{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 99.49152,
  "eval_steps": 500,
  "global_step": 19500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "balancing_entropy": 1.0585,
      "balancing_loss": 1.1163,
      "epoch": 0.0512,
      "grad_norm": 0.9375,
      "learning_rate": 3.076923076923077e-05,
      "loss": 12.4732,
      "router_z_loss": 1.3444,
      "step": 10
    },
    {
      "balancing_entropy": 1.0779,
      "balancing_loss": 1.1163,
      "epoch": 0.1024,
      "grad_norm": 0.91015625,
      "learning_rate": 6.153846153846154e-05,
      "loss": 12.2404,
      "router_z_loss": 1.3439,
      "step": 20
    },
    {
      "balancing_entropy": 1.085,
      "balancing_loss": 1.1169,
      "epoch": 0.1536,
      "grad_norm": 0.85546875,
      "learning_rate": 9.230769230769232e-05,
      "loss": 11.7854,
      "router_z_loss": 1.3334,
      "step": 30
    },
    {
      "balancing_entropy": 1.0844,
      "balancing_loss": 1.1186,
      "epoch": 0.2048,
      "grad_norm": 0.8359375,
      "learning_rate": 0.00012307692307692307,
      "loss": 11.117,
      "router_z_loss": 1.2929,
      "step": 40
    },
    {
      "balancing_entropy": 1.0915,
      "balancing_loss": 1.1168,
      "epoch": 0.256,
      "grad_norm": 1.1640625,
      "learning_rate": 0.00015384615384615385,
      "loss": 10.173,
      "router_z_loss": 1.234,
      "step": 50
    },
    {
      "balancing_entropy": 1.0902,
      "balancing_loss": 1.1168,
      "epoch": 0.3072,
      "grad_norm": 0.6796875,
      "learning_rate": 0.00018461538461538463,
      "loss": 9.2854,
      "router_z_loss": 1.1822,
      "step": 60
    },
    {
      "balancing_entropy": 1.0941,
      "balancing_loss": 1.1157,
      "epoch": 0.3584,
      "grad_norm": 1.1875,
      "learning_rate": 0.0002153846153846154,
      "loss": 8.626,
      "router_z_loss": 1.1239,
      "step": 70
    },
    {
      "balancing_entropy": 1.0933,
      "balancing_loss": 1.1186,
      "epoch": 0.4096,
      "grad_norm": 7.875,
      "learning_rate": 0.00024615384615384614,
      "loss": 8.2062,
      "router_z_loss": 0.98,
      "step": 80
    },
    {
      "balancing_entropy": 1.0709,
      "balancing_loss": 1.119,
      "epoch": 0.4608,
      "grad_norm": 1.6875,
      "learning_rate": 0.00027692307692307695,
      "loss": 7.9262,
      "router_z_loss": 0.8035,
      "step": 90
    },
    {
      "balancing_entropy": 1.0497,
      "balancing_loss": 1.12,
      "epoch": 0.512,
      "grad_norm": 0.96875,
      "learning_rate": 0.0003076923076923077,
      "loss": 7.5813,
      "router_z_loss": 0.6652,
      "step": 100
    },
    {
      "balancing_entropy": 1.0792,
      "balancing_loss": 1.1095,
      "epoch": 0.5632,
      "grad_norm": 0.69921875,
      "learning_rate": 0.00033846153846153846,
      "loss": 7.1597,
      "router_z_loss": 0.5917,
      "step": 110
    },
    {
      "balancing_entropy": 1.0919,
      "balancing_loss": 1.1034,
      "epoch": 0.6144,
      "grad_norm": 0.396484375,
      "learning_rate": 0.00036923076923076927,
      "loss": 6.6704,
      "router_z_loss": 0.4439,
      "step": 120
    },
    {
      "balancing_entropy": 1.085,
      "balancing_loss": 1.1135,
      "epoch": 0.6656,
      "grad_norm": 1.0390625,
      "learning_rate": 0.0004,
      "loss": 6.3153,
      "router_z_loss": 0.3157,
      "step": 130
    },
    {
      "balancing_entropy": 1.0874,
      "balancing_loss": 1.12,
      "epoch": 0.7168,
      "grad_norm": 0.51171875,
      "learning_rate": 0.0004307692307692308,
      "loss": 6.0417,
      "router_z_loss": 0.2864,
      "step": 140
    },
    {
      "balancing_entropy": 1.082,
      "balancing_loss": 1.1019,
      "epoch": 0.768,
      "grad_norm": 0.90234375,
      "learning_rate": 0.0004615384615384616,
      "loss": 5.8099,
      "router_z_loss": 0.2192,
      "step": 150
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.108,
      "epoch": 0.8192,
      "grad_norm": 0.5625,
      "learning_rate": 0.0004923076923076923,
      "loss": 5.6118,
      "router_z_loss": 0.2121,
      "step": 160
    },
    {
      "balancing_entropy": 1.0919,
      "balancing_loss": 1.1049,
      "epoch": 0.8704,
      "grad_norm": 0.66015625,
      "learning_rate": 0.0005230769230769231,
      "loss": 5.4182,
      "router_z_loss": 0.1715,
      "step": 170
    },
    {
      "balancing_entropy": 1.0925,
      "balancing_loss": 1.1085,
      "epoch": 0.9216,
      "grad_norm": 0.875,
      "learning_rate": 0.0005538461538461539,
      "loss": 5.2536,
      "router_z_loss": 0.1392,
      "step": 180
    },
    {
      "balancing_entropy": 1.087,
      "balancing_loss": 1.099,
      "epoch": 0.9728,
      "grad_norm": 0.60546875,
      "learning_rate": 0.0005846153846153846,
      "loss": 5.1422,
      "router_z_loss": 0.1122,
      "step": 190
    },
    {
      "balancing_entropy": 1.0531,
      "balancing_loss": 1.0969,
      "epoch": 1.02048,
      "grad_norm": 0.703125,
      "learning_rate": 0.0006153846153846154,
      "loss": 4.9685,
      "router_z_loss": 0.0989,
      "step": 200
    },
    {
      "balancing_entropy": 1.0936,
      "balancing_loss": 1.0961,
      "epoch": 1.07168,
      "grad_norm": 0.50390625,
      "learning_rate": 0.0006461538461538462,
      "loss": 4.8081,
      "router_z_loss": 0.0862,
      "step": 210
    },
    {
      "balancing_entropy": 1.0895,
      "balancing_loss": 1.0895,
      "epoch": 1.12288,
      "grad_norm": 0.94921875,
      "learning_rate": 0.0006769230769230769,
      "loss": 4.7005,
      "router_z_loss": 0.088,
      "step": 220
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.0874,
      "epoch": 1.17408,
      "grad_norm": 0.72265625,
      "learning_rate": 0.0007076923076923077,
      "loss": 4.5767,
      "router_z_loss": 0.0796,
      "step": 230
    },
    {
      "balancing_entropy": 1.0835,
      "balancing_loss": 1.0894,
      "epoch": 1.22528,
      "grad_norm": 0.7578125,
      "learning_rate": 0.0007384615384615385,
      "loss": 4.4502,
      "router_z_loss": 0.078,
      "step": 240
    },
    {
      "balancing_entropy": 1.0882,
      "balancing_loss": 1.0813,
      "epoch": 1.27648,
      "grad_norm": 0.474609375,
      "learning_rate": 0.0007692307692307692,
      "loss": 4.3645,
      "router_z_loss": 0.079,
      "step": 250
    },
    {
      "balancing_entropy": 1.0839,
      "balancing_loss": 1.0749,
      "epoch": 1.32768,
      "grad_norm": 1.5234375,
      "learning_rate": 0.0008,
      "loss": 4.2608,
      "router_z_loss": 0.0696,
      "step": 260
    },
    {
      "balancing_entropy": 1.0859,
      "balancing_loss": 1.0747,
      "epoch": 1.37888,
      "grad_norm": 0.80078125,
      "learning_rate": 0.0008307692307692308,
      "loss": 4.1836,
      "router_z_loss": 0.071,
      "step": 270
    },
    {
      "balancing_entropy": 1.0857,
      "balancing_loss": 1.0699,
      "epoch": 1.43008,
      "grad_norm": 0.65625,
      "learning_rate": 0.0008615384615384615,
      "loss": 4.1049,
      "router_z_loss": 0.0683,
      "step": 280
    },
    {
      "balancing_entropy": 1.0855,
      "balancing_loss": 1.0694,
      "epoch": 1.48128,
      "grad_norm": 1.265625,
      "learning_rate": 0.0008923076923076924,
      "loss": 4.0122,
      "router_z_loss": 0.06,
      "step": 290
    },
    {
      "balancing_entropy": 1.082,
      "balancing_loss": 1.0749,
      "epoch": 1.53248,
      "grad_norm": 1.0859375,
      "learning_rate": 0.0009230769230769232,
      "loss": 3.9697,
      "router_z_loss": 0.062,
      "step": 300
    },
    {
      "balancing_entropy": 1.0865,
      "balancing_loss": 1.0718,
      "epoch": 1.58368,
      "grad_norm": 0.58203125,
      "learning_rate": 0.0009538461538461538,
      "loss": 3.8966,
      "router_z_loss": 0.0678,
      "step": 310
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.0702,
      "epoch": 1.6348799999999999,
      "grad_norm": 0.5703125,
      "learning_rate": 0.0009846153846153846,
      "loss": 3.8502,
      "router_z_loss": 0.0603,
      "step": 320
    },
    {
      "balancing_entropy": 1.0859,
      "balancing_loss": 1.0703,
      "epoch": 1.68608,
      "grad_norm": 0.546875,
      "learning_rate": 0.0010153846153846155,
      "loss": 3.7763,
      "router_z_loss": 0.062,
      "step": 330
    },
    {
      "balancing_entropy": 1.0814,
      "balancing_loss": 1.0721,
      "epoch": 1.7372800000000002,
      "grad_norm": 0.62109375,
      "learning_rate": 0.0010461538461538462,
      "loss": 3.7262,
      "router_z_loss": 0.0612,
      "step": 340
    },
    {
      "balancing_entropy": 1.0835,
      "balancing_loss": 1.0691,
      "epoch": 1.7884799999999998,
      "grad_norm": 0.82421875,
      "learning_rate": 0.0010769230769230769,
      "loss": 3.6827,
      "router_z_loss": 0.0585,
      "step": 350
    },
    {
      "balancing_entropy": 1.0856,
      "balancing_loss": 1.0733,
      "epoch": 1.83968,
      "grad_norm": 0.484375,
      "learning_rate": 0.0011076923076923078,
      "loss": 3.6388,
