing_loss": 1.2075,
      "epoch": 3.9318400000000002,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 3.1371,
      "router_z_loss": 0.1475,
      "step": 770
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.2114,
      "epoch": 3.98304,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 3.1317,
      "router_z_loss": 0.0947,
      "step": 780
    },
    {
      "balancing_entropy": 1.0441,
      "balancing_loss": 1.2069,
      "epoch": 4.03072,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 3.1571,
      "router_z_loss": 0.0373,
      "step": 790
    },
    {
      "balancing_entropy": 1.0807,
      "balancing_loss": 1.2077,
      "epoch": 4.08192,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 3.1193,
      "router_z_loss": 0.0612,
      "step": 800
    },
    {
      "balancing_entropy": 1.0807,
      "balancing_loss": 1.2232,
      "epoch": 4.13312,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 3.1017,
      "router_z_loss": 0.0594,
      "step": 810
    },
    {
      "balancing_entropy": 1.0758,
      "balancing_loss": 1.2202,
      "epoch": 4.18432,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 3.0898,
      "router_z_loss": 0.0464,
      "step": 820
    },
    {
      "balancing_entropy": 1.0275,
      "balancing_loss": 1.2298,
      "epoch": 4.23552,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 3.0905,
      "router_z_loss": 0.1455,
      "step": 830
    },
    {
      "balancing_entropy": 1.0744,
      "balancing_loss": 1.2049,
      "epoch": 4.28672,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 3.0958,
      "router_z_loss": 0.1656,
      "step": 840
    },
    {
      "balancing_entropy": 1.0805,
      "balancing_loss": 1.209,
      "epoch": 4.33792,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 3.0849,
      "router_z_loss": 0.0463,
      "step": 850
    },
    {
      "balancing_entropy": 1.055,
      "balancing_loss": 1.2152,
      "epoch": 4.38912,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 3.0969,
      "router_z_loss": 0.043,
      "step": 860
    },
    {
      "balancing_entropy": 1.0943,
      "balancing_loss": 1.218,
      "epoch": 4.44032,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 3.0615,
      "router_z_loss": 0.0811,
      "step": 870
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.224,
      "epoch": 4.49152,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 3.0521,
      "router_z_loss": 0.0392,
      "step": 880
    },
    {
      "balancing_entropy": 1.078,
      "balancing_loss": 1.2127,
      "epoch": 4.54272,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 3.0552,
      "router_z_loss": 0.0647,
      "step": 890
    },
    {
      "balancing_entropy": 1.083,
      "balancing_loss": 1.2199,
      "epoch": 4.59392,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 3.0741,
      "router_z_loss": 0.0809,
      "step": 900
    },
    {
      "balancing_entropy": 1.0847,
      "balancing_loss": 1.2119,
      "epoch": 4.64512,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 3.0683,
      "router_z_loss": 0.0431,
      "step": 910
    },
    {
      "balancing_entropy": 1.0217,
      "balancing_loss": 1.2271,
      "epoch": 4.69632,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 3.0516,
      "router_z_loss": 0.0573,
      "step": 920
    },
    {
      "balancing_entropy": 1.065,
      "balancing_loss": 1.2214,
      "epoch": 4.74752,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 3.046,
      "router_z_loss": 0.1873,
      "step": 930
    },
    {
      "balancing_entropy": 1.0573,
      "balancing_loss": 1.2146,
      "epoch": 4.79872,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 3.0496,
      "router_z_loss": 0.1174,
      "step": 940
    },
    {
      "balancing_entropy": 1.0707,
      "balancing_loss": 1.2243,
      "epoch": 4.84992,
      "grad_norm": 0.240234375,
      "learning_rate": 0.003,
      "loss": 3.045,
      "router_z_loss": 0.0647,
      "step": 950
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.2329,
      "epoch": 4.90112,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 3.0302,
      "router_z_loss": 0.0379,
      "step": 960
    },
    {
      "balancing_entropy": 1.0589,
      "balancing_loss": 1.2253,
      "epoch": 4.95232,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 3.0354,
      "router_z_loss": 0.0517,
      "step": 970
    },
    {
      "balancing_entropy": 1.0322,
      "balancing_loss": 1.242,
      "epoch": 5.0,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 3.0135,
      "router_z_loss": 0.0908,
      "step": 980
    },
    {
      "balancing_entropy": 1.0759,
      "balancing_loss": 1.2109,
      "epoch": 5.0512,
      "grad_norm": 0.232421875,
      "learning_rate": 0.003,
      "loss": 3.0219,
      "router_z_loss": 0.0746,
      "step": 990
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.2324,
      "epoch": 5.1024,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.9878,
      "router_z_loss": 0.0662,
      "step": 1000
    },
    {
      "balancing_entropy": 1.0547,
      "balancing_loss": 1.2246,
      "epoch": 5.1536,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 3.0087,
      "router_z_loss": 0.0501,
      "step": 1010
    },
    {
      "balancing_entropy": 1.0818,
      "balancing_loss": 1.2128,
      "epoch": 5.2048,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.9748,
      "router_z_loss": 0.0934,
      "step": 1020
    },
    {
      "balancing_entropy": 1.0771,
      "balancing_loss": 1.2181,
      "epoch": 5.256,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.9883,
      "router_z_loss": 0.0593,
      "step": 1030
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.2303,
      "epoch": 5.3072,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.9864,
      "router_z_loss": 0.0405,
      "step": 1040
    },
    {
      "balancing_entropy": 1.0745,
      "balancing_loss": 1.2331,
      "epoch": 5.3584,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.9861,
      "router_z_loss": 0.0379,
      "step": 1050
    },
    {
      "balancing_entropy": 1.0595,
      "balancing_loss": 1.2112,
      "epoch": 5.4096,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.9816,
      "router_z_loss": 0.0616,
      "step": 1060
    },
    {
      "balancing_entropy": 1.0734,
      "balancing_loss": 1.2293,
      "epoch": 5.4608,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.9698,
      "router_z_loss": 0.0707,
      "step": 1070
    },
    {
      "balancing_entropy": 1.0892,
      "balancing_loss": 1.2403,
      "epoch": 5.5120000000000005,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.9686,
      "router_z_loss": 0.0359,
      "step": 1080
    },
    {
      "balancing_entropy": 1.0653,
      "balancing_loss": 1.2301,
      "epoch": 5.5632,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.9606,
      "router_z_loss": 0.048,
      "step": 1090
    },
    {
      "balancing_entropy": 1.0353,
      "balancing_loss": 1.2307,
      "epoch": 5.6144,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.9571,
      "router_z_loss": 0.0775,
      "step": 1100
    },
    {
      "balancing_entropy": 1.0649,
      "balancing_loss": 1.2082,
      "epoch": 5.6655999999999995,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.9631,
      "router_z_loss": 0.0946,
      "step": 1110
    },
    {
      "balancing_entropy": 1.0718,
      "balancing_loss": 1.234,
      "epoch": 5.7168,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.9616,
      "router_z_loss": 0.0608,
      "step": 1120
    },
    {
      "balancing_entropy": 1.065,
      "balancing_loss": 1.2308,
      "epoch": 5.768,
      "grad_norm": 0.474609375,
      "learning_rate": 0.003,
      "loss": 2.9642,
      "router_z_loss": 0.0417,
      "step": 1130
    },
    {
      "balancing_entropy": 1.0602,
      "balancing_loss": 1.2173,
      "epoch": 5.8192,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.9677,
      "router_z_loss": 0.0664,
      "step": 1140
    },
    {
      "balancing_entropy": 1.0847,
      "balancing_loss": 1.2302,
      "epoch": 5.8704,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.9486,
      "router_z_loss": 0.0926,
      "step": 1150
    },
    {
      "balancing_entropy": 1.0813,
      "balancing_loss": 1.2241,
      "epoch": 5.9216,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.9477,
      "router_z_loss": 0.0516,
      "step": 1160
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.2342,
      "epoch": 5.9728,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.9505,
      "router_z_loss": 0.0424,
      "step": 1170
    },
    {
      "balancing_entropy": 1.0092,
      "balancing_loss": 1.2365,
      "epoch": 6.02048,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.9375,
      "router_z_loss": 0.065,
      "step": 1180
    },
    {
      "balancing_entropy": 1.0798,
      "balancing_loss": 1.2277,
      "epoch": 6.07168,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.927,
      "router_z_loss": 0.0901,
      "step": 1190
    },
    {
      "balancing_entropy": 1.0715,
      "balancing_loss": 1.2252,
      "epoch": 6.12288,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.9386,
      "router_z_loss": 0.0553,
      "step": 1200
    },
    {
      "balancing_entropy": 1.0674,
      "balancing_loss": 1.2161,
      "epoch": 6.17408,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.9215,
      "router_z_loss": 0.0608,
      "step": 1210
    },
    {
      "balancing_entropy": 1.0828,
      "balancing_loss": 1.2348,
      "epoch": 6.22528,
      "grad_norm": 0.2294921875,
      "learning_rate": 0.003,
      "loss": 2.8964,
      "router_z_loss": 0.0557,
      "step": 1220
    },
    {
      "balancing_entropy": 1.092,
      "balancing_loss": 1.2257,
      "epoch": 6.27648,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.9038,
      "router_z_loss": 0.0428,
      "step": 1230
    },
    {
      "balancing_entropy": 1.0884,
      "balancing_loss": 1.2347,
      "epoch": 6.32768,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.9132,
      "router_z_loss": 0.0392,
      "step": 1240
    },
    {
      "balancing_entropy": 1.0621,
      "balancing_loss": 1.2372,
      "epoch": 6.37888,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.9006,
      "router_z_loss": 0.0303,
      "step": 1250
    },
    {
      "balancing_entropy": 1.0622,
      "balancing_loss": 1.2266,
      "epoch": 6.43008,
      "grad_norm": 0.23828125,
      "learning_rate": 0.003,
      "loss": 2.9117,
      "router_z_loss": 0.1032,
      "step": 1260
    },
    {
      "balancing_entropy": 1.0454,
      "balancing_loss": 1.2436,
      "epoch": 6.48128,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.899,
      "router_z_loss": 0.0911,
      "step": 1270
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.2234,
      "epoch": 6.53248,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.8986,
      "router_z_loss": 0.0701,
      "step": 1280
    },
    {
      "balancing_entropy": 1.0793,
      "balancing_loss": 1.2227,
      "epoch": 6.58368,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.8993,
      "router_z_loss": 0.0574,
      "step": 1290
    },
    {
      "balancing_entropy": 1.0836,
      "balancing_loss": 1.2371,
      "epoch": 6.63488,
      "grad_norm": 0.431640625,
      "learning_rate": 0.003,
      "loss": 2.9103,
      "router_z_loss": 0.077,
      "step": 1300
    },
    {
      "balancing_entropy": 1.0503,
      "balancing_loss": 1.23,
      "epoch": 6.6860800000000005,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.8977,
      "router_z_loss": 0.0555,
      "step": 1310
    },
    {
      "balancing_entropy": 1.0821,
      "balancing_loss": 1.2409,
      "epoch": 6.73728,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.906,
      "router_z_loss": 0.0993,
      "step": 1320
    },
    {
      "balancing_entropy": 1.0818,
      "balancing_loss": 1.2275,
      "epoch": 6.78848,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.893,
      "router_z_loss": 0.0745,
      "step": 1330
    },
    {
      "balancing_entropy": 1.0786,
      "balancing_loss": 1.2213,
      "epoch": 6.8396799999999995,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.8838,
      "router_z_loss": 0.0351,
      "step": 1340
    },
    {
      "balancing_entropy": 1.0282,
      "balancing_loss": 1.2295,
      "epoch": 6.89088,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.8834,
      "router_z_loss": 0.0636,
      "step": 1350
    },
    {
      "balancing_entropy": 1.0668,
      "balancing_loss": 1.238,
      "epoch": 6.94208,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.8926,
      "router_z_loss": 0.1251,
      "step": 1360
    },
    {
      "balancing_entropy": 1.0445,
      "balancing_loss": 1.2217,
      "epoch": 6.99328,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.8888,
      "router_z_loss": 0.07,
      "step": 1370
    },
    {
      "balancing_entropy": 1.024,
      "balancing_loss": 1.2171,
      "epoch": 7.04096,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.8777,
      "router_z_loss": 0.0632,
      "step": 1380
    },
    {
      "balancing_entropy": 1.0833,
      "balancing_loss": 1.2333,
      "epoch": 7.09216,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.8674,
      "router_z_loss": 0.0706,
      "step": 1390
    },
    {
      "balancing_entropy": 1.0496,
      "balancing_loss": 1.232,
      "epoch": 7.14336,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.863,
      "router_z_loss": 0.0459,
      "step": 1400
    },
    {
      "balancing_entropy": 1.0642,
      "balancing_loss": 1.2162,
      "epoch": 7.19456,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.8603,
      "router_z_loss": 0.0901,
      "step": 1410
    },
    {
      "balancing_entropy": 1.0862,
      "balancing_loss": 1.2391,
      "epoch": 7.24576,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.8476,
      "router_z_loss": 0.0649,
      "step": 1420
    },
    {
      "balancing_entropy": 1.0764,
      "balancing_loss": 1.2311,
      "epoch": 7.29696,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.8419,
      "router_z_loss": 0.0443,
      "step": 1430
    },
    {
      "balancing_entropy": 1.0655,
      "balancing_loss": 1.2478,
      "epoch": 7.34816,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.8571,
      "router_z_loss": 0.0582,
      "step": 1440
    },
    {
      "balancing_entropy": 1.0569,
      "balancing_loss": 1.195,
      "epoch": 7.39936,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.8575,
      "router_z_loss": 0.0791,
      "step": 1450
    },
    {
      "balancing_entropy": 1.0564,
      "balancing_loss": 1.2449,
      "epoch": 7.45056,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.8689,
      "router_z_loss": 0.108,
      "step": 1460
    },
    {
      "balancing_entropy": 1.0504,
      "balancing_loss": 1.2363,
      "epoch": 7.50176,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.8507,
      "router_z_loss": 0.0729,
      "step": 1470
    },
    {
      "balancing_entropy": 1.0665,
      "balancing_loss": 1.2227,
      "epoch": 7.55296,
      "grad_norm": 1.625,
      "learning_rate": 0.003,
      "loss": 2.8677,
      "router_z_loss": 0.0711,
      "step": 1480
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.2089,
      "epoch": 7.60416,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.9223,
      "router_z_loss": 0.0744,
      "step": 1490
    },
    {
      "balancing_entropy": 1.0852,
      "balancing_loss": 1.242,
      "epoch": 7.65536,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.8734,
      "router_z_loss": 0.091,
      "step": 1500
    },
    {
      "balancing_entropy": 1.0692,
      "balancing_loss": 1.2335,
      "epoch": 7.70656,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.8505,
      "router_z_loss": 0.0449,
      "step": 1510
    },
    {
      "balancing_entropy": 1.0404,
      "balancing_loss": 1.2398,
      "epoch": 7.75776,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.8478,
      "router_z_loss": 0.0739,
      "step": 1520
    },
    {
      "balancing_entropy": 1.0825,
      "balancing_loss": 1.219,
      "epoch": 7.80896,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.8468,
      "router_z_loss": 0.0837,
      "step": 1530
    },
    {
      "balancing_entropy": 1.059,
      "balancing_loss": 1.2312,
      "epoch": 7.8601600000000005,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.8304,
      "router_z_loss": 0.0305,
      "step": 1540
    },
    {
      "balancing_entropy": 1.0784,
      "balancing_loss": 1.2255,
      "epoch": 7.91136,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.849,
      "router_z_loss": 0.0534,
      "step": 1550
    },
    {
      "balancing_entropy": 1.0611,
      "balancing_loss": 1.2447,
      "epoch": 7.96256,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.8395,
      "router_z_loss": 0.0493,
      "step": 1560
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.2341,
      "epoch": 8.01024,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.8335,
      "router_z_loss": 0.0539,
      "step": 1570
    },
    {
      "balancing_entropy": 1.0571,
      "balancing_loss": 1.2166,
      "epoch": 8.06144,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.8305,
      "router_z_loss": 0.0559,
      "step": 1580
    },
    {
      "balancing_entropy": 1.0554,
      "balancing_loss": 1.2371,
      "epoch": 8.11264,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.823,
      "router_z_loss": 0.0782,
      "step": 1590
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.2222,
      "epoch": 8.16384,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.8208,
      "router_z_loss": 0.074,
      "step": 1600
    },
    {
      "balancing_entropy": 1.0616,
      "balancing_loss": 1.2397,
      "epoch": 8.21504,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.8146,
      "router_z_loss": 0.0327,
      "step": 1610
    },
    {
      "balancing_entropy": 1.0768,
      "balancing_loss": 1.2232,
      "epoch": 8.26624,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.8123,
      "router_z_loss": 0.0446,
      "step": 1620
    },
    {
      "balancing_entropy": 1.0635,
      "balancing_loss": 1.2431,
      "epoch": 8.31744,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.8155,
      "router_z_loss": 0.0606,
      "step": 1630
    },
    {
      "balancing_entropy": 1.0541,
      "balancing_loss": 1.2328,
      "epoch": 8.36864,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.8236,
      "router_z_loss": 0.0435,
      "step": 1640
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.2242,
      "epoch": 8.41984,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.8151,
      "router_z_loss": 0.0637,
      "step": 1650
    },
    {
      "balancing_entropy": 1.0887,
      "balancing_loss": 1.2272,
      "epoch": 8.47104,
      "grad_norm": 0.23046875,
      "learning_rate": 0.003,
      "loss": 2.8119,
      "router_z_loss": 0.0808,
      "step": 1660
    },
    {
      "balancing_entropy": 1.0831,
      "balancing_loss": 1.219,
      "epoch": 8.52224,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.8067,
      "router_z_loss": 0.0371,
      "step": 1670
    },
    {
      "balancing_entropy": 1.055,
      "balancing_loss": 1.251,
      "epoch": 8.57344,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.8022,
      "router_z_loss": 0.0485,
      "step": 1680
    },
    {
      "balancing_entropy": 1.0799,
      "balancing_loss": 1.2117,
      "epoch": 8.62464,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.8082,
      "router_z_loss": 0.0718,
      "step": 1690
    },
    {
      "balancing_entropy": 1.0585,
      "balancing_loss": 1.2443,
      "epoch": 8.67584,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.8109,
      "router_z_loss": 0.046,
      "step": 1700
    },
    {
      "balancing_entropy": 1.0469,
      "balancing_loss": 1.2398,
      "epoch": 8.72704,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.8035,
      "router_z_loss": 0.0989,
      "step": 1710
    },
    {
      "balancing_entropy": 1.066,
      "balancing_loss": 1.2007,
      "epoch": 8.77824,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.7924,
      "router_z_loss": 0.1063,
      "step": 1720
    },
    {
      "balancing_entropy": 1.0373,
      "balancing_loss": 1.2404,
      "epoch": 8.82944,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.8068,
      "router_z_loss": 0.0707,
      "step": 1730
    },
    {
      "balancing_entropy": 1.0649,
      "balancing_loss": 1.2449,
      "epoch": 8.88064,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.8035,
      "router_z_loss": 0.0815,
      "step": 1740
    },
    {
      "balancing_entropy": 1.0539,
      "balancing_loss": 1.1982,
      "epoch": 8.93184,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.8094,
      "router_z_loss": 0.0732,
      "step": 1750
    },
    {
      "balancing_entropy": 1.0465,
      "balancing_loss": 1.2441,
      "epoch": 8.98304,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.812,
      "router_z_loss": 0.0539,
      "step": 1760
    },
    {
      "balancing_entropy": 1.0079,
      "balancing_loss": 1.215,
      "epoch": 9.03072,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.7914,
      "router_z_loss": 0.0692,
      "step": 1770
    },
    {
      "balancing_entropy": 1.0386,
      "balancing_loss": 1.2393,
      "epoch": 9.08192,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.7865,
      "router_z_loss": 0.0615,
      "step": 1780
    },
    {
      "balancing_entropy": 1.0542,
      "balancing_loss": 1.2145,
      "epoch": 9.13312,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.7724,
      "router_z_loss": 0.0902,
      "step": 1790
    },
    {
      "balancing_entropy": 1.0848,
      "balancing_loss": 1.2326,
      "epoch": 9.18432,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.78,
      "router_z_loss": 0.0728,
      "step": 1800
    },
    {
      "balancing_entropy": 1.0578,
      "balancing_loss": 1.242,
      "epoch": 9.23552,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.7801,
      "router_z_loss": 0.0321,
      "step": 1810
    },
    {
      "balancing_entropy": 1.0597,
      "balancing_loss": 1.2229,
      "epoch": 9.28672,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.7656,
      "router_z_loss": 0.068,
      "step": 1820
    },
    {
      "balancing_entropy": 1.0666,
      "balancing_loss": 1.2386,
      "epoch": 9.33792,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.7779,
      "router_z_loss": 0.0918,
      "step": 1830
    },
    {
      "balancing_entropy": 1.0426,
      "balancing_loss": 1.2359,
      "epoch": 9.38912,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.7728,
      "router_z_loss": 0.0372,
      "step": 1840
    },
    {
      "balancing_entropy": 1.0643,
      "balancing_loss": 1.2131,
      "epoch": 9.44032,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.7758,
      "router_z_loss": 0.0737,
      "step": 1850
    },
    {
      "balancing_entropy": 1.0481,
      "balancing_loss": 1.2294,
      "epoch": 9.49152,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.7756,
      "router_z_loss": 0.088,
      "step": 1860
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.2263,
      "epoch": 9.54272,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.79,
      "router_z_loss": 0.0651,
      "step": 1870
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.2291,
      "epoch": 9.59392,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.7809,
      "router_z_loss": 0.0768,
      "step": 1880
    },
    {
      "balancing_entropy": 1.0552,
      "balancing_loss": 1.2261,
      "epoch": 9.64512,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.7724,
      "router_z_loss": 0.0594,
      "step": 1890
    },
    {
      "balancing_entropy": 1.0305,
      "balancing_loss": 1.2191,
      "epoch": 9.69632,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.7852,
      "router_z_loss": 0.066,
      "step": 1900
    },
    {
      "balancing_entropy": 1.0111,
      "balancing_loss": 1.2561,
      "epoch": 9.74752,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.7744,
      "router_z_loss": 0.1239,
      "step": 1910
    },
    {
      "balancing_entropy": 1.0573,
      "balancing_loss": 1.1821,
      "epoch": 9.79872,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.7776,
      "router_z_loss": 0.1146,
      "step": 1920
    },
    {
      "balancing_entropy": 1.0724,
      "balancing_loss": 1.2459,
      "epoch": 9.849920000000001,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.7639,
      "router_z_loss": 0.0652,
      "step": 1930
    },
    {
      "balancing_entropy": 1.0604,
      "balancing_loss": 1.2401,
      "epoch": 9.90112,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.7655,
      "router_z_loss": 0.0388,
      "step": 1940
    },
    {
      "balancing_entropy": 1.0481,
      "balancing_loss": 1.2626,
      "epoch": 9.95232,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.7775,
      "router_z_loss": 0.0849,
      "step": 1950
    },
    {
      "balancing_entropy": 1.0131,
      "balancing_loss": 1.2257,
      "epoch": 10.0,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.7791,
      "router_z_loss": 0.1003,
      "step": 1960
    },
    {
      "balancing_entropy": 1.0574,
