{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 25.512,
  "eval_steps": 500,
  "global_step": 5000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "balancing_entropy": 1.0907,
      "balancing_loss": 1.1417,
      "epoch": 0.0512,
      "grad_norm": 1.015625,
      "learning_rate": 3.076923076923077e-05,
      "loss": 12.4636,
      "router_z_loss": 1.3718,
      "step": 10
    },
    {
      "balancing_entropy": 1.0804,
      "balancing_loss": 1.1423,
      "epoch": 0.1024,
      "grad_norm": 0.91015625,
      "learning_rate": 6.153846153846154e-05,
      "loss": 12.2249,
      "router_z_loss": 1.3676,
      "step": 20
    },
    {
      "balancing_entropy": 1.0952,
      "balancing_loss": 1.1429,
      "epoch": 0.1536,
      "grad_norm": 0.8515625,
      "learning_rate": 9.230769230769232e-05,
      "loss": 11.7705,
      "router_z_loss": 1.3434,
      "step": 30
    },
    {
      "balancing_entropy": 1.0779,
      "balancing_loss": 1.144,
      "epoch": 0.2048,
      "grad_norm": 0.828125,
      "learning_rate": 0.00012307692307692307,
      "loss": 11.111,
      "router_z_loss": 1.2754,
      "step": 40
    },
    {
      "balancing_entropy": 1.0905,
      "balancing_loss": 1.1422,
      "epoch": 0.256,
      "grad_norm": 1.046875,
      "learning_rate": 0.00015384615384615385,
      "loss": 10.1841,
      "router_z_loss": 1.2004,
      "step": 50
    },
    {
      "balancing_entropy": 1.0915,
      "balancing_loss": 1.1419,
      "epoch": 0.3072,
      "grad_norm": 0.63671875,
      "learning_rate": 0.00018461538461538463,
      "loss": 9.2971,
      "router_z_loss": 1.1388,
      "step": 60
    },
    {
      "balancing_entropy": 1.0865,
      "balancing_loss": 1.1409,
      "epoch": 0.3584,
      "grad_norm": 0.73828125,
      "learning_rate": 0.0002153846153846154,
      "loss": 8.6456,
      "router_z_loss": 1.1099,
      "step": 70
    },
    {
      "balancing_entropy": 1.0621,
      "balancing_loss": 1.1474,
      "epoch": 0.4096,
      "grad_norm": 2.71875,
      "learning_rate": 0.00024615384615384614,
      "loss": 8.2519,
      "router_z_loss": 1.0139,
      "step": 80
    },
    {
      "balancing_entropy": 1.0621,
      "balancing_loss": 1.1455,
      "epoch": 0.4608,
      "grad_norm": 1.0703125,
      "learning_rate": 0.00027692307692307695,
      "loss": 7.9354,
      "router_z_loss": 0.8722,
      "step": 90
    },
    {
      "balancing_entropy": 1.0783,
      "balancing_loss": 1.1375,
      "epoch": 0.512,
      "grad_norm": 0.765625,
      "learning_rate": 0.0003076923076923077,
      "loss": 7.5548,
      "router_z_loss": 0.7341,
      "step": 100
    },
    {
      "balancing_entropy": 1.0898,
      "balancing_loss": 1.1312,
      "epoch": 0.5632,
      "grad_norm": 0.91015625,
      "learning_rate": 0.00033846153846153846,
      "loss": 7.0768,
      "router_z_loss": 0.6111,
      "step": 110
    },
    {
      "balancing_entropy": 1.0923,
      "balancing_loss": 1.131,
      "epoch": 0.6144,
      "grad_norm": 0.65234375,
      "learning_rate": 0.00036923076923076927,
      "loss": 6.5868,
      "router_z_loss": 0.4137,
      "step": 120
    },
    {
      "balancing_entropy": 1.0932,
      "balancing_loss": 1.1305,
      "epoch": 0.6656,
      "grad_norm": 0.458984375,
      "learning_rate": 0.0004,
      "loss": 6.2033,
      "router_z_loss": 0.2772,
      "step": 130
    },
    {
      "balancing_entropy": 1.0953,
      "balancing_loss": 1.1305,
      "epoch": 0.7168,
      "grad_norm": 1.421875,
      "learning_rate": 0.0004307692307692308,
      "loss": 5.9507,
      "router_z_loss": 0.2233,
      "step": 140
    },
    {
      "balancing_entropy": 1.0941,
      "balancing_loss": 1.1261,
      "epoch": 0.768,
      "grad_norm": 1.0234375,
      "learning_rate": 0.0004615384615384616,
      "loss": 5.7447,
      "router_z_loss": 0.1919,
      "step": 150
    },
    {
      "balancing_entropy": 1.0942,
      "balancing_loss": 1.1258,
      "epoch": 0.8192,
      "grad_norm": 0.625,
      "learning_rate": 0.0004923076923076923,
      "loss": 5.5495,
      "router_z_loss": 0.14,
      "step": 160
    },
    {
      "balancing_entropy": 1.0947,
      "balancing_loss": 1.1259,
      "epoch": 0.8704,
      "grad_norm": 0.61328125,
      "learning_rate": 0.0005230769230769231,
      "loss": 5.3746,
      "router_z_loss": 0.1123,
      "step": 170
    },
    {
      "balancing_entropy": 1.0951,
      "balancing_loss": 1.1255,
      "epoch": 0.9216,
      "grad_norm": 1.1328125,
      "learning_rate": 0.0005538461538461539,
      "loss": 5.2159,
      "router_z_loss": 0.0839,
      "step": 180
    },
    {
      "balancing_entropy": 1.0916,
      "balancing_loss": 1.1175,
      "epoch": 0.9728,
      "grad_norm": 1.4140625,
      "learning_rate": 0.0005846153846153846,
      "loss": 5.0335,
      "router_z_loss": 0.071,
      "step": 190
    },
    {
      "balancing_entropy": 1.0491,
      "balancing_loss": 1.1153,
      "epoch": 1.02048,
      "grad_norm": 0.7578125,
      "learning_rate": 0.0006153846153846154,
      "loss": 4.9021,
      "router_z_loss": 0.064,
      "step": 200
    },
    {
      "balancing_entropy": 1.0888,
      "balancing_loss": 1.1112,
      "epoch": 1.07168,
      "grad_norm": 1.046875,
      "learning_rate": 0.0006461538461538462,
      "loss": 4.7524,
      "router_z_loss": 0.0585,
      "step": 210
    },
    {
      "balancing_entropy": 1.0875,
      "balancing_loss": 1.1083,
      "epoch": 1.12288,
      "grad_norm": 1.2109375,
      "learning_rate": 0.0006769230769230769,
      "loss": 4.6408,
      "router_z_loss": 0.0605,
      "step": 220
    },
    {
      "balancing_entropy": 1.0884,
      "balancing_loss": 1.1063,
      "epoch": 1.17408,
      "grad_norm": 0.9453125,
      "learning_rate": 0.0007076923076923077,
      "loss": 4.53,
      "router_z_loss": 0.0596,
      "step": 230
    },
    {
      "balancing_entropy": 1.0853,
      "balancing_loss": 1.1053,
      "epoch": 1.22528,
      "grad_norm": 1.0546875,
      "learning_rate": 0.0007384615384615385,
      "loss": 4.3935,
      "router_z_loss": 0.054,
      "step": 240
    },
    {
      "balancing_entropy": 1.0895,
      "balancing_loss": 1.1085,
      "epoch": 1.27648,
      "grad_norm": 1.0390625,
      "learning_rate": 0.0007692307692307692,
      "loss": 4.3076,
      "router_z_loss": 0.064,
      "step": 250
    },
    {
      "balancing_entropy": 1.0829,
      "balancing_loss": 1.1033,
      "epoch": 1.32768,
      "grad_norm": 0.7265625,
      "learning_rate": 0.0008,
      "loss": 4.2385,
      "router_z_loss": 0.0581,
      "step": 260
    },
    {
      "balancing_entropy": 1.0855,
      "balancing_loss": 1.1075,
      "epoch": 1.37888,
      "grad_norm": 0.6484375,
      "learning_rate": 0.0008307692307692308,
      "loss": 4.1286,
      "router_z_loss": 0.0563,
      "step": 270
    },
    {
      "balancing_entropy": 1.0869,
      "balancing_loss": 1.0988,
      "epoch": 1.43008,
      "grad_norm": 0.734375,
      "learning_rate": 0.0008615384615384615,
      "loss": 4.0637,
      "router_z_loss": 0.0596,
      "step": 280
    },
    {
      "balancing_entropy": 1.0885,
      "balancing_loss": 1.1019,
      "epoch": 1.48128,
      "grad_norm": 1.015625,
      "learning_rate": 0.0008923076923076924,
      "loss": 3.9922,
      "router_z_loss": 0.0532,
      "step": 290
    },
    {
      "balancing_entropy": 1.084,
      "balancing_loss": 1.103,
      "epoch": 1.53248,
      "grad_norm": 0.9765625,
      "learning_rate": 0.0009230769230769232,
      "loss": 3.9192,
      "router_z_loss": 0.0501,
      "step": 300
    },
    {
      "balancing_entropy": 1.086,
      "balancing_loss": 1.0974,
      "epoch": 1.58368,
      "grad_norm": 0.57421875,
      "learning_rate": 0.0009538461538461538,
      "loss": 3.8677,
      "router_z_loss": 0.0553,
      "step": 310
    },
    {
      "balancing_entropy": 1.0875,
      "balancing_loss": 1.1006,
      "epoch": 1.6348799999999999,
      "grad_norm": 0.78125,
      "learning_rate": 0.0009846153846153846,
      "loss": 3.8262,
      "router_z_loss": 0.0528,
      "step": 320
    },
    {
      "balancing_entropy": 1.0874,
      "balancing_loss": 1.1033,
      "epoch": 1.68608,
      "grad_norm": 0.53515625,
      "learning_rate": 0.0010153846153846155,
      "loss": 3.7606,
      "router_z_loss": 0.053,
      "step": 330
    },
    {
      "balancing_entropy": 1.0766,
      "balancing_loss": 1.1084,
      "epoch": 1.7372800000000002,
      "grad_norm": 0.7890625,
      "learning_rate": 0.0010461538461538462,
      "loss": 3.6967,
      "router_z_loss": 0.0483,
      "step": 340
    },
    {
      "balancing_entropy": 1.0786,
      "balancing_loss": 1.0986,
      "epoch": 1.7884799999999998,
      "grad_norm": 0.80078125,
      "learning_rate": 0.0010769230769230769,
      "loss": 3.684,
      "router_z_loss": 0.0669,
      "step": 350
    },
    {
      "balancing_entropy": 1.0872,
      "balancing_loss": 1.0975,
      "epoch": 1.83968,
      "grad_norm": 0.75,
      "learning_rate": 0.0011076923076923078,
      "loss": 3.6041,
      "router_z_loss": 0.0526,
      "step": 360
    },
    {
      "balancing_entropy": 1.0866,
      "balancing_loss": 1.106,
      "epoch": 1.8908800000000001,
      "grad_norm": 0.88671875,
      "learning_rate": 0.0011384615384615385,
      "loss": 3.5964,
      "router_z_loss": 0.0513,
      "step": 370
    },
    {
      "balancing_entropy": 1.0801,
      "balancing_loss": 1.0959,
      "epoch": 1.94208,
      "grad_norm": 0.453125,
      "learning_rate": 0.0011692307692307692,
      "loss": 3.5742,
      "router_z_loss": 0.0519,
      "step": 380
    },
    {
      "balancing_entropy": 1.0785,
      "balancing_loss": 1.1051,
      "epoch": 1.99328,
      "grad_norm": 0.65234375,
      "learning_rate": 0.0012000000000000001,
      "loss": 3.5006,
      "router_z_loss": 0.0563,
      "step": 390
    },
    {
      "balancing_entropy": 1.0483,
      "balancing_loss": 1.0966,
      "epoch": 2.04096,
      "grad_norm": 1.0390625,
      "learning_rate": 0.0012307692307692308,
      "loss": 3.4658,
      "router_z_loss": 0.0533,
      "step": 400
    },
    {
      "balancing_entropy": 1.0882,
      "balancing_loss": 1.1001,
      "epoch": 2.09216,
      "grad_norm": 0.4609375,
      "learning_rate": 0.0012615384615384615,
      "loss": 3.4427,
      "router_z_loss": 0.0443,
      "step": 410
    },
    {
      "balancing_entropy": 1.0891,
      "balancing_loss": 1.0997,
      "epoch": 2.14336,
      "grad_norm": 0.76171875,
      "learning_rate": 0.0012923076923076924,
      "loss": 3.4206,
      "router_z_loss": 0.0417,
      "step": 420
    },
    {
      "balancing_entropy": 1.088,
      "balancing_loss": 1.0989,
      "epoch": 2.19456,
      "grad_norm": 0.8125,
      "learning_rate": 0.0013230769230769231,
      "loss": 3.3869,
      "router_z_loss": 0.0436,
      "step": 430
    },
    {
      "balancing_entropy": 1.0683,
      "balancing_loss": 1.1073,
      "epoch": 2.24576,
      "grad_norm": 0.62890625,
      "learning_rate": 0.0013538461538461538,
      "loss": 3.3571,
      "router_z_loss": 0.0491,
      "step": 440
    },
    {
      "balancing_entropy": 1.0751,
      "balancing_loss": 1.0967,
      "epoch": 2.29696,
      "grad_norm": 0.55859375,
      "learning_rate": 0.0013846153846153847,
      "loss": 3.3558,
      "router_z_loss": 0.053,
      "step": 450
    },
    {
      "balancing_entropy": 1.0669,
      "balancing_loss": 1.1059,
      "epoch": 2.34816,
      "grad_norm": 0.609375,
      "learning_rate": 0.0014153846153846154,
      "loss": 3.3114,
      "router_z_loss": 0.0561,
      "step": 460
    },
    {
      "balancing_entropy": 1.0747,
      "balancing_loss": 1.0934,
      "epoch": 2.39936,
      "grad_norm": 0.490234375,
      "learning_rate": 0.0014461538461538461,
      "loss": 3.3037,
      "router_z_loss": 0.0516,
      "step": 470
    },
    {
      "balancing_entropy": 1.0815,
      "balancing_loss": 1.0988,
      "epoch": 2.45056,
      "grad_norm": 0.66796875,
      "learning_rate": 0.001476923076923077,
      "loss": 3.2948,
      "router_z_loss": 0.0525,
      "step": 480
    },
    {
      "balancing_entropy": 1.0909,
      "balancing_loss": 1.1038,
      "epoch": 2.50176,
      "grad_norm": 0.49609375,
      "learning_rate": 0.0015076923076923078,
      "loss": 3.2504,
      "router_z_loss": 0.0423,
      "step": 490
    },
    {
      "balancing_entropy": 1.0772,
      "balancing_loss": 1.1019,
      "epoch": 2.55296,
      "grad_norm": 0.4921875,
      "learning_rate": 0.0015384615384615385,
      "loss": 3.2454,
      "router_z_loss": 0.0377,
      "step": 500
    },
    {
      "balancing_entropy": 1.082,
      "balancing_loss": 1.0907,
      "epoch": 2.6041600000000003,
      "grad_norm": 0.7578125,
      "learning_rate": 0.0015692307692307694,
      "loss": 3.2209,
      "router_z_loss": 0.0374,
      "step": 510
    },
    {
      "balancing_entropy": 1.0875,
      "balancing_loss": 1.104,
      "epoch": 2.65536,
      "grad_norm": 0.443359375,
      "learning_rate": 0.0016,
      "loss": 3.2235,
      "router_z_loss": 0.0341,
      "step": 520
    },
    {
      "balancing_entropy": 1.0829,
      "balancing_loss": 1.1012,
      "epoch": 2.70656,
      "grad_norm": 0.8359375,
      "learning_rate": 0.0016307692307692308,
      "loss": 3.1925,
      "router_z_loss": 0.0327,
      "step": 530
    },
    {
      "balancing_entropy": 1.0765,
      "balancing_loss": 1.1007,
      "epoch": 2.75776,
      "grad_norm": 0.326171875,
      "learning_rate": 0.0016615384615384617,
      "loss": 3.2021,
      "router_z_loss": 0.0391,
      "step": 540
    },
    {
      "balancing_entropy": 1.0868,
      "balancing_loss": 1.0972,
      "epoch": 2.80896,
      "grad_norm": 0.4375,
      "learning_rate": 0.0016923076923076924,