      "router_z_loss": 0.0605,
      "step": 360
    },
    {
      "balancing_entropy": 1.0823,
      "balancing_loss": 1.0579,
      "epoch": 1.8908800000000001,
      "grad_norm": 0.6796875,
      "learning_rate": 0.0011384615384615385,
      "loss": 3.6038,
      "router_z_loss": 0.0595,
      "step": 370
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.0716,
      "epoch": 1.94208,
      "grad_norm": 0.53125,
      "learning_rate": 0.0011692307692307692,
      "loss": 3.5645,
      "router_z_loss": 0.0691,
      "step": 380
    },
    {
      "balancing_entropy": 1.0764,
      "balancing_loss": 1.0704,
      "epoch": 1.99328,
      "grad_norm": 1.421875,
      "learning_rate": 0.0012000000000000001,
      "loss": 3.5027,
      "router_z_loss": 0.0658,
      "step": 390
    },
    {
      "balancing_entropy": 1.0343,
      "balancing_loss": 1.0689,
      "epoch": 2.04096,
      "grad_norm": 0.53125,
      "learning_rate": 0.0012307692307692308,
      "loss": 3.501,
      "router_z_loss": 0.0709,
      "step": 400
    },
    {
      "balancing_entropy": 1.0791,
      "balancing_loss": 1.0659,
      "epoch": 2.09216,
      "grad_norm": 0.5546875,
      "learning_rate": 0.0012615384615384615,
      "loss": 3.4561,
      "router_z_loss": 0.0655,
      "step": 410
    },
    {
      "balancing_entropy": 1.0815,
      "balancing_loss": 1.0648,
      "epoch": 2.14336,
      "grad_norm": 0.9453125,
      "learning_rate": 0.0012923076923076924,
      "loss": 3.4314,
      "router_z_loss": 0.0617,
      "step": 420
    },
    {
      "balancing_entropy": 1.0855,
      "balancing_loss": 1.0706,
      "epoch": 2.19456,
      "grad_norm": 0.59375,
      "learning_rate": 0.0013230769230769231,
      "loss": 3.3993,
      "router_z_loss": 0.0597,
      "step": 430
    },
    {
      "balancing_entropy": 1.0717,
      "balancing_loss": 1.0613,
      "epoch": 2.24576,
      "grad_norm": 0.77734375,
      "learning_rate": 0.0013538461538461538,
      "loss": 3.3538,
      "router_z_loss": 0.0525,
      "step": 440
    },
    {
      "balancing_entropy": 1.0788,
      "balancing_loss": 1.0573,
      "epoch": 2.29696,
      "grad_norm": 0.6171875,
      "learning_rate": 0.0013846153846153847,
      "loss": 3.3785,
      "router_z_loss": 0.062,
      "step": 450
    },
    {
      "balancing_entropy": 1.0796,
      "balancing_loss": 1.0691,
      "epoch": 2.34816,
      "grad_norm": 0.99609375,
      "learning_rate": 0.0014153846153846154,
      "loss": 3.3269,
      "router_z_loss": 0.0611,
      "step": 460
    },
    {
      "balancing_entropy": 1.0865,
      "balancing_loss": 1.0642,
      "epoch": 2.39936,
      "grad_norm": 0.494140625,
      "learning_rate": 0.0014461538461538461,
      "loss": 3.3115,
      "router_z_loss": 0.0573,
      "step": 470
    },
    {
      "balancing_entropy": 1.0769,
      "balancing_loss": 1.0715,
      "epoch": 2.45056,
      "grad_norm": 0.6171875,
      "learning_rate": 0.001476923076923077,
      "loss": 3.2882,
      "router_z_loss": 0.0497,
      "step": 480
    },
    {
      "balancing_entropy": 1.0801,
      "balancing_loss": 1.061,
      "epoch": 2.50176,
      "grad_norm": 0.54296875,
      "learning_rate": 0.0015076923076923078,
      "loss": 3.2401,
      "router_z_loss": 0.0465,
      "step": 490
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.0564,
      "epoch": 2.55296,
      "grad_norm": 0.498046875,
      "learning_rate": 0.0015384615384615385,
      "loss": 3.2675,
      "router_z_loss": 0.0505,
      "step": 500
    },
    {
      "balancing_entropy": 1.0722,
      "balancing_loss": 1.0688,
      "epoch": 2.6041600000000003,
      "grad_norm": 0.427734375,
      "learning_rate": 0.0015692307692307694,
      "loss": 3.22,
      "router_z_loss": 0.0504,
      "step": 510
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.0589,
      "epoch": 2.65536,
      "grad_norm": 0.4296875,
      "learning_rate": 0.0016,
      "loss": 3.2273,
      "router_z_loss": 0.0603,
      "step": 520
    },
    {
      "balancing_entropy": 1.071,
      "balancing_loss": 1.0632,
      "epoch": 2.70656,
      "grad_norm": 0.73828125,
      "learning_rate": 0.0016307692307692308,
      "loss": 3.1923,
      "router_z_loss": 0.0643,
      "step": 530
    },
    {
      "balancing_entropy": 1.0669,
      "balancing_loss": 1.0707,
      "epoch": 2.75776,
      "grad_norm": 0.5234375,
      "learning_rate": 0.0016615384615384617,
      "loss": 3.1995,
      "router_z_loss": 0.0867,
      "step": 540
    },
    {
      "balancing_entropy": 1.0739,
      "balancing_loss": 1.0586,
      "epoch": 2.80896,
      "grad_norm": 0.435546875,
      "learning_rate": 0.0016923076923076924,
      "loss": 3.1735,
      "router_z_loss": 0.0697,
      "step": 550
    },
    {
      "balancing_entropy": 1.0794,
      "balancing_loss": 1.0549,
      "epoch": 2.86016,
      "grad_norm": 0.40234375,
      "learning_rate": 0.001723076923076923,
      "loss": 3.1604,
      "router_z_loss": 0.0523,
      "step": 560
    },
    {
      "balancing_entropy": 1.0743,
      "balancing_loss": 1.0687,
      "epoch": 2.91136,
      "grad_norm": 0.38671875,
      "learning_rate": 0.001753846153846154,
      "loss": 3.143,
      "router_z_loss": 0.0485,
      "step": 570
    },
    {
      "balancing_entropy": 1.0737,
      "balancing_loss": 1.0571,
      "epoch": 2.96256,
      "grad_norm": 0.423828125,
      "learning_rate": 0.0017846153846153847,
      "loss": 3.1331,
      "router_z_loss": 0.0478,
      "step": 580
    },
    {
      "balancing_entropy": 1.0323,
      "balancing_loss": 1.0633,
      "epoch": 3.01024,
      "grad_norm": 0.443359375,
      "learning_rate": 0.0018153846153846154,
      "loss": 3.1222,
      "router_z_loss": 0.0446,
      "step": 590
    },
    {
      "balancing_entropy": 1.0575,
      "balancing_loss": 1.0525,
      "epoch": 3.06144,
      "grad_norm": 0.40625,
      "learning_rate": 0.0018461538461538463,
      "loss": 3.1147,
      "router_z_loss": 0.0764,
      "step": 600
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.0612,
      "epoch": 3.11264,
      "grad_norm": 0.349609375,
      "learning_rate": 0.001876923076923077,
      "loss": 3.1053,
      "router_z_loss": 0.0655,
      "step": 610
    },
    {
      "balancing_entropy": 1.0805,
      "balancing_loss": 1.055,
      "epoch": 3.16384,
      "grad_norm": 0.341796875,
      "learning_rate": 0.0019076923076923075,
      "loss": 3.0862,
      "router_z_loss": 0.0448,
      "step": 620
    },
    {
      "balancing_entropy": 1.0813,
      "balancing_loss": 1.0645,
      "epoch": 3.21504,
      "grad_norm": 0.396484375,
      "learning_rate": 0.0019384615384615386,
      "loss": 3.0768,
      "router_z_loss": 0.0414,
      "step": 630
    },
    {
      "balancing_entropy": 1.0683,
      "balancing_loss": 1.0578,
      "epoch": 3.26624,
      "grad_norm": 0.41796875,
      "learning_rate": 0.001969230769230769,
      "loss": 3.0784,
      "router_z_loss": 0.0448,
      "step": 640
    },
    {
      "balancing_entropy": 1.0791,
      "balancing_loss": 1.0519,
      "epoch": 3.31744,
      "grad_norm": 0.33203125,
      "learning_rate": 0.002,
      "loss": 3.0721,
      "router_z_loss": 0.0507,
      "step": 650
    },
    {
      "balancing_entropy": 1.0775,
      "balancing_loss": 1.0777,
      "epoch": 3.36864,
      "grad_norm": 0.61328125,
      "learning_rate": 0.002030769230769231,
      "loss": 3.0531,
      "router_z_loss": 0.0286,
      "step": 660
    },
    {
      "balancing_entropy": 1.0715,
      "balancing_loss": 1.0463,
      "epoch": 3.4198399999999998,
      "grad_norm": 0.34765625,
      "learning_rate": 0.0020615384615384614,
      "loss": 3.0686,
      "router_z_loss": 0.0466,
      "step": 670
    },
    {
      "balancing_entropy": 1.0499,
      "balancing_loss": 1.0661,
      "epoch": 3.47104,
      "grad_norm": 0.412109375,
      "learning_rate": 0.0020923076923076924,
      "loss": 3.0466,
      "router_z_loss": 0.0705,
      "step": 680
    },
    {
      "balancing_entropy": 1.0802,
      "balancing_loss": 1.0543,
      "epoch": 3.52224,
      "grad_norm": 1.28125,
      "learning_rate": 0.0021230769230769233,
      "loss": 3.116,
      "router_z_loss": 0.0514,
      "step": 690
    },
    {
      "balancing_entropy": 1.0568,
      "balancing_loss": 1.0623,
      "epoch": 3.5734399999999997,
      "grad_norm": 0.8125,
      "learning_rate": 0.0021538461538461538,
      "loss": 3.1147,
      "router_z_loss": 0.1138,
      "step": 700
    },
    {
      "balancing_entropy": 1.0641,
      "balancing_loss": 1.0526,
      "epoch": 3.62464,
      "grad_norm": 0.435546875,
      "learning_rate": 0.0021846153846153847,
      "loss": 3.073,
      "router_z_loss": 0.1333,
      "step": 710
    },
    {
      "balancing_entropy": 1.0352,
      "balancing_loss": 1.0477,
      "epoch": 3.67584,
      "grad_norm": 0.3515625,
      "learning_rate": 0.0022153846153846156,
      "loss": 3.0469,
      "router_z_loss": 0.0801,