      "balancing_loss": 1.2207,
      "epoch": 10.0512,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.7631,
      "router_z_loss": 0.104,
      "step": 1970
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.2208,
      "epoch": 10.1024,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.7561,
      "router_z_loss": 0.0488,
      "step": 1980
    },
    {
      "balancing_entropy": 1.0694,
      "balancing_loss": 1.2312,
      "epoch": 10.1536,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.7577,
      "router_z_loss": 0.0324,
      "step": 1990
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.2137,
      "epoch": 10.2048,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.7517,
      "router_z_loss": 0.0721,
      "step": 2000
    },
    {
      "balancing_entropy": 1.0542,
      "balancing_loss": 1.2469,
      "epoch": 10.256,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.7416,
      "router_z_loss": 0.1109,
      "step": 2010
    },
    {
      "balancing_entropy": 1.0444,
      "balancing_loss": 1.2409,
      "epoch": 10.3072,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.7557,
      "router_z_loss": 0.0795,
      "step": 2020
    },
    {
      "balancing_entropy": 1.036,
      "balancing_loss": 1.2308,
      "epoch": 10.3584,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.7476,
      "router_z_loss": 0.0638,
      "step": 2030
    },
    {
      "balancing_entropy": 1.0714,
      "balancing_loss": 1.2454,
      "epoch": 10.4096,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.7558,
      "router_z_loss": 0.1008,
      "step": 2040
    },
    {
      "balancing_entropy": 1.0836,
      "balancing_loss": 1.239,
      "epoch": 10.4608,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.743,
      "router_z_loss": 0.0425,
      "step": 2050
    },
    {
      "balancing_entropy": 1.0662,
      "balancing_loss": 1.2395,
      "epoch": 10.512,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.7459,
      "router_z_loss": 0.0447,
      "step": 2060
    },
    {
      "balancing_entropy": 1.0707,
      "balancing_loss": 1.2166,
      "epoch": 10.5632,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.7421,
      "router_z_loss": 0.0599,
      "step": 2070
    },
    {
      "balancing_entropy": 1.0685,
      "balancing_loss": 1.2211,
      "epoch": 10.6144,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.7458,
      "router_z_loss": 0.0463,
      "step": 2080
    },
    {
      "balancing_entropy": 1.0421,
      "balancing_loss": 1.2226,
      "epoch": 10.6656,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.7491,
      "router_z_loss": 0.0593,
      "step": 2090
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.2412,
      "epoch": 10.7168,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.7392,
      "router_z_loss": 0.0473,
      "step": 2100
    },
    {
      "balancing_entropy": 1.0669,
      "balancing_loss": 1.2476,
      "epoch": 10.768,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.7482,
      "router_z_loss": 0.0342,
      "step": 2110
    },
    {
      "balancing_entropy": 1.0659,
      "balancing_loss": 1.1986,
      "epoch": 10.8192,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.7449,
      "router_z_loss": 0.0627,
      "step": 2120
    },
    {
      "balancing_entropy": 1.0496,
      "balancing_loss": 1.2457,
      "epoch": 10.8704,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.7448,
      "router_z_loss": 0.0579,
      "step": 2130
    },
    {
      "balancing_entropy": 1.0687,
      "balancing_loss": 1.228,
      "epoch": 10.9216,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.7384,
      "router_z_loss": 0.0581,
      "step": 2140
    },
    {
      "balancing_entropy": 1.0329,
      "balancing_loss": 1.2189,
      "epoch": 10.9728,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.7497,
      "router_z_loss": 0.0429,
      "step": 2150
    },
    {
      "balancing_entropy": 1.0155,
      "balancing_loss": 1.2403,
      "epoch": 11.02048,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.7391,
      "router_z_loss": 0.1094,
      "step": 2160
    },
    {
      "balancing_entropy": 1.0749,
      "balancing_loss": 1.2232,
      "epoch": 11.07168,
      "grad_norm": 0.2333984375,
      "learning_rate": 0.003,
      "loss": 2.716,
      "router_z_loss": 0.0668,
      "step": 2170
    },
    {
      "balancing_entropy": 1.0601,
      "balancing_loss": 1.2294,
      "epoch": 11.12288,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.7231,
      "router_z_loss": 0.0509,
      "step": 2180
    },
    {
      "balancing_entropy": 1.0539,
      "balancing_loss": 1.2124,
      "epoch": 11.17408,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.7137,
      "router_z_loss": 0.0739,
      "step": 2190
    },
    {
      "balancing_entropy": 1.0388,
      "balancing_loss": 1.2488,
      "epoch": 11.22528,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.7202,
      "router_z_loss": 0.0924,
      "step": 2200
    },
    {
      "balancing_entropy": 1.0759,
      "balancing_loss": 1.2326,
      "epoch": 11.27648,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.7158,
      "router_z_loss": 0.0677,
      "step": 2210
    },
    {
      "balancing_entropy": 1.0362,
      "balancing_loss": 1.206,
      "epoch": 11.32768,
      "grad_norm": 0.244140625,
      "learning_rate": 0.003,
      "loss": 2.7207,
      "router_z_loss": 0.0536,
      "step": 2220
    },
    {
      "balancing_entropy": 1.0395,
      "balancing_loss": 1.2671,
      "epoch": 11.37888,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.7271,
      "router_z_loss": 0.1195,
      "step": 2230
    },
    {
      "balancing_entropy": 1.0587,
      "balancing_loss": 1.2216,
      "epoch": 11.43008,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.7167,
      "router_z_loss": 0.1148,
      "step": 2240
    },
    {
      "balancing_entropy": 1.0726,
      "balancing_loss": 1.2282,
      "epoch": 11.48128,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.7324,
      "router_z_loss": 0.0523,
      "step": 2250
    },
    {
      "balancing_entropy": 1.0611,
      "balancing_loss": 1.2328,
      "epoch": 11.53248,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.7244,
      "router_z_loss": 0.0358,
      "step": 2260
    },
    {
      "balancing_entropy": 1.0618,
      "balancing_loss": 1.2291,
      "epoch": 11.58368,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.7266,
      "router_z_loss": 0.0478,
      "step": 2270
    },
    {
      "balancing_entropy": 1.0737,
      "balancing_loss": 1.2528,
      "epoch": 11.63488,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.717,
      "router_z_loss": 0.0723,
      "step": 2280
    },
    {
      "balancing_entropy": 1.0525,
      "balancing_loss": 1.2256,
      "epoch": 11.68608,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.7295,
      "router_z_loss": 0.0536,
      "step": 2290
    },
    {
      "balancing_entropy": 1.0696,
      "balancing_loss": 1.2533,
      "epoch": 11.73728,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.7155,
      "router_z_loss": 0.0829,
      "step": 2300
    },
    {
      "balancing_entropy": 1.0407,
      "balancing_loss": 1.2109,
      "epoch": 11.78848,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.7236,
      "router_z_loss": 0.0658,
      "step": 2310
    },
    {
      "balancing_entropy": 1.0446,
      "balancing_loss": 1.2483,
      "epoch": 11.83968,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.7242,
      "router_z_loss": 0.0743,
      "step": 2320
    },
    {
      "balancing_entropy": 1.0687,
      "balancing_loss": 1.2108,
      "epoch": 11.89088,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.7115,
      "router_z_loss": 0.0837,
      "step": 2330
    },
    {
      "balancing_entropy": 1.0482,
      "balancing_loss": 1.2005,
      "epoch": 11.94208,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.7198,
      "router_z_loss": 0.0417,
      "step": 2340
    },
    {
      "balancing_entropy": 1.0353,
      "balancing_loss": 1.2577,
      "epoch": 11.99328,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.7178,
      "router_z_loss": 0.0688,
      "step": 2350
    },
    {
      "balancing_entropy": 1.0094,
      "balancing_loss": 1.2189,
      "epoch": 12.04096,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.7163,
      "router_z_loss": 0.1012,
      "step": 2360
    },
    {
      "balancing_entropy": 1.0583,
      "balancing_loss": 1.23,
      "epoch": 12.09216,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.7122,
      "router_z_loss": 0.0436,
      "step": 2370
    },
    {
      "balancing_entropy": 1.0666,
      "balancing_loss": 1.2038,
      "epoch": 12.14336,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.696,
      "router_z_loss": 0.066,
      "step": 2380
    },
    {
      "balancing_entropy": 1.0115,
      "balancing_loss": 1.2575,
      "epoch": 12.19456,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.7051,
      "router_z_loss": 0.0862,
      "step": 2390
    },
    {
      "balancing_entropy": 1.0712,
      "balancing_loss": 1.2018,
      "epoch": 12.24576,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.7042,
      "router_z_loss": 0.1334,
      "step": 2400
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.2252,
      "epoch": 12.29696,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.7001,
      "router_z_loss": 0.0552,
      "step": 2410
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.2509,
      "epoch": 12.34816,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.7015,
      "router_z_loss": 0.0335,
      "step": 2420
    },
    {
      "balancing_entropy": 1.0528,
      "balancing_loss": 1.2065,
      "epoch": 12.39936,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.6971,
      "router_z_loss": 0.0553,
      "step": 2430
    },
    {
      "balancing_entropy": 1.0521,
      "balancing_loss": 1.236,
      "epoch": 12.45056,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.7101,
      "router_z_loss": 0.0825,
      "step": 2440
    },
    {
      "balancing_entropy": 1.0527,
      "balancing_loss": 1.2436,
      "epoch": 12.50176,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.7016,
      "router_z_loss": 0.0948,
      "step": 2450
    },
    {
      "balancing_entropy": 1.0586,
      "balancing_loss": 1.216,
      "epoch": 12.55296,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.7013,
      "router_z_loss": 0.086,
      "step": 2460
    },
    {
      "balancing_entropy": 1.0509,
      "balancing_loss": 1.2593,
      "epoch": 12.60416,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.7032,
      "router_z_loss": 0.0781,
      "step": 2470
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.2371,
      "epoch": 12.65536,
      "grad_norm": 0.44140625,
      "learning_rate": 0.003,
      "loss": 2.6836,
      "router_z_loss": 0.0604,
      "step": 2480
    },
    {
      "balancing_entropy": 1.068,
      "balancing_loss": 1.2369,
      "epoch": 12.70656,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.6921,
      "router_z_loss": 0.0799,
      "step": 2490
    },
    {
      "balancing_entropy": 1.0571,
      "balancing_loss": 1.2169,
      "epoch": 12.75776,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.6892,
      "router_z_loss": 0.0969,
      "step": 2500
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.2267,
      "epoch": 12.80896,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.6989,
      "router_z_loss": 0.0739,
      "step": 2510
    },
    {
      "balancing_entropy": 1.0266,
      "balancing_loss": 1.2222,
      "epoch": 12.86016,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.7004,
      "router_z_loss": 0.0413,
      "step": 2520
    },
    {
      "balancing_entropy": 1.0311,
      "balancing_loss": 1.2324,
      "epoch": 12.91136,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.7119,
      "router_z_loss": 0.1331,
      "step": 2530
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.2269,
      "epoch": 12.96256,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.6923,
      "router_z_loss": 0.0913,
      "step": 2540
    },
    {
      "balancing_entropy": 1.0131,
      "balancing_loss": 1.2392,
      "epoch": 13.01024,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.7,
      "router_z_loss": 0.0612,
      "step": 2550
    },
    {
      "balancing_entropy": 1.0808,
      "balancing_loss": 1.2068,
      "epoch": 13.06144,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.6785,
      "router_z_loss": 0.0893,
      "step": 2560
    },
    {
      "balancing_entropy": 1.0672,
      "balancing_loss": 1.2667,
      "epoch": 13.11264,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.6777,
      "router_z_loss": 0.0463,
      "step": 2570
    },
    {
      "balancing_entropy": 1.072,
      "balancing_loss": 1.2245,
      "epoch": 13.16384,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.6783,
      "router_z_loss": 0.0475,
      "step": 2580
    },
    {
      "balancing_entropy": 1.0629,
      "balancing_loss": 1.2236,
      "epoch": 13.21504,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.6749,
      "router_z_loss": 0.0481,
      "step": 2590
    },
    {
      "balancing_entropy": 1.0382,
      "balancing_loss": 1.268,
      "epoch": 13.26624,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.6749,
      "router_z_loss": 0.076,
      "step": 2600
    },
    {
      "balancing_entropy": 1.0367,
      "balancing_loss": 1.1943,
      "epoch": 13.31744,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.6825,
      "router_z_loss": 0.1174,
      "step": 2610
    },
    {
      "balancing_entropy": 1.059,
      "balancing_loss": 1.2373,
      "epoch": 13.36864,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.6829,
      "router_z_loss": 0.085,
      "step": 2620
    },
    {
      "balancing_entropy": 1.0563,
      "balancing_loss": 1.1916,
      "epoch": 13.41984,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.6842,
      "router_z_loss": 0.0742,
      "step": 2630
    },
    {
      "balancing_entropy": 1.0532,
      "balancing_loss": 1.2616,
      "epoch": 13.47104,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.6948,
      "router_z_loss": 0.0908,
      "step": 2640
    },
    {
      "balancing_entropy": 1.062,
      "balancing_loss": 1.2134,
      "epoch": 13.52224,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.6854,
      "router_z_loss": 0.0785,
      "step": 2650
    },
    {
      "balancing_entropy": 1.049,
      "balancing_loss": 1.2272,
      "epoch": 13.57344,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.6904,
      "router_z_loss": 0.0709,
      "step": 2660
    },
    {
      "balancing_entropy": 1.078,
      "balancing_loss": 1.251,
      "epoch": 13.62464,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.6751,
      "router_z_loss": 0.0643,
      "step": 2670
    },
    {
      "balancing_entropy": 1.0688,
      "balancing_loss": 1.2217,
      "epoch": 13.67584,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.6849,
      "router_z_loss": 0.0539,
      "step": 2680
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.2278,
      "epoch": 13.72704,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.6893,
      "router_z_loss": 0.0548,
      "step": 2690
    },
    {
      "balancing_entropy": 1.052,
      "balancing_loss": 1.2468,
      "epoch": 13.77824,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.6825,
      "router_z_loss": 0.0989,
      "step": 2700
    },
    {
      "balancing_entropy": 1.0825,
      "balancing_loss": 1.2403,
      "epoch": 13.82944,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.6738,
      "router_z_loss": 0.0504,
      "step": 2710
    },
    {
      "balancing_entropy": 1.0716,
      "balancing_loss": 1.213,
      "epoch": 13.88064,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.6773,
      "router_z_loss": 0.0514,
      "step": 2720
    },
    {
      "balancing_entropy": 1.0617,
      "balancing_loss": 1.2474,
      "epoch": 13.93184,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.6757,
      "router_z_loss": 0.1044,
      "step": 2730
    },
    {
      "balancing_entropy": 1.0627,
      "balancing_loss": 1.2321,
      "epoch": 13.98304,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.7021,
      "router_z_loss": 0.066,
      "step": 2740
    },
    {
      "balancing_entropy": 0.9989,
      "balancing_loss": 1.2216,
      "epoch": 14.03072,
      "grad_norm": 0.2333984375,
      "learning_rate": 0.003,
      "loss": 2.7125,
      "router_z_loss": 0.1309,
      "step": 2750
    },
    {
      "balancing_entropy": 1.0749,
      "balancing_loss": 1.2188,
      "epoch": 14.08192,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.6776,
      "router_z_loss": 0.2207,
      "step": 2760
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.2447,
      "epoch": 14.13312,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.6541,
      "router_z_loss": 0.0833,
      "step": 2770
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.2378,
      "epoch": 14.18432,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.6547,
      "router_z_loss": 0.0405,
      "step": 2780
    },
    {
      "balancing_entropy": 1.0818,
      "balancing_loss": 1.2319,
      "epoch": 14.23552,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.6615,
      "router_z_loss": 0.0651,
      "step": 2790
    },
    {
      "balancing_entropy": 1.0517,
      "balancing_loss": 1.2434,
      "epoch": 14.28672,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.6479,
      "router_z_loss": 0.0703,
      "step": 2800
    },
    {
      "balancing_entropy": 1.0477,
      "balancing_loss": 1.2201,
      "epoch": 14.33792,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.6769,
      "router_z_loss": 0.0973,
      "step": 2810
    },
    {
      "balancing_entropy": 1.074,
      "balancing_loss": 1.2213,
      "epoch": 14.38912,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.6687,
      "router_z_loss": 0.128,
      "step": 2820
    },
    {
      "balancing_entropy": 1.0726,
      "balancing_loss": 1.2318,
      "epoch": 14.44032,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.6773,
      "router_z_loss": 0.0659,
      "step": 2830
    },
    {
      "balancing_entropy": 1.0663,
      "balancing_loss": 1.2133,
      "epoch": 14.49152,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.6722,
      "router_z_loss": 0.0555,
      "step": 2840
    },
    {
      "balancing_entropy": 1.0674,
      "balancing_loss": 1.2356,
      "epoch": 14.54272,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.6645,
      "router_z_loss": 0.0701,
      "step": 2850
    },
    {
      "balancing_entropy": 1.0785,
      "balancing_loss": 1.2071,
      "epoch": 14.59392,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.675,
      "router_z_loss": 0.0558,
      "step": 2860
    },
    {
      "balancing_entropy": 1.0399,
      "balancing_loss": 1.2211,
      "epoch": 14.64512,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.6749,
      "router_z_loss": 0.0546,
      "step": 2870
    },
    {
      "balancing_entropy": 1.0149,
      "balancing_loss": 1.246,
      "epoch": 14.69632,
      "grad_norm": 0.486328125,
      "learning_rate": 0.003,
      "loss": 2.6651,
      "router_z_loss": 0.1092,
      "step": 2880
    },
    {
      "balancing_entropy": 1.0484,
      "balancing_loss": 1.2129,
      "epoch": 14.74752,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.6724,
      "router_z_loss": 0.1245,
      "step": 2890
    },
    {
      "balancing_entropy": 1.0359,
      "balancing_loss": 1.2455,
      "epoch": 14.79872,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.6654,
      "router_z_loss": 0.0736,
      "step": 2900
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.1983,
      "epoch": 14.849920000000001,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.6621,
      "router_z_loss": 0.1315,
      "step": 2910
    },
    {
      "balancing_entropy": 1.0437,
      "balancing_loss": 1.2026,
      "epoch": 14.90112,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.6647,
      "router_z_loss": 0.0926,
      "step": 2920
    },
    {
      "balancing_entropy": 1.0606,
      "balancing_loss": 1.254,
      "epoch": 14.95232,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.6741,
      "router_z_loss": 0.1247,
      "step": 2930
    },
    {
      "balancing_entropy": 1.0135,
      "balancing_loss": 1.1961,
      "epoch": 15.0,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.6668,
      "router_z_loss": 0.0922,
      "step": 2940
    },
    {
      "balancing_entropy": 1.0636,
      "balancing_loss": 1.2484,
      "epoch": 15.0512,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.6574,
      "router_z_loss": 0.1037,
      "step": 2950
    },
    {
      "balancing_entropy": 1.0585,
      "balancing_loss": 1.1787,
      "epoch": 15.1024,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.6528,
      "router_z_loss": 0.0746,
      "step": 2960
    },
    {
      "balancing_entropy": 1.023,
      "balancing_loss": 1.287,
      "epoch": 15.1536,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.6544,
      "router_z_loss": 0.0859,
      "step": 2970
    },
    {
      "balancing_entropy": 1.0601,
      "balancing_loss": 1.184,
      "epoch": 15.2048,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.6384,
      "router_z_loss": 0.1885,
      "step": 2980
    },
    {
      "balancing_entropy": 1.0693,
      "balancing_loss": 1.2328,
      "epoch": 15.256,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.648,
      "router_z_loss": 0.125,
      "step": 2990
    },
    {
      "balancing_entropy": 1.0761,
      "balancing_loss": 1.2246,
      "epoch": 15.3072,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.6409,
      "router_z_loss": 0.0301,
      "step": 3000
    },
    {
      "balancing_entropy": 1.0697,
      "balancing_loss": 1.2194,
      "epoch": 15.3584,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.6464,
      "router_z_loss": 0.0297,
      "step": 3010
    },
    {
      "balancing_entropy": 1.0736,
      "balancing_loss": 1.2462,
      "epoch": 15.4096,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.6438,
      "router_z_loss": 0.0692,
      "step": 3020
    },
    {
      "balancing_entropy": 1.0506,
      "balancing_loss": 1.2558,
      "epoch": 15.4608,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.6439,
      "router_z_loss": 0.055,
      "step": 3030
    },
    {
      "balancing_entropy": 0.9902,
      "balancing_loss": 1.2133,
      "epoch": 15.512,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.653,
      "router_z_loss": 0.1002,
      "step": 3040
    },
    {
      "balancing_entropy": 1.0617,
      "balancing_loss": 1.2295,
      "epoch": 15.5632,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.6559,
      "router_z_loss": 0.2241,
      "step": 3050
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.2201,
      "epoch": 15.6144,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.6533,
      "router_z_loss": 0.1893,
      "step": 3060
    },
    {
      "balancing_entropy": 1.0692,
      "balancing_loss": 1.2278,
      "epoch": 15.6656,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.6463,
      "router_z_loss": 0.0893,
      "step": 3070
    },
    {
      "balancing_entropy": 1.0815,
      "balancing_loss": 1.2479,
      "epoch": 15.7168,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.6547,
      "router_z_loss": 0.0528,
      "step": 3080
    },
    {
      "balancing_entropy": 1.0718,
      "balancing_loss": 1.2181,
      "epoch": 15.768,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.6337,
      "router_z_loss": 0.0347,
      "step": 3090
    },
    {
      "balancing_entropy": 1.0651,
      "balancing_loss": 1.2335,
      "epoch": 15.8192,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.6429,
      "router_z_loss": 0.0669,
      "step": 3100
    },
    {
      "balancing_entropy": 1.0586,
      "balancing_loss": 1.2455,
      "epoch": 15.8704,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.6488,
      "router_z_loss": 0.0492,
      "step": 3110
    },
    {
      "balancing_entropy": 1.0474,
      "balancing_loss": 1.2137,
      "epoch": 15.9216,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.6494,
      "router_z_loss": 0.0954,
      "step": 3120
    },
    {
      "balancing_entropy": 1.0309,
      "balancing_loss": 1.2511,
      "epoch": 15.9728,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.6463,
      "router_z_loss": 0.1141,
      "step": 3130
    },
    {
      "balancing_entropy": 1.0344,
      "balancing_loss": 1.22,
      "epoch": 16.02048,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.6349,
      "router_z_loss": 0.084,
      "step": 3140
    },
    {
      "balancing_entropy": 1.0784,
      "balancing_loss": 1.2464,
      "epoch": 16.07168,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.6244,
      "router_z_loss": 0.042,
      "step": 3150
    },
    {
      "balancing_entropy": 1.0592,
      "balancing_loss": 1.2256,
      "epoch": 16.12288,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.6238,
      "router_z_loss": 0.0551,
      "step": 3160
    },
    {
      "balancing_entropy": 1.0619,
      "balancing_loss": 1.2303,