      "loss": 3.1871,
      "router_z_loss": 0.0459,
      "step": 550
    },
    {
      "balancing_entropy": 1.0733,
      "balancing_loss": 1.0959,
      "epoch": 2.86016,
      "grad_norm": 0.48046875,
      "learning_rate": 0.001723076923076923,
      "loss": 3.1595,
      "router_z_loss": 0.0345,
      "step": 560
    },
    {
      "balancing_entropy": 1.077,
      "balancing_loss": 1.1003,
      "epoch": 2.91136,
      "grad_norm": 0.482421875,
      "learning_rate": 0.001753846153846154,
      "loss": 3.1634,
      "router_z_loss": 0.0334,
      "step": 570
    },
    {
      "balancing_entropy": 1.088,
      "balancing_loss": 1.0995,
      "epoch": 2.96256,
      "grad_norm": 0.390625,
      "learning_rate": 0.0017846153846153847,
      "loss": 3.1342,
      "router_z_loss": 0.0411,
      "step": 580
    },
    {
      "balancing_entropy": 1.0335,
      "balancing_loss": 1.1026,
      "epoch": 3.01024,
      "grad_norm": 0.58984375,
      "learning_rate": 0.0018153846153846154,
      "loss": 3.13,
      "router_z_loss": 0.0393,
      "step": 590
    },
    {
      "balancing_entropy": 1.0877,
      "balancing_loss": 1.1057,
      "epoch": 3.06144,
      "grad_norm": 0.37109375,
      "learning_rate": 0.0018461538461538463,
      "loss": 3.1095,
      "router_z_loss": 0.0361,
      "step": 600
    },
    {
      "balancing_entropy": 1.0627,
      "balancing_loss": 1.0867,
      "epoch": 3.11264,
      "grad_norm": 0.5859375,
      "learning_rate": 0.001876923076923077,
      "loss": 3.1136,
      "router_z_loss": 0.0396,
      "step": 610
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.1135,
      "epoch": 3.16384,
      "grad_norm": 0.294921875,
      "learning_rate": 0.0019076923076923075,
      "loss": 3.0938,
      "router_z_loss": 0.0543,
      "step": 620
    },
    {
      "balancing_entropy": 1.0881,
      "balancing_loss": 1.092,
      "epoch": 3.21504,
      "grad_norm": 0.40234375,
      "learning_rate": 0.0019384615384615386,
      "loss": 3.0719,
      "router_z_loss": 0.0355,
      "step": 630
    },
    {
      "balancing_entropy": 1.08,
      "balancing_loss": 1.0957,
      "epoch": 3.26624,
      "grad_norm": 0.443359375,
      "learning_rate": 0.001969230769230769,
      "loss": 3.0781,
      "router_z_loss": 0.033,
      "step": 640
    },
    {
      "balancing_entropy": 1.0757,
      "balancing_loss": 1.0982,
      "epoch": 3.31744,
      "grad_norm": 0.443359375,
      "learning_rate": 0.002,
      "loss": 3.0823,
      "router_z_loss": 0.0554,
      "step": 650
    },
    {
      "balancing_entropy": 1.0748,
      "balancing_loss": 1.0967,
      "epoch": 3.36864,
      "grad_norm": 0.30078125,
      "learning_rate": 0.002030769230769231,
      "loss": 3.0585,
      "router_z_loss": 0.0474,
      "step": 660
    },
    {
      "balancing_entropy": 1.0867,
      "balancing_loss": 1.0972,
      "epoch": 3.4198399999999998,
      "grad_norm": 0.37109375,
      "learning_rate": 0.0020615384615384614,
      "loss": 3.0622,
      "router_z_loss": 0.0297,
      "step": 670
    },
    {
      "balancing_entropy": 1.0796,
      "balancing_loss": 1.1021,
      "epoch": 3.47104,
      "grad_norm": 0.5234375,
      "learning_rate": 0.0020923076923076924,
      "loss": 3.0576,
      "router_z_loss": 0.033,
      "step": 680
    },
    {
      "balancing_entropy": 1.0829,
      "balancing_loss": 1.096,
      "epoch": 3.52224,
      "grad_norm": 0.345703125,
      "learning_rate": 0.0021230769230769233,
      "loss": 3.0549,
      "router_z_loss": 0.0452,
      "step": 690
    },
    {
      "balancing_entropy": 1.0569,
      "balancing_loss": 1.1113,
      "epoch": 3.5734399999999997,
      "grad_norm": 0.58203125,
      "learning_rate": 0.0021538461538461538,
      "loss": 3.0211,
      "router_z_loss": 0.0321,
      "step": 700
    },
    {
      "balancing_entropy": 1.0693,
      "balancing_loss": 1.0886,
      "epoch": 3.62464,
      "grad_norm": 0.498046875,
      "learning_rate": 0.0021846153846153847,
      "loss": 3.0315,
      "router_z_loss": 0.0772,
      "step": 710
    },
    {
      "balancing_entropy": 1.0871,
      "balancing_loss": 1.1029,
      "epoch": 3.67584,
      "grad_norm": 0.2734375,
      "learning_rate": 0.0022153846153846156,
      "loss": 3.0335,
      "router_z_loss": 0.0542,
      "step": 720
    },
    {
      "balancing_entropy": 1.0822,
      "balancing_loss": 1.0957,
      "epoch": 3.72704,
      "grad_norm": 0.412109375,
      "learning_rate": 0.002246153846153846,
      "loss": 3.023,
      "router_z_loss": 0.0253,
      "step": 730
    },
    {
      "balancing_entropy": 1.0735,
      "balancing_loss": 1.095,
      "epoch": 3.7782400000000003,
      "grad_norm": 0.35546875,
      "learning_rate": 0.002276923076923077,
      "loss": 3.0195,
      "router_z_loss": 0.0519,
      "step": 740
    },
    {
      "balancing_entropy": 1.0694,
      "balancing_loss": 1.1025,
      "epoch": 3.82944,
      "grad_norm": 0.4140625,
      "learning_rate": 0.002307692307692308,
      "loss": 3.0002,
      "router_z_loss": 0.0451,
      "step": 750
    },
    {
      "balancing_entropy": 1.0771,
      "balancing_loss": 1.0982,
      "epoch": 3.88064,
      "grad_norm": 0.328125,
      "learning_rate": 0.0023384615384615384,
      "loss": 3.0,
      "router_z_loss": 0.0436,
      "step": 760
    },
    {
      "balancing_entropy": 1.0773,
      "balancing_loss": 1.0907,
      "epoch": 3.9318400000000002,
      "grad_norm": 0.30859375,
      "learning_rate": 0.0023692307692307693,
      "loss": 2.989,
      "router_z_loss": 0.0405,
      "step": 770
    },
    {
      "balancing_entropy": 1.0698,
      "balancing_loss": 1.112,
      "epoch": 3.98304,
      "grad_norm": 0.337890625,
      "learning_rate": 0.0024000000000000002,
      "loss": 2.9775,
      "router_z_loss": 0.0363,
      "step": 780
    },
    {
      "balancing_entropy": 1.0339,
      "balancing_loss": 1.0875,
      "epoch": 4.03072,
      "grad_norm": 0.44140625,
      "learning_rate": 0.0024307692307692307,
      "loss": 2.9959,
      "router_z_loss": 0.0587,
      "step": 790
    },
    {
      "balancing_entropy": 1.0776,
      "balancing_loss": 1.0943,
      "epoch": 4.08192,
      "grad_norm": 0.267578125,
      "learning_rate": 0.0024615384615384616,
      "loss": 2.965,
      "router_z_loss": 0.0367,
      "step": 800
    },
    {
      "balancing_entropy": 1.0624,
      "balancing_loss": 1.1087,
      "epoch": 4.13312,
      "grad_norm": 0.373046875,
      "learning_rate": 0.0024923076923076925,
      "loss": 2.9649,
      "router_z_loss": 0.0522,
      "step": 810
    },
    {
      "balancing_entropy": 1.0813,
      "balancing_loss": 1.1016,
      "epoch": 4.18432,
      "grad_norm": 0.287109375,
      "learning_rate": 0.002523076923076923,
      "loss": 2.953,
      "router_z_loss": 0.055,
      "step": 820
    },
    {
      "balancing_entropy": 1.0713,
      "balancing_loss": 1.1215,
      "epoch": 4.23552,
      "grad_norm": 0.33203125,
      "learning_rate": 0.002553846153846154,
      "loss": 2.9571,
      "router_z_loss": 0.068,
      "step": 830
    },
    {
      "balancing_entropy": 1.0691,
      "balancing_loss": 1.0899,
      "epoch": 4.28672,
      "grad_norm": 0.30078125,
      "learning_rate": 0.002584615384615385,
      "loss": 2.9438,
      "router_z_loss": 0.0722,
      "step": 840
    },
    {
      "balancing_entropy": 1.0766,
      "balancing_loss": 1.1022,
      "epoch": 4.33792,
      "grad_norm": 0.4140625,
      "learning_rate": 0.0026153846153846153,
      "loss": 2.959,
      "router_z_loss": 0.0367,
      "step": 850
    },
    {
      "balancing_entropy": 1.0748,
      "balancing_loss": 1.0939,
      "epoch": 4.38912,
      "grad_norm": 0.369140625,
      "learning_rate": 0.0026461538461538463,
      "loss": 2.9531,
      "router_z_loss": 0.0574,
      "step": 860
    },
    {
      "balancing_entropy": 1.0741,
      "balancing_loss": 1.0957,
      "epoch": 4.44032,
      "grad_norm": 0.30078125,
      "learning_rate": 0.002676923076923077,
      "loss": 2.9278,
      "router_z_loss": 0.067,
      "step": 870
    },
    {
      "balancing_entropy": 1.0788,
      "balancing_loss": 1.0995,
      "epoch": 4.49152,
      "grad_norm": 0.376953125,
      "learning_rate": 0.0027076923076923077,
      "loss": 2.9105,
      "router_z_loss": 0.0408,
      "step": 880
    },
    {
      "balancing_entropy": 1.0742,
      "balancing_loss": 1.0967,
      "epoch": 4.54272,
      "grad_norm": 0.36328125,
      "learning_rate": 0.0027384615384615386,
      "loss": 2.9266,
      "router_z_loss": 0.0403,
      "step": 890
    },
    {
      "balancing_entropy": 1.0653,
      "balancing_loss": 1.095,
      "epoch": 4.59392,
      "grad_norm": 0.2470703125,
      "learning_rate": 0.0027692307692307695,
      "loss": 2.9362,
      "router_z_loss": 0.0709,
      "step": 900
    },
    {
      "balancing_entropy": 1.0863,
      "balancing_loss": 1.115,
      "epoch": 4.64512,
      "grad_norm": 0.29296875,
      "learning_rate": 0.0028,
      "loss": 2.9221,
      "router_z_loss": 0.0481,
      "step": 910
    },
    {
      "balancing_entropy": 1.0608,
      "balancing_loss": 1.1215,
      "epoch": 4.69632,
      "grad_norm": 0.34375,
      "learning_rate": 0.002830769230769231,
      "loss": 2.919,
      "router_z_loss": 0.0382,
      "step": 920
    },
    {
      "balancing_entropy": 1.059,
      "balancing_loss": 1.0758,
      "epoch": 4.74752,
      "grad_norm": 0.251953125,
      "learning_rate": 0.002861538461538462,
      "loss": 2.9262,
      "router_z_loss": 0.1472,
      "step": 930
    },
    {
      "balancing_entropy": 1.0388,
      "balancing_loss": 1.1026,
      "epoch": 4.79872,
      "grad_norm": 0.298828125,
      "learning_rate": 0.0028923076923076923,
      "loss": 2.9038,
      "router_z_loss": 0.1057,
      "step": 940
    },
    {
      "balancing_entropy": 1.0384,
      "balancing_loss": 1.1249,
      "epoch": 4.84992,
      "grad_norm": 0.294921875,
      "learning_rate": 0.002923076923076923,
      "loss": 2.9105,
      "router_z_loss": 0.0981,
      "step": 950
    },
    {
      "balancing_entropy": 1.0729,
      "balancing_loss": 1.1026,
      "epoch": 4.90112,
      "grad_norm": 0.28125,
      "learning_rate": 0.002953846153846154,
      "loss": 2.8872,
      "router_z_loss": 0.1501,
      "step": 960
    },
    {
      "balancing_entropy": 1.0843,
      "balancing_loss": 1.1123,
      "epoch": 4.95232,
      "grad_norm": 0.271484375,
      "learning_rate": 0.0029846153846153846,
      "loss": 2.9128,
      "router_z_loss": 0.0492,
      "step": 970
    },
    {
      "balancing_entropy": 1.0237,
      "balancing_loss": 1.1246,
      "epoch": 5.0,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.8729,
      "router_z_loss": 0.0501,
      "step": 980
    },
    {
      "balancing_entropy": 1.0714,
      "balancing_loss": 1.0844,
      "epoch": 5.0512,
      "grad_norm": 0.25,
      "learning_rate": 0.003,
      "loss": 2.8925,
      "router_z_loss": 0.1031,
      "step": 990
    },
    {
      "balancing_entropy": 1.0844,
      "balancing_loss": 1.1099,
      "epoch": 5.1024,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.8403,
      "router_z_loss": 0.079,
      "step": 1000
    },
    {
      "balancing_entropy": 1.0786,
      "balancing_loss": 1.096,
      "epoch": 5.1536,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.8647,
      "router_z_loss": 0.0456,
      "step": 1010
    },
    {
      "balancing_entropy": 1.0676,
      "balancing_loss": 1.108,
      "epoch": 5.2048,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.8194,
      "router_z_loss": 0.0442,
      "step": 1020
    },
    {
      "balancing_entropy": 1.0733,
      "balancing_loss": 1.1015,
      "epoch": 5.256,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.8276,
      "router_z_loss": 0.0402,
      "step": 1030
    },
    {
      "balancing_entropy": 1.0855,
      "balancing_loss": 1.1128,
      "epoch": 5.3072,
      "grad_norm": 0.2421875,
      "learning_rate": 0.003,
      "loss": 2.8414,
      "router_z_loss": 0.0399,
      "step": 1040
    },
    {
      "balancing_entropy": 1.0807,
      "balancing_loss": 1.104,
      "epoch": 5.3584,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.8114,
      "router_z_loss": 0.0427,
      "step": 1050
    },
    {
      "balancing_entropy": 1.0881,
      "balancing_loss": 1.1114,
      "epoch": 5.4096,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.8032,
      "router_z_loss": 0.0286,
      "step": 1060
    },
    {
      "balancing_entropy": 1.0567,
      "balancing_loss": 1.1257,
      "epoch": 5.4608,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.796,
      "router_z_loss": 0.0563,
      "step": 1070
    },
    {
      "balancing_entropy": 1.0871,
      "balancing_loss": 1.098,
      "epoch": 5.5120000000000005,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.7893,
      "router_z_loss": 0.0573,
      "step": 1080
    },
    {
      "balancing_entropy": 1.0589,
      "balancing_loss": 1.1092,
      "epoch": 5.5632,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.7915,
      "router_z_loss": 0.0361,
      "step": 1090
    },
    {
      "balancing_entropy": 1.0779,
      "balancing_loss": 1.121,
      "epoch": 5.6144,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.7814,
      "router_z_loss": 0.0575,
      "step": 1100
    },
    {
      "balancing_entropy": 1.0739,
      "balancing_loss": 1.1173,
      "epoch": 5.6655999999999995,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.7735,
      "router_z_loss": 0.0452,
      "step": 1110
    },
    {
      "balancing_entropy": 1.0798,
      "balancing_loss": 1.108,
      "epoch": 5.7168,
      "grad_norm": 0.224609375,
      "learning_rate": 0.003,
      "loss": 2.7718,
      "router_z_loss": 0.0369,
      "step": 1120