      "step": 720
    },
    {
      "balancing_entropy": 1.0777,
      "balancing_loss": 1.0576,
      "epoch": 3.72704,
      "grad_norm": 0.345703125,
      "learning_rate": 0.002246153846153846,
      "loss": 3.0371,
      "router_z_loss": 0.0675,
      "step": 730
    },
    {
      "balancing_entropy": 1.0618,
      "balancing_loss": 1.0607,
      "epoch": 3.7782400000000003,
      "grad_norm": 0.46875,
      "learning_rate": 0.002276923076923077,
      "loss": 3.0152,
      "router_z_loss": 0.0448,
      "step": 740
    },
    {
      "balancing_entropy": 1.0732,
      "balancing_loss": 1.0662,
      "epoch": 3.82944,
      "grad_norm": 0.361328125,
      "learning_rate": 0.002307692307692308,
      "loss": 3.0069,
      "router_z_loss": 0.0399,
      "step": 750
    },
    {
      "balancing_entropy": 1.0766,
      "balancing_loss": 1.0487,
      "epoch": 3.88064,
      "grad_norm": 0.40234375,
      "learning_rate": 0.0023384615384615384,
      "loss": 2.9992,
      "router_z_loss": 0.0415,
      "step": 760
    },
    {
      "balancing_entropy": 1.0522,
      "balancing_loss": 1.0701,
      "epoch": 3.9318400000000002,
      "grad_norm": 0.38671875,
      "learning_rate": 0.0023692307692307693,
      "loss": 2.9871,
      "router_z_loss": 0.0607,
      "step": 770
    },
    {
      "balancing_entropy": 1.0814,
      "balancing_loss": 1.0668,
      "epoch": 3.98304,
      "grad_norm": 0.3359375,
      "learning_rate": 0.0024000000000000002,
      "loss": 2.9765,
      "router_z_loss": 0.0568,
      "step": 780
    },
    {
      "balancing_entropy": 1.0144,
      "balancing_loss": 1.0596,
      "epoch": 4.03072,
      "grad_norm": 1.4453125,
      "learning_rate": 0.0024307692307692307,
      "loss": 2.9983,
      "router_z_loss": 0.0458,
      "step": 790
    },
    {
      "balancing_entropy": 1.0668,
      "balancing_loss": 1.0612,
      "epoch": 4.08192,
      "grad_norm": 0.52734375,
      "learning_rate": 0.0024615384615384616,
      "loss": 3.0143,
      "router_z_loss": 0.1153,
      "step": 800
    },
    {
      "balancing_entropy": 1.0725,
      "balancing_loss": 1.0575,
      "epoch": 4.13312,
      "grad_norm": 0.4609375,
      "learning_rate": 0.0024923076923076925,
      "loss": 2.9904,
      "router_z_loss": 0.0798,
      "step": 810
    },
    {
      "balancing_entropy": 1.0738,
      "balancing_loss": 1.0687,
      "epoch": 4.18432,
      "grad_norm": 0.318359375,
      "learning_rate": 0.002523076923076923,
      "loss": 2.9654,
      "router_z_loss": 0.0619,
      "step": 820
    },
    {
      "balancing_entropy": 1.0587,
      "balancing_loss": 1.0688,
      "epoch": 4.23552,
      "grad_norm": 0.337890625,
      "learning_rate": 0.002553846153846154,
      "loss": 2.9491,
      "router_z_loss": 0.0479,
      "step": 830
    },
    {
      "balancing_entropy": 1.0806,
      "balancing_loss": 1.0718,
      "epoch": 4.28672,
      "grad_norm": 0.353515625,
      "learning_rate": 0.002584615384615385,
      "loss": 2.9554,
      "router_z_loss": 0.0434,
      "step": 840
    },
    {
      "balancing_entropy": 1.032,
      "balancing_loss": 1.0697,
      "epoch": 4.33792,
      "grad_norm": 0.3046875,
      "learning_rate": 0.0026153846153846153,
      "loss": 2.9431,
      "router_z_loss": 0.0797,
      "step": 850
    },
    {
      "balancing_entropy": 1.0786,
      "balancing_loss": 1.0799,
      "epoch": 4.38912,
      "grad_norm": 0.43359375,
      "learning_rate": 0.0026461538461538463,
      "loss": 2.9489,
      "router_z_loss": 0.1067,
      "step": 860
    },
    {
      "balancing_entropy": 1.0722,
      "balancing_loss": 1.0613,
      "epoch": 4.44032,
      "grad_norm": 0.37109375,
      "learning_rate": 0.002676923076923077,
      "loss": 2.9348,
      "router_z_loss": 0.0499,
      "step": 870
    },
    {
      "balancing_entropy": 1.052,
      "balancing_loss": 1.075,
      "epoch": 4.49152,
      "grad_norm": 0.5078125,
      "learning_rate": 0.0027076923076923077,
      "loss": 2.9207,
      "router_z_loss": 0.0769,
      "step": 880
    },
    {
      "balancing_entropy": 1.0655,
      "balancing_loss": 1.0746,
      "epoch": 4.54272,
      "grad_norm": 0.26953125,
      "learning_rate": 0.0027384615384615386,
      "loss": 2.9192,
      "router_z_loss": 0.073,
      "step": 890
    },
    {
      "balancing_entropy": 1.08,
      "balancing_loss": 1.0727,
      "epoch": 4.59392,
      "grad_norm": 0.2392578125,
      "learning_rate": 0.0027692307692307695,
      "loss": 2.9239,
      "router_z_loss": 0.0672,
      "step": 900
    },
    {
      "balancing_entropy": 1.0647,
      "balancing_loss": 1.0661,
      "epoch": 4.64512,
      "grad_norm": 0.3359375,
      "learning_rate": 0.0028,
      "loss": 2.9272,
      "router_z_loss": 0.0472,
      "step": 910
    },
    {
      "balancing_entropy": 1.0563,
      "balancing_loss": 1.0951,
      "epoch": 4.69632,
      "grad_norm": 0.322265625,
      "learning_rate": 0.002830769230769231,
      "loss": 2.9024,
      "router_z_loss": 0.081,
      "step": 920
    },
    {
      "balancing_entropy": 1.0773,
      "balancing_loss": 1.0657,
      "epoch": 4.74752,
      "grad_norm": 0.330078125,
      "learning_rate": 0.002861538461538462,
      "loss": 2.9081,
      "router_z_loss": 0.0793,
      "step": 930
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.075,
      "epoch": 4.79872,
      "grad_norm": 0.25390625,
      "learning_rate": 0.0028923076923076923,
      "loss": 2.8901,
      "router_z_loss": 0.0592,
      "step": 940
    },
    {
      "balancing_entropy": 1.0841,
      "balancing_loss": 1.0828,
      "epoch": 4.84992,
      "grad_norm": 0.2890625,
      "learning_rate": 0.002923076923076923,
      "loss": 2.9034,
      "router_z_loss": 0.0611,
      "step": 950
    },
    {
      "balancing_entropy": 1.0724,
      "balancing_loss": 1.0751,
      "epoch": 4.90112,
      "grad_norm": 0.255859375,
      "learning_rate": 0.002953846153846154,
      "loss": 2.8906,
      "router_z_loss": 0.055,
      "step": 960
    },
    {
      "balancing_entropy": 1.0526,
      "balancing_loss": 1.0799,
      "epoch": 4.95232,
      "grad_norm": 0.31640625,
      "learning_rate": 0.0029846153846153846,
      "loss": 2.8845,
      "router_z_loss": 0.0599,
      "step": 970
    },
    {
      "balancing_entropy": 1.0305,
      "balancing_loss": 1.0729,
      "epoch": 5.0,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.8583,
      "router_z_loss": 0.0754,
      "step": 980
    },
    {
      "balancing_entropy": 1.0387,
      "balancing_loss": 1.0792,
      "epoch": 5.0512,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.8699,
      "router_z_loss": 0.1356,
      "step": 990
    },
    {
      "balancing_entropy": 1.0813,
      "balancing_loss": 1.083,
      "epoch": 5.1024,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.8367,
      "router_z_loss": 0.109,
      "step": 1000
    },
    {
      "balancing_entropy": 1.0891,
      "balancing_loss": 1.0879,
      "epoch": 5.1536,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.8535,
      "router_z_loss": 0.0484,
      "step": 1010
    },
    {
      "balancing_entropy": 1.0686,
      "balancing_loss": 1.0802,
      "epoch": 5.2048,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.8188,
      "router_z_loss": 0.0519,
      "step": 1020
    },
    {
      "balancing_entropy": 1.0653,
      "balancing_loss": 1.0687,
      "epoch": 5.256,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.8249,
      "router_z_loss": 0.0658,
      "step": 1030
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.0903,
      "epoch": 5.3072,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.8114,
      "router_z_loss": 0.0667,
      "step": 1040
    },
    {
      "balancing_entropy": 1.0907,
      "balancing_loss": 1.0812,
      "epoch": 5.3584,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.8064,
      "router_z_loss": 0.0454,
      "step": 1050
    },
    {
      "balancing_entropy": 1.0864,
      "balancing_loss": 1.0799,
      "epoch": 5.4096,
      "grad_norm": 0.240234375,
      "learning_rate": 0.003,
      "loss": 2.8045,
      "router_z_loss": 0.0366,
      "step": 1060
    },
    {
      "balancing_entropy": 1.0637,
      "balancing_loss": 1.0829,
      "epoch": 5.4608,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.7872,
      "router_z_loss": 0.0425,
      "step": 1070
    },
    {
      "balancing_entropy": 1.0752,
      "balancing_loss": 1.0863,
      "epoch": 5.5120000000000005,
      "grad_norm": 0.2236328125,
      "learning_rate": 0.003,
      "loss": 2.7975,
      "router_z_loss": 0.0532,
      "step": 1080
    },
    {
      "balancing_entropy": 1.0737,
      "balancing_loss": 1.0807,
      "epoch": 5.5632,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.7764,
      "router_z_loss": 0.0571,
      "step": 1090
    },
    {
      "balancing_entropy": 1.0698,
      "balancing_loss": 1.0909,
      "epoch": 5.6144,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.7735,