      "epoch": 16.17408,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.6279,
      "router_z_loss": 0.0589,
      "step": 3170
    },
    {
      "balancing_entropy": 1.0812,
      "balancing_loss": 1.2261,
      "epoch": 16.22528,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.6318,
      "router_z_loss": 0.0519,
      "step": 3180
    },
    {
      "balancing_entropy": 1.03,
      "balancing_loss": 1.2656,
      "epoch": 16.27648,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.6256,
      "router_z_loss": 0.0491,
      "step": 3190
    },
    {
      "balancing_entropy": 1.0756,
      "balancing_loss": 1.1689,
      "epoch": 16.32768,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.6419,
      "router_z_loss": 0.1281,
      "step": 3200
    },
    {
      "balancing_entropy": 1.0642,
      "balancing_loss": 1.248,
      "epoch": 16.37888,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.6317,
      "router_z_loss": 0.0601,
      "step": 3210
    },
    {
      "balancing_entropy": 1.0612,
      "balancing_loss": 1.2456,
      "epoch": 16.43008,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.6357,
      "router_z_loss": 0.0578,
      "step": 3220
    },
    {
      "balancing_entropy": 1.048,
      "balancing_loss": 1.2213,
      "epoch": 16.48128,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.6195,
      "router_z_loss": 0.1048,
      "step": 3230
    },
    {
      "balancing_entropy": 1.0713,
      "balancing_loss": 1.2395,
      "epoch": 16.53248,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.633,
      "router_z_loss": 0.0774,
      "step": 3240
    },
    {
      "balancing_entropy": 1.0132,
      "balancing_loss": 1.2313,
      "epoch": 16.58368,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.6357,
      "router_z_loss": 0.0725,
      "step": 3250
    },
    {
      "balancing_entropy": 1.0526,
      "balancing_loss": 1.2236,
      "epoch": 16.63488,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.6336,
      "router_z_loss": 0.11,
      "step": 3260
    },
    {
      "balancing_entropy": 1.05,
      "balancing_loss": 1.2443,
      "epoch": 16.68608,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.6375,
      "router_z_loss": 0.0993,
      "step": 3270
    },
    {
      "balancing_entropy": 1.0332,
      "balancing_loss": 1.2352,
      "epoch": 16.73728,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.64,
      "router_z_loss": 0.0874,
      "step": 3280
    },
    {
      "balancing_entropy": 1.0526,
      "balancing_loss": 1.2042,
      "epoch": 16.78848,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.6323,
      "router_z_loss": 0.1389,
      "step": 3290
    },
    {
      "balancing_entropy": 1.0476,
      "balancing_loss": 1.2635,
      "epoch": 16.83968,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.629,
      "router_z_loss": 0.119,
      "step": 3300
    },
    {
      "balancing_entropy": 1.074,
      "balancing_loss": 1.206,
      "epoch": 16.89088,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.6372,
      "router_z_loss": 0.0777,
      "step": 3310
    },
    {
      "balancing_entropy": 1.0726,
      "balancing_loss": 1.2518,
      "epoch": 16.94208,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.6373,
      "router_z_loss": 0.0598,
      "step": 3320
    },
    {
      "balancing_entropy": 1.0552,
      "balancing_loss": 1.2272,
      "epoch": 16.99328,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.6297,
      "router_z_loss": 0.0648,
      "step": 3330
    },
    {
      "balancing_entropy": 1.0285,
      "balancing_loss": 1.2017,
      "epoch": 17.04096,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.6167,
      "router_z_loss": 0.1014,
      "step": 3340
    },
    {
      "balancing_entropy": 1.0302,
      "balancing_loss": 1.2256,
      "epoch": 17.09216,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.6173,
      "router_z_loss": 0.0664,
      "step": 3350
    },
    {
      "balancing_entropy": 1.0595,
      "balancing_loss": 1.2019,
      "epoch": 17.14336,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.6108,
      "router_z_loss": 0.0734,
      "step": 3360
    },
    {
      "balancing_entropy": 1.0613,
      "balancing_loss": 1.24,
      "epoch": 17.19456,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.6185,
      "router_z_loss": 0.0613,
      "step": 3370
    },
    {
      "balancing_entropy": 1.0741,
      "balancing_loss": 1.2129,
      "epoch": 17.24576,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.6257,
      "router_z_loss": 0.0523,
      "step": 3380
    },
    {
      "balancing_entropy": 1.0585,
      "balancing_loss": 1.2102,
      "epoch": 17.29696,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.6085,
      "router_z_loss": 0.0655,
      "step": 3390
    },
    {
      "balancing_entropy": 1.0502,
      "balancing_loss": 1.2319,
      "epoch": 17.34816,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.6266,
      "router_z_loss": 0.0792,
      "step": 3400
    },
    {
      "balancing_entropy": 1.0543,
      "balancing_loss": 1.2191,
      "epoch": 17.39936,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.6198,
      "router_z_loss": 0.117,
      "step": 3410
    },
    {
      "balancing_entropy": 1.0449,
      "balancing_loss": 1.2672,
      "epoch": 17.45056,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.6009,
      "router_z_loss": 0.0492,
      "step": 3420
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.2471,
      "epoch": 17.50176,
      "grad_norm": 0.671875,
      "learning_rate": 0.003,
      "loss": 2.6142,
      "router_z_loss": 0.0794,
      "step": 3430
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.1779,
      "epoch": 17.55296,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.6246,
      "router_z_loss": 0.0901,
      "step": 3440
    },
    {
      "balancing_entropy": 0.9994,
      "balancing_loss": 1.2712,
      "epoch": 17.60416,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.6271,
      "router_z_loss": 0.1172,
      "step": 3450
    },
    {
      "balancing_entropy": 1.0577,
      "balancing_loss": 1.2347,
      "epoch": 17.65536,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.6178,
      "router_z_loss": 0.1044,
      "step": 3460
    },
    {
      "balancing_entropy": 1.0734,
      "balancing_loss": 1.1762,
      "epoch": 17.70656,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.6184,
      "router_z_loss": 0.0991,
      "step": 3470
    },
    {
      "balancing_entropy": 1.037,
      "balancing_loss": 1.2391,
      "epoch": 17.75776,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.623,
      "router_z_loss": 0.0835,
      "step": 3480
    },
    {
      "balancing_entropy": 1.0335,
      "balancing_loss": 1.2778,
      "epoch": 17.80896,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.6103,
      "router_z_loss": 0.151,
      "step": 3490
    },
    {
      "balancing_entropy": 1.053,
      "balancing_loss": 1.2151,
      "epoch": 17.86016,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.6082,
      "router_z_loss": 0.1173,
      "step": 3500
    },
    {
      "balancing_entropy": 1.0487,
      "balancing_loss": 1.2356,
      "epoch": 17.91136,
      "grad_norm": 0.25,
      "learning_rate": 0.003,
      "loss": 2.6138,
      "router_z_loss": 0.0628,
      "step": 3510
    },
    {
      "balancing_entropy": 1.0593,
      "balancing_loss": 1.2222,
      "epoch": 17.96256,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.6243,
      "router_z_loss": 0.0918,
      "step": 3520
    },
    {
      "balancing_entropy": 0.972,
      "balancing_loss": 1.2415,
      "epoch": 18.01024,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.6236,
      "router_z_loss": 0.0986,
      "step": 3530
    },
    {
      "balancing_entropy": 1.0628,
      "balancing_loss": 1.2086,
      "epoch": 18.06144,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.6176,
      "router_z_loss": 0.1218,
      "step": 3540
    },
    {
      "balancing_entropy": 1.0475,
      "balancing_loss": 1.1874,
      "epoch": 18.11264,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.6076,
      "router_z_loss": 0.1032,
      "step": 3550
    },
    {
      "balancing_entropy": 1.0632,
      "balancing_loss": 1.208,
      "epoch": 18.16384,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.5964,
      "router_z_loss": 0.0611,
      "step": 3560
    },
    {
      "balancing_entropy": 1.0609,
      "balancing_loss": 1.2662,
      "epoch": 18.21504,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5995,
      "router_z_loss": 0.0846,
      "step": 3570
    },
    {
      "balancing_entropy": 1.0635,
      "balancing_loss": 1.2138,
      "epoch": 18.26624,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.597,
      "router_z_loss": 0.0675,
      "step": 3580
    },
    {
      "balancing_entropy": 1.035,
      "balancing_loss": 1.2456,
      "epoch": 18.31744,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.6067,
      "router_z_loss": 0.0492,
      "step": 3590
    },
    {
      "balancing_entropy": 1.0629,
      "balancing_loss": 1.2134,
      "epoch": 18.36864,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.601,
      "router_z_loss": 0.1015,
      "step": 3600
    },
    {
      "balancing_entropy": 1.0529,
      "balancing_loss": 1.2017,
      "epoch": 18.41984,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.6104,
      "router_z_loss": 0.0622,
      "step": 3610
    },
    {
      "balancing_entropy": 1.0731,
      "balancing_loss": 1.2207,
      "epoch": 18.47104,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.6089,
      "router_z_loss": 0.0376,
      "step": 3620
    },
    {
      "balancing_entropy": 0.9913,
      "balancing_loss": 1.2519,
      "epoch": 18.52224,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.618,
      "router_z_loss": 0.0574,
      "step": 3630
    },
    {
      "balancing_entropy": 1.0552,
      "balancing_loss": 1.1954,
      "epoch": 18.57344,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.6049,
      "router_z_loss": 0.1246,
      "step": 3640
    },
    {
      "balancing_entropy": 1.0356,
      "balancing_loss": 1.2293,
      "epoch": 18.62464,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.6119,
      "router_z_loss": 0.1262,
      "step": 3650
    },
    {
      "balancing_entropy": 1.0446,
      "balancing_loss": 1.2323,
      "epoch": 18.67584,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.6214,
      "router_z_loss": 0.089,
      "step": 3660
    },
    {
      "balancing_entropy": 1.0721,
      "balancing_loss": 1.2235,
      "epoch": 18.72704,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.6061,
      "router_z_loss": 0.0732,
      "step": 3670
    },
    {
      "balancing_entropy": 1.056,
      "balancing_loss": 1.2372,
      "epoch": 18.77824,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.6073,
      "router_z_loss": 0.0773,
      "step": 3680
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.1987,
      "epoch": 18.829439999999998,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.5972,
      "router_z_loss": 0.0647,
      "step": 3690
    },
    {
      "balancing_entropy": 1.0538,
      "balancing_loss": 1.2308,
      "epoch": 18.88064,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.611,
      "router_z_loss": 0.052,
      "step": 3700
    },
    {
      "balancing_entropy": 1.0485,
      "balancing_loss": 1.198,
      "epoch": 18.93184,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.6124,
      "router_z_loss": 0.0682,
      "step": 3710
    },
    {
      "balancing_entropy": 1.0524,
      "balancing_loss": 1.2405,
      "epoch": 18.98304,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.6072,
      "router_z_loss": 0.0547,
      "step": 3720
    },
    {
      "balancing_entropy": 1.0318,
      "balancing_loss": 1.2147,
      "epoch": 19.03072,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.6039,
      "router_z_loss": 0.0663,
      "step": 3730
    },
    {
      "balancing_entropy": 1.0413,
      "balancing_loss": 1.2264,
      "epoch": 19.08192,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.5849,
      "router_z_loss": 0.0597,
      "step": 3740
    },
    {
      "balancing_entropy": 1.0505,
      "balancing_loss": 1.2051,
      "epoch": 19.13312,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.606,
      "router_z_loss": 0.0903,
      "step": 3750
    },
    {
      "balancing_entropy": 1.0437,
      "balancing_loss": 1.211,
      "epoch": 19.18432,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5952,
      "router_z_loss": 0.0976,
      "step": 3760
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.2385,
      "epoch": 19.23552,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.5954,
      "router_z_loss": 0.0483,
      "step": 3770
    },
    {
      "balancing_entropy": 1.0322,
      "balancing_loss": 1.243,
      "epoch": 19.28672,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.5871,
      "router_z_loss": 0.0616,
      "step": 3780
    },
    {
      "balancing_entropy": 1.0563,
      "balancing_loss": 1.1785,
      "epoch": 19.33792,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.6019,
      "router_z_loss": 0.1292,
      "step": 3790
    },
    {
      "balancing_entropy": 1.0543,
      "balancing_loss": 1.2353,
      "epoch": 19.38912,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.5982,
      "router_z_loss": 0.1065,
      "step": 3800
    },
    {
      "balancing_entropy": 1.068,
      "balancing_loss": 1.2045,
      "epoch": 19.44032,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.5763,
      "router_z_loss": 0.0808,
      "step": 3810
    },
    {
      "balancing_entropy": 1.0217,
      "balancing_loss": 1.2199,
      "epoch": 19.49152,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5948,
      "router_z_loss": 0.0832,
      "step": 3820
    },
    {
      "balancing_entropy": 1.0313,
      "balancing_loss": 1.2063,
      "epoch": 19.54272,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.6099,
      "router_z_loss": 0.0773,
      "step": 3830
    },
    {
      "balancing_entropy": 1.0608,
      "balancing_loss": 1.2102,
      "epoch": 19.59392,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.5961,
      "router_z_loss": 0.1106,
      "step": 3840
    },
    {
      "balancing_entropy": 1.0516,
      "balancing_loss": 1.236,
      "epoch": 19.64512,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.589,
      "router_z_loss": 0.0557,
      "step": 3850
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.2475,
      "epoch": 19.69632,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.5937,
      "router_z_loss": 0.0543,
      "step": 3860
    },
    {
      "balancing_entropy": 1.0246,
      "balancing_loss": 1.2167,
      "epoch": 19.74752,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.5989,
      "router_z_loss": 0.1096,
      "step": 3870
    },
    {
      "balancing_entropy": 1.039,
      "balancing_loss": 1.2235,
      "epoch": 19.79872,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.6077,
      "router_z_loss": 0.1068,
      "step": 3880
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.2569,
      "epoch": 19.84992,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5947,
      "router_z_loss": 0.1335,
      "step": 3890
    },
    {
      "balancing_entropy": 1.0417,
      "balancing_loss": 1.1976,
      "epoch": 19.90112,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.6103,
      "router_z_loss": 0.1001,
      "step": 3900
    },
    {
      "balancing_entropy": 1.0586,
      "balancing_loss": 1.1817,
      "epoch": 19.95232,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.5978,
      "router_z_loss": 0.1062,
      "step": 3910
    },
    {
      "balancing_entropy": 0.9851,
      "balancing_loss": 1.2478,
      "epoch": 20.0,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.609,
      "router_z_loss": 0.0946,
      "step": 3920
    },
    {
      "balancing_entropy": 1.0511,
      "balancing_loss": 1.2145,
      "epoch": 20.0512,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.5776,
      "router_z_loss": 0.0911,
      "step": 3930
    },
    {
      "balancing_entropy": 1.0321,
      "balancing_loss": 1.2391,
      "epoch": 20.1024,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.5848,
      "router_z_loss": 0.0762,
      "step": 3940
    },
    {
      "balancing_entropy": 1.0162,
      "balancing_loss": 1.241,
      "epoch": 20.1536,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.5685,
      "router_z_loss": 0.1766,
      "step": 3950
    },
    {
      "balancing_entropy": 0.9881,
      "balancing_loss": 1.2115,
      "epoch": 20.2048,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5882,
      "router_z_loss": 0.1489,
      "step": 3960
    },
    {
      "balancing_entropy": 1.0477,
      "balancing_loss": 1.2013,
      "epoch": 20.256,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.5847,
      "router_z_loss": 0.1039,
      "step": 3970
    },
    {
      "balancing_entropy": 1.0411,
      "balancing_loss": 1.2153,
      "epoch": 20.3072,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.6084,
      "router_z_loss": 0.1064,
      "step": 3980
    },
    {
      "balancing_entropy": 1.0121,
      "balancing_loss": 1.1809,
      "epoch": 20.3584,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.6035,
      "router_z_loss": 0.1876,
      "step": 3990
    },
    {
      "balancing_entropy": 1.0302,
      "balancing_loss": 1.2155,
      "epoch": 20.4096,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.593,
      "router_z_loss": 0.233,
      "step": 4000
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.1642,
      "epoch": 20.4608,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.5842,
      "router_z_loss": 0.1734,
      "step": 4010
    },
    {
      "balancing_entropy": 1.0326,
      "balancing_loss": 1.2143,
      "epoch": 20.512,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.5979,
      "router_z_loss": 0.1506,
      "step": 4020
    },
    {
      "balancing_entropy": 1.0014,
      "balancing_loss": 1.2416,
      "epoch": 20.5632,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.5917,
      "router_z_loss": 0.175,
      "step": 4030
    },
    {
      "balancing_entropy": 1.0355,
      "balancing_loss": 1.2161,
      "epoch": 20.6144,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5967,
      "router_z_loss": 0.1571,
      "step": 4040
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.204,
      "epoch": 20.6656,
      "grad_norm": 0.69921875,
      "learning_rate": 0.003,
      "loss": 2.6037,
      "router_z_loss": 0.149,
      "step": 4050
    },
    {
      "balancing_entropy": 1.0497,
      "balancing_loss": 1.1924,
      "epoch": 20.7168,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.6131,
      "router_z_loss": 0.1503,
      "step": 4060
    },
    {
      "balancing_entropy": 1.0277,
      "balancing_loss": 1.2246,
      "epoch": 20.768,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.5937,
      "router_z_loss": 0.2485,
      "step": 4070
    },
    {
      "balancing_entropy": 1.0302,
      "balancing_loss": 1.2019,
      "epoch": 20.8192,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.582,
      "router_z_loss": 0.1301,
      "step": 4080
    },
    {
      "balancing_entropy": 0.9962,
      "balancing_loss": 1.2726,
      "epoch": 20.8704,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.5953,
      "router_z_loss": 0.11,
      "step": 4090
    },
    {
      "balancing_entropy": 1.0631,
      "balancing_loss": 1.1613,
      "epoch": 20.9216,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.5892,
      "router_z_loss": 0.1478,
      "step": 4100
    },
    {
      "balancing_entropy": 0.9029,
      "balancing_loss": 1.2495,
      "epoch": 20.9728,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.5966,
      "router_z_loss": 0.1894,
      "step": 4110
    },
    {
      "balancing_entropy": 1.0413,
      "balancing_loss": 1.1697,
      "epoch": 21.02048,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.586,
      "router_z_loss": 0.1477,
      "step": 4120
    },
    {
      "balancing_entropy": 1.0243,
      "balancing_loss": 1.2141,
      "epoch": 21.07168,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.5969,
      "router_z_loss": 0.0895,
      "step": 4130
    },
    {
      "balancing_entropy": 1.0394,
      "balancing_loss": 1.1541,
      "epoch": 21.12288,
      "grad_norm": 0.2314453125,
      "learning_rate": 0.003,
      "loss": 2.5699,
      "router_z_loss": 0.2672,
      "step": 4140
    },
    {
      "balancing_entropy": 0.9787,
      "balancing_loss": 1.2677,
      "epoch": 21.17408,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5752,
      "router_z_loss": 0.2141,
      "step": 4150
    },
    {
      "balancing_entropy": 0.9865,
      "balancing_loss": 1.189,
      "epoch": 21.22528,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.5666,
      "router_z_loss": 0.1704,
      "step": 4160
    },
    {
      "balancing_entropy": 1.0147,
      "balancing_loss": 1.219,
      "epoch": 21.27648,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.5961,
      "router_z_loss": 0.2057,
      "step": 4170
    },
    {
      "balancing_entropy": 1.0319,
      "balancing_loss": 1.2013,
      "epoch": 21.32768,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.5795,
      "router_z_loss": 0.1485,
      "step": 4180
    },
    {
      "balancing_entropy": 1.0258,
      "balancing_loss": 1.1955,
      "epoch": 21.37888,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5724,
      "router_z_loss": 0.0823,
      "step": 4190
    },
    {
      "balancing_entropy": 1.0296,
      "balancing_loss": 1.2161,
      "epoch": 21.43008,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.5757,
      "router_z_loss": 0.1304,
      "step": 4200
    },
    {
      "balancing_entropy": 1.0485,
      "balancing_loss": 1.2117,
      "epoch": 21.48128,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.5793,
      "router_z_loss": 0.1061,
      "step": 4210
    },
    {
      "balancing_entropy": 0.9992,
      "balancing_loss": 1.224,
      "epoch": 21.53248,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.5733,
      "router_z_loss": 0.094,
      "step": 4220
    },
    {
      "balancing_entropy": 1.0205,
      "balancing_loss": 1.2159,
      "epoch": 21.58368,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5704,
      "router_z_loss": 0.1503,
      "step": 4230
    },
    {
      "balancing_entropy": 0.9124,
      "balancing_loss": 1.2608,
      "epoch": 21.63488,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.5802,
      "router_z_loss": 0.1369,
      "step": 4240
    },
    {
      "balancing_entropy": 1.0437,
      "balancing_loss": 1.2152,
      "epoch": 21.68608,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.5733,
      "router_z_loss": 0.2373,
      "step": 4250
    },
    {
      "balancing_entropy": 0.9847,
      "balancing_loss": 1.2403,
      "epoch": 21.73728,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.5852,
      "router_z_loss": 0.1088,
      "step": 4260
    },
    {
      "balancing_entropy": 1.0365,
      "balancing_loss": 1.2083,
      "epoch": 21.78848,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.578,
      "router_z_loss": 0.0906,
      "step": 4270
    },
    {
      "balancing_entropy": 1.021,
      "balancing_loss": 1.2238,
      "epoch": 21.83968,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.5626,
      "router_z_loss": 0.0954,
      "step": 4280
    },
    {
      "balancing_entropy": 1.0081,
      "balancing_loss": 1.2224,
      "epoch": 21.89088,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5815,
      "router_z_loss": 0.1029,
      "step": 4290
    },
    {
      "balancing_entropy": 0.9593,
      "balancing_loss": 1.2776,
      "epoch": 21.94208,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.5757,
      "router_z_loss": 0.1627,
      "step": 4300
    },
    {
      "balancing_entropy": 0.9669,
      "balancing_loss": 1.2634,
      "epoch": 21.99328,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.5824,
      "router_z_loss": 0.1565,
      "step": 4310
    },
    {
      "balancing_entropy": 0.9849,
      "balancing_loss": 1.2016,
      "epoch": 22.04096,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.5611,
      "router_z_loss": 0.1846,
      "step": 4320
    },
    {
      "balancing_entropy": 0.9565,
      "balancing_loss": 1.2578,
      "epoch": 22.09216,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.5807,
      "router_z_loss": 0.1227,
      "step": 4330
    },
    {
      "balancing_entropy": 0.9791,
      "balancing_loss": 1.2715,
      "epoch": 22.14336,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.5684,
      "router_z_loss": 0.1828,
      "step": 4340
    },
    {
      "balancing_entropy": 1.0657,
      "balancing_loss": 1.1666,
      "epoch": 22.19456,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5695,
      "router_z_loss": 0.1933,
      "step": 4350
    },
    {
      "balancing_entropy": 1.0365,
      "balancing_loss": 1.1995,
      "epoch": 22.24576,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.5713,
      "router_z_loss": 0.1079,
      "step": 4360
    },
    {
      "balancing_entropy": 1.0026,