    },
    {
      "balancing_entropy": 1.0562,
      "balancing_loss": 1.1132,
      "epoch": 5.768,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.7737,
      "router_z_loss": 0.0553,
      "step": 1130
    },
    {
      "balancing_entropy": 1.049,
      "balancing_loss": 1.1171,
      "epoch": 5.8192,
      "grad_norm": 1.625,
      "learning_rate": 0.003,
      "loss": 2.8174,
      "router_z_loss": 0.154,
      "step": 1140
    },
    {
      "balancing_entropy": 1.0838,
      "balancing_loss": 1.0896,
      "epoch": 5.8704,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.8409,
      "router_z_loss": 0.3276,
      "step": 1150
    },
    {
      "balancing_entropy": 1.0754,
      "balancing_loss": 1.0922,
      "epoch": 5.9216,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.8197,
      "router_z_loss": 0.2162,
      "step": 1160
    },
    {
      "balancing_entropy": 1.084,
      "balancing_loss": 1.1066,
      "epoch": 5.9728,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.7698,
      "router_z_loss": 0.0843,
      "step": 1170
    },
    {
      "balancing_entropy": 1.0194,
      "balancing_loss": 1.1086,
      "epoch": 6.02048,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.7488,
      "router_z_loss": 0.0517,
      "step": 1180
    },
    {
      "balancing_entropy": 1.0724,
      "balancing_loss": 1.1055,
      "epoch": 6.07168,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.73,
      "router_z_loss": 0.0522,
      "step": 1190
    },
    {
      "balancing_entropy": 1.0264,
      "balancing_loss": 1.1178,
      "epoch": 6.12288,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.727,
      "router_z_loss": 0.0499,
      "step": 1200
    },
    {
      "balancing_entropy": 1.0858,
      "balancing_loss": 1.1043,
      "epoch": 6.17408,
      "grad_norm": 0.240234375,
      "learning_rate": 0.003,
      "loss": 2.7092,
      "router_z_loss": 0.0643,
      "step": 1210
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.127,
      "epoch": 6.22528,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.706,
      "router_z_loss": 0.0392,
      "step": 1220
    },
    {
      "balancing_entropy": 1.0836,
      "balancing_loss": 1.1099,
      "epoch": 6.27648,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.7024,
      "router_z_loss": 0.0519,
      "step": 1230
    },
    {
      "balancing_entropy": 1.0788,
      "balancing_loss": 1.1085,
      "epoch": 6.32768,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.7044,
      "router_z_loss": 0.0326,
      "step": 1240
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.1099,
      "epoch": 6.37888,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.6874,
      "router_z_loss": 0.0592,
      "step": 1250
    },
    {
      "balancing_entropy": 1.0006,
      "balancing_loss": 1.1272,
      "epoch": 6.43008,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.7091,
      "router_z_loss": 0.1513,
      "step": 1260
    },
    {
      "balancing_entropy": 1.0775,
      "balancing_loss": 1.108,
      "epoch": 6.48128,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.6821,
      "router_z_loss": 0.106,
      "step": 1270
    },
    {
      "balancing_entropy": 1.0368,
      "balancing_loss": 1.1091,
      "epoch": 6.53248,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.6736,
      "router_z_loss": 0.0363,
      "step": 1280
    },
    {
      "balancing_entropy": 0.9919,
      "balancing_loss": 1.1228,
      "epoch": 6.58368,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.6844,
      "router_z_loss": 0.0494,
      "step": 1290
    },
    {
      "balancing_entropy": 1.0121,
      "balancing_loss": 1.1238,
      "epoch": 6.63488,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.675,
      "router_z_loss": 0.1156,
      "step": 1300
    },
    {
      "balancing_entropy": 1.0146,
      "balancing_loss": 1.0942,
      "epoch": 6.6860800000000005,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.6669,
      "router_z_loss": 0.1085,
      "step": 1310
    },
    {
      "balancing_entropy": 1.051,
      "balancing_loss": 1.1207,
      "epoch": 6.73728,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.6719,
      "router_z_loss": 0.0844,
      "step": 1320
    },
    {
      "balancing_entropy": 1.0765,
      "balancing_loss": 1.1175,
      "epoch": 6.78848,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.6575,
      "router_z_loss": 0.0487,
      "step": 1330
    },
    {
      "balancing_entropy": 1.0473,
      "balancing_loss": 1.1171,
      "epoch": 6.8396799999999995,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.6567,
      "router_z_loss": 0.035,
      "step": 1340
    },
    {
      "balancing_entropy": 1.018,
      "balancing_loss": 1.1165,
      "epoch": 6.89088,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.6607,
      "router_z_loss": 0.0385,
      "step": 1350
    },
    {
      "balancing_entropy": 1.0763,
      "balancing_loss": 1.1156,
      "epoch": 6.94208,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.6576,
      "router_z_loss": 0.058,
      "step": 1360
    },
    {
      "balancing_entropy": 1.0217,
      "balancing_loss": 1.1241,
      "epoch": 6.99328,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.6491,
      "router_z_loss": 0.0383,
      "step": 1370
    },
    {
      "balancing_entropy": 1.0177,
      "balancing_loss": 1.1123,
      "epoch": 7.04096,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.643,
      "router_z_loss": 0.0347,
      "step": 1380
    },
    {
      "balancing_entropy": 1.0424,
      "balancing_loss": 1.1078,
      "epoch": 7.09216,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.6393,
      "router_z_loss": 0.0356,
      "step": 1390
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.1167,
      "epoch": 7.14336,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.6262,
      "router_z_loss": 0.03,
      "step": 1400
    },
    {
      "balancing_entropy": 1.0365,
      "balancing_loss": 1.1101,
      "epoch": 7.19456,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.6217,
      "router_z_loss": 0.0242,
      "step": 1410
    },
    {
      "balancing_entropy": 1.0052,
      "balancing_loss": 1.1091,
      "epoch": 7.24576,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.6217,
      "router_z_loss": 0.0437,
      "step": 1420
    },
    {
      "balancing_entropy": 1.0175,
      "balancing_loss": 1.1232,
      "epoch": 7.29696,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.6106,
      "router_z_loss": 0.0484,
      "step": 1430
    },
    {
      "balancing_entropy": 1.0869,
      "balancing_loss": 1.1058,
      "epoch": 7.34816,
      "grad_norm": 0.240234375,
      "learning_rate": 0.003,
      "loss": 2.6125,
      "router_z_loss": 0.0338,
      "step": 1440
    },
    {
      "balancing_entropy": 1.0047,
      "balancing_loss": 1.1259,
      "epoch": 7.39936,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.621,
      "router_z_loss": 0.0265,
      "step": 1450
    },
    {
      "balancing_entropy": 1.028,
      "balancing_loss": 1.1028,
      "epoch": 7.45056,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.6191,
      "router_z_loss": 0.0488,
      "step": 1460
    },
    {
      "balancing_entropy": 1.0321,
      "balancing_loss": 1.1041,
      "epoch": 7.50176,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.606,
      "router_z_loss": 0.0444,
      "step": 1470
    },
    {
      "balancing_entropy": 1.019,
      "balancing_loss": 1.1233,
      "epoch": 7.55296,
      "grad_norm": 0.2431640625,
      "learning_rate": 0.003,
      "loss": 2.6033,
      "router_z_loss": 0.0494,
      "step": 1480
    },
    {
      "balancing_entropy": 1.0257,
      "balancing_loss": 1.1108,
      "epoch": 7.60416,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.6137,
      "router_z_loss": 0.0487,
      "step": 1490
    },
    {
      "balancing_entropy": 1.0655,
      "balancing_loss": 1.1121,
      "epoch": 7.65536,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.6068,
      "router_z_loss": 0.0404,
      "step": 1500
    },
    {
      "balancing_entropy": 1.0208,
      "balancing_loss": 1.1163,
      "epoch": 7.70656,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.5948,
      "router_z_loss": 0.0314,
      "step": 1510
    },
    {
      "balancing_entropy": 1.0572,
      "balancing_loss": 1.1362,
      "epoch": 7.75776,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.6933,
      "router_z_loss": 0.0492,
      "step": 1520
    },
    {
      "balancing_entropy": 1.0672,
      "balancing_loss": 1.1019,
      "epoch": 7.80896,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.6223,
      "router_z_loss": 0.084,
      "step": 1530
    },
    {
      "balancing_entropy": 1.0573,
      "balancing_loss": 1.113,
      "epoch": 7.8601600000000005,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.594,
      "router_z_loss": 0.0642,
      "step": 1540
    },
    {
      "balancing_entropy": 1.0319,
      "balancing_loss": 1.1006,
      "epoch": 7.91136,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.5974,
      "router_z_loss": 0.0273,
      "step": 1550
    },
    {
      "balancing_entropy": 1.0552,
      "balancing_loss": 1.1065,
      "epoch": 7.96256,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.588,
      "router_z_loss": 0.0336,
      "step": 1560
    },
    {
      "balancing_entropy": 0.9985,
      "balancing_loss": 1.1163,
      "epoch": 8.01024,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.5919,
      "router_z_loss": 0.0381,
      "step": 1570
    },
    {
      "balancing_entropy": 1.0203,
      "balancing_loss": 1.1204,
      "epoch": 8.06144,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.5741,
      "router_z_loss": 0.0349,
      "step": 1580
    },
    {
      "balancing_entropy": 1.0091,
      "balancing_loss": 1.1065,
      "epoch": 8.11264,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.5686,
      "router_z_loss": 0.0629,
      "step": 1590
    },
    {
      "balancing_entropy": 1.0637,
      "balancing_loss": 1.1064,
      "epoch": 8.16384,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.6053,
      "router_z_loss": 0.0668,
      "step": 1600
    },
    {
      "balancing_entropy": 1.0708,
      "balancing_loss": 1.1157,
      "epoch": 8.21504,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.6375,
      "router_z_loss": 0.0948,
      "step": 1610
    },
    {
      "balancing_entropy": 1.0865,
      "balancing_loss": 1.1038,
      "epoch": 8.26624,
      "grad_norm": 0.2470703125,
      "learning_rate": 0.003,
      "loss": 2.571,
      "router_z_loss": 0.0649,
      "step": 1620
    },
    {
      "balancing_entropy": 0.9924,
      "balancing_loss": 1.1054,
      "epoch": 8.31744,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.568,
      "router_z_loss": 0.0954,
      "step": 1630
    },
    {
      "balancing_entropy": 1.0115,
      "balancing_loss": 1.1369,
      "epoch": 8.36864,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.5658,
      "router_z_loss": 0.1433,
      "step": 1640
    },
    {
      "balancing_entropy": 1.0557,
      "balancing_loss": 1.1104,
      "epoch": 8.41984,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.555,
      "router_z_loss": 0.0909,
      "step": 1650
    },
    {
      "balancing_entropy": 1.0746,
      "balancing_loss": 1.1058,
      "epoch": 8.47104,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.5576,
      "router_z_loss": 0.0468,
      "step": 1660
    },
    {
      "balancing_entropy": 1.0082,
      "balancing_loss": 1.1245,
      "epoch": 8.52224,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.5513,
      "router_z_loss": 0.0396,
      "step": 1670
    },
    {
      "balancing_entropy": 1.0213,
      "balancing_loss": 1.1082,
      "epoch": 8.57344,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.5526,
      "router_z_loss": 0.0496,
      "step": 1680
    },
    {
      "balancing_entropy": 1.0084,
      "balancing_loss": 1.1091,
      "epoch": 8.62464,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.5538,
      "router_z_loss": 0.0603,
      "step": 1690
    },
    {
      "balancing_entropy": 1.0497,
      "balancing_loss": 1.1201,
      "epoch": 8.67584,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.5532,
      "router_z_loss": 0.0448,
      "step": 1700
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.1298,
      "epoch": 8.72704,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5535,
      "router_z_loss": 0.029,
      "step": 1710
    },
    {
      "balancing_entropy": 1.0572,
      "balancing_loss": 1.1157,
      "epoch": 8.77824,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.5401,
      "router_z_loss": 0.028,
      "step": 1720
    },
    {
      "balancing_entropy": 1.0082,
      "balancing_loss": 1.1206,
      "epoch": 8.82944,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.5493,
      "router_z_loss": 0.0464,
      "step": 1730
    },
    {
      "balancing_entropy": 0.9829,
      "balancing_loss": 1.1311,
      "epoch": 8.88064,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.5463,
      "router_z_loss": 0.0731,
      "step": 1740
    },
    {
      "balancing_entropy": 0.9818,
      "balancing_loss": 1.0975,
      "epoch": 8.93184,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.5466,
      "router_z_loss": 0.0922,
      "step": 1750
    },
    {
      "balancing_entropy": 0.9662,
      "balancing_loss": 1.1181,
      "epoch": 8.98304,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.5468,
      "router_z_loss": 0.1385,
      "step": 1760
    },
    {
      "balancing_entropy": 0.9867,
      "balancing_loss": 1.1122,
      "epoch": 9.03072,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.535,