      "router_z_loss": 0.0515,
      "step": 1100
    },
    {
      "balancing_entropy": 1.0721,
      "balancing_loss": 1.0799,
      "epoch": 5.6655999999999995,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.7688,
      "router_z_loss": 0.0527,
      "step": 1110
    },
    {
      "balancing_entropy": 1.0667,
      "balancing_loss": 1.0884,
      "epoch": 5.7168,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.7628,
      "router_z_loss": 0.0552,
      "step": 1120
    },
    {
      "balancing_entropy": 1.0569,
      "balancing_loss": 1.0946,
      "epoch": 5.768,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.7667,
      "router_z_loss": 0.0488,
      "step": 1130
    },
    {
      "balancing_entropy": 1.0555,
      "balancing_loss": 1.0803,
      "epoch": 5.8192,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.7646,
      "router_z_loss": 0.0611,
      "step": 1140
    },
    {
      "balancing_entropy": 1.0782,
      "balancing_loss": 1.0768,
      "epoch": 5.8704,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.7563,
      "router_z_loss": 0.0708,
      "step": 1150
    },
    {
      "balancing_entropy": 1.0659,
      "balancing_loss": 1.0826,
      "epoch": 5.9216,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.7447,
      "router_z_loss": 0.058,
      "step": 1160
    },
    {
      "balancing_entropy": 1.041,
      "balancing_loss": 1.0806,
      "epoch": 5.9728,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.7429,
      "router_z_loss": 0.0487,
      "step": 1170
    },
    {
      "balancing_entropy": 1.0092,
      "balancing_loss": 1.0898,
      "epoch": 6.02048,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.7351,
      "router_z_loss": 0.082,
      "step": 1180
    },
    {
      "balancing_entropy": 1.043,
      "balancing_loss": 1.0951,
      "epoch": 6.07168,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.7244,
      "router_z_loss": 0.0824,
      "step": 1190
    },
    {
      "balancing_entropy": 1.0828,
      "balancing_loss": 1.0714,
      "epoch": 6.12288,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.7207,
      "router_z_loss": 0.0499,
      "step": 1200
    },
    {
      "balancing_entropy": 1.0905,
      "balancing_loss": 1.0857,
      "epoch": 6.17408,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.7021,
      "router_z_loss": 0.0365,
      "step": 1210
    },
    {
      "balancing_entropy": 1.0613,
      "balancing_loss": 1.1023,
      "epoch": 6.22528,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.7016,
      "router_z_loss": 0.034,
      "step": 1220
    },
    {
      "balancing_entropy": 1.0492,
      "balancing_loss": 1.084,
      "epoch": 6.27648,
      "grad_norm": 0.431640625,
      "learning_rate": 0.003,
      "loss": 2.6958,
      "router_z_loss": 0.0554,
      "step": 1230
    },
    {
      "balancing_entropy": 1.0414,
      "balancing_loss": 1.0773,
      "epoch": 6.32768,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.7023,
      "router_z_loss": 0.0623,
      "step": 1240
    },
    {
      "balancing_entropy": 1.0565,
      "balancing_loss": 1.0767,
      "epoch": 6.37888,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.6969,
      "router_z_loss": 0.0749,
      "step": 1250
    },
    {
      "balancing_entropy": 1.0644,
      "balancing_loss": 1.0875,
      "epoch": 6.43008,
      "grad_norm": 0.24609375,
      "learning_rate": 0.003,
      "loss": 2.6912,
      "router_z_loss": 0.0692,
      "step": 1260
    },
    {
      "balancing_entropy": 1.0544,
      "balancing_loss": 1.088,
      "epoch": 6.48128,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.6754,
      "router_z_loss": 0.0453,
      "step": 1270
    },
    {
      "balancing_entropy": 1.0581,
      "balancing_loss": 1.0938,
      "epoch": 6.53248,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.6733,
      "router_z_loss": 0.0598,
      "step": 1280
    },
    {
      "balancing_entropy": 1.0656,
      "balancing_loss": 1.0786,
      "epoch": 6.58368,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.6887,
      "router_z_loss": 0.0627,
      "step": 1290
    },
    {
      "balancing_entropy": 1.0514,
      "balancing_loss": 1.0875,
      "epoch": 6.63488,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.6938,
      "router_z_loss": 0.0472,
      "step": 1300
    },
    {
      "balancing_entropy": 1.0774,
      "balancing_loss": 1.0843,
      "epoch": 6.6860800000000005,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.6633,
      "router_z_loss": 0.0539,
      "step": 1310
    },
    {
      "balancing_entropy": 1.0351,
      "balancing_loss": 1.092,
      "epoch": 6.73728,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.6705,
      "router_z_loss": 0.0526,
      "step": 1320
    },
    {
      "balancing_entropy": 1.0679,
      "balancing_loss": 1.082,
      "epoch": 6.78848,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.6617,
      "router_z_loss": 0.0543,
      "step": 1330
    },
    {
      "balancing_entropy": 1.0549,
      "balancing_loss": 1.0898,
      "epoch": 6.8396799999999995,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.657,
      "router_z_loss": 0.0385,
      "step": 1340
    },
    {
      "balancing_entropy": 1.0456,
      "balancing_loss": 1.0965,
      "epoch": 6.89088,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.6591,
      "router_z_loss": 0.044,
      "step": 1350
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.0829,
      "epoch": 6.94208,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.6571,
      "router_z_loss": 0.0586,
      "step": 1360
    },
    {
      "balancing_entropy": 1.0539,
      "balancing_loss": 1.1001,
      "epoch": 6.99328,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.6604,
      "router_z_loss": 0.0386,
      "step": 1370
    },
    {
      "balancing_entropy": 1.0103,
      "balancing_loss": 1.0824,
      "epoch": 7.04096,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.6434,
      "router_z_loss": 0.0576,
      "step": 1380
    },
    {
      "balancing_entropy": 1.0341,
      "balancing_loss": 1.0806,
      "epoch": 7.09216,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.6309,
      "router_z_loss": 0.0659,
      "step": 1390
    },
    {
      "balancing_entropy": 1.0726,
      "balancing_loss": 1.0874,
      "epoch": 7.14336,
      "grad_norm": 0.22265625,
      "learning_rate": 0.003,
      "loss": 2.6387,
      "router_z_loss": 0.0499,
      "step": 1400
    },
    {
      "balancing_entropy": 1.0499,
      "balancing_loss": 1.1085,
      "epoch": 7.19456,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.6277,
      "router_z_loss": 0.034,
      "step": 1410
    },
    {
      "balancing_entropy": 1.0411,
      "balancing_loss": 1.0938,
      "epoch": 7.24576,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.6198,
      "router_z_loss": 0.0658,
      "step": 1420
    },
    {
      "balancing_entropy": 1.041,
      "balancing_loss": 1.0742,
      "epoch": 7.29696,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.6141,
      "router_z_loss": 0.0604,
      "step": 1430
    },
    {
      "balancing_entropy": 1.0445,
      "balancing_loss": 1.0994,
      "epoch": 7.34816,
      "grad_norm": 0.2353515625,
      "learning_rate": 0.003,
      "loss": 2.6256,
      "router_z_loss": 0.056,
      "step": 1440
    },
    {
      "balancing_entropy": 1.065,
      "balancing_loss": 1.0861,
      "epoch": 7.39936,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.6202,
      "router_z_loss": 0.0574,
      "step": 1450
    },
    {
      "balancing_entropy": 1.044,
      "balancing_loss": 1.0868,
      "epoch": 7.45056,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.6258,
      "router_z_loss": 0.0431,
      "step": 1460
    },
    {
      "balancing_entropy": 1.0489,
      "balancing_loss": 1.0945,
      "epoch": 7.50176,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.6103,
      "router_z_loss": 0.052,
      "step": 1470
    },
    {
      "balancing_entropy": 1.0481,
      "balancing_loss": 1.1073,
      "epoch": 7.55296,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.6089,
      "router_z_loss": 0.0584,
      "step": 1480
    },
    {
      "balancing_entropy": 1.0396,
      "balancing_loss": 1.0956,
      "epoch": 7.60416,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.6075,
      "router_z_loss": 0.0445,
      "step": 1490
    },
    {
      "balancing_entropy": 1.0533,
      "balancing_loss": 1.0949,
      "epoch": 7.65536,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.6102,
      "router_z_loss": 0.055,
      "step": 1500
    },
    {
      "balancing_entropy": 1.0524,
      "balancing_loss": 1.1026,
      "epoch": 7.70656,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.6003,
      "router_z_loss": 0.0448,
      "step": 1510
    },
    {
      "balancing_entropy": 1.0381,
      "balancing_loss": 1.0928,
      "epoch": 7.75776,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.607,