      "balancing_loss": 1.2372,
      "epoch": 22.29696,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5842,
      "router_z_loss": 0.1116,
      "step": 4370
    },
    {
      "balancing_entropy": 1.0694,
      "balancing_loss": 1.1688,
      "epoch": 22.34816,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.5734,
      "router_z_loss": 0.2374,
      "step": 4380
    },
    {
      "balancing_entropy": 1.0337,
      "balancing_loss": 1.2216,
      "epoch": 22.39936,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.5577,
      "router_z_loss": 0.1848,
      "step": 4390
    },
    {
      "balancing_entropy": 1.0471,
      "balancing_loss": 1.1988,
      "epoch": 22.45056,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.5754,
      "router_z_loss": 0.0668,
      "step": 4400
    },
    {
      "balancing_entropy": 1.03,
      "balancing_loss": 1.204,
      "epoch": 22.50176,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.5685,
      "router_z_loss": 0.1062,
      "step": 4410
    },
    {
      "balancing_entropy": 0.9922,
      "balancing_loss": 1.2452,
      "epoch": 22.55296,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.5697,
      "router_z_loss": 0.1469,
      "step": 4420
    },
    {
      "balancing_entropy": 1.067,
      "balancing_loss": 1.1878,
      "epoch": 22.60416,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.5647,
      "router_z_loss": 0.1135,
      "step": 4430
    },
    {
      "balancing_entropy": 1.0009,
      "balancing_loss": 1.2566,
      "epoch": 22.65536,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.5683,
      "router_z_loss": 0.0935,
      "step": 4440
    },
    {
      "balancing_entropy": 1.0504,
      "balancing_loss": 1.2117,
      "epoch": 22.70656,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.5655,
      "router_z_loss": 0.1545,
      "step": 4450
    },
    {
      "balancing_entropy": 1.0152,
      "balancing_loss": 1.2118,
      "epoch": 22.75776,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5708,
      "router_z_loss": 0.0916,
      "step": 4460
    },
    {
      "balancing_entropy": 1.0105,
      "balancing_loss": 1.2271,
      "epoch": 22.80896,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.5644,
      "router_z_loss": 0.0823,
      "step": 4470
    },
    {
      "balancing_entropy": 1.0223,
      "balancing_loss": 1.2223,
      "epoch": 22.86016,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.5757,
      "router_z_loss": 0.1129,
      "step": 4480
    },
    {
      "balancing_entropy": 1.07,
      "balancing_loss": 1.1863,
      "epoch": 22.91136,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.5746,
      "router_z_loss": 0.1746,
      "step": 4490
    },
    {
      "balancing_entropy": 1.0387,
      "balancing_loss": 1.2284,
      "epoch": 22.96256,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.5654,
      "router_z_loss": 0.1176,
      "step": 4500
    },
    {
      "balancing_entropy": 0.9634,
      "balancing_loss": 1.2658,
      "epoch": 23.01024,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.5697,
      "router_z_loss": 0.0439,
      "step": 4510
    },
    {
      "balancing_entropy": 1.0749,
      "balancing_loss": 1.1525,
      "epoch": 23.06144,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.5565,
      "router_z_loss": 0.0961,
      "step": 4520
    },
    {
      "balancing_entropy": 0.9974,
      "balancing_loss": 1.1937,
      "epoch": 23.11264,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.5598,
      "router_z_loss": 0.1565,
      "step": 4530
    },
    {
      "balancing_entropy": 1.0454,
      "balancing_loss": 1.1769,
      "epoch": 23.16384,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.5591,
      "router_z_loss": 0.2049,
      "step": 4540
    },
    {
      "balancing_entropy": 0.9825,
      "balancing_loss": 1.2432,
      "epoch": 23.21504,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.5619,
      "router_z_loss": 0.1429,
      "step": 4550
    },
    {
      "balancing_entropy": 1.0405,
      "balancing_loss": 1.2134,
      "epoch": 23.26624,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.5689,
      "router_z_loss": 0.1092,
      "step": 4560
    },
    {
      "balancing_entropy": 1.0233,
      "balancing_loss": 1.2377,
      "epoch": 23.31744,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.564,
      "router_z_loss": 0.1182,
      "step": 4570
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.1934,
      "epoch": 23.36864,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.5637,
      "router_z_loss": 0.2159,
      "step": 4580
    },
    {
      "balancing_entropy": 1.0647,
      "balancing_loss": 1.1726,
      "epoch": 23.41984,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.5603,
      "router_z_loss": 0.1366,
      "step": 4590
    },
    {
      "balancing_entropy": 1.0384,
      "balancing_loss": 1.2281,
      "epoch": 23.47104,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.559,
      "router_z_loss": 0.1482,
      "step": 4600
    },
    {
      "balancing_entropy": 1.0067,
      "balancing_loss": 1.2545,
      "epoch": 23.52224,
      "grad_norm": 0.87890625,
      "learning_rate": 0.003,
      "loss": 2.5641,
      "router_z_loss": 0.08,
      "step": 4610
    },
    {
      "balancing_entropy": 1.0588,
      "balancing_loss": 1.1638,
      "epoch": 23.57344,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.5723,
      "router_z_loss": 0.1553,
      "step": 4620
    },
    {
      "balancing_entropy": 1.0587,
      "balancing_loss": 1.1726,
      "epoch": 23.62464,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5686,
      "router_z_loss": 0.2212,
      "step": 4630
    },
    {
      "balancing_entropy": 0.9734,
      "balancing_loss": 1.2809,
      "epoch": 23.67584,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.565,
      "router_z_loss": 0.1255,
      "step": 4640
    },
    {
      "balancing_entropy": 1.031,
      "balancing_loss": 1.1974,
      "epoch": 23.72704,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.5521,
      "router_z_loss": 0.1004,
      "step": 4650
    },
    {
      "balancing_entropy": 1.0443,
      "balancing_loss": 1.2089,
      "epoch": 23.77824,
      "grad_norm": 0.7265625,
      "learning_rate": 0.003,
      "loss": 2.5693,
      "router_z_loss": 0.1078,
      "step": 4660
    },
    {
      "balancing_entropy": 1.0391,
      "balancing_loss": 1.213,
      "epoch": 23.829439999999998,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.5669,
      "router_z_loss": 0.094,
      "step": 4670
    },
    {
      "balancing_entropy": 1.0558,
      "balancing_loss": 1.1919,
      "epoch": 23.88064,
      "grad_norm": 0.2177734375,
      "learning_rate": 0.003,
      "loss": 2.561,
      "router_z_loss": 0.178,
      "step": 4680
    },
    {
      "balancing_entropy": 1.0035,
      "balancing_loss": 1.2766,
      "epoch": 23.93184,
      "grad_norm": 1.1875,
      "learning_rate": 0.003,
      "loss": 2.5648,
      "router_z_loss": 0.0989,
      "step": 4690
    },
    {
      "balancing_entropy": 1.0735,
      "balancing_loss": 1.1629,
      "epoch": 23.98304,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.5669,
      "router_z_loss": 0.1112,
      "step": 4700
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.1575,
      "epoch": 24.03072,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.5597,
      "router_z_loss": 0.2458,
      "step": 4710
    },
    {
      "balancing_entropy": 1.0597,
      "balancing_loss": 1.17,
      "epoch": 24.08192,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.5464,
      "router_z_loss": 0.1917,
      "step": 4720
    },
    {
      "balancing_entropy": 1.0232,
      "balancing_loss": 1.2342,
      "epoch": 24.13312,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.5496,
      "router_z_loss": 0.0842,
      "step": 4730
    },
    {
      "balancing_entropy": 1.0355,
      "balancing_loss": 1.2115,
      "epoch": 24.18432,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.545,
      "router_z_loss": 0.0748,
      "step": 4740
    },
    {
      "balancing_entropy": 0.9651,
      "balancing_loss": 1.2515,
      "epoch": 24.23552,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.5483,
      "router_z_loss": 0.0943,
      "step": 4750
    },
    {
      "balancing_entropy": 0.9469,
      "balancing_loss": 1.2727,
      "epoch": 24.28672,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.5351,
      "router_z_loss": 0.093,
      "step": 4760
    },
    {
      "balancing_entropy": 1.0749,
      "balancing_loss": 1.1643,
      "epoch": 24.33792,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.5483,
      "router_z_loss": 0.162,
      "step": 4770
    },
    {
      "balancing_entropy": 0.9798,
      "balancing_loss": 1.2572,
      "epoch": 24.38912,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5512,
      "router_z_loss": 0.1103,
      "step": 4780
    },
    {
      "balancing_entropy": 0.9973,
      "balancing_loss": 1.2503,
      "epoch": 24.44032,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.5592,
      "router_z_loss": 0.1056,
      "step": 4790
    },
    {
      "balancing_entropy": 0.9426,
      "balancing_loss": 1.2111,
      "epoch": 24.49152,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.5576,
      "router_z_loss": 0.1515,
      "step": 4800
    },
    {
      "balancing_entropy": 0.951,
      "balancing_loss": 1.2731,
      "epoch": 24.54272,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.5551,
      "router_z_loss": 0.2713,
      "step": 4810
    },
    {
      "balancing_entropy": 1.028,
      "balancing_loss": 1.2113,
      "epoch": 24.59392,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.5501,
      "router_z_loss": 0.1149,
      "step": 4820
    },
    {
      "balancing_entropy": 0.9668,
      "balancing_loss": 1.2331,
      "epoch": 24.64512,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.5586,
      "router_z_loss": 0.1221,
      "step": 4830
    },
    {
      "balancing_entropy": 0.9577,
      "balancing_loss": 1.2087,
      "epoch": 24.69632,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.5529,
      "router_z_loss": 0.1688,
      "step": 4840
    },
    {
      "balancing_entropy": 1.0736,
      "balancing_loss": 1.1689,
      "epoch": 24.74752,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.5496,
      "router_z_loss": 0.0956,
      "step": 4850
    },
    {
      "balancing_entropy": 1.0232,
      "balancing_loss": 1.231,
      "epoch": 24.79872,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.5565,
      "router_z_loss": 0.0762,
      "step": 4860
    },
    {
      "balancing_entropy": 1.0415,
      "balancing_loss": 1.2166,
      "epoch": 24.84992,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5725,
      "router_z_loss": 0.0951,
      "step": 4870
    },
    {
      "balancing_entropy": 0.9951,
      "balancing_loss": 1.1921,
      "epoch": 24.90112,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.5648,
      "router_z_loss": 0.1121,
      "step": 4880
    },
    {
      "balancing_entropy": 1.0294,
      "balancing_loss": 1.2059,
      "epoch": 24.95232,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.5466,
      "router_z_loss": 0.1814,
      "step": 4890
    },
    {
      "balancing_entropy": 0.9944,
      "balancing_loss": 1.2241,
      "epoch": 25.0,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.5578,
      "router_z_loss": 0.1968,
      "step": 4900
    },
    {
      "balancing_entropy": 1.018,
      "balancing_loss": 1.2418,
      "epoch": 25.0512,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.5281,
      "router_z_loss": 0.119,
      "step": 4910
    },
    {
      "balancing_entropy": 1.0682,
      "balancing_loss": 1.1768,
      "epoch": 25.1024,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.5376,
      "router_z_loss": 0.1125,
      "step": 4920
    },
    {
      "balancing_entropy": 0.9571,
      "balancing_loss": 1.2473,
      "epoch": 25.1536,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5364,
      "router_z_loss": 0.1315,
      "step": 4930
    },
    {
      "balancing_entropy": 1.0327,
      "balancing_loss": 1.2349,
      "epoch": 25.2048,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.5361,
      "router_z_loss": 0.1471,
      "step": 4940
    },
    {
      "balancing_entropy": 1.0205,
      "balancing_loss": 1.2187,
      "epoch": 25.256,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.5434,
      "router_z_loss": 0.0483,
      "step": 4950
    },
    {
      "balancing_entropy": 1.0748,
      "balancing_loss": 1.1742,
      "epoch": 25.3072,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.5379,
      "router_z_loss": 0.1134,
      "step": 4960
    },
    {
      "balancing_entropy": 1.001,
      "balancing_loss": 1.2347,
      "epoch": 25.3584,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.5448,
      "router_z_loss": 0.1269,
      "step": 4970
    },
    {
      "balancing_entropy": 1.0027,
      "balancing_loss": 1.2848,
      "epoch": 25.4096,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.5479,
      "router_z_loss": 0.097,
      "step": 4980
    },
    {
      "balancing_entropy": 1.0575,
      "balancing_loss": 1.1532,
      "epoch": 25.4608,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5527,
      "router_z_loss": 0.1361,
      "step": 4990
    },
    {
      "balancing_entropy": 1.0169,
      "balancing_loss": 1.2198,
      "epoch": 25.512,
      "grad_norm": 0.78515625,
      "learning_rate": 0.003,
      "loss": 2.5528,
      "router_z_loss": 0.1604,
      "step": 5000
    },
    {
      "balancing_entropy": 1.077,
      "balancing_loss": 1.166,
      "epoch": 25.5632,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.559,
      "router_z_loss": 0.1975,
      "step": 5010
    },
    {
      "balancing_entropy": 1.026,
      "balancing_loss": 1.2613,
      "epoch": 25.6144,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.5459,
      "router_z_loss": 0.2406,
      "step": 5020
    },
    {
      "balancing_entropy": 1.0066,
      "balancing_loss": 1.2081,
      "epoch": 25.6656,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.5438,
      "router_z_loss": 0.1515,
      "step": 5030
    },
    {
      "balancing_entropy": 1.0784,
      "balancing_loss": 1.1743,
      "epoch": 25.7168,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.552,
      "router_z_loss": 0.2445,
      "step": 5040
    },
    {
      "balancing_entropy": 1.0156,
      "balancing_loss": 1.2143,
      "epoch": 25.768,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.5294,
      "router_z_loss": 0.1364,
      "step": 5050
    },
    {
      "balancing_entropy": 0.9466,
      "balancing_loss": 1.2184,
      "epoch": 25.8192,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.5557,
      "router_z_loss": 0.0882,
      "step": 5060
    },
    {
      "balancing_entropy": 0.9734,
      "balancing_loss": 1.2635,
      "epoch": 25.8704,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5484,
      "router_z_loss": 0.1278,
      "step": 5070
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.1738,
      "epoch": 25.9216,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.5386,
      "router_z_loss": 0.1626,
      "step": 5080
    },
    {
      "balancing_entropy": 1.0121,
      "balancing_loss": 1.2311,
      "epoch": 25.9728,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.5609,
      "router_z_loss": 0.0788,
      "step": 5090
    },
    {
      "balancing_entropy": 0.9331,
      "balancing_loss": 1.2591,
      "epoch": 26.02048,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.5431,
      "router_z_loss": 0.0648,
      "step": 5100
    },
    {
      "balancing_entropy": 1.0459,
      "balancing_loss": 1.1916,
      "epoch": 26.07168,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.5215,
      "router_z_loss": 0.1181,
      "step": 5110
    },
    {
      "balancing_entropy": 0.995,
      "balancing_loss": 1.2344,
      "epoch": 26.12288,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.5354,
      "router_z_loss": 0.0963,
      "step": 5120
    },
    {
      "balancing_entropy": 1.0637,
      "balancing_loss": 1.1954,
      "epoch": 26.17408,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.5332,
      "router_z_loss": 0.1335,
      "step": 5130
    },
    {
      "balancing_entropy": 0.9631,
      "balancing_loss": 1.2504,
      "epoch": 26.22528,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5421,
      "router_z_loss": 0.1095,
      "step": 5140
    },
    {
      "balancing_entropy": 1.0261,
      "balancing_loss": 1.2265,
      "epoch": 26.27648,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.5343,
      "router_z_loss": 0.1197,
      "step": 5150
    },
    {
      "balancing_entropy": 0.9865,
      "balancing_loss": 1.239,
      "epoch": 26.32768,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5333,
      "router_z_loss": 0.1729,
      "step": 5160
    },
    {
      "balancing_entropy": 0.9952,
      "balancing_loss": 1.2345,
      "epoch": 26.37888,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.5293,
      "router_z_loss": 0.1426,
      "step": 5170
    },
    {
      "balancing_entropy": 1.0021,
      "balancing_loss": 1.2266,
      "epoch": 26.43008,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.5328,
      "router_z_loss": 0.0724,
      "step": 5180
    },
    {
      "balancing_entropy": 0.9967,
      "balancing_loss": 1.2134,
      "epoch": 26.48128,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.5285,
      "router_z_loss": 0.1026,
      "step": 5190
    },
    {
      "balancing_entropy": 0.9476,
      "balancing_loss": 1.2601,
      "epoch": 26.53248,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.5424,
      "router_z_loss": 0.1076,
      "step": 5200
    },
    {
      "balancing_entropy": 1.0725,
      "balancing_loss": 1.1762,
      "epoch": 26.58368,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.5322,
      "router_z_loss": 0.1946,
      "step": 5210
    },
    {
      "balancing_entropy": 1.0144,
      "balancing_loss": 1.2319,
      "epoch": 26.63488,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.5498,
      "router_z_loss": 0.0611,
      "step": 5220
    },
    {
      "balancing_entropy": 1.0014,
      "balancing_loss": 1.2226,
      "epoch": 26.68608,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.5337,
      "router_z_loss": 0.1444,
      "step": 5230
    },
    {
      "balancing_entropy": 1.0202,
      "balancing_loss": 1.1865,
      "epoch": 26.73728,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.534,
      "router_z_loss": 0.1765,
      "step": 5240
    },
    {
      "balancing_entropy": 1.0135,
      "balancing_loss": 1.2199,
      "epoch": 26.78848,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.541,
      "router_z_loss": 0.1048,
      "step": 5250
    },
    {
      "balancing_entropy": 0.9916,
      "balancing_loss": 1.2186,
      "epoch": 26.83968,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.5338,
      "router_z_loss": 0.0834,
      "step": 5260
    },
    {
      "balancing_entropy": 0.9726,
      "balancing_loss": 1.2386,
      "epoch": 26.89088,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.5322,
      "router_z_loss": 0.0916,
      "step": 5270
    },
    {
      "balancing_entropy": 1.0301,
      "balancing_loss": 1.2096,
      "epoch": 26.94208,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.5451,
      "router_z_loss": 0.1529,
      "step": 5280
    },
    {
      "balancing_entropy": 1.0223,
      "balancing_loss": 1.2259,
      "epoch": 26.99328,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.5348,
      "router_z_loss": 0.0762,
      "step": 5290
    },
    {
      "balancing_entropy": 0.9783,
      "balancing_loss": 1.2035,
      "epoch": 27.04096,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.5334,
      "router_z_loss": 0.0721,
      "step": 5300
    },
    {
      "balancing_entropy": 1.0125,
      "balancing_loss": 1.2254,
      "epoch": 27.09216,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5237,
      "router_z_loss": 0.1215,
      "step": 5310
    },
    {
      "balancing_entropy": 1.0165,
      "balancing_loss": 1.2461,
      "epoch": 27.14336,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.5263,
      "router_z_loss": 0.0659,
      "step": 5320
    },
    {
      "balancing_entropy": 0.977,
      "balancing_loss": 1.2352,
      "epoch": 27.19456,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.5267,
      "router_z_loss": 0.1226,
      "step": 5330
    },
    {
      "balancing_entropy": 1.043,
      "balancing_loss": 1.1684,
      "epoch": 27.24576,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.5239,
      "router_z_loss": 0.1906,
      "step": 5340
    },
    {
      "balancing_entropy": 0.9432,
      "balancing_loss": 1.2189,
      "epoch": 27.29696,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.5222,
      "router_z_loss": 0.142,
      "step": 5350
    },
    {
      "balancing_entropy": 1.036,
      "balancing_loss": 1.2245,
      "epoch": 27.34816,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5221,
      "router_z_loss": 0.2204,
      "step": 5360
    },
    {
      "balancing_entropy": 0.994,
      "balancing_loss": 1.263,
      "epoch": 27.39936,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.5269,
      "router_z_loss": 0.0634,
      "step": 5370
    },
    {
      "balancing_entropy": 1.065,
      "balancing_loss": 1.183,
      "epoch": 27.45056,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.5231,
      "router_z_loss": 0.1618,
      "step": 5380
    },
    {
      "balancing_entropy": 0.8511,
      "balancing_loss": 1.3194,
      "epoch": 27.50176,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.5391,
      "router_z_loss": 0.1151,
      "step": 5390
    },
    {
      "balancing_entropy": 1.0776,
      "balancing_loss": 1.1854,
      "epoch": 27.55296,
      "grad_norm": 0.2177734375,
      "learning_rate": 0.003,
      "loss": 2.5224,
      "router_z_loss": 0.2078,
      "step": 5400
    },
    {
      "balancing_entropy": 0.999,
      "balancing_loss": 1.2237,
      "epoch": 27.60416,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.5275,
      "router_z_loss": 0.1647,
      "step": 5410
    },
    {
      "balancing_entropy": 1.0371,
      "balancing_loss": 1.2064,
      "epoch": 27.65536,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.5306,
      "router_z_loss": 0.0626,
      "step": 5420
    },
    {
      "balancing_entropy": 0.9999,
      "balancing_loss": 1.2207,
      "epoch": 27.70656,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.5354,
      "router_z_loss": 0.1045,
      "step": 5430
    },
    {
      "balancing_entropy": 1.0467,
      "balancing_loss": 1.2342,
      "epoch": 27.75776,
      "grad_norm": 0.82421875,
      "learning_rate": 0.003,
      "loss": 2.5314,
      "router_z_loss": 0.1551,
      "step": 5440
    },
    {
      "balancing_entropy": 1.0744,
      "balancing_loss": 1.1705,
      "epoch": 27.80896,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.5422,
      "router_z_loss": 0.0712,
      "step": 5450
    },
    {
      "balancing_entropy": 0.8913,
      "balancing_loss": 1.3732,
      "epoch": 27.86016,
      "grad_norm": 1.3515625,
      "learning_rate": 0.003,
      "loss": 2.5584,
      "router_z_loss": 0.159,
      "step": 5460
    },
    {
      "balancing_entropy": 0.9926,
      "balancing_loss": 1.2549,
      "epoch": 27.91136,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.5447,
      "router_z_loss": 0.1178,
      "step": 5470
    },
    {
      "balancing_entropy": 1.0365,
      "balancing_loss": 1.1944,
      "epoch": 27.96256,
      "grad_norm": 0.2421875,
      "learning_rate": 0.003,
      "loss": 2.5317,
      "router_z_loss": 0.0939,
      "step": 5480
    },
    {
      "balancing_entropy": 1.0435,
      "balancing_loss": 1.1721,
      "epoch": 28.01024,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.5288,
      "router_z_loss": 0.2073,
      "step": 5490
    },
    {
      "balancing_entropy": 1.0165,
      "balancing_loss": 1.266,
      "epoch": 28.06144,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.5117,
      "router_z_loss": 0.1172,
      "step": 5500
    },
    {
      "balancing_entropy": 1.0075,
      "balancing_loss": 1.1955,
      "epoch": 28.11264,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.5114,
      "router_z_loss": 0.0611,
      "step": 5510
    },
    {
      "balancing_entropy": 1.0124,
      "balancing_loss": 1.2298,
      "epoch": 28.16384,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.5115,
      "router_z_loss": 0.0564,
      "step": 5520
    },
    {
      "balancing_entropy": 0.9704,
      "balancing_loss": 1.2399,
      "epoch": 28.21504,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.5057,
      "router_z_loss": 0.0907,
      "step": 5530
    },
    {
      "balancing_entropy": 1.032,
      "balancing_loss": 1.2102,
      "epoch": 28.26624,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.5168,
      "router_z_loss": 0.1353,
      "step": 5540
    },
    {
      "balancing_entropy": 0.9804,
      "balancing_loss": 1.2108,
      "epoch": 28.31744,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.5292,
      "router_z_loss": 0.136,
      "step": 5550
    },
    {