      "router_z_loss": 0.1179,
      "step": 1770
    },
    {
      "balancing_entropy": 1.0826,
      "balancing_loss": 1.1173,
      "epoch": 9.08192,
      "grad_norm": 0.244140625,
      "learning_rate": 0.003,
      "loss": 2.5266,
      "router_z_loss": 0.0579,
      "step": 1780
    },
    {
      "balancing_entropy": 1.0822,
      "balancing_loss": 1.1131,
      "epoch": 9.13312,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.5131,
      "router_z_loss": 0.0238,
      "step": 1790
    },
    {
      "balancing_entropy": 1.0277,
      "balancing_loss": 1.1195,
      "epoch": 9.18432,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.5199,
      "router_z_loss": 0.0258,
      "step": 1800
    },
    {
      "balancing_entropy": 1.0267,
      "balancing_loss": 1.1139,
      "epoch": 9.23552,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.5239,
      "router_z_loss": 0.0403,
      "step": 1810
    },
    {
      "balancing_entropy": 1.0592,
      "balancing_loss": 1.1117,
      "epoch": 9.28672,
      "grad_norm": 0.2470703125,
      "learning_rate": 0.003,
      "loss": 2.5115,
      "router_z_loss": 0.0464,
      "step": 1820
    },
    {
      "balancing_entropy": 1.0239,
      "balancing_loss": 1.1188,
      "epoch": 9.33792,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.5205,
      "router_z_loss": 0.0388,
      "step": 1830
    },
    {
      "balancing_entropy": 1.0223,
      "balancing_loss": 1.1197,
      "epoch": 9.38912,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.5123,
      "router_z_loss": 0.0429,
      "step": 1840
    },
    {
      "balancing_entropy": 1.019,
      "balancing_loss": 1.1136,
      "epoch": 9.44032,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.516,
      "router_z_loss": 0.0368,
      "step": 1850
    },
    {
      "balancing_entropy": 1.0178,
      "balancing_loss": 1.1196,
      "epoch": 9.49152,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.5151,
      "router_z_loss": 0.0431,
      "step": 1860
    },
    {
      "balancing_entropy": 1.0814,
      "balancing_loss": 1.1081,
      "epoch": 9.54272,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5211,
      "router_z_loss": 0.0501,
      "step": 1870
    },
    {
      "balancing_entropy": 0.9928,
      "balancing_loss": 1.1233,
      "epoch": 9.59392,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.5209,
      "router_z_loss": 0.0337,
      "step": 1880
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.1074,
      "epoch": 9.64512,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.5146,
      "router_z_loss": 0.0517,
      "step": 1890
    },
    {
      "balancing_entropy": 1.0687,
      "balancing_loss": 1.1196,
      "epoch": 9.69632,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.5184,
      "router_z_loss": 0.0367,
      "step": 1900
    },
    {
      "balancing_entropy": 1.0487,
      "balancing_loss": 1.1062,
      "epoch": 9.74752,
      "grad_norm": 0.24609375,
      "learning_rate": 0.003,
      "loss": 2.503,
      "router_z_loss": 0.0273,
      "step": 1910
    },
    {
      "balancing_entropy": 0.9857,
      "balancing_loss": 1.1337,
      "epoch": 9.79872,
      "grad_norm": 0.67578125,
      "learning_rate": 0.003,
      "loss": 2.5109,
      "router_z_loss": 0.0492,
      "step": 1920
    },
    {
      "balancing_entropy": 0.9915,
      "balancing_loss": 1.1136,
      "epoch": 9.849920000000001,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.5021,
      "router_z_loss": 0.0534,
      "step": 1930
    },
    {
      "balancing_entropy": 1.0633,
      "balancing_loss": 1.1074,
      "epoch": 9.90112,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.4996,
      "router_z_loss": 0.0403,
      "step": 1940
    },
    {
      "balancing_entropy": 1.0538,
      "balancing_loss": 1.1198,
      "epoch": 9.95232,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.5085,
      "router_z_loss": 0.0385,
      "step": 1950
    },
    {
      "balancing_entropy": 0.9821,
      "balancing_loss": 1.1029,
      "epoch": 10.0,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.5148,
      "router_z_loss": 0.0379,
      "step": 1960
    },
    {
      "balancing_entropy": 1.0062,
      "balancing_loss": 1.1191,
      "epoch": 10.0512,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4998,
      "router_z_loss": 0.0559,
      "step": 1970
    },
    {
      "balancing_entropy": 1.0499,
      "balancing_loss": 1.1164,
      "epoch": 10.1024,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4899,
      "router_z_loss": 0.0644,
      "step": 1980
    },
    {
      "balancing_entropy": 1.0214,
      "balancing_loss": 1.1198,
      "epoch": 10.1536,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.4876,
      "router_z_loss": 0.0488,
      "step": 1990
    },
    {
      "balancing_entropy": 1.0141,
      "balancing_loss": 1.1119,
      "epoch": 10.2048,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.4804,
      "router_z_loss": 0.0389,
      "step": 2000
    },
    {
      "balancing_entropy": 1.0504,
      "balancing_loss": 1.1147,
      "epoch": 10.256,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.4727,
      "router_z_loss": 0.0339,
      "step": 2010
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.1201,
      "epoch": 10.3072,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4821,
      "router_z_loss": 0.0339,
      "step": 2020
    },
    {
      "balancing_entropy": 1.0804,
      "balancing_loss": 1.1231,
      "epoch": 10.3584,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.4744,
      "router_z_loss": 0.0293,
      "step": 2030
    },
    {
      "balancing_entropy": 1.0201,
      "balancing_loss": 1.1236,
      "epoch": 10.4096,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.4822,
      "router_z_loss": 0.0252,
      "step": 2040
    },
    {
      "balancing_entropy": 1.0087,
      "balancing_loss": 1.1096,
      "epoch": 10.4608,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.4801,
      "router_z_loss": 0.0467,
      "step": 2050
    },
    {
      "balancing_entropy": 1.0379,
      "balancing_loss": 1.1063,
      "epoch": 10.512,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.4788,
      "router_z_loss": 0.0539,
      "step": 2060
    },
    {
      "balancing_entropy": 1.0472,
      "balancing_loss": 1.1284,
      "epoch": 10.5632,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.4724,
      "router_z_loss": 0.0393,
      "step": 2070
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.1089,
      "epoch": 10.6144,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.4766,
      "router_z_loss": 0.037,
      "step": 2080
    },
    {
      "balancing_entropy": 1.0502,
      "balancing_loss": 1.1247,
      "epoch": 10.6656,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.4805,
      "router_z_loss": 0.0467,
      "step": 2090
    },
    {
      "balancing_entropy": 1.0292,
      "balancing_loss": 1.1186,
      "epoch": 10.7168,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4705,
      "router_z_loss": 0.0413,
      "step": 2100
    },
    {
      "balancing_entropy": 1.0256,
      "balancing_loss": 1.1288,
      "epoch": 10.768,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.4778,
      "router_z_loss": 0.0372,
      "step": 2110
    },
    {
      "balancing_entropy": 0.9658,
      "balancing_loss": 1.1134,
      "epoch": 10.8192,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.4722,
      "router_z_loss": 0.0691,
      "step": 2120
    },
    {
      "balancing_entropy": 1.0158,
      "balancing_loss": 1.1165,
      "epoch": 10.8704,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.4797,
      "router_z_loss": 0.1036,
      "step": 2130
    },
    {
      "balancing_entropy": 1.0527,
      "balancing_loss": 1.1179,
      "epoch": 10.9216,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.4655,
      "router_z_loss": 0.063,
      "step": 2140
    },
    {
      "balancing_entropy": 0.9936,
      "balancing_loss": 1.1211,
      "epoch": 10.9728,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.472,
      "router_z_loss": 0.0412,
      "step": 2150
    },
    {
      "balancing_entropy": 1.0006,
      "balancing_loss": 1.1217,
      "epoch": 11.02048,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4673,
      "router_z_loss": 0.0796,
      "step": 2160
    },
    {
      "balancing_entropy": 1.0244,
      "balancing_loss": 1.1241,
      "epoch": 11.07168,
      "grad_norm": 0.62109375,
      "learning_rate": 0.003,
      "loss": 2.451,
      "router_z_loss": 0.0559,
      "step": 2170
    },
    {
      "balancing_entropy": 1.0295,
      "balancing_loss": 1.1085,
      "epoch": 11.12288,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.4513,
      "router_z_loss": 0.037,
      "step": 2180
    },
    {
      "balancing_entropy": 1.0277,
      "balancing_loss": 1.1275,
      "epoch": 11.17408,
      "grad_norm": 0.474609375,
      "learning_rate": 0.003,
      "loss": 2.4505,
      "router_z_loss": 0.0443,
      "step": 2190
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.1076,
      "epoch": 11.22528,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.4475,
      "router_z_loss": 0.0465,
      "step": 2200
    },
    {
      "balancing_entropy": 1.0119,
      "balancing_loss": 1.136,
      "epoch": 11.27648,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.4474,
      "router_z_loss": 0.045,
      "step": 2210
    },
    {
      "balancing_entropy": 1.0212,
      "balancing_loss": 1.1097,
      "epoch": 11.32768,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.4514,
      "router_z_loss": 0.0385,
      "step": 2220
    },
    {
      "balancing_entropy": 1.0304,
      "balancing_loss": 1.1152,
      "epoch": 11.37888,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.4514,
      "router_z_loss": 0.0473,
      "step": 2230
    },
    {
      "balancing_entropy": 0.9808,
      "balancing_loss": 1.1144,
      "epoch": 11.43008,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.4499,
      "router_z_loss": 0.0555,
      "step": 2240
    },
    {
      "balancing_entropy": 0.9706,
      "balancing_loss": 1.1172,
      "epoch": 11.48128,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.4532,
      "router_z_loss": 0.091,
      "step": 2250
    },
    {
      "balancing_entropy": 1.0049,
      "balancing_loss": 1.1161,
      "epoch": 11.53248,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.4548,
      "router_z_loss": 0.0768,
      "step": 2260
    },
    {
      "balancing_entropy": 0.9899,
      "balancing_loss": 1.1088,
      "epoch": 11.58368,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.4523,
      "router_z_loss": 0.0716,
      "step": 2270
    },
    {
      "balancing_entropy": 1.0056,
      "balancing_loss": 1.1271,
      "epoch": 11.63488,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.4482,
      "router_z_loss": 0.0687,
      "step": 2280
    },
    {
      "balancing_entropy": 1.0395,
      "balancing_loss": 1.0932,
      "epoch": 11.68608,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.4506,
      "router_z_loss": 0.0497,
      "step": 2290
    },
    {
      "balancing_entropy": 1.0304,
      "balancing_loss": 1.1269,
      "epoch": 11.73728,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.4434,
      "router_z_loss": 0.0444,
      "step": 2300
    },
    {
      "balancing_entropy": 1.0463,
      "balancing_loss": 1.1166,
      "epoch": 11.78848,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.4431,
      "router_z_loss": 0.0363,
      "step": 2310
    },
    {
      "balancing_entropy": 1.0447,
      "balancing_loss": 1.1075,
      "epoch": 11.83968,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.4445,
      "router_z_loss": 0.0276,
      "step": 2320
    },
    {
      "balancing_entropy": 1.0158,
      "balancing_loss": 1.1308,
      "epoch": 11.89088,
      "grad_norm": 0.6875,
      "learning_rate": 0.003,
      "loss": 2.4363,
      "router_z_loss": 0.0328,
      "step": 2330
    },
    {
      "balancing_entropy": 1.015,
      "balancing_loss": 1.1161,
      "epoch": 11.94208,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.4422,
      "router_z_loss": 0.036,
      "step": 2340
    },
    {
      "balancing_entropy": 1.0664,
      "balancing_loss": 1.1223,
      "epoch": 11.99328,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4426,
      "router_z_loss": 0.0341,
      "step": 2350
    },
    {
      "balancing_entropy": 0.944,
      "balancing_loss": 1.1002,
      "epoch": 12.04096,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.4367,
      "router_z_loss": 0.0372,
      "step": 2360
    },
    {
      "balancing_entropy": 1.0224,
      "balancing_loss": 1.1204,
      "epoch": 12.09216,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.4344,
      "router_z_loss": 0.0938,
      "step": 2370
    },
    {
      "balancing_entropy": 1.0214,
      "balancing_loss": 1.1327,
      "epoch": 12.14336,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.4187,
      "router_z_loss": 0.0592,
      "step": 2380
    },
    {
      "balancing_entropy": 1.0196,
      "balancing_loss": 1.1121,
      "epoch": 12.19456,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.4272,
      "router_z_loss": 0.0354,
      "step": 2390
    },
    {
      "balancing_entropy": 1.0091,
      "balancing_loss": 1.1254,
      "epoch": 12.24576,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.4314,
      "router_z_loss": 0.0379,
      "step": 2400
    },
    {
      "balancing_entropy": 0.992,
      "balancing_loss": 1.1081,
      "epoch": 12.29696,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.4272,
      "router_z_loss": 0.0551,
      "step": 2410
    },
    {
      "balancing_entropy": 0.9963,
      "balancing_loss": 1.1102,