      "router_z_loss": 0.0381,
      "step": 1520
    },
    {
      "balancing_entropy": 1.0761,
      "balancing_loss": 1.0888,
      "epoch": 7.80896,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.6012,
      "router_z_loss": 0.0483,
      "step": 1530
    },
    {
      "balancing_entropy": 1.0843,
      "balancing_loss": 1.0961,
      "epoch": 7.8601600000000005,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.5861,
      "router_z_loss": 0.0457,
      "step": 1540
    },
    {
      "balancing_entropy": 1.0163,
      "balancing_loss": 1.1091,
      "epoch": 7.91136,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.5987,
      "router_z_loss": 0.0486,
      "step": 1550
    },
    {
      "balancing_entropy": 1.0666,
      "balancing_loss": 1.0832,
      "epoch": 7.96256,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.5929,
      "router_z_loss": 0.0643,
      "step": 1560
    },
    {
      "balancing_entropy": 1.0472,
      "balancing_loss": 1.0895,
      "epoch": 8.01024,
      "grad_norm": 0.232421875,
      "learning_rate": 0.003,
      "loss": 2.5825,
      "router_z_loss": 0.0445,
      "step": 1570
    },
    {
      "balancing_entropy": 1.0787,
      "balancing_loss": 1.1011,
      "epoch": 8.06144,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.5758,
      "router_z_loss": 0.0345,
      "step": 1580
    },
    {
      "balancing_entropy": 1.0564,
      "balancing_loss": 1.0812,
      "epoch": 8.11264,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.5731,
      "router_z_loss": 0.0408,
      "step": 1590
    },
    {
      "balancing_entropy": 1.032,
      "balancing_loss": 1.1041,
      "epoch": 8.16384,
      "grad_norm": 0.255859375,
      "learning_rate": 0.003,
      "loss": 2.582,
      "router_z_loss": 0.0574,
      "step": 1600
    },
    {
      "balancing_entropy": 1.084,
      "balancing_loss": 1.0891,
      "epoch": 8.21504,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.5631,
      "router_z_loss": 0.0515,
      "step": 1610
    },
    {
      "balancing_entropy": 1.0599,
      "balancing_loss": 1.0946,
      "epoch": 8.26624,
      "grad_norm": 0.220703125,
      "learning_rate": 0.003,
      "loss": 2.5519,
      "router_z_loss": 0.0378,
      "step": 1620
    },
    {
      "balancing_entropy": 1.0849,
      "balancing_loss": 1.0976,
      "epoch": 8.31744,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.5632,
      "router_z_loss": 0.034,
      "step": 1630
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.0904,
      "epoch": 8.36864,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.5617,
      "router_z_loss": 0.0446,
      "step": 1640
    },
    {
      "balancing_entropy": 1.0442,
      "balancing_loss": 1.1168,
      "epoch": 8.41984,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.554,
      "router_z_loss": 0.0427,
      "step": 1650
    },
    {
      "balancing_entropy": 1.0504,
      "balancing_loss": 1.0854,
      "epoch": 8.47104,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.5621,
      "router_z_loss": 0.053,
      "step": 1660
    },
    {
      "balancing_entropy": 1.0607,
      "balancing_loss": 1.0911,
      "epoch": 8.52224,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.5476,
      "router_z_loss": 0.0582,
      "step": 1670
    },
    {
      "balancing_entropy": 1.0466,
      "balancing_loss": 1.0911,
      "epoch": 8.57344,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.5493,
      "router_z_loss": 0.0521,
      "step": 1680
    },
    {
      "balancing_entropy": 1.0463,
      "balancing_loss": 1.099,
      "epoch": 8.62464,
      "grad_norm": 0.26953125,
      "learning_rate": 0.003,
      "loss": 2.5503,
      "router_z_loss": 0.0583,
      "step": 1690
    },
    {
      "balancing_entropy": 1.0808,
      "balancing_loss": 1.0893,
      "epoch": 8.67584,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.55,
      "router_z_loss": 0.0582,
      "step": 1700
    },
    {
      "balancing_entropy": 1.0538,
      "balancing_loss": 1.0892,
      "epoch": 8.72704,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.5424,
      "router_z_loss": 0.0463,
      "step": 1710
    },
    {
      "balancing_entropy": 1.0521,
      "balancing_loss": 1.088,
      "epoch": 8.77824,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.5362,
      "router_z_loss": 0.0393,
      "step": 1720
    },
    {
      "balancing_entropy": 1.0079,
      "balancing_loss": 1.1011,
      "epoch": 8.82944,
      "grad_norm": 0.419921875,
      "learning_rate": 0.003,
      "loss": 2.5468,
      "router_z_loss": 0.0474,
      "step": 1730
    },
    {
      "balancing_entropy": 1.0677,
      "balancing_loss": 1.0865,
      "epoch": 8.88064,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.5435,
      "router_z_loss": 0.0641,
      "step": 1740
    },
    {
      "balancing_entropy": 1.0642,
      "balancing_loss": 1.1003,
      "epoch": 8.93184,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.5419,
      "router_z_loss": 0.0505,
      "step": 1750
    },
    {
      "balancing_entropy": 1.088,
      "balancing_loss": 1.0868,
      "epoch": 8.98304,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.543,
      "router_z_loss": 0.0476,
      "step": 1760
    },
    {
      "balancing_entropy": 1.0402,
      "balancing_loss": 1.0972,
      "epoch": 9.03072,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.5303,
      "router_z_loss": 0.0394,
      "step": 1770
    },
    {
      "balancing_entropy": 1.0795,
      "balancing_loss": 1.0897,
      "epoch": 9.08192,
      "grad_norm": 0.2333984375,
      "learning_rate": 0.003,
      "loss": 2.5223,
      "router_z_loss": 0.0393,
      "step": 1780
    },
    {
      "balancing_entropy": 1.0681,
      "balancing_loss": 1.0986,
      "epoch": 9.13312,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.5125,
      "router_z_loss": 0.043,
      "step": 1790
    },
    {
      "balancing_entropy": 1.0574,
      "balancing_loss": 1.0985,
      "epoch": 9.18432,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.5122,
      "router_z_loss": 0.0493,
      "step": 1800
    },
    {
      "balancing_entropy": 1.0644,
      "balancing_loss": 1.078,
      "epoch": 9.23552,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.5214,
      "router_z_loss": 0.0424,
      "step": 1810
    },
    {
      "balancing_entropy": 1.0774,
      "balancing_loss": 1.0954,
      "epoch": 9.28672,
      "grad_norm": 0.2431640625,
      "learning_rate": 0.003,
      "loss": 2.5096,
      "router_z_loss": 0.0481,
      "step": 1820
    },
    {
      "balancing_entropy": 1.0653,
      "balancing_loss": 1.0845,
      "epoch": 9.33792,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.5181,
      "router_z_loss": 0.0429,
      "step": 1830
    },
    {
      "balancing_entropy": 1.0383,
      "balancing_loss": 1.0964,
      "epoch": 9.38912,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.5093,
      "router_z_loss": 0.0455,
      "step": 1840
    },
    {
      "balancing_entropy": 1.0394,
      "balancing_loss": 1.0928,
      "epoch": 9.44032,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.5122,
      "router_z_loss": 0.0598,
      "step": 1850
    },
    {
      "balancing_entropy": 1.083,
      "balancing_loss": 1.0864,
      "epoch": 9.49152,
      "grad_norm": 0.251953125,
      "learning_rate": 0.003,
      "loss": 2.511,
      "router_z_loss": 0.0528,
      "step": 1860
    },
    {
      "balancing_entropy": 1.0453,
      "balancing_loss": 1.087,
      "epoch": 9.54272,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.5176,
      "router_z_loss": 0.0431,
      "step": 1870
    },
    {
      "balancing_entropy": 1.0602,
      "balancing_loss": 1.087,
      "epoch": 9.59392,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.5084,
      "router_z_loss": 0.0584,
      "step": 1880
    },
    {
      "balancing_entropy": 1.0505,
      "balancing_loss": 1.1014,
      "epoch": 9.64512,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.5108,
      "router_z_loss": 0.05,
      "step": 1890
    },
    {
      "balancing_entropy": 1.0711,
      "balancing_loss": 1.082,
      "epoch": 9.69632,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.5167,
      "router_z_loss": 0.0429,
      "step": 1900
    },
    {
      "balancing_entropy": 1.0926,
      "balancing_loss": 1.0978,
      "epoch": 9.74752,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.5001,
      "router_z_loss": 0.0373,
      "step": 1910
    },
    {
      "balancing_entropy": 1.07,
      "balancing_loss": 1.0855,
      "epoch": 9.79872,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.5046,
      "router_z_loss": 0.0332,
      "step": 1920
    },
    {
      "balancing_entropy": 1.0465,
      "balancing_loss": 1.0993,
      "epoch": 9.849920000000001,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.4964,
      "router_z_loss": 0.0378,
      "step": 1930
    },
    {