      "balancing_entropy": 1.0349,
      "balancing_loss": 1.2387,
      "epoch": 28.36864,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.5271,
      "router_z_loss": 0.1635,
      "step": 5560
    },
    {
      "balancing_entropy": 1.0296,
      "balancing_loss": 1.2159,
      "epoch": 28.41984,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.5169,
      "router_z_loss": 0.0628,
      "step": 5570
    },
    {
      "balancing_entropy": 1.0151,
      "balancing_loss": 1.2386,
      "epoch": 28.47104,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.5246,
      "router_z_loss": 0.1078,
      "step": 5580
    },
    {
      "balancing_entropy": 1.0023,
      "balancing_loss": 1.2005,
      "epoch": 28.52224,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.5351,
      "router_z_loss": 0.113,
      "step": 5590
    },
    {
      "balancing_entropy": 0.9952,
      "balancing_loss": 1.221,
      "epoch": 28.57344,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.5204,
      "router_z_loss": 0.1573,
      "step": 5600
    },
    {
      "balancing_entropy": 1.0225,
      "balancing_loss": 1.2265,
      "epoch": 28.62464,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.5192,
      "router_z_loss": 0.1211,
      "step": 5610
    },
    {
      "balancing_entropy": 1.0604,
      "balancing_loss": 1.1755,
      "epoch": 28.67584,
      "grad_norm": 0.2275390625,
      "learning_rate": 0.003,
      "loss": 2.5255,
      "router_z_loss": 0.0963,
      "step": 5620
    },
    {
      "balancing_entropy": 1.0143,
      "balancing_loss": 1.2054,
      "epoch": 28.72704,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.521,
      "router_z_loss": 0.1062,
      "step": 5630
    },
    {
      "balancing_entropy": 0.9743,
      "balancing_loss": 1.232,
      "epoch": 28.77824,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.5271,
      "router_z_loss": 0.1061,
      "step": 5640
    },
    {
      "balancing_entropy": 0.9759,
      "balancing_loss": 1.2412,
      "epoch": 28.829439999999998,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.5372,
      "router_z_loss": 0.1559,
      "step": 5650
    },
    {
      "balancing_entropy": 1.066,
      "balancing_loss": 1.1819,
      "epoch": 28.88064,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.5265,
      "router_z_loss": 0.195,
      "step": 5660
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.2245,
      "epoch": 28.93184,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.5288,
      "router_z_loss": 0.0921,
      "step": 5670
    },
    {
      "balancing_entropy": 1.003,
      "balancing_loss": 1.2316,
      "epoch": 28.98304,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.5282,
      "router_z_loss": 0.0836,
      "step": 5680
    },
    {
      "balancing_entropy": 1.0068,
      "balancing_loss": 1.206,
      "epoch": 29.03072,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5093,
      "router_z_loss": 0.1084,
      "step": 5690
    },
    {
      "balancing_entropy": 0.9326,
      "balancing_loss": 1.2446,
      "epoch": 29.08192,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.5079,
      "router_z_loss": 0.1168,
      "step": 5700
    },
    {
      "balancing_entropy": 1.0441,
      "balancing_loss": 1.2118,
      "epoch": 29.13312,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.5003,
      "router_z_loss": 0.1684,
      "step": 5710
    },
    {
      "balancing_entropy": 1.0026,
      "balancing_loss": 1.2227,
      "epoch": 29.18432,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5122,
      "router_z_loss": 0.0558,
      "step": 5720
    },
    {
      "balancing_entropy": 1.0164,
      "balancing_loss": 1.2084,
      "epoch": 29.23552,
      "grad_norm": 0.2451171875,
      "learning_rate": 0.003,
      "loss": 2.5079,
      "router_z_loss": 0.1086,
      "step": 5730
    },
    {
      "balancing_entropy": 1.0379,
      "balancing_loss": 1.2161,
      "epoch": 29.28672,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.507,
      "router_z_loss": 0.0998,
      "step": 5740
    },
    {
      "balancing_entropy": 0.9586,
      "balancing_loss": 1.2817,
      "epoch": 29.33792,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.521,
      "router_z_loss": 0.094,
      "step": 5750
    },
    {
      "balancing_entropy": 1.0645,
      "balancing_loss": 1.1716,
      "epoch": 29.38912,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.5276,
      "router_z_loss": 0.1162,
      "step": 5760
    },
    {
      "balancing_entropy": 1.0091,
      "balancing_loss": 1.2267,
      "epoch": 29.44032,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.5228,
      "router_z_loss": 0.0755,
      "step": 5770
    },
    {
      "balancing_entropy": 0.9928,
      "balancing_loss": 1.2111,
      "epoch": 29.49152,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.5212,
      "router_z_loss": 0.0779,
      "step": 5780
    },
    {
      "balancing_entropy": 1.0058,
      "balancing_loss": 1.2605,
      "epoch": 29.54272,
      "grad_norm": 0.24609375,
      "learning_rate": 0.003,
      "loss": 2.5097,
      "router_z_loss": 0.1347,
      "step": 5790
    },
    {
      "balancing_entropy": 1.0099,
      "balancing_loss": 1.2331,
      "epoch": 29.59392,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.5002,
      "router_z_loss": 0.0809,
      "step": 5800
    },
    {
      "balancing_entropy": 0.9904,
      "balancing_loss": 1.2414,
      "epoch": 29.64512,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.5167,
      "router_z_loss": 0.1001,
      "step": 5810
    },
    {
      "balancing_entropy": 1.0095,
      "balancing_loss": 1.2011,
      "epoch": 29.69632,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.5187,
      "router_z_loss": 0.1251,
      "step": 5820
    },
    {
      "balancing_entropy": 0.997,
      "balancing_loss": 1.2321,
      "epoch": 29.74752,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.5295,
      "router_z_loss": 0.0637,
      "step": 5830
    },
    {
      "balancing_entropy": 1.0698,
      "balancing_loss": 1.1773,
      "epoch": 29.79872,
      "grad_norm": 0.228515625,
      "learning_rate": 0.003,
      "loss": 2.5165,
      "router_z_loss": 0.1439,
      "step": 5840
    },
    {
      "balancing_entropy": 1.0245,
      "balancing_loss": 1.2258,
      "epoch": 29.84992,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.5116,
      "router_z_loss": 0.0864,
      "step": 5850
    },
    {
      "balancing_entropy": 1.0392,
      "balancing_loss": 1.2146,
      "epoch": 29.90112,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.5168,
      "router_z_loss": 0.0768,
      "step": 5860
    },
    {
      "balancing_entropy": 1.0273,
      "balancing_loss": 1.2248,
      "epoch": 29.95232,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.5312,
      "router_z_loss": 0.1213,
      "step": 5870
    },
    {
      "balancing_entropy": 1.0018,
      "balancing_loss": 1.2023,
      "epoch": 30.0,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.5253,
      "router_z_loss": 0.137,
      "step": 5880
    },
    {
      "balancing_entropy": 1.0746,
      "balancing_loss": 1.1906,
      "epoch": 30.0512,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4908,
      "router_z_loss": 0.069,
      "step": 5890
    },
    {
      "balancing_entropy": 0.9439,
      "balancing_loss": 1.2749,
      "epoch": 30.1024,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.5076,
      "router_z_loss": 0.1159,
      "step": 5900
    },
    {
      "balancing_entropy": 1.0138,
      "balancing_loss": 1.1943,
      "epoch": 30.1536,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.5006,
      "router_z_loss": 0.0942,
      "step": 5910
    },
    {
      "balancing_entropy": 1.031,
      "balancing_loss": 1.2221,
      "epoch": 30.2048,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.5075,
      "router_z_loss": 0.1441,
      "step": 5920
    },
    {
      "balancing_entropy": 1.0148,
      "balancing_loss": 1.221,
      "epoch": 30.256,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.5094,
      "router_z_loss": 0.1192,
      "step": 5930
    },
    {
      "balancing_entropy": 0.977,
      "balancing_loss": 1.255,
      "epoch": 30.3072,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.5087,
      "router_z_loss": 0.1457,
      "step": 5940
    },
    {
      "balancing_entropy": 0.9345,
      "balancing_loss": 1.2376,
      "epoch": 30.3584,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5082,
      "router_z_loss": 0.1508,
      "step": 5950
    },
    {
      "balancing_entropy": 1.0673,
      "balancing_loss": 1.1793,
      "epoch": 30.4096,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.4985,
      "router_z_loss": 0.211,
      "step": 5960
    },
    {
      "balancing_entropy": 0.968,
      "balancing_loss": 1.2649,
      "epoch": 30.4608,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.5019,
      "router_z_loss": 0.1084,
      "step": 5970
    },
    {
      "balancing_entropy": 0.9791,
      "balancing_loss": 1.2753,
      "epoch": 30.512,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.5,
      "router_z_loss": 0.1647,
      "step": 5980
    },
    {
      "balancing_entropy": 0.9739,
      "balancing_loss": 1.2585,
      "epoch": 30.5632,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.5206,
      "router_z_loss": 0.1124,
      "step": 5990
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.1612,
      "epoch": 30.6144,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.5169,
      "router_z_loss": 0.1589,
      "step": 6000
    },
    {
      "balancing_entropy": 1.0859,
      "balancing_loss": 1.1926,
      "epoch": 30.6656,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.5129,
      "router_z_loss": 0.204,
      "step": 6010
    },
    {
      "balancing_entropy": 0.9584,
      "balancing_loss": 1.268,
      "epoch": 30.7168,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.5126,
      "router_z_loss": 0.0646,
      "step": 6020
    },
    {
      "balancing_entropy": 0.978,
      "balancing_loss": 1.2072,
      "epoch": 30.768,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.4998,
      "router_z_loss": 0.1236,
      "step": 6030
    },
    {
      "balancing_entropy": 1.0203,
      "balancing_loss": 1.2263,
      "epoch": 30.8192,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.5182,
      "router_z_loss": 0.1434,
      "step": 6040
    },
    {
      "balancing_entropy": 1.0374,
      "balancing_loss": 1.2018,
      "epoch": 30.8704,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.5388,
      "router_z_loss": 0.0845,
      "step": 6050
    },
    {
      "balancing_entropy": 1.0712,
      "balancing_loss": 1.2126,
      "epoch": 30.9216,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.518,
      "router_z_loss": 0.1644,
      "step": 6060
    },
    {
      "balancing_entropy": 0.9829,
      "balancing_loss": 1.2649,
      "epoch": 30.9728,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.5256,
      "router_z_loss": 0.1155,
      "step": 6070
    },
    {
      "balancing_entropy": 0.9542,
      "balancing_loss": 1.2097,
      "epoch": 31.02048,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.4999,
      "router_z_loss": 0.1698,
      "step": 6080
    },
    {
      "balancing_entropy": 1.0142,
      "balancing_loss": 1.24,
      "epoch": 31.07168,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.5008,
      "router_z_loss": 0.1522,
      "step": 6090
    },
    {
      "balancing_entropy": 1.0408,
      "balancing_loss": 1.1921,
      "epoch": 31.12288,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.498,
      "router_z_loss": 0.0596,
      "step": 6100
    },
    {
      "balancing_entropy": 1.0797,
      "balancing_loss": 1.1883,
      "epoch": 31.17408,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.4941,
      "router_z_loss": 0.054,
      "step": 6110
    },
    {
      "balancing_entropy": 0.9462,
      "balancing_loss": 1.2868,
      "epoch": 31.22528,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5053,
      "router_z_loss": 0.11,
      "step": 6120
    },
    {
      "balancing_entropy": 1.0634,
      "balancing_loss": 1.1815,
      "epoch": 31.27648,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.4973,
      "router_z_loss": 0.3393,
      "step": 6130
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.1819,
      "epoch": 31.32768,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.4956,
      "router_z_loss": 0.1933,
      "step": 6140
    },
    {
      "balancing_entropy": 1.0276,
      "balancing_loss": 1.2246,
      "epoch": 31.37888,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.4921,
      "router_z_loss": 0.1343,
      "step": 6150
    },
    {
      "balancing_entropy": 0.9801,
      "balancing_loss": 1.2439,
      "epoch": 31.43008,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.4988,
      "router_z_loss": 0.0717,
      "step": 6160
    },
    {
      "balancing_entropy": 1.01,
      "balancing_loss": 1.2099,
      "epoch": 31.48128,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.5023,
      "router_z_loss": 0.1139,
      "step": 6170
    },
    {
      "balancing_entropy": 0.9538,
      "balancing_loss": 1.2631,
      "epoch": 31.53248,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5094,
      "router_z_loss": 0.117,
      "step": 6180
    },
    {
      "balancing_entropy": 1.0173,
      "balancing_loss": 1.2063,
      "epoch": 31.58368,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.5072,
      "router_z_loss": 0.0773,
      "step": 6190
    },
    {
      "balancing_entropy": 1.0453,
      "balancing_loss": 1.2145,
      "epoch": 31.63488,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.509,
      "router_z_loss": 0.0842,
      "step": 6200
    },
    {
      "balancing_entropy": 1.0154,
      "balancing_loss": 1.2203,
      "epoch": 31.68608,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.5134,
      "router_z_loss": 0.0788,
      "step": 6210
    },
    {
      "balancing_entropy": 1.0678,
      "balancing_loss": 1.2076,
      "epoch": 31.73728,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.5112,
      "router_z_loss": 0.201,
      "step": 6220
    },
    {
      "balancing_entropy": 0.9898,
      "balancing_loss": 1.2684,
      "epoch": 31.78848,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.5124,
      "router_z_loss": 0.0667,
      "step": 6230
    },
    {
      "balancing_entropy": 0.9865,
      "balancing_loss": 1.2235,
      "epoch": 31.83968,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.4994,
      "router_z_loss": 0.1089,
      "step": 6240
    },
    {
      "balancing_entropy": 1.0389,
      "balancing_loss": 1.1674,
      "epoch": 31.89088,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.5022,
      "router_z_loss": 0.1547,
      "step": 6250
    },
    {
      "balancing_entropy": 0.9352,
      "balancing_loss": 1.2935,
      "epoch": 31.94208,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.5057,
      "router_z_loss": 0.1024,
      "step": 6260
    },
    {
      "balancing_entropy": 1.0803,
      "balancing_loss": 1.1791,
      "epoch": 31.99328,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.4968,
      "router_z_loss": 0.0773,
      "step": 6270
    },
    {
      "balancing_entropy": 0.8912,
      "balancing_loss": 1.2577,
      "epoch": 32.04096,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.5092,
      "router_z_loss": 0.092,
      "step": 6280
    },
    {
      "balancing_entropy": 1.0043,
      "balancing_loss": 1.2264,
      "epoch": 32.09216,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.5149,
      "router_z_loss": 0.1609,
      "step": 6290
    },
    {
      "balancing_entropy": 1.0564,
      "balancing_loss": 1.1885,
      "epoch": 32.14336,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.4981,
      "router_z_loss": 0.2408,
      "step": 6300
    },
    {
      "balancing_entropy": 1.0126,
      "balancing_loss": 1.2156,
      "epoch": 32.19456,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.4913,
      "router_z_loss": 0.2105,
      "step": 6310
    },
    {
      "balancing_entropy": 0.971,
      "balancing_loss": 1.2182,
      "epoch": 32.24576,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.4842,
      "router_z_loss": 0.1137,
      "step": 6320
    },
    {
      "balancing_entropy": 1.0092,
      "balancing_loss": 1.2225,
      "epoch": 32.29696,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.4864,
      "router_z_loss": 0.1502,
      "step": 6330
    },
    {
      "balancing_entropy": 1.031,
      "balancing_loss": 1.2098,
      "epoch": 32.34816,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.4851,
      "router_z_loss": 0.0767,
      "step": 6340
    },
    {
      "balancing_entropy": 0.9623,
      "balancing_loss": 1.2481,
      "epoch": 32.39936,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.4959,
      "router_z_loss": 0.0931,
      "step": 6350
    },
    {
      "balancing_entropy": 0.9381,
      "balancing_loss": 1.2516,
      "epoch": 32.45056,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.4933,
      "router_z_loss": 0.1348,
      "step": 6360
    },
    {
      "balancing_entropy": 0.9791,
      "balancing_loss": 1.2474,
      "epoch": 32.50176,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.4973,
      "router_z_loss": 0.1201,
      "step": 6370
    },
    {
      "balancing_entropy": 0.9839,
      "balancing_loss": 1.1968,
      "epoch": 32.55296,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.5049,
      "router_z_loss": 0.0851,
      "step": 6380
    },
    {
      "balancing_entropy": 1.0254,
      "balancing_loss": 1.1932,
      "epoch": 32.60416,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.5049,
      "router_z_loss": 0.1661,
      "step": 6390
    },
    {
      "balancing_entropy": 1.0161,
      "balancing_loss": 1.2647,
      "epoch": 32.65536,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.5029,
      "router_z_loss": 0.0793,
      "step": 6400
    },
    {
      "balancing_entropy": 1.0547,
      "balancing_loss": 1.1559,
      "epoch": 32.70656,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.497,
      "router_z_loss": 0.0652,
      "step": 6410
    },
    {
      "balancing_entropy": 0.9019,
      "balancing_loss": 1.2773,
      "epoch": 32.75776,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.5082,
      "router_z_loss": 0.0944,
      "step": 6420
    },
    {
      "balancing_entropy": 1.0494,
      "balancing_loss": 1.1813,
      "epoch": 32.80896,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.4992,
      "router_z_loss": 0.177,
      "step": 6430
    },
    {
      "balancing_entropy": 1.0345,
      "balancing_loss": 1.2226,
      "epoch": 32.86016,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.4972,
      "router_z_loss": 0.0502,
      "step": 6440
    },
    {
      "balancing_entropy": 1.0273,
      "balancing_loss": 1.2094,
      "epoch": 32.91136,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.5093,
      "router_z_loss": 0.0448,
      "step": 6450
    },
    {
      "balancing_entropy": 0.8941,
      "balancing_loss": 1.2738,
      "epoch": 32.96256,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.4931,
      "router_z_loss": 0.0989,
      "step": 6460
    },
    {
      "balancing_entropy": 0.9442,
      "balancing_loss": 1.2255,
      "epoch": 33.01024,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.4935,
      "router_z_loss": 0.17,
      "step": 6470
    },
    {
      "balancing_entropy": 0.9822,
      "balancing_loss": 1.2141,
      "epoch": 33.06144,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.4839,
      "router_z_loss": 0.093,
      "step": 6480
    },
    {
      "balancing_entropy": 1.0088,
      "balancing_loss": 1.2611,
      "epoch": 33.11264,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.4708,
      "router_z_loss": 0.0634,
      "step": 6490
    },
    {
      "balancing_entropy": 0.9436,
      "balancing_loss": 1.2661,
      "epoch": 33.16384,
      "grad_norm": 0.98828125,
      "learning_rate": 0.003,
      "loss": 2.4909,
      "router_z_loss": 0.0684,
      "step": 6500
    },
    {
      "balancing_entropy": 1.0753,
      "balancing_loss": 1.1478,
      "epoch": 33.21504,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.4942,
      "router_z_loss": 0.1545,
      "step": 6510
    },
    {
      "balancing_entropy": 1.0531,
      "balancing_loss": 1.1934,
      "epoch": 33.26624,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.4786,
      "router_z_loss": 0.2238,
      "step": 6520
    },
    {
      "balancing_entropy": 0.9836,
      "balancing_loss": 1.2396,
      "epoch": 33.31744,
      "grad_norm": 0.486328125,
      "learning_rate": 0.003,
      "loss": 2.4863,
      "router_z_loss": 0.1015,
      "step": 6530
    },
    {
      "balancing_entropy": 1.0162,
      "balancing_loss": 1.2161,
      "epoch": 33.36864,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.4913,
      "router_z_loss": 0.1146,
      "step": 6540
    },
    {
      "balancing_entropy": 0.9221,
      "balancing_loss": 1.2858,
      "epoch": 33.41984,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.4898,
      "router_z_loss": 0.1116,
      "step": 6550
    },
    {
      "balancing_entropy": 1.0489,
      "balancing_loss": 1.1584,
      "epoch": 33.47104,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.4939,
      "router_z_loss": 0.2863,
      "step": 6560
    },
    {
      "balancing_entropy": 0.9836,
      "balancing_loss": 1.2539,
      "epoch": 33.52224,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.4981,
      "router_z_loss": 0.0845,
      "step": 6570
    },
    {
      "balancing_entropy": 1.0087,
      "balancing_loss": 1.2164,
      "epoch": 33.57344,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.4988,
      "router_z_loss": 0.0648,
      "step": 6580
    },
    {
      "balancing_entropy": 1.0511,
      "balancing_loss": 1.1775,
      "epoch": 33.62464,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.4929,
      "router_z_loss": 0.1654,
      "step": 6590
    },
    {
      "balancing_entropy": 1.0245,
      "balancing_loss": 1.2416,
      "epoch": 33.67584,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5008,
      "router_z_loss": 0.0894,
      "step": 6600
    },
    {
      "balancing_entropy": 1.0143,
      "balancing_loss": 1.2132,
      "epoch": 33.72704,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.5365,
      "router_z_loss": 0.0945,
      "step": 6610
    },
    {
      "balancing_entropy": 1.0789,
      "balancing_loss": 1.1519,
      "epoch": 33.77824,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.5166,
      "router_z_loss": 0.2166,
      "step": 6620
    },
    {
      "balancing_entropy": 1.0358,
      "balancing_loss": 1.1948,
      "epoch": 33.82944,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.494,
      "router_z_loss": 0.1334,
      "step": 6630
    },
    {
      "balancing_entropy": 1.0233,
      "balancing_loss": 1.2264,
      "epoch": 33.88064,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.4973,
      "router_z_loss": 0.1616,
      "step": 6640
    },
    {
      "balancing_entropy": 1.0408,
      "balancing_loss": 1.2096,
      "epoch": 33.93184,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.49,
      "router_z_loss": 0.249,
      "step": 6650
    },
    {
      "balancing_entropy": 0.9271,
      "balancing_loss": 1.2233,
      "epoch": 33.98304,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.4935,
      "router_z_loss": 0.1634,
      "step": 6660
    },
    {
      "balancing_entropy": 0.8785,
      "balancing_loss": 1.2565,
      "epoch": 34.03072,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.4851,
      "router_z_loss": 0.1496,
      "step": 6670
    },
    {
      "balancing_entropy": 1.0191,
      "balancing_loss": 1.1714,
      "epoch": 34.08192,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.4824,
      "router_z_loss": 0.1585,
      "step": 6680
    },
    {
      "balancing_entropy": 1.0263,
      "balancing_loss": 1.2034,
      "epoch": 34.13312,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.481,
      "router_z_loss": 0.0724,
      "step": 6690
    },
    {
      "balancing_entropy": 0.9794,
      "balancing_loss": 1.1919,
      "epoch": 34.18432,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.4834,
      "router_z_loss": 0.108,
      "step": 6700
    },
    {
      "balancing_entropy": 0.9212,
      "balancing_loss": 1.2948,
      "epoch": 34.23552,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.4843,
      "router_z_loss": 0.0786,
      "step": 6710
    },
    {
      "balancing_entropy": 1.0539,
      "balancing_loss": 1.1856,
      "epoch": 34.28672,
      "grad_norm": 0.244140625,
      "learning_rate": 0.003,
      "loss": 2.4823,
      "router_z_loss": 0.1742,
      "step": 6720
    },
    {
      "balancing_entropy": 1.0294,
      "balancing_loss": 1.2079,
      "epoch": 34.33792,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.4836,
      "router_z_loss": 0.0832,
      "step": 6730
    },
    {
      "balancing_entropy": 0.9458,
      "balancing_loss": 1.2288,
      "epoch": 34.38912,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.4899,
      "router_z_loss": 0.0837,
      "step": 6740
    },