      "epoch": 12.34816,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.4226,
      "router_z_loss": 0.08,
      "step": 2420
    },
    {
      "balancing_entropy": 0.99,
      "balancing_loss": 1.143,
      "epoch": 12.39936,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.4212,
      "router_z_loss": 0.0813,
      "step": 2430
    },
    {
      "balancing_entropy": 1.0642,
      "balancing_loss": 1.1199,
      "epoch": 12.45056,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.4275,
      "router_z_loss": 0.0756,
      "step": 2440
    },
    {
      "balancing_entropy": 1.0786,
      "balancing_loss": 1.1085,
      "epoch": 12.50176,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.4182,
      "router_z_loss": 0.0467,
      "step": 2450
    },
    {
      "balancing_entropy": 1.0734,
      "balancing_loss": 1.1245,
      "epoch": 12.55296,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.4249,
      "router_z_loss": 0.03,
      "step": 2460
    },
    {
      "balancing_entropy": 1.0276,
      "balancing_loss": 1.1172,
      "epoch": 12.60416,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.425,
      "router_z_loss": 0.0253,
      "step": 2470
    },
    {
      "balancing_entropy": 1.003,
      "balancing_loss": 1.1415,
      "epoch": 12.65536,
      "grad_norm": 0.67578125,
      "learning_rate": 0.003,
      "loss": 2.4115,
      "router_z_loss": 0.0395,
      "step": 2480
    },
    {
      "balancing_entropy": 0.9809,
      "balancing_loss": 1.1129,
      "epoch": 12.70656,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.4177,
      "router_z_loss": 0.0565,
      "step": 2490
    },
    {
      "balancing_entropy": 0.9666,
      "balancing_loss": 1.1129,
      "epoch": 12.75776,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.4214,
      "router_z_loss": 0.0859,
      "step": 2500
    },
    {
      "balancing_entropy": 0.9593,
      "balancing_loss": 1.1282,
      "epoch": 12.80896,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.4237,
      "router_z_loss": 0.0815,
      "step": 2510
    },
    {
      "balancing_entropy": 1.0344,
      "balancing_loss": 1.1142,
      "epoch": 12.86016,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.4229,
      "router_z_loss": 0.0826,
      "step": 2520
    },
    {
      "balancing_entropy": 1.0866,
      "balancing_loss": 1.118,
      "epoch": 12.91136,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.4288,
      "router_z_loss": 0.0598,
      "step": 2530
    },
    {
      "balancing_entropy": 0.9846,
      "balancing_loss": 1.13,
      "epoch": 12.96256,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.4165,
      "router_z_loss": 0.0325,
      "step": 2540
    },
    {
      "balancing_entropy": 0.9752,
      "balancing_loss": 1.1213,
      "epoch": 13.01024,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.4166,
      "router_z_loss": 0.0626,
      "step": 2550
    },
    {
      "balancing_entropy": 0.9882,
      "balancing_loss": 1.1267,
      "epoch": 13.06144,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.4076,
      "router_z_loss": 0.082,
      "step": 2560
    },
    {
      "balancing_entropy": 1.0058,
      "balancing_loss": 1.1276,
      "epoch": 13.11264,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.4044,
      "router_z_loss": 0.0598,
      "step": 2570
    },
    {
      "balancing_entropy": 1.0118,
      "balancing_loss": 1.1163,
      "epoch": 13.16384,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.404,
      "router_z_loss": 0.0412,
      "step": 2580
    },
    {
      "balancing_entropy": 1.0178,
      "balancing_loss": 1.1228,
      "epoch": 13.21504,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.3999,
      "router_z_loss": 0.0427,
      "step": 2590
    },
    {
      "balancing_entropy": 1.05,
      "balancing_loss": 1.1239,
      "epoch": 13.26624,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.3926,
      "router_z_loss": 0.0324,
      "step": 2600
    },
    {
      "balancing_entropy": 1.011,
      "balancing_loss": 1.118,
      "epoch": 13.31744,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.4006,
      "router_z_loss": 0.0336,
      "step": 2610
    },
    {
      "balancing_entropy": 1.0196,
      "balancing_loss": 1.1216,
      "epoch": 13.36864,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.3974,
      "router_z_loss": 0.0571,
      "step": 2620
    },
    {
      "balancing_entropy": 1.0709,
      "balancing_loss": 1.1224,
      "epoch": 13.41984,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.4011,
      "router_z_loss": 0.042,
      "step": 2630
    },
    {
      "balancing_entropy": 1.0066,
      "balancing_loss": 1.1253,
      "epoch": 13.47104,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.4045,
      "router_z_loss": 0.0282,
      "step": 2640
    },
    {
      "balancing_entropy": 0.9824,
      "balancing_loss": 1.1105,
      "epoch": 13.52224,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.4038,
      "router_z_loss": 0.051,
      "step": 2650
    },
    {
      "balancing_entropy": 0.9804,
      "balancing_loss": 1.1256,
      "epoch": 13.57344,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.4069,
      "router_z_loss": 0.0928,
      "step": 2660
    },
    {
      "balancing_entropy": 1.0275,
      "balancing_loss": 1.1105,
      "epoch": 13.62464,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.3963,
      "router_z_loss": 0.0814,
      "step": 2670
    },
    {
      "balancing_entropy": 1.064,
      "balancing_loss": 1.1226,
      "epoch": 13.67584,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.4008,
      "router_z_loss": 0.0438,
      "step": 2680
    },
    {
      "balancing_entropy": 1.0209,
      "balancing_loss": 1.1104,
      "epoch": 13.72704,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4065,
      "router_z_loss": 0.0316,
      "step": 2690
    },
    {
      "balancing_entropy": 1.0391,
      "balancing_loss": 1.1213,
      "epoch": 13.77824,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.3987,
      "router_z_loss": 0.0457,
      "step": 2700
    },
    {
      "balancing_entropy": 1.0793,
      "balancing_loss": 1.1361,
      "epoch": 13.82944,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.4013,
      "router_z_loss": 0.0471,
      "step": 2710
    },
    {
      "balancing_entropy": 1.047,
      "balancing_loss": 1.1196,
      "epoch": 13.88064,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.3992,
      "router_z_loss": 0.0285,
      "step": 2720
    },
    {
      "balancing_entropy": 1.019,
      "balancing_loss": 1.1264,
      "epoch": 13.93184,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.3928,
      "router_z_loss": 0.0323,
      "step": 2730
    },
    {
      "balancing_entropy": 0.9814,
      "balancing_loss": 1.1301,
      "epoch": 13.98304,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.3967,
      "router_z_loss": 0.0577,
      "step": 2740
    },
    {
      "balancing_entropy": 0.9678,
      "balancing_loss": 1.11,
      "epoch": 14.03072,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.3944,
      "router_z_loss": 0.0776,
      "step": 2750
    },
    {
      "balancing_entropy": 0.9596,
      "balancing_loss": 1.1404,
      "epoch": 14.08192,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.3871,
      "router_z_loss": 0.0686,
      "step": 2760
    },
    {
      "balancing_entropy": 0.9881,
      "balancing_loss": 1.1024,
      "epoch": 14.13312,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.3799,
      "router_z_loss": 0.0882,
      "step": 2770
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.1325,
      "epoch": 14.18432,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.38,
      "router_z_loss": 0.0758,
      "step": 2780
    },
    {
      "balancing_entropy": 1.0355,
      "balancing_loss": 1.1196,
      "epoch": 14.23552,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.3779,
      "router_z_loss": 0.0412,
      "step": 2790
    },
    {
      "balancing_entropy": 0.9692,
      "balancing_loss": 1.1301,
      "epoch": 14.28672,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.3772,
      "router_z_loss": 0.0479,
      "step": 2800
    },
    {
      "balancing_entropy": 0.998,
      "balancing_loss": 1.102,
      "epoch": 14.33792,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.3862,
      "router_z_loss": 0.0857,
      "step": 2810
    },
    {
      "balancing_entropy": 0.949,
      "balancing_loss": 1.1283,
      "epoch": 14.38912,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.3829,
      "router_z_loss": 0.0815,
      "step": 2820
    },
    {
      "balancing_entropy": 0.9766,
      "balancing_loss": 1.1072,
      "epoch": 14.44032,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.3862,
      "router_z_loss": 0.0777,
      "step": 2830
    },
    {
      "balancing_entropy": 0.9713,
      "balancing_loss": 1.1353,
      "epoch": 14.49152,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.3861,
      "router_z_loss": 0.1315,
      "step": 2840
    },
    {
      "balancing_entropy": 1.0357,
      "balancing_loss": 1.1127,
      "epoch": 14.54272,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.3811,
      "router_z_loss": 0.069,
      "step": 2850
    },
    {
      "balancing_entropy": 1.0359,
      "balancing_loss": 1.126,
      "epoch": 14.59392,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.3896,
      "router_z_loss": 0.0412,
      "step": 2860
    },
    {
      "balancing_entropy": 0.9815,
      "balancing_loss": 1.1256,
      "epoch": 14.64512,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.3858,
      "router_z_loss": 0.0503,
      "step": 2870
    },
    {
      "balancing_entropy": 0.9612,
      "balancing_loss": 1.1326,
      "epoch": 14.69632,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.3799,
      "router_z_loss": 0.1111,
      "step": 2880
    },
    {
      "balancing_entropy": 0.9821,
      "balancing_loss": 1.106,
      "epoch": 14.74752,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.387,
      "router_z_loss": 0.0991,
      "step": 2890
    },
    {
      "balancing_entropy": 1.0261,
      "balancing_loss": 1.1278,
      "epoch": 14.79872,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.3844,
      "router_z_loss": 0.0964,
      "step": 2900
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.1216,
      "epoch": 14.849920000000001,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.3813,
      "router_z_loss": 0.0442,
      "step": 2910
    },
    {
      "balancing_entropy": 1.0436,
      "balancing_loss": 1.1264,
      "epoch": 14.90112,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.3793,
      "router_z_loss": 0.0359,
      "step": 2920
    },
    {
      "balancing_entropy": 1.0291,
      "balancing_loss": 1.1204,
      "epoch": 14.95232,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.387,
      "router_z_loss": 0.0347,
      "step": 2930
    },
    {
      "balancing_entropy": 1.039,
      "balancing_loss": 1.1301,
      "epoch": 15.0,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.3943,
      "router_z_loss": 0.0519,
      "step": 2940
    },
    {
      "balancing_entropy": 1.0275,
      "balancing_loss": 1.1044,
      "epoch": 15.0512,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.3809,
      "router_z_loss": 0.0406,
      "step": 2950
    },
    {
      "balancing_entropy": 1.0025,
      "balancing_loss": 1.124,
      "epoch": 15.1024,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.3748,
      "router_z_loss": 0.0633,
      "step": 2960
    },
    {
      "balancing_entropy": 1.021,
      "balancing_loss": 1.1146,
      "epoch": 15.1536,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.3658,
      "router_z_loss": 0.0651,
      "step": 2970
    },
    {
      "balancing_entropy": 0.979,
      "balancing_loss": 1.1244,
      "epoch": 15.2048,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.3574,
      "router_z_loss": 0.0579,
      "step": 2980
    },
    {
      "balancing_entropy": 1.0096,
      "balancing_loss": 1.1257,
      "epoch": 15.256,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.3704,
      "router_z_loss": 0.0645,
      "step": 2990
    },
    {
      "balancing_entropy": 1.0429,
      "balancing_loss": 1.1309,
      "epoch": 15.3072,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.3594,
      "router_z_loss": 0.0378,
      "step": 3000
    },
    {
      "balancing_entropy": 1.0138,
      "balancing_loss": 1.1322,
      "epoch": 15.3584,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.3669,
      "router_z_loss": 0.0305,
      "step": 3010
    },
    {
      "balancing_entropy": 1.0301,
      "balancing_loss": 1.1319,
      "epoch": 15.4096,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.3596,
      "router_z_loss": 0.0437,
      "step": 3020
    },
    {
      "balancing_entropy": 1.0507,
      "balancing_loss": 1.1013,
      "epoch": 15.4608,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.3651,
      "router_z_loss": 0.0399,
      "step": 3030
    },
    {
      "balancing_entropy": 0.9875,
      "balancing_loss": 1.1303,
      "epoch": 15.512,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.3655,
      "router_z_loss": 0.0457,
      "step": 3040
    },
    {
      "balancing_entropy": 1.0375,
      "balancing_loss": 1.1243,
      "epoch": 15.5632,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.3694,
      "router_z_loss": 0.0559,
      "step": 3050
    },
    {
      "balancing_entropy": 0.9945,
      "balancing_loss": 1.1211,
      "epoch": 15.6144,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.3678,
      "router_z_loss": 0.05,
      "step": 3060
    },
    {
      "balancing_entropy": 1.0451,
      "balancing_loss": 1.1335,