      "balancing_entropy": 1.053,
      "balancing_loss": 1.1113,
      "epoch": 9.90112,
      "grad_norm": 0.66015625,
      "learning_rate": 0.003,
      "loss": 2.499,
      "router_z_loss": 0.0383,
      "step": 1940
    },
    {
      "balancing_entropy": 1.0583,
      "balancing_loss": 1.0769,
      "epoch": 9.95232,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.5098,
      "router_z_loss": 0.0609,
      "step": 1950
    },
    {
      "balancing_entropy": 1.0337,
      "balancing_loss": 1.0762,
      "epoch": 10.0,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.5056,
      "router_z_loss": 0.0448,
      "step": 1960
    },
    {
      "balancing_entropy": 1.0587,
      "balancing_loss": 1.0942,
      "epoch": 10.0512,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.4957,
      "router_z_loss": 0.0389,
      "step": 1970
    },
    {
      "balancing_entropy": 1.0687,
      "balancing_loss": 1.0801,
      "epoch": 10.1024,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.4861,
      "router_z_loss": 0.0446,
      "step": 1980
    },
    {
      "balancing_entropy": 1.0266,
      "balancing_loss": 1.1002,
      "epoch": 10.1536,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.4833,
      "router_z_loss": 0.0409,
      "step": 1990
    },
    {
      "balancing_entropy": 1.0685,
      "balancing_loss": 1.0642,
      "epoch": 10.2048,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.4785,
      "router_z_loss": 0.0453,
      "step": 2000
    },
    {
      "balancing_entropy": 1.0387,
      "balancing_loss": 1.1013,
      "epoch": 10.256,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.4739,
      "router_z_loss": 0.0422,
      "step": 2010
    },
    {
      "balancing_entropy": 1.0183,
      "balancing_loss": 1.0957,
      "epoch": 10.3072,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.4841,
      "router_z_loss": 0.0564,
      "step": 2020
    },
    {
      "balancing_entropy": 1.0598,
      "balancing_loss": 1.0796,
      "epoch": 10.3584,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.4744,
      "router_z_loss": 0.0617,
      "step": 2030
    },
    {
      "balancing_entropy": 1.087,
      "balancing_loss": 1.0948,
      "epoch": 10.4096,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.4819,
      "router_z_loss": 0.0526,
      "step": 2040
    },
    {
      "balancing_entropy": 1.033,
      "balancing_loss": 1.1048,
      "epoch": 10.4608,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.473,
      "router_z_loss": 0.0423,
      "step": 2050
    },
    {
      "balancing_entropy": 1.0287,
      "balancing_loss": 1.0962,
      "epoch": 10.512,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.4773,
      "router_z_loss": 0.0646,
      "step": 2060
    },
    {
      "balancing_entropy": 1.0231,
      "balancing_loss": 1.1177,
      "epoch": 10.5632,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.4709,
      "router_z_loss": 0.0847,
      "step": 2070
    },
    {
      "balancing_entropy": 1.0554,
      "balancing_loss": 1.0933,
      "epoch": 10.6144,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.4725,
      "router_z_loss": 0.0819,
      "step": 2080
    },
    {
      "balancing_entropy": 1.0375,
      "balancing_loss": 1.1016,
      "epoch": 10.6656,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.4815,
      "router_z_loss": 0.0581,
      "step": 2090
    },
    {
      "balancing_entropy": 1.0431,
      "balancing_loss": 1.0678,
      "epoch": 10.7168,
      "grad_norm": 0.24609375,
      "learning_rate": 0.003,
      "loss": 2.4713,
      "router_z_loss": 0.0655,
      "step": 2100
    },
    {
      "balancing_entropy": 1.0615,
      "balancing_loss": 1.0946,
      "epoch": 10.768,
      "grad_norm": 0.25,
      "learning_rate": 0.003,
      "loss": 2.4749,
      "router_z_loss": 0.0561,
      "step": 2110
    },
    {
      "balancing_entropy": 1.0691,
      "balancing_loss": 1.0829,
      "epoch": 10.8192,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.4748,
      "router_z_loss": 0.0569,
      "step": 2120
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.0817,
      "epoch": 10.8704,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.4715,
      "router_z_loss": 0.0534,
      "step": 2130
    },
    {
      "balancing_entropy": 1.0483,
      "balancing_loss": 1.0831,
      "epoch": 10.9216,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.4666,
      "router_z_loss": 0.0732,
      "step": 2140
    },
    {
      "balancing_entropy": 1.054,
      "balancing_loss": 1.0951,
      "epoch": 10.9728,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.4698,
      "router_z_loss": 0.0889,
      "step": 2150
    },
    {
      "balancing_entropy": 1.0184,
      "balancing_loss": 1.0997,
      "epoch": 11.02048,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.4697,
      "router_z_loss": 0.0572,
      "step": 2160
    },
    {
      "balancing_entropy": 1.0478,
      "balancing_loss": 1.0841,
      "epoch": 11.07168,
      "grad_norm": 0.2470703125,
      "learning_rate": 0.003,
      "loss": 2.451,
      "router_z_loss": 0.0465,
      "step": 2170
    },
    {
      "balancing_entropy": 1.0671,
      "balancing_loss": 1.0789,
      "epoch": 11.12288,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.4497,
      "router_z_loss": 0.0487,
      "step": 2180
    },
    {
      "balancing_entropy": 1.0507,
      "balancing_loss": 1.0759,
      "epoch": 11.17408,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.4455,
      "router_z_loss": 0.0489,
      "step": 2190
    },
    {
      "balancing_entropy": 1.076,
      "balancing_loss": 1.1024,
      "epoch": 11.22528,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.4478,
      "router_z_loss": 0.0501,
      "step": 2200
    },
    {
      "balancing_entropy": 1.0774,
      "balancing_loss": 1.0923,
      "epoch": 11.27648,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.4483,
      "router_z_loss": 0.0512,
      "step": 2210
    },
    {
      "balancing_entropy": 1.0734,
      "balancing_loss": 1.0978,
      "epoch": 11.32768,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.4482,
      "router_z_loss": 0.0448,
      "step": 2220
    },
    {
      "balancing_entropy": 1.0689,
      "balancing_loss": 1.0972,
      "epoch": 11.37888,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.4477,
      "router_z_loss": 0.0429,
      "step": 2230
    },
    {
      "balancing_entropy": 1.063,
      "balancing_loss": 1.0751,
      "epoch": 11.43008,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.4525,
      "router_z_loss": 0.0487,
      "step": 2240
    },
    {
      "balancing_entropy": 1.0563,
      "balancing_loss": 1.0944,
      "epoch": 11.48128,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.4566,
      "router_z_loss": 0.0605,
      "step": 2250
    },
    {
      "balancing_entropy": 1.0743,
      "balancing_loss": 1.0924,
      "epoch": 11.53248,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.4532,
      "router_z_loss": 0.06,
      "step": 2260
    },
    {
      "balancing_entropy": 1.073,
      "balancing_loss": 1.0811,
      "epoch": 11.58368,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.4508,
      "router_z_loss": 0.0407,
      "step": 2270
    },
    {
      "balancing_entropy": 1.0753,
      "balancing_loss": 1.1044,
      "epoch": 11.63488,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.4407,
      "router_z_loss": 0.0398,
      "step": 2280
    },
    {
      "balancing_entropy": 1.0681,
      "balancing_loss": 1.0876,
      "epoch": 11.68608,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.4508,
      "router_z_loss": 0.0464,
      "step": 2290
    },
    {
      "balancing_entropy": 1.0885,
      "balancing_loss": 1.0733,
      "epoch": 11.73728,
      "grad_norm": 0.244140625,
      "learning_rate": 0.003,
      "loss": 2.4419,
      "router_z_loss": 0.042,
      "step": 2300
    },
    {
      "balancing_entropy": 1.051,
      "balancing_loss": 1.126,
      "epoch": 11.78848,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.4405,
      "router_z_loss": 0.0387,
      "step": 2310
    },
    {
      "balancing_entropy": 1.0763,
      "balancing_loss": 1.0771,
      "epoch": 11.83968,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.4487,
      "router_z_loss": 0.0513,
      "step": 2320
    },
    {
      "balancing_entropy": 1.0589,
      "balancing_loss": 1.0981,
      "epoch": 11.89088,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.4415,
      "router_z_loss": 0.0583,
      "step": 2330
    },
    {
      "balancing_entropy": 1.0783,
      "balancing_loss": 1.106,
      "epoch": 11.94208,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.4404,
      "router_z_loss": 0.0447,
      "step": 2340
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.0927,
      "epoch": 11.99328,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.4415,
      "router_z_loss": 0.0408,
      "step": 2350
    },
    {