    {
      "balancing_entropy": 1.0158,
      "balancing_loss": 1.1935,
      "epoch": 34.44032,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.4874,
      "router_z_loss": 0.1701,
      "step": 6750
    },
    {
      "balancing_entropy": 0.9638,
      "balancing_loss": 1.2639,
      "epoch": 34.49152,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.4887,
      "router_z_loss": 0.0705,
      "step": 6760
    },
    {
      "balancing_entropy": 1.035,
      "balancing_loss": 1.2054,
      "epoch": 34.54272,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.488,
      "router_z_loss": 0.0652,
      "step": 6770
    },
    {
      "balancing_entropy": 1.018,
      "balancing_loss": 1.2608,
      "epoch": 34.59392,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.4984,
      "router_z_loss": 0.1047,
      "step": 6780
    },
    {
      "balancing_entropy": 1.0612,
      "balancing_loss": 1.1882,
      "epoch": 34.64512,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.485,
      "router_z_loss": 0.0817,
      "step": 6790
    },
    {
      "balancing_entropy": 0.9474,
      "balancing_loss": 1.2684,
      "epoch": 34.69632,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.4934,
      "router_z_loss": 0.1234,
      "step": 6800
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.2007,
      "epoch": 34.74752,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.4878,
      "router_z_loss": 0.1107,
      "step": 6810
    },
    {
      "balancing_entropy": 0.9841,
      "balancing_loss": 1.2117,
      "epoch": 34.79872,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.4934,
      "router_z_loss": 0.0583,
      "step": 6820
    },
    {
      "balancing_entropy": 0.9766,
      "balancing_loss": 1.2576,
      "epoch": 34.84992,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.494,
      "router_z_loss": 0.0784,
      "step": 6830
    },
    {
      "balancing_entropy": 1.0768,
      "balancing_loss": 1.178,
      "epoch": 34.90112,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.4915,
      "router_z_loss": 0.2461,
      "step": 6840
    },
    {
      "balancing_entropy": 0.957,
      "balancing_loss": 1.2793,
      "epoch": 34.95232,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.4933,
      "router_z_loss": 0.1175,
      "step": 6850
    },
    {
      "balancing_entropy": 1.0054,
      "balancing_loss": 1.1831,
      "epoch": 35.0,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.4911,
      "router_z_loss": 0.0949,
      "step": 6860
    },
    {
      "balancing_entropy": 0.9133,
      "balancing_loss": 1.2876,
      "epoch": 35.0512,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.4868,
      "router_z_loss": 0.1459,
      "step": 6870
    },
    {
      "balancing_entropy": 1.0743,
      "balancing_loss": 1.19,
      "epoch": 35.1024,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.482,
      "router_z_loss": 0.1453,
      "step": 6880
    },
    {
      "balancing_entropy": 1.0154,
      "balancing_loss": 1.2053,
      "epoch": 35.1536,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.4731,
      "router_z_loss": 0.0692,
      "step": 6890
    },
    {
      "balancing_entropy": 1.0529,
      "balancing_loss": 1.2061,
      "epoch": 35.2048,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.472,
      "router_z_loss": 0.0912,
      "step": 6900
    },
    {
      "balancing_entropy": 1.0194,
      "balancing_loss": 1.2332,
      "epoch": 35.256,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.4852,
      "router_z_loss": 0.1159,
      "step": 6910
    },
    {
      "balancing_entropy": 0.9722,
      "balancing_loss": 1.2675,
      "epoch": 35.3072,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4736,
      "router_z_loss": 0.1479,
      "step": 6920
    },
    {
      "balancing_entropy": 0.9293,
      "balancing_loss": 1.3072,
      "epoch": 35.3584,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.4847,
      "router_z_loss": 0.1305,
      "step": 6930
    },
    {
      "balancing_entropy": 0.9235,
      "balancing_loss": 1.274,
      "epoch": 35.4096,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4819,
      "router_z_loss": 0.114,
      "step": 6940
    },
    {
      "balancing_entropy": 1.0721,
      "balancing_loss": 1.1606,
      "epoch": 35.4608,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4703,
      "router_z_loss": 0.1435,
      "step": 6950
    },
    {
      "balancing_entropy": 1.0116,
      "balancing_loss": 1.2544,
      "epoch": 35.512,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.481,
      "router_z_loss": 0.0749,
      "step": 6960
    },
    {
      "balancing_entropy": 1.0326,
      "balancing_loss": 1.1779,
      "epoch": 35.5632,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.4823,
      "router_z_loss": 0.0854,
      "step": 6970
    },
    {
      "balancing_entropy": 0.9798,
      "balancing_loss": 1.2514,
      "epoch": 35.6144,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.4935,
      "router_z_loss": 0.1015,
      "step": 6980
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.2262,
      "epoch": 35.6656,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.488,
      "router_z_loss": 0.1129,
      "step": 6990
    },
    {
      "balancing_entropy": 0.997,
      "balancing_loss": 1.2444,
      "epoch": 35.7168,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.493,
      "router_z_loss": 0.059,
      "step": 7000
    },
    {
      "balancing_entropy": 1.0382,
      "balancing_loss": 1.2178,
      "epoch": 35.768,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.4845,
      "router_z_loss": 0.1225,
      "step": 7010
    },
    {
      "balancing_entropy": 0.9465,
      "balancing_loss": 1.2952,
      "epoch": 35.8192,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.4884,
      "router_z_loss": 0.1087,
      "step": 7020
    },
    {
      "balancing_entropy": 0.9475,
      "balancing_loss": 1.2135,
      "epoch": 35.8704,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.4865,
      "router_z_loss": 0.0883,
      "step": 7030
    },
    {
      "balancing_entropy": 0.908,
      "balancing_loss": 1.2531,
      "epoch": 35.9216,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.4874,
      "router_z_loss": 0.1506,
      "step": 7040
    },
    {
      "balancing_entropy": 1.0706,
      "balancing_loss": 1.1971,
      "epoch": 35.9728,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.4801,
      "router_z_loss": 0.1899,
      "step": 7050
    },
    {
      "balancing_entropy": 0.9472,
      "balancing_loss": 1.248,
      "epoch": 36.02048,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.4746,
      "router_z_loss": 0.0727,
      "step": 7060
    },
    {
      "balancing_entropy": 1.0349,
      "balancing_loss": 1.1973,
      "epoch": 36.07168,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.465,
      "router_z_loss": 0.073,
      "step": 7070
    },
    {
      "balancing_entropy": 0.998,
      "balancing_loss": 1.2204,
      "epoch": 36.12288,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.4721,
      "router_z_loss": 0.0774,
      "step": 7080
    },
    {
      "balancing_entropy": 0.9549,
      "balancing_loss": 1.2767,
      "epoch": 36.174080000000004,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.4739,
      "router_z_loss": 0.1052,
      "step": 7090
    },
    {
      "balancing_entropy": 1.0446,
      "balancing_loss": 1.1762,
      "epoch": 36.22528,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.4735,
      "router_z_loss": 0.144,
      "step": 7100
    },
    {
      "balancing_entropy": 0.9847,
      "balancing_loss": 1.2652,
      "epoch": 36.27648,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.4767,
      "router_z_loss": 0.1323,
      "step": 7110
    },
    {
      "balancing_entropy": 1.0457,
      "balancing_loss": 1.1906,
      "epoch": 36.32768,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.4902,
      "router_z_loss": 0.1379,
      "step": 7120
    },
    {
      "balancing_entropy": 0.9455,
      "balancing_loss": 1.2573,
      "epoch": 36.37888,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.4739,
      "router_z_loss": 0.201,
      "step": 7130
    },
    {
      "balancing_entropy": 0.9733,
      "balancing_loss": 1.2264,
      "epoch": 36.43008,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.4769,
      "router_z_loss": 0.1468,
      "step": 7140
    },
    {
      "balancing_entropy": 0.9405,
      "balancing_loss": 1.2477,
      "epoch": 36.48128,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.4755,
      "router_z_loss": 0.1543,
      "step": 7150
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.2084,
      "epoch": 36.53248,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.4903,
      "router_z_loss": 0.0667,
      "step": 7160
    },
    {
      "balancing_entropy": 0.9425,
      "balancing_loss": 1.279,
      "epoch": 36.58368,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.4853,
      "router_z_loss": 0.1872,
      "step": 7170
    },
    {
      "balancing_entropy": 1.0138,
      "balancing_loss": 1.2224,
      "epoch": 36.63488,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.4761,
      "router_z_loss": 0.0872,
      "step": 7180
    },
    {
      "balancing_entropy": 1.078,
      "balancing_loss": 1.1766,
      "epoch": 36.68608,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.469,
      "router_z_loss": 0.0979,
      "step": 7190
    },
    {
      "balancing_entropy": 0.9676,
      "balancing_loss": 1.2819,
      "epoch": 36.73728,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.4892,
      "router_z_loss": 0.0835,
      "step": 7200
    },
    {
      "balancing_entropy": 0.9847,
      "balancing_loss": 1.2387,
      "epoch": 36.78848,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.4741,
      "router_z_loss": 0.1145,
      "step": 7210
    },
    {
      "balancing_entropy": 1.052,
      "balancing_loss": 1.2263,
      "epoch": 36.83968,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.4775,
      "router_z_loss": 0.1408,
      "step": 7220
    },
    {
      "balancing_entropy": 1.0042,
      "balancing_loss": 1.2153,
      "epoch": 36.89088,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.4808,
      "router_z_loss": 0.0879,
      "step": 7230
    },
    {
      "balancing_entropy": 0.9979,
      "balancing_loss": 1.2147,
      "epoch": 36.94208,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.4964,
      "router_z_loss": 0.1058,
      "step": 7240
    },
    {
      "balancing_entropy": 0.9671,
      "balancing_loss": 1.3124,
      "epoch": 36.99328,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.4953,
      "router_z_loss": 0.1698,
      "step": 7250
    },
    {
      "balancing_entropy": 0.8958,
      "balancing_loss": 1.2687,
      "epoch": 37.04096,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.466,
      "router_z_loss": 0.075,
      "step": 7260
    },
    {
      "balancing_entropy": 1.0796,
      "balancing_loss": 1.1684,
      "epoch": 37.09216,
      "grad_norm": 0.24609375,
      "learning_rate": 0.003,
      "loss": 2.4558,
      "router_z_loss": 0.1764,
      "step": 7270
    },
    {
      "balancing_entropy": 1.0109,
      "balancing_loss": 1.2154,
      "epoch": 37.14336,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.4569,
      "router_z_loss": 0.2071,
      "step": 7280
    },
    {
      "balancing_entropy": 0.9364,
      "balancing_loss": 1.2501,
      "epoch": 37.19456,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.467,
      "router_z_loss": 0.1127,
      "step": 7290
    },
    {
      "balancing_entropy": 0.9746,
      "balancing_loss": 1.2533,
      "epoch": 37.24576,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.4721,
      "router_z_loss": 0.0873,
      "step": 7300
    },
    {
      "balancing_entropy": 0.9072,
      "balancing_loss": 1.2629,
      "epoch": 37.29696,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.469,
      "router_z_loss": 0.0786,
      "step": 7310
    },
    {
      "balancing_entropy": 0.9984,
      "balancing_loss": 1.2014,
      "epoch": 37.34816,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.4683,
      "router_z_loss": 0.1524,
      "step": 7320
    },
    {
      "balancing_entropy": 1.0488,
      "balancing_loss": 1.2368,
      "epoch": 37.39936,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.4731,
      "router_z_loss": 0.1773,
      "step": 7330
    },
    {
      "balancing_entropy": 1.0012,
      "balancing_loss": 1.188,
      "epoch": 37.45056,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.4744,
      "router_z_loss": 0.056,
      "step": 7340
    },
    {
      "balancing_entropy": 1.041,
      "balancing_loss": 1.216,
      "epoch": 37.50176,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.4665,
      "router_z_loss": 0.2156,
      "step": 7350
    },
    {
      "balancing_entropy": 0.9965,
      "balancing_loss": 1.249,
      "epoch": 37.55296,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.483,
      "router_z_loss": 0.0689,
      "step": 7360
    },
    {
      "balancing_entropy": 0.9304,
      "balancing_loss": 1.2482,
      "epoch": 37.60416,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.4723,
      "router_z_loss": 0.2078,
      "step": 7370
    },
    {
      "balancing_entropy": 1.0265,
      "balancing_loss": 1.195,
      "epoch": 37.65536,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.4731,
      "router_z_loss": 0.205,
      "step": 7380
    },
    {
      "balancing_entropy": 0.9973,
      "balancing_loss": 1.2385,
      "epoch": 37.70656,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.4788,
      "router_z_loss": 0.0804,
      "step": 7390
    },
    {
      "balancing_entropy": 0.9999,
      "balancing_loss": 1.1948,
      "epoch": 37.75776,
      "grad_norm": 0.2373046875,
      "learning_rate": 0.003,
      "loss": 2.4679,
      "router_z_loss": 0.0968,
      "step": 7400
    },
    {
      "balancing_entropy": 0.977,
      "balancing_loss": 1.2672,
      "epoch": 37.80896,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.4744,
      "router_z_loss": 0.0755,
      "step": 7410
    },
    {
      "balancing_entropy": 1.0182,
      "balancing_loss": 1.2161,
      "epoch": 37.86016,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.468,
      "router_z_loss": 0.0481,
      "step": 7420
    },
    {
      "balancing_entropy": 0.998,
      "balancing_loss": 1.2674,
      "epoch": 37.91136,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.4774,
      "router_z_loss": 0.1288,
      "step": 7430
    },
    {
      "balancing_entropy": 0.9905,
      "balancing_loss": 1.2255,
      "epoch": 37.96256,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.4786,
      "router_z_loss": 0.0705,
      "step": 7440
    },
    {
      "balancing_entropy": 0.9834,
      "balancing_loss": 1.221,
      "epoch": 38.01024,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.4809,
      "router_z_loss": 0.0937,
      "step": 7450
    },
    {
      "balancing_entropy": 1.0194,
      "balancing_loss": 1.2077,
      "epoch": 38.06144,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.4634,
      "router_z_loss": 0.104,
      "step": 7460
    },
    {
      "balancing_entropy": 0.9442,
      "balancing_loss": 1.2756,
      "epoch": 38.11264,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.4688,
      "router_z_loss": 0.094,
      "step": 7470
    },
    {
      "balancing_entropy": 1.0332,
      "balancing_loss": 1.1673,
      "epoch": 38.16384,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.4652,
      "router_z_loss": 0.1157,
      "step": 7480
    },
    {
      "balancing_entropy": 0.9739,
      "balancing_loss": 1.2724,
      "epoch": 38.21504,
      "grad_norm": 0.78515625,
      "learning_rate": 0.003,
      "loss": 2.4695,
      "router_z_loss": 0.0833,
      "step": 7490
    },
    {
      "balancing_entropy": 1.0406,
      "balancing_loss": 1.1641,
      "epoch": 38.26624,
      "grad_norm": 0.2470703125,
      "learning_rate": 0.003,
      "loss": 2.467,
      "router_z_loss": 0.1122,
      "step": 7500
    },
    {
      "balancing_entropy": 1.0463,
      "balancing_loss": 1.2145,
      "epoch": 38.31744,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.4646,
      "router_z_loss": 0.1247,
      "step": 7510
    },
    {
      "balancing_entropy": 0.9746,
      "balancing_loss": 1.2634,
      "epoch": 38.36864,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.4671,
      "router_z_loss": 0.0579,
      "step": 7520
    },
    {
      "balancing_entropy": 0.9366,
      "balancing_loss": 1.2668,
      "epoch": 38.41984,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4636,
      "router_z_loss": 0.0961,
      "step": 7530
    },
    {
      "balancing_entropy": 1.0851,
      "balancing_loss": 1.1831,
      "epoch": 38.47104,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.457,
      "router_z_loss": 0.1443,
      "step": 7540
    },
    {
      "balancing_entropy": 1.0494,
      "balancing_loss": 1.2107,
      "epoch": 38.52224,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.4614,
      "router_z_loss": 0.0901,
      "step": 7550
    },
    {
      "balancing_entropy": 0.9922,
      "balancing_loss": 1.2317,
      "epoch": 38.57344,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.4647,
      "router_z_loss": 0.0647,
      "step": 7560
    },
    {
      "balancing_entropy": 0.9923,
      "balancing_loss": 1.2329,
      "epoch": 38.62464,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.4709,
      "router_z_loss": 0.0934,
      "step": 7570
    },
    {
      "balancing_entropy": 0.9592,
      "balancing_loss": 1.2611,
      "epoch": 38.67584,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.463,
      "router_z_loss": 0.0676,
      "step": 7580
    },
    {
      "balancing_entropy": 1.0474,
      "balancing_loss": 1.1749,
      "epoch": 38.72704,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.4809,
      "router_z_loss": 0.1022,
      "step": 7590
    },
    {
      "balancing_entropy": 1.0063,
      "balancing_loss": 1.2149,
      "epoch": 38.77824,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.4897,
      "router_z_loss": 0.0867,
      "step": 7600
    },
    {
      "balancing_entropy": 0.9842,
      "balancing_loss": 1.2775,
      "epoch": 38.82944,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.4808,
      "router_z_loss": 0.1016,
      "step": 7610
    },
    {
      "balancing_entropy": 1.0383,
      "balancing_loss": 1.1712,
      "epoch": 38.88064,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.4911,
      "router_z_loss": 0.1962,
      "step": 7620
    },
    {
      "balancing_entropy": 1.089,
      "balancing_loss": 1.1761,
      "epoch": 38.93184,
      "grad_norm": 0.2158203125,
      "learning_rate": 0.003,
      "loss": 2.4622,
      "router_z_loss": 0.1681,
      "step": 7630
    },
    {
      "balancing_entropy": 0.9218,
      "balancing_loss": 1.2378,
      "epoch": 38.98304,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.4724,
      "router_z_loss": 0.0657,
      "step": 7640
    },
    {
      "balancing_entropy": 0.9348,
      "balancing_loss": 1.2828,
      "epoch": 39.03072,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.4699,
      "router_z_loss": 0.1683,
      "step": 7650
    },
    {
      "balancing_entropy": 0.9469,
      "balancing_loss": 1.2718,
      "epoch": 39.08192,
      "grad_norm": 0.66015625,
      "learning_rate": 0.003,
      "loss": 2.4556,
      "router_z_loss": 0.0744,
      "step": 7660
    },
    {
      "balancing_entropy": 1.0363,
      "balancing_loss": 1.1587,
      "epoch": 39.13312,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.4654,
      "router_z_loss": 0.0914,
      "step": 7670
    },
    {
      "balancing_entropy": 1.0175,
      "balancing_loss": 1.2175,
      "epoch": 39.18432,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.4577,
      "router_z_loss": 0.1371,
      "step": 7680
    },
    {
      "balancing_entropy": 1.0438,
      "balancing_loss": 1.2008,
      "epoch": 39.23552,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.4468,
      "router_z_loss": 0.1055,
      "step": 7690
    },
    {
      "balancing_entropy": 1.0029,
      "balancing_loss": 1.2662,
      "epoch": 39.28672,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.4614,
      "router_z_loss": 0.0914,
      "step": 7700
    },
    {
      "balancing_entropy": 1.0118,
      "balancing_loss": 1.2066,
      "epoch": 39.33792,
      "grad_norm": 0.228515625,
      "learning_rate": 0.003,
      "loss": 2.4587,
      "router_z_loss": 0.0995,
      "step": 7710
    },
    {
      "balancing_entropy": 1.0495,
      "balancing_loss": 1.2188,
      "epoch": 39.38912,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.4644,
      "router_z_loss": 0.1631,
      "step": 7720
    },
    {
      "balancing_entropy": 0.9703,
      "balancing_loss": 1.2414,
      "epoch": 39.44032,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.4606,
      "router_z_loss": 0.0869,
      "step": 7730
    },
    {
      "balancing_entropy": 0.9428,
      "balancing_loss": 1.2689,
      "epoch": 39.49152,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.4673,
      "router_z_loss": 0.1625,
      "step": 7740
    },
    {
      "balancing_entropy": 0.9341,
      "balancing_loss": 1.2781,
      "epoch": 39.54272,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4685,
      "router_z_loss": 0.1358,
      "step": 7750
    },
    {
      "balancing_entropy": 1.0457,
      "balancing_loss": 1.1859,
      "epoch": 39.59392,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.4699,
      "router_z_loss": 0.2257,
      "step": 7760
    },
    {
      "balancing_entropy": 0.9676,
      "balancing_loss": 1.273,
      "epoch": 39.64512,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.4644,
      "router_z_loss": 0.1148,
      "step": 7770
    },
    {
      "balancing_entropy": 0.9831,
      "balancing_loss": 1.2091,
      "epoch": 39.69632,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.4668,
      "router_z_loss": 0.1217,
      "step": 7780
    },
    {
      "balancing_entropy": 1.0438,
      "balancing_loss": 1.1749,
      "epoch": 39.74752,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.4642,
      "router_z_loss": 0.2402,
      "step": 7790
    },
    {
      "balancing_entropy": 1.0263,
      "balancing_loss": 1.2105,
      "epoch": 39.79872,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.4795,
      "router_z_loss": 0.0693,
      "step": 7800
    },
    {
      "balancing_entropy": 0.9949,
      "balancing_loss": 1.2611,
      "epoch": 39.84992,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.4591,
      "router_z_loss": 0.1338,
      "step": 7810
    },
    {
      "balancing_entropy": 0.9848,
      "balancing_loss": 1.2165,
      "epoch": 39.90112,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.4681,
      "router_z_loss": 0.0881,
      "step": 7820
    },
    {
      "balancing_entropy": 1.014,
      "balancing_loss": 1.1783,
      "epoch": 39.95232,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.4605,
      "router_z_loss": 0.118,
      "step": 7830
    },
    {
      "balancing_entropy": 0.9582,
      "balancing_loss": 1.2332,
      "epoch": 40.0,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.4673,
      "router_z_loss": 0.1084,
      "step": 7840
    },
    {
      "balancing_entropy": 1.0265,
      "balancing_loss": 1.2152,
      "epoch": 40.0512,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.4645,
      "router_z_loss": 0.0996,
      "step": 7850
    },
    {
      "balancing_entropy": 1.01,
      "balancing_loss": 1.1943,
      "epoch": 40.1024,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.4564,
      "router_z_loss": 0.2651,
      "step": 7860
    },
    {
      "balancing_entropy": 0.9901,
      "balancing_loss": 1.2149,
      "epoch": 40.1536,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.455,
      "router_z_loss": 0.1373,
      "step": 7870
    },
    {
      "balancing_entropy": 1.0055,
      "balancing_loss": 1.2096,
      "epoch": 40.2048,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.4467,
      "router_z_loss": 0.137,
      "step": 7880
    },
    {
      "balancing_entropy": 0.9228,
      "balancing_loss": 1.2653,
      "epoch": 40.256,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.453,
      "router_z_loss": 0.0796,
      "step": 7890
    },
    {
      "balancing_entropy": 0.9737,
      "balancing_loss": 1.2019,
      "epoch": 40.3072,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.4583,
      "router_z_loss": 0.1026,
      "step": 7900
    },
    {
      "balancing_entropy": 0.9783,
      "balancing_loss": 1.2308,
      "epoch": 40.3584,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.466,
      "router_z_loss": 0.1236,
      "step": 7910
    },
    {
      "balancing_entropy": 1.0341,
      "balancing_loss": 1.2209,
      "epoch": 40.4096,
      "grad_norm": 0.74609375,
      "learning_rate": 0.003,
      "loss": 2.461,
      "router_z_loss": 0.1863,
      "step": 7920
    },
    {
      "balancing_entropy": 1.0288,
      "balancing_loss": 1.176,
      "epoch": 40.4608,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.4606,
      "router_z_loss": 0.1405,
      "step": 7930