      "epoch": 15.6656,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.3664,
      "router_z_loss": 0.0664,
      "step": 3070
    },
    {
      "balancing_entropy": 1.0456,
      "balancing_loss": 1.1062,
      "epoch": 15.7168,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.3696,
      "router_z_loss": 0.0416,
      "step": 3080
    },
    {
      "balancing_entropy": 0.9843,
      "balancing_loss": 1.1241,
      "epoch": 15.768,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.3568,
      "router_z_loss": 0.0287,
      "step": 3090
    },
    {
      "balancing_entropy": 1.0232,
      "balancing_loss": 1.1357,
      "epoch": 15.8192,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.361,
      "router_z_loss": 0.051,
      "step": 3100
    },
    {
      "balancing_entropy": 1.0529,
      "balancing_loss": 1.1139,
      "epoch": 15.8704,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.3668,
      "router_z_loss": 0.0397,
      "step": 3110
    },
    {
      "balancing_entropy": 0.9659,
      "balancing_loss": 1.1444,
      "epoch": 15.9216,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.3664,
      "router_z_loss": 0.0447,
      "step": 3120
    },
    {
      "balancing_entropy": 1.0293,
      "balancing_loss": 1.101,
      "epoch": 15.9728,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.3643,
      "router_z_loss": 0.0909,
      "step": 3130
    },
    {
      "balancing_entropy": 1.0103,
      "balancing_loss": 1.1374,
      "epoch": 16.02048,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.3554,
      "router_z_loss": 0.0587,
      "step": 3140
    },
    {
      "balancing_entropy": 1.0656,
      "balancing_loss": 1.1193,
      "epoch": 16.07168,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.3467,
      "router_z_loss": 0.0355,
      "step": 3150
    },
    {
      "balancing_entropy": 1.0487,
      "balancing_loss": 1.1401,
      "epoch": 16.12288,
      "grad_norm": 0.474609375,
      "learning_rate": 0.003,
      "loss": 2.3467,
      "router_z_loss": 0.0391,
      "step": 3160
    },
    {
      "balancing_entropy": 1.0587,
      "balancing_loss": 1.1202,
      "epoch": 16.17408,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.3492,
      "router_z_loss": 0.0381,
      "step": 3170
    },
    {
      "balancing_entropy": 1.031,
      "balancing_loss": 1.1273,
      "epoch": 16.22528,
      "grad_norm": 0.76953125,
      "learning_rate": 0.003,
      "loss": 2.3497,
      "router_z_loss": 0.0332,
      "step": 3180
    },
    {
      "balancing_entropy": 1.0183,
      "balancing_loss": 1.1255,
      "epoch": 16.27648,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.3481,
      "router_z_loss": 0.0449,
      "step": 3190
    },
    {
      "balancing_entropy": 1.0169,
      "balancing_loss": 1.1111,
      "epoch": 16.32768,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.3593,
      "router_z_loss": 0.0457,
      "step": 3200
    },
    {
      "balancing_entropy": 1.0312,
      "balancing_loss": 1.1087,
      "epoch": 16.37888,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.3548,
      "router_z_loss": 0.0534,
      "step": 3210
    },
    {
      "balancing_entropy": 1.0515,
      "balancing_loss": 1.1123,
      "epoch": 16.43008,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.3556,
      "router_z_loss": 0.0416,
      "step": 3220
    },
    {
      "balancing_entropy": 1.0469,
      "balancing_loss": 1.1226,
      "epoch": 16.48128,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.3427,
      "router_z_loss": 0.0486,
      "step": 3230
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.1283,
      "epoch": 16.53248,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.3534,
      "router_z_loss": 0.0483,
      "step": 3240
    },
    {
      "balancing_entropy": 1.0217,
      "balancing_loss": 1.114,
      "epoch": 16.58368,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.356,
      "router_z_loss": 0.0623,
      "step": 3250
    },
    {
      "balancing_entropy": 1.0573,
      "balancing_loss": 1.1287,
      "epoch": 16.63488,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.3487,
      "router_z_loss": 0.0699,
      "step": 3260
    },
    {
      "balancing_entropy": 1.0375,
      "balancing_loss": 1.134,
      "epoch": 16.68608,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.3549,
      "router_z_loss": 0.042,
      "step": 3270
    },
    {
      "balancing_entropy": 1.04,
      "balancing_loss": 1.1358,
      "epoch": 16.73728,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.3604,
      "router_z_loss": 0.0331,
      "step": 3280
    },
    {
      "balancing_entropy": 0.9718,
      "balancing_loss": 1.117,
      "epoch": 16.78848,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.3529,
      "router_z_loss": 0.0523,
      "step": 3290
    },
    {
      "balancing_entropy": 1.0086,
      "balancing_loss": 1.1137,
      "epoch": 16.83968,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.352,
      "router_z_loss": 0.0803,
      "step": 3300
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.14,
      "epoch": 16.89088,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.3564,
      "router_z_loss": 0.0418,
      "step": 3310
    },
    {
      "balancing_entropy": 1.0892,
      "balancing_loss": 1.1227,
      "epoch": 16.94208,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.3526,
      "router_z_loss": 0.0308,
      "step": 3320
    },
    {
      "balancing_entropy": 1.0845,
      "balancing_loss": 1.1338,
      "epoch": 16.99328,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.353,
      "router_z_loss": 0.0393,
      "step": 3330
    },
    {
      "balancing_entropy": 1.0336,
      "balancing_loss": 1.1332,
      "epoch": 17.04096,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.3431,
      "router_z_loss": 0.0401,
      "step": 3340
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.1366,
      "epoch": 17.09216,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.339,
      "router_z_loss": 0.0411,
      "step": 3350
    },
    {
      "balancing_entropy": 0.9933,
      "balancing_loss": 1.1317,
      "epoch": 17.14336,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.331,
      "router_z_loss": 0.0535,
      "step": 3360
    },
    {
      "balancing_entropy": 1.0034,
      "balancing_loss": 1.1198,
      "epoch": 17.19456,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.3471,
      "router_z_loss": 0.0803,
      "step": 3370
    },
    {
      "balancing_entropy": 1.056,
      "balancing_loss": 1.1265,
      "epoch": 17.24576,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.341,
      "router_z_loss": 0.0498,
      "step": 3380
    },
    {
      "balancing_entropy": 1.0427,
      "balancing_loss": 1.1375,
      "epoch": 17.29696,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.3335,
      "router_z_loss": 0.0433,
      "step": 3390
    },
    {
      "balancing_entropy": 1.0325,
      "balancing_loss": 1.1195,
      "epoch": 17.34816,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.344,
      "router_z_loss": 0.0626,
      "step": 3400
    },
    {
      "balancing_entropy": 1.0441,
      "balancing_loss": 1.1184,
      "epoch": 17.39936,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.3477,
      "router_z_loss": 0.042,
      "step": 3410
    },
    {
      "balancing_entropy": 1.033,
      "balancing_loss": 1.1332,
      "epoch": 17.45056,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.3299,
      "router_z_loss": 0.0311,
      "step": 3420
    },
    {
      "balancing_entropy": 1.0314,
      "balancing_loss": 1.1342,
      "epoch": 17.50176,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.338,
      "router_z_loss": 0.0438,
      "step": 3430
    },
    {
      "balancing_entropy": 1.0458,
      "balancing_loss": 1.1274,
      "epoch": 17.55296,
      "grad_norm": 0.431640625,
      "learning_rate": 0.003,
      "loss": 2.3454,
      "router_z_loss": 0.0437,
      "step": 3440
    },
    {
      "balancing_entropy": 1.046,
      "balancing_loss": 1.1204,
      "epoch": 17.60416,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.3413,
      "router_z_loss": 0.0402,
      "step": 3450
    },
    {
      "balancing_entropy": 1.0101,
      "balancing_loss": 1.117,
      "epoch": 17.65536,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.339,
      "router_z_loss": 0.0383,
      "step": 3460
    },
    {
      "balancing_entropy": 1.0508,
      "balancing_loss": 1.1372,
      "epoch": 17.70656,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.3426,
      "router_z_loss": 0.0731,
      "step": 3470
    },
    {
      "balancing_entropy": 1.0491,
      "balancing_loss": 1.1222,
      "epoch": 17.75776,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.3409,
      "router_z_loss": 0.0513,
      "step": 3480
    },
    {
      "balancing_entropy": 1.0475,
      "balancing_loss": 1.1388,
      "epoch": 17.80896,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.3349,
      "router_z_loss": 0.0452,
      "step": 3490
    },
    {
      "balancing_entropy": 1.0364,
      "balancing_loss": 1.1411,
      "epoch": 17.86016,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.3313,
      "router_z_loss": 0.0513,
      "step": 3500
    },
    {
      "balancing_entropy": 1.017,
      "balancing_loss": 1.1236,
      "epoch": 17.91136,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.3363,
      "router_z_loss": 0.05,
      "step": 3510
    },
    {
      "balancing_entropy": 1.0213,
      "balancing_loss": 1.1387,
      "epoch": 17.96256,
      "grad_norm": 0.474609375,
      "learning_rate": 0.003,
      "loss": 2.3405,
      "router_z_loss": 0.0683,
      "step": 3520
    },
    {
      "balancing_entropy": 0.9772,
      "balancing_loss": 1.13,
      "epoch": 18.01024,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.3404,
      "router_z_loss": 0.0417,
      "step": 3530
    },
    {
      "balancing_entropy": 1.0304,
      "balancing_loss": 1.1341,
      "epoch": 18.06144,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.3337,
      "router_z_loss": 0.0499,
      "step": 3540
    },
    {
      "balancing_entropy": 1.0296,
      "balancing_loss": 1.1121,
      "epoch": 18.11264,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.3267,
      "router_z_loss": 0.0403,
      "step": 3550
    },
    {
      "balancing_entropy": 1.066,
      "balancing_loss": 1.1384,
      "epoch": 18.16384,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.3238,
      "router_z_loss": 0.04,
      "step": 3560
    },
    {
      "balancing_entropy": 1.0454,
      "balancing_loss": 1.1299,
      "epoch": 18.21504,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.3202,
      "router_z_loss": 0.0293,
      "step": 3570
    },
    {
      "balancing_entropy": 1.0453,
      "balancing_loss": 1.1245,
      "epoch": 18.26624,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.3229,
      "router_z_loss": 0.037,
      "step": 3580
    },
    {
      "balancing_entropy": 0.9897,
      "balancing_loss": 1.1545,
      "epoch": 18.31744,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.3246,
      "router_z_loss": 0.0563,
      "step": 3590
    },
    {
      "balancing_entropy": 1.0073,
      "balancing_loss": 1.1369,
      "epoch": 18.36864,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.3236,
      "router_z_loss": 0.0663,
      "step": 3600
    },
    {
      "balancing_entropy": 1.0421,
      "balancing_loss": 1.1096,
      "epoch": 18.41984,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.3281,
      "router_z_loss": 0.0518,
      "step": 3610
    },
    {
      "balancing_entropy": 1.0917,
      "balancing_loss": 1.1235,
      "epoch": 18.47104,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.3293,
      "router_z_loss": 0.0377,
      "step": 3620
    },
    {
      "balancing_entropy": 1.0374,
      "balancing_loss": 1.1221,
      "epoch": 18.52224,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.3308,
      "router_z_loss": 0.0354,
      "step": 3630
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.1258,
      "epoch": 18.57344,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.327,
      "router_z_loss": 0.0604,
      "step": 3640
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.1386,
      "epoch": 18.62464,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.3345,
      "router_z_loss": 0.0517,
      "step": 3650
    },
    {
      "balancing_entropy": 1.0527,
      "balancing_loss": 1.1294,
      "epoch": 18.67584,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.3365,
      "router_z_loss": 0.035,
      "step": 3660
    },
    {
      "balancing_entropy": 1.0151,
      "balancing_loss": 1.1301,
      "epoch": 18.72704,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.3304,
      "router_z_loss": 0.0367,
      "step": 3670
    },
    {
      "balancing_entropy": 1.0685,
      "balancing_loss": 1.1199,
      "epoch": 18.77824,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.3308,
      "router_z_loss": 0.0512,
      "step": 3680
    },
    {
      "balancing_entropy": 1.0582,
      "balancing_loss": 1.1277,
      "epoch": 18.829439999999998,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.3248,
      "router_z_loss": 0.0356,
      "step": 3690
    },
    {
      "balancing_entropy": 1.0744,
      "balancing_loss": 1.1132,
      "epoch": 18.88064,
      "grad_norm": 0.2392578125,
      "learning_rate": 0.003,
      "loss": 2.3329,
      "router_z_loss": 0.0338,
      "step": 3700
    },
    {
      "balancing_entropy": 1.0843,
      "balancing_loss": 1.1303,
      "epoch": 18.93184,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.33,
      "router_z_loss": 0.0418,
      "step": 3710
    },
    {
      "balancing_entropy": 1.0871,