      "balancing_entropy": 1.0479,
      "balancing_loss": 1.0908,
      "epoch": 12.04096,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.4397,
      "router_z_loss": 0.0443,
      "step": 2360
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.094,
      "epoch": 12.09216,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4342,
      "router_z_loss": 0.0404,
      "step": 2370
    },
    {
      "balancing_entropy": 1.0663,
      "balancing_loss": 1.0783,
      "epoch": 12.14336,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.4184,
      "router_z_loss": 0.0443,
      "step": 2380
    },
    {
      "balancing_entropy": 1.0588,
      "balancing_loss": 1.0854,
      "epoch": 12.19456,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4274,
      "router_z_loss": 0.0567,
      "step": 2390
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.0862,
      "epoch": 12.24576,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.4326,
      "router_z_loss": 0.0566,
      "step": 2400
    },
    {
      "balancing_entropy": 1.0421,
      "balancing_loss": 1.1051,
      "epoch": 12.29696,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.4289,
      "router_z_loss": 0.0501,
      "step": 2410
    },
    {
      "balancing_entropy": 1.0671,
      "balancing_loss": 1.1032,
      "epoch": 12.34816,
      "grad_norm": 0.68359375,
      "learning_rate": 0.003,
      "loss": 2.4214,
      "router_z_loss": 0.0514,
      "step": 2420
    },
    {
      "balancing_entropy": 1.0722,
      "balancing_loss": 1.0775,
      "epoch": 12.39936,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.4264,
      "router_z_loss": 0.054,
      "step": 2430
    },
    {
      "balancing_entropy": 1.0487,
      "balancing_loss": 1.1037,
      "epoch": 12.45056,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.4257,
      "router_z_loss": 0.062,
      "step": 2440
    },
    {
      "balancing_entropy": 1.0597,
      "balancing_loss": 1.0797,
      "epoch": 12.50176,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.423,
      "router_z_loss": 0.062,
      "step": 2450
    },
    {
      "balancing_entropy": 1.066,
      "balancing_loss": 1.0912,
      "epoch": 12.55296,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.4212,
      "router_z_loss": 0.0688,
      "step": 2460
    },
    {
      "balancing_entropy": 1.088,
      "balancing_loss": 1.0989,
      "epoch": 12.60416,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.429,
      "router_z_loss": 0.0545,
      "step": 2470
    },
    {
      "balancing_entropy": 1.0693,
      "balancing_loss": 1.0926,
      "epoch": 12.65536,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.4139,
      "router_z_loss": 0.0364,
      "step": 2480
    },
    {
      "balancing_entropy": 1.0336,
      "balancing_loss": 1.1007,
      "epoch": 12.70656,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.4234,
      "router_z_loss": 0.053,
      "step": 2490
    },
    {
      "balancing_entropy": 1.0263,
      "balancing_loss": 1.0903,
      "epoch": 12.75776,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.4202,
      "router_z_loss": 0.0776,
      "step": 2500
    },
    {
      "balancing_entropy": 1.0686,
      "balancing_loss": 1.0713,
      "epoch": 12.80896,
      "grad_norm": 0.25,
      "learning_rate": 0.003,
      "loss": 2.4181,
      "router_z_loss": 0.0698,
      "step": 2510
    },
    {
      "balancing_entropy": 1.0655,
      "balancing_loss": 1.1029,
      "epoch": 12.86016,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.4206,
      "router_z_loss": 0.0502,
      "step": 2520
    },
    {
      "balancing_entropy": 1.057,
      "balancing_loss": 1.0774,
      "epoch": 12.91136,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.4302,
      "router_z_loss": 0.0388,
      "step": 2530
    },
    {
      "balancing_entropy": 1.0628,
      "balancing_loss": 1.1026,
      "epoch": 12.96256,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.4193,
      "router_z_loss": 0.0392,
      "step": 2540
    },
    {
      "balancing_entropy": 1.0271,
      "balancing_loss": 1.0836,
      "epoch": 13.01024,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.4201,
      "router_z_loss": 0.0366,
      "step": 2550
    },
    {
      "balancing_entropy": 1.0647,
      "balancing_loss": 1.109,
      "epoch": 13.06144,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.4081,
      "router_z_loss": 0.0421,
      "step": 2560
    },
    {
      "balancing_entropy": 1.0704,
      "balancing_loss": 1.0896,
      "epoch": 13.11264,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.4001,
      "router_z_loss": 0.0417,
      "step": 2570
    },
    {
      "balancing_entropy": 1.0761,
      "balancing_loss": 1.085,
      "epoch": 13.16384,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.4012,
      "router_z_loss": 0.0507,
      "step": 2580
    },
    {
      "balancing_entropy": 1.0803,
      "balancing_loss": 1.0989,
      "epoch": 13.21504,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.4026,
      "router_z_loss": 0.0532,
      "step": 2590
    },
    {
      "balancing_entropy": 1.0752,
      "balancing_loss": 1.1013,
      "epoch": 13.26624,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.3987,
      "router_z_loss": 0.0439,
      "step": 2600
    },
    {
      "balancing_entropy": 1.0778,
      "balancing_loss": 1.0934,
      "epoch": 13.31744,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.4011,
      "router_z_loss": 0.0414,
      "step": 2610
    },
    {
      "balancing_entropy": 1.0792,
      "balancing_loss": 1.089,
      "epoch": 13.36864,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.3998,
      "router_z_loss": 0.061,
      "step": 2620
    },
    {
      "balancing_entropy": 1.058,
      "balancing_loss": 1.0987,
      "epoch": 13.41984,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.4061,
      "router_z_loss": 0.057,
      "step": 2630
    },
    {
      "balancing_entropy": 1.0718,
      "balancing_loss": 1.0896,
      "epoch": 13.47104,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.4141,
      "router_z_loss": 0.0505,
      "step": 2640
    },
    {
      "balancing_entropy": 1.0522,
      "balancing_loss": 1.0958,
      "epoch": 13.52224,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.4243,
      "router_z_loss": 0.0815,
      "step": 2650
    },
    {
      "balancing_entropy": 1.0473,
      "balancing_loss": 1.086,
      "epoch": 13.57344,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.42,
      "router_z_loss": 0.1014,
      "step": 2660
    },
    {
      "balancing_entropy": 1.0636,
      "balancing_loss": 1.0705,
      "epoch": 13.62464,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.4023,
      "router_z_loss": 0.092,
      "step": 2670
    },
    {
      "balancing_entropy": 0.9863,
      "balancing_loss": 1.0971,
      "epoch": 13.67584,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.4122,
      "router_z_loss": 0.0712,
      "step": 2680
    },
    {
      "balancing_entropy": 1.0408,
      "balancing_loss": 1.1069,
      "epoch": 13.72704,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.4151,
      "router_z_loss": 0.0886,
      "step": 2690
    },
    {
      "balancing_entropy": 1.0364,
      "balancing_loss": 1.0814,
      "epoch": 13.77824,
      "grad_norm": 0.78515625,
      "learning_rate": 0.003,
      "loss": 2.4033,
      "router_z_loss": 0.1072,
      "step": 2700
    },
    {
      "balancing_entropy": 1.0832,
      "balancing_loss": 1.0745,
      "epoch": 13.82944,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.4022,
      "router_z_loss": 0.0608,
      "step": 2710
    },
    {
      "balancing_entropy": 1.045,
      "balancing_loss": 1.1129,
      "epoch": 13.88064,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.4086,
      "router_z_loss": 0.0395,
      "step": 2720
    },
    {
      "balancing_entropy": 1.0285,
      "balancing_loss": 1.0791,
      "epoch": 13.93184,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.4034,
      "router_z_loss": 0.0636,
      "step": 2730
    },
    {
      "balancing_entropy": 1.0443,
      "balancing_loss": 1.0906,
      "epoch": 13.98304,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.4057,
      "router_z_loss": 0.0902,
      "step": 2740
    },
    {
      "balancing_entropy": 1.0447,
      "balancing_loss": 1.096,
      "epoch": 14.03072,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.4,
      "router_z_loss": 0.0445,
      "step": 2750
    },
    {
      "balancing_entropy": 1.0827,
      "balancing_loss": 1.0814,
      "epoch": 14.08192,
      "grad_norm": 0.23046875,
      "learning_rate": 0.003,
      "loss": 2.3865,
      "router_z_loss": 0.0476,
      "step": 2760
    },
    {
      "balancing_entropy": 1.0617,
      "balancing_loss": 1.099,
      "epoch": 14.13312,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.3807,
      "router_z_loss": 0.0631,
      "step": 2770
    },
    {