    },
    {
      "balancing_entropy": 0.9792,
      "balancing_loss": 1.2389,
      "epoch": 40.512,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.4669,
      "router_z_loss": 0.0994,
      "step": 7940
    },
    {
      "balancing_entropy": 0.9834,
      "balancing_loss": 1.2536,
      "epoch": 40.5632,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.47,
      "router_z_loss": 0.1266,
      "step": 7950
    },
    {
      "balancing_entropy": 0.9836,
      "balancing_loss": 1.2272,
      "epoch": 40.6144,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.4577,
      "router_z_loss": 0.079,
      "step": 7960
    },
    {
      "balancing_entropy": 1.0087,
      "balancing_loss": 1.2394,
      "epoch": 40.6656,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.4627,
      "router_z_loss": 0.1275,
      "step": 7970
    },
    {
      "balancing_entropy": 0.9642,
      "balancing_loss": 1.2847,
      "epoch": 40.7168,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.4638,
      "router_z_loss": 0.0944,
      "step": 7980
    },
    {
      "balancing_entropy": 1.0103,
      "balancing_loss": 1.2108,
      "epoch": 40.768,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.4703,
      "router_z_loss": 0.0941,
      "step": 7990
    },
    {
      "balancing_entropy": 0.9347,
      "balancing_loss": 1.2483,
      "epoch": 40.8192,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.4713,
      "router_z_loss": 0.0967,
      "step": 8000
    },
    {
      "balancing_entropy": 0.985,
      "balancing_loss": 1.2513,
      "epoch": 40.8704,
      "grad_norm": 0.26953125,
      "learning_rate": 0.0029850000000000002,
      "loss": 2.4644,
      "router_z_loss": 0.1464,
      "step": 8010
    },
    {
      "balancing_entropy": 1.0081,
      "balancing_loss": 1.2518,
      "epoch": 40.9216,
      "grad_norm": 0.271484375,
      "learning_rate": 0.00297,
      "loss": 2.4644,
      "router_z_loss": 0.0726,
      "step": 8020
    },
    {
      "balancing_entropy": 0.9227,
      "balancing_loss": 1.2679,
      "epoch": 40.9728,
      "grad_norm": 0.7421875,
      "learning_rate": 0.002955,
      "loss": 2.4611,
      "router_z_loss": 0.0763,
      "step": 8030
    },
    {
      "balancing_entropy": 0.8089,
      "balancing_loss": 1.2141,
      "epoch": 41.02048,
      "grad_norm": 1.0234375,
      "learning_rate": 0.00294,
      "loss": 2.4646,
      "router_z_loss": 0.1471,
      "step": 8040
    },
    {
      "balancing_entropy": 0.6148,
      "balancing_loss": 1.2597,
      "epoch": 41.07168,
      "grad_norm": 0.609375,
      "learning_rate": 0.002925,
      "loss": 2.4533,
      "router_z_loss": 0.2208,
      "step": 8050
    },
    {
      "balancing_entropy": 0.7379,
      "balancing_loss": 1.2229,
      "epoch": 41.12288,
      "grad_norm": 0.73828125,
      "learning_rate": 0.00291,
      "loss": 2.4563,
      "router_z_loss": 0.2596,
      "step": 8060
    },
    {
      "balancing_entropy": 0.8774,
      "balancing_loss": 1.1816,
      "epoch": 41.174080000000004,
      "grad_norm": 0.44921875,
      "learning_rate": 0.002895,
      "loss": 2.455,
      "router_z_loss": 0.2394,
      "step": 8070
    },
    {
      "balancing_entropy": 0.848,
      "balancing_loss": 1.224,
      "epoch": 41.22528,
      "grad_norm": 0.72265625,
      "learning_rate": 0.0028799999999999997,
      "loss": 2.4433,
      "router_z_loss": 0.2768,
      "step": 8080
    },
    {
      "balancing_entropy": 0.7739,
      "balancing_loss": 1.2156,
      "epoch": 41.27648,
      "grad_norm": 0.55078125,
      "learning_rate": 0.002865,
      "loss": 2.4493,
      "router_z_loss": 0.1597,
      "step": 8090
    },
    {
      "balancing_entropy": 0.86,
      "balancing_loss": 1.2636,
      "epoch": 41.32768,
      "grad_norm": 0.359375,
      "learning_rate": 0.00285,
      "loss": 2.4451,
      "router_z_loss": 0.1765,
      "step": 8100
    },
    {
      "balancing_entropy": 0.8144,
      "balancing_loss": 1.2129,
      "epoch": 41.37888,
      "grad_norm": 0.57421875,
      "learning_rate": 0.002835,
      "loss": 2.4481,
      "router_z_loss": 0.1658,
      "step": 8110
    },
    {
      "balancing_entropy": 0.8106,
      "balancing_loss": 1.22,
      "epoch": 41.43008,
      "grad_norm": 0.58984375,
      "learning_rate": 0.00282,
      "loss": 2.4604,
      "router_z_loss": 0.1355,
      "step": 8120
    },
    {
      "balancing_entropy": 0.7999,
      "balancing_loss": 1.1825,
      "epoch": 41.48128,
      "grad_norm": 0.37890625,
      "learning_rate": 0.002805,
      "loss": 2.4431,
      "router_z_loss": 0.1147,
      "step": 8130
    },
    {
      "balancing_entropy": 0.7034,
      "balancing_loss": 1.2579,
      "epoch": 41.53248,
      "grad_norm": 0.62890625,
      "learning_rate": 0.00279,
      "loss": 2.4448,
      "router_z_loss": 0.1998,
      "step": 8140
    },
    {
      "balancing_entropy": 0.9196,
      "balancing_loss": 1.1834,
      "epoch": 41.58368,
      "grad_norm": 0.490234375,
      "learning_rate": 0.002775,
      "loss": 2.4389,
      "router_z_loss": 0.2185,
      "step": 8150
    },
    {
      "balancing_entropy": 0.8884,
      "balancing_loss": 1.2543,
      "epoch": 41.63488,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0027600000000000003,
      "loss": 2.4472,
      "router_z_loss": 0.1293,
      "step": 8160
    },
    {
      "balancing_entropy": 0.7892,
      "balancing_loss": 1.2589,
      "epoch": 41.68608,
      "grad_norm": 0.34765625,
      "learning_rate": 0.002745,
      "loss": 2.453,
      "router_z_loss": 0.1362,
      "step": 8170
    },
    {
      "balancing_entropy": 0.8657,
      "balancing_loss": 1.2232,
      "epoch": 41.73728,
      "grad_norm": 0.51171875,
      "learning_rate": 0.0027300000000000002,
      "loss": 2.4471,
      "router_z_loss": 0.2019,
      "step": 8180
    },
    {
      "balancing_entropy": 0.9077,
      "balancing_loss": 1.2508,
      "epoch": 41.78848,
      "grad_norm": 0.447265625,
      "learning_rate": 0.002715,
      "loss": 2.4403,
      "router_z_loss": 0.1763,
      "step": 8190
    },
    {
      "balancing_entropy": 0.8898,
      "balancing_loss": 1.252,
      "epoch": 41.83968,
      "grad_norm": 0.671875,
      "learning_rate": 0.0027,
      "loss": 2.4532,
      "router_z_loss": 0.0968,
      "step": 8200
    },
    {
      "balancing_entropy": 0.7995,
      "balancing_loss": 1.2154,
      "epoch": 41.89088,
      "grad_norm": 0.55078125,
      "learning_rate": 0.0026850000000000003,
      "loss": 2.4571,
      "router_z_loss": 0.1144,
      "step": 8210
    },
    {
      "balancing_entropy": 0.8251,
      "balancing_loss": 1.2152,
      "epoch": 41.94208,
      "grad_norm": 0.373046875,
      "learning_rate": 0.00267,
      "loss": 2.4427,
      "router_z_loss": 0.1099,
      "step": 8220
    },
    {
      "balancing_entropy": 0.7362,
      "balancing_loss": 1.2561,
      "epoch": 41.99328,
      "grad_norm": 0.55859375,
      "learning_rate": 0.0026550000000000002,
      "loss": 2.4383,
      "router_z_loss": 0.1622,
      "step": 8230
    },
    {
      "balancing_entropy": 0.8404,
      "balancing_loss": 1.2164,
      "epoch": 42.04096,
      "grad_norm": 0.59765625,
      "learning_rate": 0.00264,
      "loss": 2.4332,
      "router_z_loss": 0.0872,
      "step": 8240
    },
    {
      "balancing_entropy": 0.8541,
      "balancing_loss": 1.2171,
      "epoch": 42.09216,
      "grad_norm": 0.58203125,
      "learning_rate": 0.002625,
      "loss": 2.4302,
      "router_z_loss": 0.0446,
      "step": 8250
    },
    {
      "balancing_entropy": 0.843,
      "balancing_loss": 1.1849,
      "epoch": 42.14336,
      "grad_norm": 0.55859375,
      "learning_rate": 0.00261,
      "loss": 2.4175,
      "router_z_loss": 0.1318,
      "step": 8260
    },
    {
      "balancing_entropy": 0.8273,
      "balancing_loss": 1.26,
      "epoch": 42.19456,
      "grad_norm": 0.53125,
      "learning_rate": 0.002595,
      "loss": 2.4156,
      "router_z_loss": 0.1049,
      "step": 8270
    },
    {
      "balancing_entropy": 0.8428,
      "balancing_loss": 1.225,
      "epoch": 42.24576,
      "grad_norm": 0.3828125,
      "learning_rate": 0.00258,
      "loss": 2.4205,
      "router_z_loss": 0.0831,
      "step": 8280
    },
    {
      "balancing_entropy": 0.7153,
      "balancing_loss": 1.2614,
      "epoch": 42.29696,
      "grad_norm": 1.28125,
      "learning_rate": 0.002565,
      "loss": 2.4267,
      "router_z_loss": 0.058,
      "step": 8290
    },
    {
      "balancing_entropy": 0.8971,
      "balancing_loss": 1.1777,
      "epoch": 42.34816,
      "grad_norm": 0.53125,
      "learning_rate": 0.00255,
      "loss": 2.424,
      "router_z_loss": 0.0864,
      "step": 8300
    },
    {
      "balancing_entropy": 0.7372,
      "balancing_loss": 1.2599,
      "epoch": 42.39936,
      "grad_norm": 0.59765625,
      "learning_rate": 0.002535,
      "loss": 2.4413,
      "router_z_loss": 0.1219,
      "step": 8310
    },
    {
      "balancing_entropy": 0.6794,
      "balancing_loss": 1.2586,
      "epoch": 42.45056,
      "grad_norm": 0.63671875,
      "learning_rate": 0.00252,
      "loss": 2.4275,
      "router_z_loss": 0.176,
      "step": 8320
    },
    {
      "balancing_entropy": 0.7012,
      "balancing_loss": 1.2621,
      "epoch": 42.50176,
      "grad_norm": 0.6875,
      "learning_rate": 0.002505,
      "loss": 2.4303,
      "router_z_loss": 0.1779,
      "step": 8330
    },
    {
      "balancing_entropy": 0.8968,
      "balancing_loss": 1.1948,
      "epoch": 42.55296,
      "grad_norm": 0.6953125,
      "learning_rate": 0.00249,
      "loss": 2.4278,
      "router_z_loss": 0.3305,
      "step": 8340
    },
    {
      "balancing_entropy": 0.9196,
      "balancing_loss": 1.2566,
      "epoch": 42.60416,
      "grad_norm": 0.59765625,
      "learning_rate": 0.0024749999999999998,
      "loss": 2.4307,
      "router_z_loss": 0.2568,
      "step": 8350
    },
    {
      "balancing_entropy": 0.8319,
      "balancing_loss": 1.2589,
      "epoch": 42.65536,
      "grad_norm": 0.71484375,
      "learning_rate": 0.0024600000000000004,
      "loss": 2.4177,
      "router_z_loss": 0.0923,
      "step": 8360
    },
    {
      "balancing_entropy": 0.8464,
      "balancing_loss": 1.1867,
      "epoch": 42.70656,
      "grad_norm": 0.55078125,
      "learning_rate": 0.0024449999999999997,
      "loss": 2.4188,
      "router_z_loss": 0.1331,
      "step": 8370
    },
    {
      "balancing_entropy": 0.7588,
      "balancing_loss": 1.2256,
      "epoch": 42.75776,
      "grad_norm": 0.65625,
      "learning_rate": 0.0024300000000000003,
      "loss": 2.4265,
      "router_z_loss": 0.1734,
      "step": 8380
    },
    {
      "balancing_entropy": 0.657,
      "balancing_loss": 1.2912,
      "epoch": 42.80896,
      "grad_norm": 0.515625,
      "learning_rate": 0.002415,
      "loss": 2.4274,
      "router_z_loss": 0.1688,
      "step": 8390
    },
    {
      "balancing_entropy": 0.8216,
      "balancing_loss": 1.1765,
      "epoch": 42.86016,
      "grad_norm": 0.8203125,
      "learning_rate": 0.0024000000000000002,
      "loss": 2.4349,
      "router_z_loss": 0.1528,
      "step": 8400
    },
    {
      "balancing_entropy": 0.6923,
      "balancing_loss": 1.2599,
      "epoch": 42.91136,
      "grad_norm": 0.6015625,
      "learning_rate": 0.002385,
      "loss": 2.4317,
      "router_z_loss": 0.0679,
      "step": 8410
    },
    {
      "balancing_entropy": 0.9217,
      "balancing_loss": 1.1801,
      "epoch": 42.96256,
      "grad_norm": 0.408203125,
      "learning_rate": 0.00237,
      "loss": 2.4335,
      "router_z_loss": 0.1641,
      "step": 8420
    },
    {
      "balancing_entropy": 0.7633,
      "balancing_loss": 1.2524,
      "epoch": 43.01024,
      "grad_norm": 0.67578125,
      "learning_rate": 0.0023550000000000003,
      "loss": 2.4194,
      "router_z_loss": 0.0864,
      "step": 8430
    },
    {
      "balancing_entropy": 0.9301,
      "balancing_loss": 1.216,
      "epoch": 43.06144,
      "grad_norm": 0.3046875,
      "learning_rate": 0.00234,
      "loss": 2.4084,
      "router_z_loss": 0.1031,
      "step": 8440
    },
    {
      "balancing_entropy": 0.6962,
      "balancing_loss": 1.3018,
      "epoch": 43.11264,
      "grad_norm": 1.0546875,
      "learning_rate": 0.0023250000000000002,
      "loss": 2.4005,
      "router_z_loss": 0.0913,
      "step": 8450
    },
    {
      "balancing_entropy": 0.7738,
      "balancing_loss": 1.2168,
      "epoch": 43.16384,
      "grad_norm": 0.51171875,
      "learning_rate": 0.00231,
      "loss": 2.3999,
      "router_z_loss": 0.2029,
      "step": 8460
    },
    {
      "balancing_entropy": 0.8257,
      "balancing_loss": 1.1812,
      "epoch": 43.21504,
      "grad_norm": 0.56640625,
      "learning_rate": 0.002295,
      "loss": 2.4039,
      "router_z_loss": 0.2101,
      "step": 8470
    },
    {
      "balancing_entropy": 0.8524,
      "balancing_loss": 1.2533,
      "epoch": 43.26624,
      "grad_norm": 0.70703125,
      "learning_rate": 0.00228,
      "loss": 2.403,
      "router_z_loss": 0.0489,
      "step": 8480
    },
    {
      "balancing_entropy": 0.8359,
      "balancing_loss": 1.2558,
      "epoch": 43.31744,
      "grad_norm": 0.5703125,
      "learning_rate": 0.002265,
      "loss": 2.3995,
      "router_z_loss": 0.0814,
      "step": 8490
    },
    {
      "balancing_entropy": 0.7418,
      "balancing_loss": 1.2162,
      "epoch": 43.36864,
      "grad_norm": 0.423828125,
      "learning_rate": 0.0022500000000000003,
      "loss": 2.4063,
      "router_z_loss": 0.1489,
      "step": 8500
    },
    {
      "balancing_entropy": 0.7238,
      "balancing_loss": 1.2962,
      "epoch": 43.41984,
      "grad_norm": 0.29296875,
      "learning_rate": 0.002235,
      "loss": 2.3958,
      "router_z_loss": 0.1697,
      "step": 8510
    },
    {
      "balancing_entropy": 0.8345,
      "balancing_loss": 1.219,
      "epoch": 43.47104,
      "grad_norm": 0.5859375,
      "learning_rate": 0.00222,
      "loss": 2.4069,
      "router_z_loss": 0.0662,
      "step": 8520
    },
    {
      "balancing_entropy": 0.734,
      "balancing_loss": 1.2599,
      "epoch": 43.52224,
      "grad_norm": 1.09375,
      "learning_rate": 0.002205,
      "loss": 2.414,
      "router_z_loss": 0.068,
      "step": 8530
    },
    {
      "balancing_entropy": 0.843,
      "balancing_loss": 1.1763,
      "epoch": 43.57344,
      "grad_norm": 0.63671875,
      "learning_rate": 0.00219,
      "loss": 2.4155,
      "router_z_loss": 0.0995,
      "step": 8540
    },
    {
      "balancing_entropy": 0.828,
      "balancing_loss": 1.2171,
      "epoch": 43.62464,
      "grad_norm": 0.71875,
      "learning_rate": 0.002175,
      "loss": 2.4108,
      "router_z_loss": 0.0649,
      "step": 8550
    },
    {
      "balancing_entropy": 0.6969,
      "balancing_loss": 1.2992,
      "epoch": 43.67584,
      "grad_norm": 0.65234375,
      "learning_rate": 0.00216,
      "loss": 2.402,
      "router_z_loss": 0.1246,
      "step": 8560
    },
    {
      "balancing_entropy": 0.9256,
      "balancing_loss": 1.1814,
      "epoch": 43.72704,
      "grad_norm": 0.5859375,
      "learning_rate": 0.002145,
      "loss": 2.4054,
      "router_z_loss": 0.1637,
      "step": 8570
    },
    {
      "balancing_entropy": 0.848,
      "balancing_loss": 1.3096,
      "epoch": 43.77824,
      "grad_norm": 0.34765625,
      "learning_rate": 0.00213,
      "loss": 2.3981,
      "router_z_loss": 0.0824,
      "step": 8580
    },
    {
      "balancing_entropy": 0.8694,
      "balancing_loss": 1.1731,
      "epoch": 43.82944,
      "grad_norm": 0.73046875,
      "learning_rate": 0.002115,
      "loss": 2.4046,
      "router_z_loss": 0.1055,
      "step": 8590
    },
    {
      "balancing_entropy": 0.7188,
      "balancing_loss": 1.2979,
      "epoch": 43.88064,
      "grad_norm": 0.53125,
      "learning_rate": 0.0021,
      "loss": 2.3976,
      "router_z_loss": 0.1363,
      "step": 8600
    },
    {
      "balancing_entropy": 0.7956,
      "balancing_loss": 1.2155,
      "epoch": 43.93184,
      "grad_norm": 0.66796875,
      "learning_rate": 0.002085,
      "loss": 2.4032,
      "router_z_loss": 0.102,
      "step": 8610
    },
    {
      "balancing_entropy": 0.7573,
      "balancing_loss": 1.214,
      "epoch": 43.98304,
      "grad_norm": 0.625,
      "learning_rate": 0.00207,
      "loss": 2.3949,
      "router_z_loss": 0.0869,
      "step": 8620
    },
    {
      "balancing_entropy": 0.8077,
      "balancing_loss": 1.2149,
      "epoch": 44.03072,
      "grad_norm": 0.37890625,
      "learning_rate": 0.0020550000000000004,
      "loss": 2.3897,
      "router_z_loss": 0.0705,
      "step": 8630
    },
    {
      "balancing_entropy": 0.8476,
      "balancing_loss": 1.2142,
      "epoch": 44.08192,
      "grad_norm": 0.6328125,
      "learning_rate": 0.0020399999999999997,
      "loss": 2.3936,
      "router_z_loss": 0.0624,
      "step": 8640
    },
    {
      "balancing_entropy": 0.7332,
      "balancing_loss": 1.2561,
      "epoch": 44.13312,
      "grad_norm": 0.345703125,
      "learning_rate": 0.0020250000000000003,
      "loss": 2.3847,
      "router_z_loss": 0.1036,
      "step": 8650
    },
    {
      "balancing_entropy": 0.8359,
      "balancing_loss": 1.2187,
      "epoch": 44.18432,
      "grad_norm": 0.3671875,
      "learning_rate": 0.0020099999999999996,
      "loss": 2.3802,
      "router_z_loss": 0.1095,
      "step": 8660
    },
    {
      "balancing_entropy": 0.7743,
      "balancing_loss": 1.2997,
      "epoch": 44.23552,
      "grad_norm": 0.76171875,
      "learning_rate": 0.0019950000000000002,
      "loss": 2.3784,
      "router_z_loss": 0.0906,
      "step": 8670
    },
    {
      "balancing_entropy": 0.8957,
      "balancing_loss": 1.1735,
      "epoch": 44.28672,
      "grad_norm": 0.44140625,
      "learning_rate": 0.00198,
      "loss": 2.3824,
      "router_z_loss": 0.12,
      "step": 8680
    },
    {
      "balancing_entropy": 0.7518,
      "balancing_loss": 1.3328,
      "epoch": 44.33792,
      "grad_norm": 0.7734375,
      "learning_rate": 0.001965,
      "loss": 2.3855,
      "router_z_loss": 0.0754,
      "step": 8690
    },
    {
      "balancing_entropy": 0.8893,
      "balancing_loss": 1.172,
      "epoch": 44.38912,
      "grad_norm": 0.421875,
      "learning_rate": 0.0019500000000000001,
      "loss": 2.3878,
      "router_z_loss": 0.0741,
      "step": 8700
    },
    {
      "balancing_entropy": 0.7977,
      "balancing_loss": 1.2169,
      "epoch": 44.44032,
      "grad_norm": 0.37109375,
      "learning_rate": 0.001935,
      "loss": 2.3886,
      "router_z_loss": 0.0871,
      "step": 8710
    },
    {
      "balancing_entropy": 0.8243,
      "balancing_loss": 1.2579,
      "epoch": 44.49152,
      "grad_norm": 0.515625,
      "learning_rate": 0.00192,
      "loss": 2.372,
      "router_z_loss": 0.0661,
      "step": 8720
    },
    {
      "balancing_entropy": 0.8478,
      "balancing_loss": 1.2523,
      "epoch": 44.54272,
      "grad_norm": 0.3671875,
      "learning_rate": 0.001905,
      "loss": 2.3677,
      "router_z_loss": 0.0771,
      "step": 8730
    },
    {
      "balancing_entropy": 0.8065,
      "balancing_loss": 1.2595,
      "epoch": 44.59392,
      "grad_norm": 0.380859375,
      "learning_rate": 0.00189,
      "loss": 2.3712,
      "router_z_loss": 0.0727,
      "step": 8740
    },
    {
      "balancing_entropy": 0.7898,
      "balancing_loss": 1.2551,
      "epoch": 44.64512,
      "grad_norm": 0.5390625,
      "learning_rate": 0.001875,
      "loss": 2.3817,
      "router_z_loss": 0.103,
      "step": 8750
    },
    {
      "balancing_entropy": 0.8247,
      "balancing_loss": 1.1756,
      "epoch": 44.69632,
      "grad_norm": 0.55859375,
      "learning_rate": 0.00186,
      "loss": 2.372,
      "router_z_loss": 0.1009,
      "step": 8760
    },
    {
      "balancing_entropy": 0.8376,
      "balancing_loss": 1.2526,
      "epoch": 44.74752,
      "grad_norm": 0.494140625,
      "learning_rate": 0.001845,
      "loss": 2.3855,
      "router_z_loss": 0.0666,
      "step": 8770
    },
    {
      "balancing_entropy": 0.754,
      "balancing_loss": 1.256,
      "epoch": 44.79872,
      "grad_norm": 0.5546875,
      "learning_rate": 0.00183,
      "loss": 2.376,
      "router_z_loss": 0.0834,
      "step": 8780
    },
    {
      "balancing_entropy": 0.8348,
      "balancing_loss": 1.2976,
      "epoch": 44.84992,
      "grad_norm": 0.36328125,
      "learning_rate": 0.001815,
      "loss": 2.3715,
      "router_z_loss": 0.0725,
      "step": 8790
    },
    {
      "balancing_entropy": 0.8547,
      "balancing_loss": 1.2481,
      "epoch": 44.90112,
      "grad_norm": 0.64453125,
      "learning_rate": 0.0018,
      "loss": 2.3811,
      "router_z_loss": 0.1,
      "step": 8800
    },
    {
      "balancing_entropy": 0.8243,
      "balancing_loss": 1.2115,
      "epoch": 44.95232,
      "grad_norm": 0.443359375,
      "learning_rate": 0.001785,
      "loss": 2.3796,
      "router_z_loss": 0.0958,
      "step": 8810
    },
    {
      "balancing_entropy": 0.8555,
      "balancing_loss": 1.2578,
      "epoch": 45.0,
      "grad_norm": 0.5390625,
      "learning_rate": 0.0017700000000000003,
      "loss": 2.3829,
      "router_z_loss": 0.0869,
      "step": 8820
    },
    {
      "balancing_entropy": 0.7651,
      "balancing_loss": 1.2494,
      "epoch": 45.0512,
      "grad_norm": 0.42578125,
      "learning_rate": 0.0017549999999999998,
      "loss": 2.364,
      "router_z_loss": 0.1107,
      "step": 8830
    },
    {
      "balancing_entropy": 0.8047,
      "balancing_loss": 1.213,
      "epoch": 45.1024,
      "grad_norm": 0.37890625,
      "learning_rate": 0.0017400000000000002,
      "loss": 2.3629,
      "router_z_loss": 0.1786,
      "step": 8840
    },
    {
      "balancing_entropy": 0.9152,
      "balancing_loss": 1.1772,
      "epoch": 45.1536,
      "grad_norm": 0.68359375,
      "learning_rate": 0.001725,
      "loss": 2.3586,
      "router_z_loss": 0.1242,
      "step": 8850
    },
    {
      "balancing_entropy": 0.6198,
      "balancing_loss": 1.2982,
      "epoch": 45.2048,
      "grad_norm": 0.318359375,
      "learning_rate": 0.0017100000000000001,
      "loss": 2.3585,
      "router_z_loss": 0.0991,
      "step": 8860
    },
    {
      "balancing_entropy": 0.6195,
      "balancing_loss": 1.2615,
      "epoch": 45.256,
      "grad_norm": 0.5234375,
      "learning_rate": 0.001695,
      "loss": 2.3597,
      "router_z_loss": 0.221,
      "step": 8870
    },
    {
      "balancing_entropy": 0.8695,
      "balancing_loss": 1.1775,
      "epoch": 45.3072,
      "grad_norm": 0.671875,
      "learning_rate": 0.0016800000000000003,
      "loss": 2.3559,
      "router_z_loss": 0.1646,
      "step": 8880
    },
    {
      "balancing_entropy": 0.8616,
      "balancing_loss": 1.2501,
      "epoch": 45.3584,
      "grad_norm": 0.52734375,
      "learning_rate": 0.0016649999999999998,
      "loss": 2.3587,
      "router_z_loss": 0.0439,
      "step": 8890
    },
    {
      "balancing_entropy": 0.8529,
      "balancing_loss": 1.304,
      "epoch": 45.4096,
      "grad_norm": 0.271484375,
      "learning_rate": 0.0016500000000000002,
      "loss": 2.3508,
      "router_z_loss": 0.0568,
      "step": 8900
    },
    {
      "balancing_entropy": 0.891,
      "balancing_loss": 1.1723,
      "epoch": 45.4608,
      "grad_norm": 0.6328125,
      "learning_rate": 0.0016349999999999997,
      "loss": 2.3665,
      "router_z_loss": 0.1077,
      "step": 8910
    },
    {
      "balancing_entropy": 0.8896,
      "balancing_loss": 1.2522,
      "epoch": 45.512,
      "grad_norm": 0.65625,
      "learning_rate": 0.0016200000000000001,
      "loss": 2.3521,
      "router_z_loss": 0.1102,
      "step": 8920
    },
    {
      "balancing_entropy": 0.8247,
      "balancing_loss": 1.2479,
      "epoch": 45.5632,
      "grad_norm": 0.47265625,
      "learning_rate": 0.0016049999999999999,
      "loss": 2.3583,
      "router_z_loss": 0.0446,
      "step": 8930
    },
    {
      "balancing_entropy": 0.8356,
      "balancing_loss": 1.1758,
      "epoch": 45.6144,
      "grad_norm": 0.44921875,
      "learning_rate": 0.00159,
      "loss": 2.3489,
      "router_z_loss": 0.0759,
      "step": 8940
    },
    {
      "balancing_entropy": 0.6825,
      "balancing_loss": 1.2948,
      "epoch": 45.6656,
      "grad_norm": 0.46484375,
      "learning_rate": 0.001575,
      "loss": 2.3581,
      "router_z_loss": 0.1018,
      "step": 8950
    },
    {
      "balancing_entropy": 0.8924,
      "balancing_loss": 1.2518,
      "epoch": 45.7168,
      "grad_norm": 0.353515625,
      "learning_rate": 0.0015600000000000002,
      "loss": 2.3553,
      "router_z_loss": 0.0782,
      "step": 8960
    },
    {
      "balancing_entropy": 0.8789,
      "balancing_loss": 1.2523,
      "epoch": 45.768,
      "grad_norm": 0.5703125,
      "learning_rate": 0.0015450000000000001,
      "loss": 2.3576,
      "router_z_loss": 0.0674,
      "step": 8970
    },
    {
      "balancing_entropy": 0.7109,
      "balancing_loss": 1.2625,
      "epoch": 45.8192,
      "grad_norm": 0.53515625,
      "learning_rate": 0.0015300000000000001,
      "loss": 2.3538,
      "router_z_loss": 0.0766,
      "step": 8980
    },
    {
      "balancing_entropy": 0.8926,
      "balancing_loss": 1.1736,
      "epoch": 45.8704,
      "grad_norm": 0.49609375,