      "balancing_loss": 1.1295,
      "epoch": 18.98304,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.3322,
      "router_z_loss": 0.0367,
      "step": 3720
    },
    {
      "balancing_entropy": 1.0513,
      "balancing_loss": 1.123,
      "epoch": 19.03072,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.3312,
      "router_z_loss": 0.0394,
      "step": 3730
    },
    {
      "balancing_entropy": 1.0894,
      "balancing_loss": 1.1285,
      "epoch": 19.08192,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.3185,
      "router_z_loss": 0.0432,
      "step": 3740
    },
    {
      "balancing_entropy": 1.0925,
      "balancing_loss": 1.1308,
      "epoch": 19.13312,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.3206,
      "router_z_loss": 0.0477,
      "step": 3750
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.1238,
      "epoch": 19.18432,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.3233,
      "router_z_loss": 0.0395,
      "step": 3760
    },
    {
      "balancing_entropy": 1.0722,
      "balancing_loss": 1.1328,
      "epoch": 19.23552,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.324,
      "router_z_loss": 0.0377,
      "step": 3770
    },
    {
      "balancing_entropy": 1.0895,
      "balancing_loss": 1.1288,
      "epoch": 19.28672,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.3119,
      "router_z_loss": 0.0484,
      "step": 3780
    },
    {
      "balancing_entropy": 1.0912,
      "balancing_loss": 1.1208,
      "epoch": 19.33792,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.3251,
      "router_z_loss": 0.0429,
      "step": 3790
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.1277,
      "epoch": 19.38912,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.3286,
      "router_z_loss": 0.0364,
      "step": 3800
    },
    {
      "balancing_entropy": 1.0604,
      "balancing_loss": 1.128,
      "epoch": 19.44032,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.3202,
      "router_z_loss": 0.0522,
      "step": 3810
    },
    {
      "balancing_entropy": 1.0644,
      "balancing_loss": 1.1246,
      "epoch": 19.49152,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.3231,
      "router_z_loss": 0.0577,
      "step": 3820
    },
    {
      "balancing_entropy": 1.0396,
      "balancing_loss": 1.1341,
      "epoch": 19.54272,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.3307,
      "router_z_loss": 0.0406,
      "step": 3830
    },
    {
      "balancing_entropy": 1.0122,
      "balancing_loss": 1.1193,
      "epoch": 19.59392,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.3299,
      "router_z_loss": 0.0519,
      "step": 3840
    },
    {
      "balancing_entropy": 1.0458,
      "balancing_loss": 1.1171,
      "epoch": 19.64512,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.3236,
      "router_z_loss": 0.0475,
      "step": 3850
    },
    {
      "balancing_entropy": 0.9796,
      "balancing_loss": 1.1462,
      "epoch": 19.69632,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.3271,
      "router_z_loss": 0.0477,
      "step": 3860
    },
    {
      "balancing_entropy": 1.0068,
      "balancing_loss": 1.0965,
      "epoch": 19.74752,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.3292,
      "router_z_loss": 0.0887,
      "step": 3870
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.1434,
      "epoch": 19.79872,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.3255,
      "router_z_loss": 0.068,
      "step": 3880
    },
    {
      "balancing_entropy": 1.0937,
      "balancing_loss": 1.1337,
      "epoch": 19.84992,
      "grad_norm": 0.2421875,
      "learning_rate": 0.003,
      "loss": 2.3212,
      "router_z_loss": 0.0444,
      "step": 3890
    },
    {
      "balancing_entropy": 1.0607,
      "balancing_loss": 1.1281,
      "epoch": 19.90112,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.3333,
      "router_z_loss": 0.032,
      "step": 3900
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.1199,
      "epoch": 19.95232,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.3279,
      "router_z_loss": 0.0464,
      "step": 3910
    },
    {
      "balancing_entropy": 1.0229,
      "balancing_loss": 1.132,
      "epoch": 20.0,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.3329,
      "router_z_loss": 0.0448,
      "step": 3920
    },
    {
      "balancing_entropy": 1.0142,
      "balancing_loss": 1.1257,
      "epoch": 20.0512,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.3077,
      "router_z_loss": 0.049,
      "step": 3930
    },
    {
      "balancing_entropy": 1.0025,
      "balancing_loss": 1.1225,
      "epoch": 20.1024,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.3138,
      "router_z_loss": 0.0816,
      "step": 3940
    },
    {
      "balancing_entropy": 1.0195,
      "balancing_loss": 1.1433,
      "epoch": 20.1536,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.313,
      "router_z_loss": 0.087,
      "step": 3950
    },
    {
      "balancing_entropy": 1.0725,
      "balancing_loss": 1.1381,
      "epoch": 20.2048,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.3157,
      "router_z_loss": 0.0779,
      "step": 3960
    },
    {
      "balancing_entropy": 1.0654,
      "balancing_loss": 1.1203,
      "epoch": 20.256,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.3124,
      "router_z_loss": 0.0429,
      "step": 3970
    },
    {
      "balancing_entropy": 1.0739,
      "balancing_loss": 1.1384,
      "epoch": 20.3072,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.3125,
      "router_z_loss": 0.05,
      "step": 3980
    },
    {
      "balancing_entropy": 1.0768,
      "balancing_loss": 1.134,
      "epoch": 20.3584,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.313,
      "router_z_loss": 0.0368,
      "step": 3990
    },
    {
      "balancing_entropy": 1.0431,
      "balancing_loss": 1.1251,
      "epoch": 20.4096,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.3163,
      "router_z_loss": 0.0372,
      "step": 4000
    },
    {
      "balancing_entropy": 1.0229,
      "balancing_loss": 1.1299,
      "epoch": 20.4608,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.3179,
      "router_z_loss": 0.0571,
      "step": 4010
    },
    {
      "balancing_entropy": 1.0859,
      "balancing_loss": 1.1238,
      "epoch": 20.512,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.3196,
      "router_z_loss": 0.0571,
      "step": 4020
    },
    {
      "balancing_entropy": 1.0828,
      "balancing_loss": 1.1359,
      "epoch": 20.5632,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.3168,
      "router_z_loss": 0.0353,
      "step": 4030
    },
    {
      "balancing_entropy": 1.0532,
      "balancing_loss": 1.133,
      "epoch": 20.6144,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.3206,
      "router_z_loss": 0.0357,
      "step": 4040
    },
    {
      "balancing_entropy": 1.0563,
      "balancing_loss": 1.1198,
      "epoch": 20.6656,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.3145,
      "router_z_loss": 0.055,
      "step": 4050
    },
    {
      "balancing_entropy": 1.0531,
      "balancing_loss": 1.1521,
      "epoch": 20.7168,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.3216,
      "router_z_loss": 0.06,
      "step": 4060
    },
    {
      "balancing_entropy": 1.0252,
      "balancing_loss": 1.1323,
      "epoch": 20.768,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.3264,
      "router_z_loss": 0.0396,
      "step": 4070
    },
    {
      "balancing_entropy": 1.0652,
      "balancing_loss": 1.129,
      "epoch": 20.8192,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.3155,
      "router_z_loss": 0.05,
      "step": 4080
    },
    {
      "balancing_entropy": 1.0933,
      "balancing_loss": 1.134,
      "epoch": 20.8704,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.3148,
      "router_z_loss": 0.036,
      "step": 4090
    },
    {
      "balancing_entropy": 1.0958,
      "balancing_loss": 1.128,
      "epoch": 20.9216,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.3119,
      "router_z_loss": 0.0301,
      "step": 4100
    },
    {
      "balancing_entropy": 1.0887,
      "balancing_loss": 1.1354,
      "epoch": 20.9728,
      "grad_norm": 0.2197265625,
      "learning_rate": 0.003,
      "loss": 2.3083,
      "router_z_loss": 0.0324,
      "step": 4110
    },
    {
      "balancing_entropy": 1.0297,
      "balancing_loss": 1.1405,
      "epoch": 21.02048,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.3102,
      "router_z_loss": 0.0326,
      "step": 4120
    },
    {
      "balancing_entropy": 1.0745,
      "balancing_loss": 1.139,
      "epoch": 21.07168,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.3082,
      "router_z_loss": 0.0506,
      "step": 4130
    },
    {
      "balancing_entropy": 1.0805,
      "balancing_loss": 1.1195,
      "epoch": 21.12288,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.2994,
      "router_z_loss": 0.0352,
      "step": 4140
    },
    {
      "balancing_entropy": 1.0906,
      "balancing_loss": 1.1368,
      "epoch": 21.17408,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.298,
      "router_z_loss": 0.0352,
      "step": 4150
    },
    {
      "balancing_entropy": 1.0787,
      "balancing_loss": 1.1367,
      "epoch": 21.22528,
      "grad_norm": 0.234375,
      "learning_rate": 0.003,
      "loss": 2.2971,
      "router_z_loss": 0.0368,
      "step": 4160
    },
    {
      "balancing_entropy": 1.0896,
      "balancing_loss": 1.1229,
      "epoch": 21.27648,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.2993,
      "router_z_loss": 0.0319,
      "step": 4170
    },
    {
      "balancing_entropy": 1.0083,
      "balancing_loss": 1.1473,
      "epoch": 21.32768,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.3058,
      "router_z_loss": 0.0505,
      "step": 4180
    },
    {
      "balancing_entropy": 1.0559,
      "balancing_loss": 1.1424,
      "epoch": 21.37888,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.3025,
      "router_z_loss": 0.0774,
      "step": 4190
    },
    {
      "balancing_entropy": 1.0491,
      "balancing_loss": 1.1169,
      "epoch": 21.43008,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.3018,
      "router_z_loss": 0.043,
      "step": 4200
    },
    {
      "balancing_entropy": 1.0295,
      "balancing_loss": 1.1264,
      "epoch": 21.48128,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.3103,
      "router_z_loss": 0.0752,
      "step": 4210
    },
    {
      "balancing_entropy": 1.0557,
      "balancing_loss": 1.119,
      "epoch": 21.53248,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.3019,
      "router_z_loss": 0.0842,
      "step": 4220
    },
    {
      "balancing_entropy": 1.0386,
      "balancing_loss": 1.1247,
      "epoch": 21.58368,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.3054,
      "router_z_loss": 0.0546,
      "step": 4230
    },
    {
      "balancing_entropy": 1.0632,
      "balancing_loss": 1.1399,
      "epoch": 21.63488,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.307,
      "router_z_loss": 0.0554,
      "step": 4240
    },
    {
      "balancing_entropy": 1.0652,
      "balancing_loss": 1.1398,
      "epoch": 21.68608,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.2999,
      "router_z_loss": 0.0542,
      "step": 4250
    },
    {
      "balancing_entropy": 1.051,
      "balancing_loss": 1.1272,
      "epoch": 21.73728,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.3111,
      "router_z_loss": 0.0523,
      "step": 4260
    },
    {
      "balancing_entropy": 1.0284,
      "balancing_loss": 1.1418,
      "epoch": 21.78848,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.3119,
      "router_z_loss": 0.0507,
      "step": 4270
    },
    {
      "balancing_entropy": 1.0015,
      "balancing_loss": 1.11,
      "epoch": 21.83968,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.3048,
      "router_z_loss": 0.0512,
      "step": 4280
    },
    {
      "balancing_entropy": 1.0566,
      "balancing_loss": 1.1091,
      "epoch": 21.89088,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.3144,
      "router_z_loss": 0.0598,
      "step": 4290
    },
    {
      "balancing_entropy": 1.0463,
      "balancing_loss": 1.1171,
      "epoch": 21.94208,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.3107,
      "router_z_loss": 0.0398,
      "step": 4300
    },
    {
      "balancing_entropy": 1.0103,
      "balancing_loss": 1.1259,
      "epoch": 21.99328,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.3131,
      "router_z_loss": 0.0561,
      "step": 4310
    },
    {
      "balancing_entropy": 1.035,
      "balancing_loss": 1.1451,
      "epoch": 22.04096,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.2944,
      "router_z_loss": 0.0544,
      "step": 4320
    },
    {
      "balancing_entropy": 1.0442,
      "balancing_loss": 1.1572,
      "epoch": 22.09216,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.2986,
      "router_z_loss": 0.0582,
      "step": 4330
    },
    {
      "balancing_entropy": 1.0668,
      "balancing_loss": 1.1234,
      "epoch": 22.14336,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.2902,
      "router_z_loss": 0.0543,
      "step": 4340
    },
    {
      "balancing_entropy": 1.0114,
      "balancing_loss": 1.1257,
      "epoch": 22.19456,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.2964,
      "router_z_loss": 0.053,
      "step": 4350
    },
    {
      "balancing_entropy": 1.0688,
      "balancing_loss": 1.1459,
      "epoch": 22.24576,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.2852,
      "router_z_loss": 0.0609,
      "step": 4360
    },
    {
      "balancing_entropy": 1.0939,