      "learning_rate": 0.001515,
      "loss": 2.356,
      "router_z_loss": 0.1049,
      "step": 8990
    },
    {
      "balancing_entropy": 0.8498,
      "balancing_loss": 1.2559,
      "epoch": 45.9216,
      "grad_norm": 0.451171875,
      "learning_rate": 0.0015,
      "loss": 2.351,
      "router_z_loss": 0.0731,
      "step": 9000
    },
    {
      "balancing_entropy": 0.8328,
      "balancing_loss": 1.246,
      "epoch": 45.9728,
      "grad_norm": 0.5390625,
      "learning_rate": 0.001485,
      "loss": 2.3588,
      "router_z_loss": 0.0465,
      "step": 9010
    },
    {
      "balancing_entropy": 0.8651,
      "balancing_loss": 1.285,
      "epoch": 46.02048,
      "grad_norm": 0.4375,
      "learning_rate": 0.00147,
      "loss": 2.3382,
      "router_z_loss": 0.1124,
      "step": 9020
    },
    {
      "balancing_entropy": 0.7781,
      "balancing_loss": 1.2596,
      "epoch": 46.07168,
      "grad_norm": 0.419921875,
      "learning_rate": 0.001455,
      "loss": 2.3368,
      "router_z_loss": 0.0748,
      "step": 9030
    },
    {
      "balancing_entropy": 0.919,
      "balancing_loss": 1.172,
      "epoch": 46.12288,
      "grad_norm": 0.68359375,
      "learning_rate": 0.0014399999999999999,
      "loss": 2.3337,
      "router_z_loss": 0.0715,
      "step": 9040
    },
    {
      "balancing_entropy": 0.6969,
      "balancing_loss": 1.3058,
      "epoch": 46.174080000000004,
      "grad_norm": 0.388671875,
      "learning_rate": 0.001425,
      "loss": 2.3327,
      "router_z_loss": 0.0642,
      "step": 9050
    },
    {
      "balancing_entropy": 0.7495,
      "balancing_loss": 1.1762,
      "epoch": 46.22528,
      "grad_norm": 0.423828125,
      "learning_rate": 0.00141,
      "loss": 2.332,
      "router_z_loss": 0.2183,
      "step": 9060
    },
    {
      "balancing_entropy": 0.7601,
      "balancing_loss": 1.2506,
      "epoch": 46.27648,
      "grad_norm": 0.423828125,
      "learning_rate": 0.001395,
      "loss": 2.3395,
      "router_z_loss": 0.0943,
      "step": 9070
    },
    {
      "balancing_entropy": 0.8267,
      "balancing_loss": 1.2555,
      "epoch": 46.32768,
      "grad_norm": 0.578125,
      "learning_rate": 0.00138,
      "loss": 2.3255,
      "router_z_loss": 0.0728,
      "step": 9080
    },
    {
      "balancing_entropy": 0.7687,
      "balancing_loss": 1.2545,
      "epoch": 46.37888,
      "grad_norm": 0.388671875,
      "learning_rate": 0.001365,
      "loss": 2.3373,
      "router_z_loss": 0.0582,
      "step": 9090
    },
    {
      "balancing_entropy": 0.8851,
      "balancing_loss": 1.1812,
      "epoch": 46.43008,
      "grad_norm": 0.3984375,
      "learning_rate": 0.0013499999999999999,
      "loss": 2.3302,
      "router_z_loss": 0.1045,
      "step": 9100
    },
    {
      "balancing_entropy": 0.8044,
      "balancing_loss": 1.2888,
      "epoch": 46.48128,
      "grad_norm": 0.99609375,
      "learning_rate": 0.0013349999999999998,
      "loss": 2.3403,
      "router_z_loss": 0.0438,
      "step": 9110
    },
    {
      "balancing_entropy": 0.8186,
      "balancing_loss": 1.2103,
      "epoch": 46.53248,
      "grad_norm": 0.53515625,
      "learning_rate": 0.0013199999999999998,
      "loss": 2.3269,
      "router_z_loss": 0.0529,
      "step": 9120
    },
    {
      "balancing_entropy": 0.7715,
      "balancing_loss": 1.2569,
      "epoch": 46.58368,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0013050000000000002,
      "loss": 2.3281,
      "router_z_loss": 0.0845,
      "step": 9130
    },
    {
      "balancing_entropy": 0.7781,
      "balancing_loss": 1.2905,
      "epoch": 46.63488,
      "grad_norm": 0.55859375,
      "learning_rate": 0.0012900000000000001,
      "loss": 2.3289,
      "router_z_loss": 0.0546,
      "step": 9140
    },
    {
      "balancing_entropy": 0.8095,
      "balancing_loss": 1.2537,
      "epoch": 46.68608,
      "grad_norm": 0.56640625,
      "learning_rate": 0.001275,
      "loss": 2.3258,
      "router_z_loss": 0.0611,
      "step": 9150
    },
    {
      "balancing_entropy": 0.8461,
      "balancing_loss": 1.2532,
      "epoch": 46.73728,
      "grad_norm": 0.3359375,
      "learning_rate": 0.00126,
      "loss": 2.3351,
      "router_z_loss": 0.0637,
      "step": 9160
    },
    {
      "balancing_entropy": 0.8628,
      "balancing_loss": 1.2167,
      "epoch": 46.78848,
      "grad_norm": 0.32421875,
      "learning_rate": 0.0012450000000000002,
      "loss": 2.3315,
      "router_z_loss": 0.052,
      "step": 9170
    },
    {
      "balancing_entropy": 0.8756,
      "balancing_loss": 1.2138,
      "epoch": 46.83968,
      "grad_norm": 0.47265625,
      "learning_rate": 0.0012300000000000002,
      "loss": 2.3294,
      "router_z_loss": 0.0436,
      "step": 9180
    },
    {
      "balancing_entropy": 0.8248,
      "balancing_loss": 1.2548,
      "epoch": 46.89088,
      "grad_norm": 0.53125,
      "learning_rate": 0.0012150000000000002,
      "loss": 2.3305,
      "router_z_loss": 0.0511,
      "step": 9190
    },
    {
      "balancing_entropy": 0.789,
      "balancing_loss": 1.2571,
      "epoch": 46.94208,
      "grad_norm": 0.515625,
      "learning_rate": 0.0012000000000000001,
      "loss": 2.3337,
      "router_z_loss": 0.0549,
      "step": 9200
    },
    {
      "balancing_entropy": 0.7669,
      "balancing_loss": 1.2158,
      "epoch": 46.99328,
      "grad_norm": 0.8828125,
      "learning_rate": 0.001185,
      "loss": 2.3304,
      "router_z_loss": 0.057,
      "step": 9210
    },
    {
      "balancing_entropy": 0.7949,
      "balancing_loss": 1.2162,
      "epoch": 47.04096,
      "grad_norm": 0.62890625,
      "learning_rate": 0.00117,
      "loss": 2.3229,
      "router_z_loss": 0.0489,
      "step": 9220
    },
    {
      "balancing_entropy": 0.8259,
      "balancing_loss": 1.2529,
      "epoch": 47.09216,
      "grad_norm": 0.341796875,
      "learning_rate": 0.001155,
      "loss": 2.3189,
      "router_z_loss": 0.0547,
      "step": 9230
    },
    {
      "balancing_entropy": 0.8956,
      "balancing_loss": 1.2981,
      "epoch": 47.14336,
      "grad_norm": 0.400390625,
      "learning_rate": 0.00114,
      "loss": 2.3153,
      "router_z_loss": 0.0508,
      "step": 9240
    },
    {
      "balancing_entropy": 0.8627,
      "balancing_loss": 1.212,
      "epoch": 47.19456,
      "grad_norm": 0.79296875,
      "learning_rate": 0.0011250000000000001,
      "loss": 2.3179,
      "router_z_loss": 0.0365,
      "step": 9250
    },
    {
      "balancing_entropy": 0.8062,
      "balancing_loss": 1.2525,
      "epoch": 47.24576,
      "grad_norm": 0.431640625,
      "learning_rate": 0.00111,
      "loss": 2.3149,
      "router_z_loss": 0.0589,
      "step": 9260
    },
    {
      "balancing_entropy": 0.8146,
      "balancing_loss": 1.1783,
      "epoch": 47.29696,
      "grad_norm": 0.6328125,
      "learning_rate": 0.001095,
      "loss": 2.3134,
      "router_z_loss": 0.1028,
      "step": 9270
    },
    {
      "balancing_entropy": 0.7608,
      "balancing_loss": 1.2896,
      "epoch": 47.34816,
      "grad_norm": 0.3359375,
      "learning_rate": 0.00108,
      "loss": 2.3115,
      "router_z_loss": 0.0641,
      "step": 9280
    },
    {
      "balancing_entropy": 0.8193,
      "balancing_loss": 1.2937,
      "epoch": 47.39936,
      "grad_norm": 0.6328125,
      "learning_rate": 0.001065,
      "loss": 2.3105,
      "router_z_loss": 0.0441,
      "step": 9290
    },
    {
      "balancing_entropy": 0.8843,
      "balancing_loss": 1.1728,
      "epoch": 47.45056,
      "grad_norm": 0.61328125,
      "learning_rate": 0.00105,
      "loss": 2.3094,
      "router_z_loss": 0.0358,
      "step": 9300
    },
    {
      "balancing_entropy": 0.8306,
      "balancing_loss": 1.2872,
      "epoch": 47.50176,
      "grad_norm": 0.330078125,
      "learning_rate": 0.001035,
      "loss": 2.3112,
      "router_z_loss": 0.045,
      "step": 9310
    },
    {
      "balancing_entropy": 0.7419,
      "balancing_loss": 1.2918,
      "epoch": 47.55296,
      "grad_norm": 0.52734375,
      "learning_rate": 0.0010199999999999999,
      "loss": 2.31,
      "router_z_loss": 0.0479,
      "step": 9320
    },
    {
      "balancing_entropy": 0.8635,
      "balancing_loss": 1.2024,
      "epoch": 47.60416,
      "grad_norm": 0.50390625,
      "learning_rate": 0.0010049999999999998,
      "loss": 2.3106,
      "router_z_loss": 0.0493,
      "step": 9330
    },
    {
      "balancing_entropy": 0.8387,
      "balancing_loss": 1.29,
      "epoch": 47.65536,
      "grad_norm": 0.5390625,
      "learning_rate": 0.00099,
      "loss": 2.3042,
      "router_z_loss": 0.0557,
      "step": 9340
    },
    {
      "balancing_entropy": 0.7894,
      "balancing_loss": 1.2133,
      "epoch": 47.70656,
      "grad_norm": 0.89453125,
      "learning_rate": 0.0009749999999999998,
      "loss": 2.3024,
      "router_z_loss": 0.0634,
      "step": 9350
    },
    {
      "balancing_entropy": 0.8418,
      "balancing_loss": 1.2132,
      "epoch": 47.75776,
      "grad_norm": 0.625,
      "learning_rate": 0.0009599999999999999,
      "loss": 2.3155,
      "router_z_loss": 0.0638,
      "step": 9360
    },
    {
      "balancing_entropy": 0.7028,
      "balancing_loss": 1.2928,
      "epoch": 47.80896,
      "grad_norm": 0.373046875,
      "learning_rate": 0.0009449999999999999,
      "loss": 2.3096,
      "router_z_loss": 0.0543,
      "step": 9370
    },
    {
      "balancing_entropy": 0.8533,
      "balancing_loss": 1.2123,
      "epoch": 47.86016,
      "grad_norm": 0.953125,
      "learning_rate": 0.0009300000000000002,
      "loss": 2.3067,
      "router_z_loss": 0.0498,
      "step": 9380
    },
    {
      "balancing_entropy": 0.7485,
      "balancing_loss": 1.2537,
      "epoch": 47.91136,
      "grad_norm": 0.455078125,
      "learning_rate": 0.0009150000000000001,
      "loss": 2.3076,
      "router_z_loss": 0.0529,
      "step": 9390
    },
    {
      "balancing_entropy": 0.8873,
      "balancing_loss": 1.176,
      "epoch": 47.96256,
      "grad_norm": 0.3515625,
      "learning_rate": 0.0009000000000000002,
      "loss": 2.3022,
      "router_z_loss": 0.0911,
      "step": 9400
    },
    {
      "balancing_entropy": 0.8318,
      "balancing_loss": 1.3321,
      "epoch": 48.01024,
      "grad_norm": 0.5625,
      "learning_rate": 0.0008850000000000002,
      "loss": 2.2981,
      "router_z_loss": 0.0628,
      "step": 9410
    },
    {
      "balancing_entropy": 0.8385,
      "balancing_loss": 1.2914,
      "epoch": 48.06144,
      "grad_norm": 0.298828125,
      "learning_rate": 0.0008700000000000001,
      "loss": 2.2909,
      "router_z_loss": 0.0434,
      "step": 9420
    },
    {
      "balancing_entropy": 0.9519,
      "balancing_loss": 1.1714,
      "epoch": 48.11264,
      "grad_norm": 0.3671875,
      "learning_rate": 0.0008550000000000001,
      "loss": 2.2856,
      "router_z_loss": 0.0391,
      "step": 9430
    },
    {
      "balancing_entropy": 0.844,
      "balancing_loss": 1.336,
      "epoch": 48.16384,
      "grad_norm": 0.55078125,
      "learning_rate": 0.0008400000000000001,
      "loss": 2.292,
      "router_z_loss": 0.0518,
      "step": 9440
    },
    {
      "balancing_entropy": 0.8913,
      "balancing_loss": 1.1693,
      "epoch": 48.21504,
      "grad_norm": 0.4921875,
      "learning_rate": 0.0008250000000000001,
      "loss": 2.2929,
      "router_z_loss": 0.0382,
      "step": 9450
    },
    {
      "balancing_entropy": 0.8748,
      "balancing_loss": 1.2917,
      "epoch": 48.26624,
      "grad_norm": 0.38671875,
      "learning_rate": 0.0008100000000000001,
      "loss": 2.2916,
      "router_z_loss": 0.0323,
      "step": 9460
    },
    {
      "balancing_entropy": 0.8322,
      "balancing_loss": 1.2515,
      "epoch": 48.31744,
      "grad_norm": 0.33203125,
      "learning_rate": 0.000795,
      "loss": 2.2917,
      "router_z_loss": 0.033,
      "step": 9470
    },
    {
      "balancing_entropy": 0.9211,
      "balancing_loss": 1.2099,
      "epoch": 48.36864,
      "grad_norm": 0.439453125,
      "learning_rate": 0.0007800000000000001,
      "loss": 2.3038,
      "router_z_loss": 0.0365,
      "step": 9480
    },
    {
      "balancing_entropy": 0.7566,
      "balancing_loss": 1.3362,
      "epoch": 48.41984,
      "grad_norm": 0.60546875,
      "learning_rate": 0.0007650000000000001,
      "loss": 2.2996,
      "router_z_loss": 0.0396,
      "step": 9490
    },
    {
      "balancing_entropy": 0.9477,
      "balancing_loss": 1.167,
      "epoch": 48.47104,
      "grad_norm": 0.29296875,
      "learning_rate": 0.00075,
      "loss": 2.2856,
      "router_z_loss": 0.0553,
      "step": 9500
    },
    {
      "balancing_entropy": 0.7564,
      "balancing_loss": 1.2942,
      "epoch": 48.52224,
      "grad_norm": 0.578125,
      "learning_rate": 0.000735,
      "loss": 2.2869,
      "router_z_loss": 0.0947,
      "step": 9510
    },
    {
      "balancing_entropy": 0.7861,
      "balancing_loss": 1.179,
      "epoch": 48.57344,
      "grad_norm": 0.5,
      "learning_rate": 0.0007199999999999999,
      "loss": 2.2933,
      "router_z_loss": 0.0755,
      "step": 9520
    },
    {
      "balancing_entropy": 0.7025,
      "balancing_loss": 1.3375,
      "epoch": 48.62464,
      "grad_norm": 0.3203125,
      "learning_rate": 0.000705,
      "loss": 2.2822,
      "router_z_loss": 0.0676,
      "step": 9530
    },
    {
      "balancing_entropy": 0.7556,
      "balancing_loss": 1.1704,
      "epoch": 48.67584,
      "grad_norm": 0.55078125,
      "learning_rate": 0.00069,
      "loss": 2.2849,
      "router_z_loss": 0.1073,
      "step": 9540
    },
    {
      "balancing_entropy": 0.8016,
      "balancing_loss": 1.2495,
      "epoch": 48.72704,
      "grad_norm": 0.330078125,
      "learning_rate": 0.0006749999999999999,
      "loss": 2.2864,
      "router_z_loss": 0.06,
      "step": 9550
    },
    {
      "balancing_entropy": 0.9008,
      "balancing_loss": 1.2432,
      "epoch": 48.77824,
      "grad_norm": 0.5390625,
      "learning_rate": 0.0006599999999999999,
      "loss": 2.2887,
      "router_z_loss": 0.0358,
      "step": 9560
    },
    {
      "balancing_entropy": 0.8854,
      "balancing_loss": 1.1747,
      "epoch": 48.82944,
      "grad_norm": 0.5546875,
      "learning_rate": 0.000645,
      "loss": 2.2755,
      "router_z_loss": 0.0251,
      "step": 9570
    },
    {
      "balancing_entropy": 0.9317,
      "balancing_loss": 1.3128,
      "epoch": 48.88064,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0006299999999999999,
      "loss": 2.2953,
      "router_z_loss": 0.0236,
      "step": 9580
    },
    {
      "balancing_entropy": 0.9258,
      "balancing_loss": 1.1714,
      "epoch": 48.93184,
      "grad_norm": 0.66015625,
      "learning_rate": 0.0006149999999999999,
      "loss": 2.2907,
      "router_z_loss": 0.0287,
      "step": 9590
    },
    {
      "balancing_entropy": 0.8287,
      "balancing_loss": 1.3313,
      "epoch": 48.98304,
      "grad_norm": 0.447265625,
      "learning_rate": 0.0005999999999999998,
      "loss": 2.2902,
      "router_z_loss": 0.0257,
      "step": 9600
    },
    {
      "balancing_entropy": 0.8069,
      "balancing_loss": 1.1782,
      "epoch": 49.03072,
      "grad_norm": 0.39453125,
      "learning_rate": 0.0005849999999999999,
      "loss": 2.2829,
      "router_z_loss": 0.0519,
      "step": 9610
    },
    {
      "balancing_entropy": 0.8963,
      "balancing_loss": 1.1816,
      "epoch": 49.08192,
      "grad_norm": 0.6328125,
      "learning_rate": 0.0005699999999999999,
      "loss": 2.2822,
      "router_z_loss": 0.0295,
      "step": 9620
    },
    {
      "balancing_entropy": 0.7805,
      "balancing_loss": 1.341,
      "epoch": 49.13312,
      "grad_norm": 0.515625,
      "learning_rate": 0.0005550000000000002,
      "loss": 2.2768,
      "router_z_loss": 0.0314,
      "step": 9630
    },
    {
      "balancing_entropy": 0.8128,
      "balancing_loss": 1.1728,
      "epoch": 49.18432,
      "grad_norm": 0.546875,
      "learning_rate": 0.0005400000000000001,
      "loss": 2.2777,
      "router_z_loss": 0.0368,
      "step": 9640
    },
    {
      "balancing_entropy": 0.772,
      "balancing_loss": 1.3542,
      "epoch": 49.23552,
      "grad_norm": 0.5703125,
      "learning_rate": 0.0005250000000000002,
      "loss": 2.2718,
      "router_z_loss": 0.0285,
      "step": 9650
    },
    {
      "balancing_entropy": 0.8671,
      "balancing_loss": 1.1715,
      "epoch": 49.28672,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0005100000000000001,
      "loss": 2.2713,
      "router_z_loss": 0.0353,
      "step": 9660
    },
    {
      "balancing_entropy": 0.8725,
      "balancing_loss": 1.175,
      "epoch": 49.33792,
      "grad_norm": 0.515625,
      "learning_rate": 0.0004950000000000001,
      "loss": 2.2787,
      "router_z_loss": 0.0194,
      "step": 9670
    },
    {
      "balancing_entropy": 0.9204,
      "balancing_loss": 1.3564,
      "epoch": 49.38912,
      "grad_norm": 0.5859375,
      "learning_rate": 0.0004800000000000001,
      "loss": 2.2743,
      "router_z_loss": 0.0217,
      "step": 9680
    },
    {
      "balancing_entropy": 0.8554,
      "balancing_loss": 1.1729,
      "epoch": 49.44032,
      "grad_norm": 0.44140625,
      "learning_rate": 0.0004650000000000001,
      "loss": 2.2746,
      "router_z_loss": 0.0305,
      "step": 9690
    },
    {
      "balancing_entropy": 0.8614,
      "balancing_loss": 1.2042,
      "epoch": 49.49152,
      "grad_norm": 0.255859375,
      "learning_rate": 0.0004500000000000001,
      "loss": 2.2699,
      "router_z_loss": 0.019,
      "step": 9700
    },
    {
      "balancing_entropy": 0.9288,
      "balancing_loss": 1.3517,
      "epoch": 49.54272,
      "grad_norm": 0.5546875,
      "learning_rate": 0.00043500000000000006,
      "loss": 2.2757,
      "router_z_loss": 0.024,
      "step": 9710
    },
    {
      "balancing_entropy": 0.9421,
      "balancing_loss": 1.1711,
      "epoch": 49.59392,
      "grad_norm": 0.32421875,
      "learning_rate": 0.00042000000000000007,
      "loss": 2.2651,
      "router_z_loss": 0.0334,
      "step": 9720
    },
    {
      "balancing_entropy": 0.8111,
      "balancing_loss": 1.1803,
      "epoch": 49.64512,
      "grad_norm": 0.52734375,
      "learning_rate": 0.00040500000000000003,
      "loss": 2.2689,
      "router_z_loss": 0.0198,
      "step": 9730
    },
    {
      "balancing_entropy": 0.8269,
      "balancing_loss": 1.3019,
      "epoch": 49.69632,
      "grad_norm": 0.333984375,
      "learning_rate": 0.00039000000000000005,
      "loss": 2.2819,
      "router_z_loss": 0.0165,
      "step": 9740
    },
    {
      "balancing_entropy": 0.8636,
      "balancing_loss": 1.2742,
      "epoch": 49.74752,
      "grad_norm": 0.4453125,
      "learning_rate": 0.000375,
      "loss": 2.2713,
      "router_z_loss": 0.0321,
      "step": 9750
    },
    {
      "balancing_entropy": 0.8321,
      "balancing_loss": 1.1731,
      "epoch": 49.79872,
      "grad_norm": 0.53125,
      "learning_rate": 0.00035999999999999997,
      "loss": 2.2636,
      "router_z_loss": 0.023,
      "step": 9760
    },
    {
      "balancing_entropy": 0.8978,
      "balancing_loss": 1.1841,
      "epoch": 49.84992,
      "grad_norm": 0.65234375,
      "learning_rate": 0.000345,
      "loss": 2.2685,
      "router_z_loss": 0.0169,
      "step": 9770
    },
    {
      "balancing_entropy": 0.761,
      "balancing_loss": 1.1895,
      "epoch": 49.90112,
      "grad_norm": 0.67578125,
      "learning_rate": 0.00032999999999999994,
      "loss": 2.2703,
      "router_z_loss": 0.0157,
      "step": 9780
    },
    {
      "balancing_entropy": 0.8607,
      "balancing_loss": 1.1839,
      "epoch": 49.95232,
      "grad_norm": 0.2265625,
      "learning_rate": 0.00031499999999999996,
      "loss": 2.2673,
      "router_z_loss": 0.0156,
      "step": 9790
    },
    {
      "balancing_entropy": 0.7363,
      "balancing_loss": 1.1928,
      "epoch": 50.0,
      "grad_norm": 0.54296875,
      "learning_rate": 0.0002999999999999999,
      "loss": 2.2713,
      "router_z_loss": 0.0161,
      "step": 9800
    },
    {
      "balancing_entropy": 0.9795,
      "balancing_loss": 1.182,
      "epoch": 50.0512,
      "grad_norm": 0.201171875,
      "learning_rate": 0.00028499999999999993,
      "loss": 2.2657,
      "router_z_loss": 0.0164,
      "step": 9810
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.1916,
      "epoch": 50.1024,
      "grad_norm": 0.1669921875,
      "learning_rate": 0.0002699999999999999,
      "loss": 2.2583,
      "router_z_loss": 0.017,
      "step": 9820
    },
    {
      "balancing_entropy": 1.0843,
      "balancing_loss": 1.1877,
      "epoch": 50.1536,
      "grad_norm": 0.1806640625,
      "learning_rate": 0.0002549999999999999,
      "loss": 2.2661,
      "router_z_loss": 0.0155,
      "step": 9830
    },
    {
      "balancing_entropy": 1.0868,
      "balancing_loss": 1.1892,
      "epoch": 50.2048,
      "grad_norm": 0.1396484375,
      "learning_rate": 0.0002399999999999999,
      "loss": 2.2657,
      "router_z_loss": 0.0147,
      "step": 9840
    },
    {
      "balancing_entropy": 1.0922,
      "balancing_loss": 1.1956,
      "epoch": 50.256,
      "grad_norm": 0.1337890625,
      "learning_rate": 0.00022499999999999986,
      "loss": 2.2712,
      "router_z_loss": 0.014,
      "step": 9850
    },
    {
      "balancing_entropy": 1.0919,
      "balancing_loss": 1.1956,
      "epoch": 50.3072,
      "grad_norm": 0.154296875,
      "learning_rate": 0.00020999999999999985,
      "loss": 2.2601,
      "router_z_loss": 0.0139,
      "step": 9860
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.1894,
      "epoch": 50.3584,
      "grad_norm": 0.1689453125,
      "learning_rate": 0.00019499999999999983,
      "loss": 2.2719,
      "router_z_loss": 0.0133,
      "step": 9870
    },
    {
      "balancing_entropy": 1.0911,
      "balancing_loss": 1.1926,
      "epoch": 50.4096,
      "grad_norm": 0.1318359375,
      "learning_rate": 0.00018000000000000017,
      "loss": 2.2666,
      "router_z_loss": 0.0138,
      "step": 9880
    },
    {
      "balancing_entropy": 1.0923,
      "balancing_loss": 1.1953,
      "epoch": 50.4608,
      "grad_norm": 0.1337890625,
      "learning_rate": 0.00016500000000000016,
      "loss": 2.2696,
      "router_z_loss": 0.0136,
      "step": 9890
    },
    {
      "balancing_entropy": 1.0915,
      "balancing_loss": 1.1939,
      "epoch": 50.512,
      "grad_norm": 0.1376953125,
      "learning_rate": 0.00015000000000000015,
      "loss": 2.2629,
      "router_z_loss": 0.0131,
      "step": 9900
    },
    {
      "balancing_entropy": 1.0919,
      "balancing_loss": 1.1941,
      "epoch": 50.5632,
      "grad_norm": 0.126953125,
      "learning_rate": 0.0001350000000000001,
      "loss": 2.2637,
      "router_z_loss": 0.0132,
      "step": 9910
    },
    {
      "balancing_entropy": 1.0918,
      "balancing_loss": 1.1932,
      "epoch": 50.6144,
      "grad_norm": 0.1259765625,
      "learning_rate": 0.00012000000000000011,
      "loss": 2.2597,
      "router_z_loss": 0.0135,
      "step": 9920
    },
    {
      "balancing_entropy": 1.0926,
      "balancing_loss": 1.1971,
      "epoch": 50.6656,
      "grad_norm": 0.130859375,
      "learning_rate": 0.0001050000000000001,
      "loss": 2.2567,
      "router_z_loss": 0.0131,
      "step": 9930
    },
    {
      "balancing_entropy": 1.0925,
      "balancing_loss": 1.1968,
      "epoch": 50.7168,
      "grad_norm": 0.142578125,
      "learning_rate": 9.000000000000009e-05,
      "loss": 2.2656,
      "router_z_loss": 0.013,
      "step": 9940
    },
    {
      "balancing_entropy": 1.0922,
      "balancing_loss": 1.1971,
      "epoch": 50.768,
      "grad_norm": 0.12890625,
      "learning_rate": 7.500000000000007e-05,
      "loss": 2.2574,
      "router_z_loss": 0.013,
      "step": 9950
    },
    {
      "balancing_entropy": 1.0918,
      "balancing_loss": 1.1954,
      "epoch": 50.8192,
      "grad_norm": 0.12890625,
      "learning_rate": 6.0000000000000056e-05,
      "loss": 2.2584,
      "router_z_loss": 0.013,
      "step": 9960
    },
    {
      "balancing_entropy": 1.0918,
      "balancing_loss": 1.194,
      "epoch": 50.8704,
      "grad_norm": 0.130859375,
      "learning_rate": 4.5000000000000043e-05,
      "loss": 2.2738,
      "router_z_loss": 0.013,
      "step": 9970
    },
    {
      "balancing_entropy": 1.0922,
      "balancing_loss": 1.1952,
      "epoch": 50.9216,
      "grad_norm": 0.12890625,
      "learning_rate": 3.0000000000000028e-05,
      "loss": 2.271,
      "router_z_loss": 0.013,
      "step": 9980
    },
    {
      "balancing_entropy": 1.0922,
      "balancing_loss": 1.1955,
      "epoch": 50.9728,
      "grad_norm": 0.1376953125,
      "learning_rate": 1.5000000000000014e-05,
      "loss": 2.2595,
      "router_z_loss": 0.0129,
      "step": 9990
    },
    {
      "balancing_entropy": 1.0537,
      "balancing_loss": 1.1973,
      "epoch": 51.02048,
      "grad_norm": 0.1328125,
      "learning_rate": 0.0,
      "loss": 2.264,
      "router_z_loss": 0.0128,
      "step": 10000
    }
  ],
  "logging_steps": 10,
  "max_steps": 10000,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 52,
  "save_steps": 2500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 6.208056806237798e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}