      "balancing_loss": 1.1498,
      "epoch": 22.29696,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.2868,
      "router_z_loss": 0.0383,
      "step": 4370
    },
    {
      "balancing_entropy": 1.0894,
      "balancing_loss": 1.1373,
      "epoch": 22.34816,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.2948,
      "router_z_loss": 0.0366,
      "step": 4380
    },
    {
      "balancing_entropy": 1.0741,
      "balancing_loss": 1.13,
      "epoch": 22.39936,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.2896,
      "router_z_loss": 0.0337,
      "step": 4390
    },
    {
      "balancing_entropy": 1.0656,
      "balancing_loss": 1.1352,
      "epoch": 22.45056,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.2973,
      "router_z_loss": 0.0445,
      "step": 4400
    },
    {
      "balancing_entropy": 1.021,
      "balancing_loss": 1.1298,
      "epoch": 22.50176,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.3002,
      "router_z_loss": 0.0504,
      "step": 4410
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.145,
      "epoch": 22.55296,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.2945,
      "router_z_loss": 0.0491,
      "step": 4420
    },
    {
      "balancing_entropy": 1.0332,
      "balancing_loss": 1.1362,
      "epoch": 22.60416,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.2972,
      "router_z_loss": 0.0441,
      "step": 4430
    },
    {
      "balancing_entropy": 0.9803,
      "balancing_loss": 1.1438,
      "epoch": 22.65536,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.3051,
      "router_z_loss": 0.0522,
      "step": 4440
    },
    {
      "balancing_entropy": 1.0601,
      "balancing_loss": 1.1232,
      "epoch": 22.70656,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.2975,
      "router_z_loss": 0.0611,
      "step": 4450
    },
    {
      "balancing_entropy": 1.007,
      "balancing_loss": 1.1101,
      "epoch": 22.75776,
      "grad_norm": 0.67578125,
      "learning_rate": 0.003,
      "loss": 2.2998,
      "router_z_loss": 0.0452,
      "step": 4460
    },
    {
      "balancing_entropy": 1.0561,
      "balancing_loss": 1.1192,
      "epoch": 22.80896,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.2935,
      "router_z_loss": 0.0554,
      "step": 4470
    },
    {
      "balancing_entropy": 1.0627,
      "balancing_loss": 1.1333,
      "epoch": 22.86016,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.2923,
      "router_z_loss": 0.0465,
      "step": 4480
    },
    {
      "balancing_entropy": 1.06,
      "balancing_loss": 1.1433,
      "epoch": 22.91136,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.2996,
      "router_z_loss": 0.0432,
      "step": 4490
    },
    {
      "balancing_entropy": 1.0649,
      "balancing_loss": 1.1329,
      "epoch": 22.96256,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.297,
      "router_z_loss": 0.0527,
      "step": 4500
    },
    {
      "balancing_entropy": 1.0084,
      "balancing_loss": 1.163,
      "epoch": 23.01024,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.2983,
      "router_z_loss": 0.0448,
      "step": 4510
    },
    {
      "balancing_entropy": 1.0353,
      "balancing_loss": 1.1122,
      "epoch": 23.06144,
      "grad_norm": 0.265625,
      "learning_rate": 0.003,
      "loss": 2.2864,
      "router_z_loss": 0.0556,
      "step": 4520
    },
    {
      "balancing_entropy": 1.0281,
      "balancing_loss": 1.1533,
      "epoch": 23.11264,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.2755,
      "router_z_loss": 0.0569,
      "step": 4530
    },
    {
      "balancing_entropy": 1.0706,
      "balancing_loss": 1.1579,
      "epoch": 23.16384,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.288,
      "router_z_loss": 0.0515,
      "step": 4540
    },
    {
      "balancing_entropy": 1.0586,
      "balancing_loss": 1.1544,
      "epoch": 23.21504,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.2825,
      "router_z_loss": 0.0532,
      "step": 4550
    },
    {
      "balancing_entropy": 1.0204,
      "balancing_loss": 1.1167,
      "epoch": 23.26624,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.2936,
      "router_z_loss": 0.0474,
      "step": 4560
    },
    {
      "balancing_entropy": 1.0752,
      "balancing_loss": 1.1596,
      "epoch": 23.31744,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.2838,
      "router_z_loss": 0.063,
      "step": 4570
    },
    {
      "balancing_entropy": 0.9842,
      "balancing_loss": 1.1321,
      "epoch": 23.36864,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.2884,
      "router_z_loss": 0.0563,
      "step": 4580
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.1547,
      "epoch": 23.41984,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.2874,
      "router_z_loss": 0.082,
      "step": 4590
    },
    {
      "balancing_entropy": 1.0761,
      "balancing_loss": 1.14,
      "epoch": 23.47104,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.2872,
      "router_z_loss": 0.0696,
      "step": 4600
    },
    {
      "balancing_entropy": 1.0439,
      "balancing_loss": 1.1381,
      "epoch": 23.52224,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.2864,
      "router_z_loss": 0.0576,
      "step": 4610
    },
    {
      "balancing_entropy": 1.0911,
      "balancing_loss": 1.1326,
      "epoch": 23.57344,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.2905,
      "router_z_loss": 0.0555,
      "step": 4620
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.1211,
      "epoch": 23.62464,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.2932,
      "router_z_loss": 0.0454,
      "step": 4630
    },
    {
      "balancing_entropy": 1.0509,
      "balancing_loss": 1.1415,
      "epoch": 23.67584,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.2851,
      "router_z_loss": 0.0415,
      "step": 4640
    },
    {
      "balancing_entropy": 1.0814,
      "balancing_loss": 1.139,
      "epoch": 23.72704,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.2815,
      "router_z_loss": 0.0463,
      "step": 4650
    },
    {
      "balancing_entropy": 1.0366,
      "balancing_loss": 1.1473,
      "epoch": 23.77824,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.2927,
      "router_z_loss": 0.0513,
      "step": 4660
    },
    {
      "balancing_entropy": 1.0449,
      "balancing_loss": 1.1243,
      "epoch": 23.829439999999998,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.2838,
      "router_z_loss": 0.0461,
      "step": 4670
    },
    {
      "balancing_entropy": 1.0811,
      "balancing_loss": 1.1359,
      "epoch": 23.88064,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.2877,
      "router_z_loss": 0.0451,
      "step": 4680
    },
    {
      "balancing_entropy": 1.0638,
      "balancing_loss": 1.1632,
      "epoch": 23.93184,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.2878,
      "router_z_loss": 0.0387,
      "step": 4690
    },
    {
      "balancing_entropy": 1.0699,
      "balancing_loss": 1.1294,
      "epoch": 23.98304,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.2858,
      "router_z_loss": 0.0398,
      "step": 4700
    },
    {
      "balancing_entropy": 1.0435,
      "balancing_loss": 1.1169,
      "epoch": 24.03072,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.289,
      "router_z_loss": 0.0401,
      "step": 4710
    },
    {
      "balancing_entropy": 1.0646,
      "balancing_loss": 1.1232,
      "epoch": 24.08192,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.2828,
      "router_z_loss": 0.0396,
      "step": 4720
    },
    {
      "balancing_entropy": 1.0821,
      "balancing_loss": 1.1243,
      "epoch": 24.13312,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.2779,
      "router_z_loss": 0.0406,
      "step": 4730
    },
    {
      "balancing_entropy": 1.0832,
      "balancing_loss": 1.1521,
      "epoch": 24.18432,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.2733,
      "router_z_loss": 0.0439,
      "step": 4740
    },
    {
      "balancing_entropy": 1.0704,
      "balancing_loss": 1.1228,
      "epoch": 24.23552,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.2775,
      "router_z_loss": 0.0462,
      "step": 4750
    },
    {
      "balancing_entropy": 1.0795,
      "balancing_loss": 1.1359,
      "epoch": 24.28672,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.2688,
      "router_z_loss": 0.043,
      "step": 4760
    },
    {
      "balancing_entropy": 1.0511,
      "balancing_loss": 1.1325,
      "epoch": 24.33792,
      "grad_norm": 0.2421875,
      "learning_rate": 0.003,
      "loss": 2.28,
      "router_z_loss": 0.0428,
      "step": 4770
    },
    {
      "balancing_entropy": 1.037,
      "balancing_loss": 1.1263,
      "epoch": 24.38912,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.2732,
      "router_z_loss": 0.0447,
      "step": 4780
    },
    {
      "balancing_entropy": 1.0001,
      "balancing_loss": 1.1322,
      "epoch": 24.44032,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.2847,
      "router_z_loss": 0.0459,
      "step": 4790
    },
    {
      "balancing_entropy": 1.039,
      "balancing_loss": 1.1448,
      "epoch": 24.49152,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.2863,
      "router_z_loss": 0.0461,
      "step": 4800
    },
    {
      "balancing_entropy": 1.0218,
      "balancing_loss": 1.1421,
      "epoch": 24.54272,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.2808,
      "router_z_loss": 0.0468,
      "step": 4810
    },
    {
      "balancing_entropy": 1.0368,
      "balancing_loss": 1.145,
      "epoch": 24.59392,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.2825,
      "router_z_loss": 0.0523,
      "step": 4820
    },
    {
      "balancing_entropy": 1.0493,
      "balancing_loss": 1.1353,
      "epoch": 24.64512,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.2823,
      "router_z_loss": 0.0485,
      "step": 4830
    },
    {
      "balancing_entropy": 1.0598,
      "balancing_loss": 1.1165,
      "epoch": 24.69632,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.2776,
      "router_z_loss": 0.0428,
      "step": 4840
    },
    {
      "balancing_entropy": 1.0402,
      "balancing_loss": 1.1347,
      "epoch": 24.74752,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.2834,
      "router_z_loss": 0.0458,
      "step": 4850
    },
    {
      "balancing_entropy": 1.0102,
      "balancing_loss": 1.1622,
      "epoch": 24.79872,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.2829,
      "router_z_loss": 0.0509,
      "step": 4860
    },
    {
      "balancing_entropy": 0.9894,
      "balancing_loss": 1.1695,
      "epoch": 24.84992,
      "grad_norm": 0.86328125,
      "learning_rate": 0.003,
      "loss": 2.2829,
      "router_z_loss": 0.0572,
      "step": 4870
    },
    {
      "balancing_entropy": 1.044,
      "balancing_loss": 1.1283,
      "epoch": 24.90112,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.2855,
      "router_z_loss": 0.0592,
      "step": 4880
    },
    {
      "balancing_entropy": 1.084,
      "balancing_loss": 1.1366,
      "epoch": 24.95232,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.279,
      "router_z_loss": 0.0519,
      "step": 4890
    },
    {
      "balancing_entropy": 1.0463,
      "balancing_loss": 1.1289,
      "epoch": 25.0,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.281,
      "router_z_loss": 0.0457,
      "step": 4900
    },
    {
      "balancing_entropy": 1.0756,
      "balancing_loss": 1.1409,
      "epoch": 25.0512,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.2624,
      "router_z_loss": 0.0432,
      "step": 4910
    },
    {
      "balancing_entropy": 1.0128,
      "balancing_loss": 1.1419,
      "epoch": 25.1024,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.2761,
      "router_z_loss": 0.043,
      "step": 4920
    },
    {
      "balancing_entropy": 1.0785,
      "balancing_loss": 1.1306,
      "epoch": 25.1536,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.2716,
      "router_z_loss": 0.0464,
      "step": 4930
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.1443,
      "epoch": 25.2048,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.2666,
      "router_z_loss": 0.0451,
      "step": 4940
    },
    {
      "balancing_entropy": 1.0652,
      "balancing_loss": 1.1235,
      "epoch": 25.256,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.27,
      "router_z_loss": 0.0429,
      "step": 4950
    },
    {
      "balancing_entropy": 1.0634,
      "balancing_loss": 1.1462,
      "epoch": 25.3072,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.2702,
      "router_z_loss": 0.0448,
      "step": 4960
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.1268,
      "epoch": 25.3584,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.2745,
      "router_z_loss": 0.0429,
      "step": 4970
    },
    {
      "balancing_entropy": 1.0726,
      "balancing_loss": 1.1487,
      "epoch": 25.4096,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.2655,
      "router_z_loss": 0.04,
      "step": 4980
    },
    {
      "balancing_entropy": 1.0657,
      "balancing_loss": 1.1231,
      "epoch": 25.4608,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.2679,
      "router_z_loss": 0.0376,
      "step": 4990
    },
    {
      "balancing_entropy": 1.0308,
      "balancing_loss": 1.1208,
      "epoch": 25.512,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.2698,
      "router_z_loss": 0.039,
      "step": 5000
    }
  ],
  "logging_steps": 10,
  "max_steps": 19500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 5000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 3.17699824084992e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}