      "balancing_entropy": 1.0227,
      "balancing_loss": 1.0719,
      "epoch": 14.18432,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.3848,
      "router_z_loss": 0.0542,
      "step": 2780
    },
    {
      "balancing_entropy": 1.0485,
      "balancing_loss": 1.095,
      "epoch": 14.23552,
      "grad_norm": 0.279296875,
      "learning_rate": 0.003,
      "loss": 2.3882,
      "router_z_loss": 0.0635,
      "step": 2790
    },
    {
      "balancing_entropy": 1.0421,
      "balancing_loss": 1.0935,
      "epoch": 14.28672,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.3835,
      "router_z_loss": 0.0477,
      "step": 2800
    },
    {
      "balancing_entropy": 1.0391,
      "balancing_loss": 1.0766,
      "epoch": 14.33792,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.4203,
      "router_z_loss": 0.0844,
      "step": 2810
    },
    {
      "balancing_entropy": 1.0673,
      "balancing_loss": 1.0418,
      "epoch": 14.38912,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.4013,
      "router_z_loss": 0.1286,
      "step": 2820
    },
    {
      "balancing_entropy": 1.0632,
      "balancing_loss": 1.0529,
      "epoch": 14.44032,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.3921,
      "router_z_loss": 0.1725,
      "step": 2830
    },
    {
      "balancing_entropy": 1.0634,
      "balancing_loss": 1.0587,
      "epoch": 14.49152,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.3928,
      "router_z_loss": 0.1124,
      "step": 2840
    },
    {
      "balancing_entropy": 1.0361,
      "balancing_loss": 1.0746,
      "epoch": 14.54272,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.3917,
      "router_z_loss": 0.086,
      "step": 2850
    },
    {
      "balancing_entropy": 0.9905,
      "balancing_loss": 1.0882,
      "epoch": 14.59392,
      "grad_norm": 1.1171875,
      "learning_rate": 0.003,
      "loss": 2.3999,
      "router_z_loss": 0.1269,
      "step": 2860
    },
    {
      "balancing_entropy": 1.0314,
      "balancing_loss": 1.0557,
      "epoch": 14.64512,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.3955,
      "router_z_loss": 0.12,
      "step": 2870
    },
    {
      "balancing_entropy": 1.0054,
      "balancing_loss": 1.0635,
      "epoch": 14.69632,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.3867,
      "router_z_loss": 0.1578,
      "step": 2880
    },
    {
      "balancing_entropy": 1.0374,
      "balancing_loss": 1.1162,
      "epoch": 14.74752,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.3979,
      "router_z_loss": 0.1295,
      "step": 2890
    },
    {
      "balancing_entropy": 1.0568,
      "balancing_loss": 1.0836,
      "epoch": 14.79872,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.3893,
      "router_z_loss": 0.0689,
      "step": 2900
    },
    {
      "balancing_entropy": 1.0459,
      "balancing_loss": 1.0847,
      "epoch": 14.849920000000001,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.3843,
      "router_z_loss": 0.058,
      "step": 2910
    },
    {
      "balancing_entropy": 1.0215,
      "balancing_loss": 1.0878,
      "epoch": 14.90112,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.3838,
      "router_z_loss": 0.0596,
      "step": 2920
    },
    {
      "balancing_entropy": 0.9784,
      "balancing_loss": 1.1073,
      "epoch": 14.95232,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.3958,
      "router_z_loss": 0.0674,
      "step": 2930
    },
    {
      "balancing_entropy": 0.9716,
      "balancing_loss": 1.0894,
      "epoch": 15.0,
      "grad_norm": 0.73828125,
      "learning_rate": 0.003,
      "loss": 2.3901,
      "router_z_loss": 0.0621,
      "step": 2940
    },
    {
      "balancing_entropy": 0.9993,
      "balancing_loss": 1.1021,
      "epoch": 15.0512,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.3822,
      "router_z_loss": 0.0698,
      "step": 2950
    },
    {
      "balancing_entropy": 1.0238,
      "balancing_loss": 1.0934,
      "epoch": 15.1024,
      "grad_norm": 0.84375,
      "learning_rate": 0.003,
      "loss": 2.3785,
      "router_z_loss": 0.0728,
      "step": 2960
    },
    {
      "balancing_entropy": 1.0051,
      "balancing_loss": 1.1068,
      "epoch": 15.1536,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.3777,
      "router_z_loss": 0.0709,
      "step": 2970
    },
    {
      "balancing_entropy": 1.0506,
      "balancing_loss": 1.073,
      "epoch": 15.2048,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.3634,
      "router_z_loss": 0.059,
      "step": 2980
    },
    {
      "balancing_entropy": 1.0439,
      "balancing_loss": 1.097,
      "epoch": 15.256,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.3764,
      "router_z_loss": 0.0655,
      "step": 2990
    },
    {
      "balancing_entropy": 1.0423,
      "balancing_loss": 1.1,
      "epoch": 15.3072,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.3645,
      "router_z_loss": 0.0841,
      "step": 3000
    },
    {
      "balancing_entropy": 1.0086,
      "balancing_loss": 1.08,
      "epoch": 15.3584,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.3732,
      "router_z_loss": 0.0684,
      "step": 3010
    },
    {
      "balancing_entropy": 1.0134,
      "balancing_loss": 1.0831,
      "epoch": 15.4096,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.367,
      "router_z_loss": 0.066,
      "step": 3020
    },
    {
      "balancing_entropy": 0.9691,
      "balancing_loss": 1.0918,
      "epoch": 15.4608,
      "grad_norm": 0.86328125,
      "learning_rate": 0.003,
      "loss": 2.376,
      "router_z_loss": 0.0698,
      "step": 3030
    },
    {
      "balancing_entropy": 1.0062,
      "balancing_loss": 1.0903,
      "epoch": 15.512,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.3756,
      "router_z_loss": 0.0888,
      "step": 3040
    },
    {
      "balancing_entropy": 1.0067,
      "balancing_loss": 1.0806,
      "epoch": 15.5632,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.3794,
      "router_z_loss": 0.0762,
      "step": 3050
    },
    {
      "balancing_entropy": 1.0322,
      "balancing_loss": 1.1124,
      "epoch": 15.6144,
      "grad_norm": 0.9921875,
      "learning_rate": 0.003,
      "loss": 2.3862,
      "router_z_loss": 0.064,
      "step": 3060
    },
    {
      "balancing_entropy": 1.0643,
      "balancing_loss": 1.0494,
      "epoch": 15.6656,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.3955,
      "router_z_loss": 0.0811,
      "step": 3070
    },
    {
      "balancing_entropy": 1.0166,
      "balancing_loss": 1.0836,
      "epoch": 15.7168,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.3792,
      "router_z_loss": 0.0997,
      "step": 3080
    },
    {
      "balancing_entropy": 1.0339,
      "balancing_loss": 1.0757,
      "epoch": 15.768,
      "grad_norm": 1.125,
      "learning_rate": 0.003,
      "loss": 2.361,
      "router_z_loss": 0.0626,
      "step": 3090
    },
    {
      "balancing_entropy": 1.0058,
      "balancing_loss": 1.0958,
      "epoch": 15.8192,
      "grad_norm": 0.66015625,
      "learning_rate": 0.003,
      "loss": 2.3678,
      "router_z_loss": 0.06,
      "step": 3100
    },
    {
      "balancing_entropy": 1.0527,
      "balancing_loss": 1.1091,
      "epoch": 15.8704,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.3712,
      "router_z_loss": 0.04,
      "step": 3110
    },
    {
      "balancing_entropy": 0.9781,
      "balancing_loss": 1.0918,
      "epoch": 15.9216,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.3718,
      "router_z_loss": 0.0462,
      "step": 3120
    },
    {
      "balancing_entropy": 1.0233,
      "balancing_loss": 1.0849,
      "epoch": 15.9728,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.3757,
      "router_z_loss": 0.08,
      "step": 3130
    },
    {
      "balancing_entropy": 1.0291,
      "balancing_loss": 1.0749,
      "epoch": 16.02048,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.3611,
      "router_z_loss": 0.082,
      "step": 3140
    },
    {
      "balancing_entropy": 1.0406,
      "balancing_loss": 1.1055,
      "epoch": 16.07168,
      "grad_norm": 0.294921875,
      "learning_rate": 0.003,
      "loss": 2.3507,
      "router_z_loss": 0.048,
      "step": 3150
    },
    {
      "balancing_entropy": 1.0092,
      "balancing_loss": 1.0977,
      "epoch": 16.12288,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.3502,
      "router_z_loss": 0.0494,
      "step": 3160
    },
    {
      "balancing_entropy": 1.0515,
      "balancing_loss": 1.0826,
      "epoch": 16.17408,
      "grad_norm": 0.2275390625,
      "learning_rate": 0.003,
      "loss": 2.3559,
      "router_z_loss": 0.0837,
      "step": 3170
    },
    {
      "balancing_entropy": 1.0461,
      "balancing_loss": 1.0922,
      "epoch": 16.22528,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.354,
      "router_z_loss": 0.0541,
      "step": 3180
    },
    {
      "balancing_entropy": 0.9938,
      "balancing_loss": 1.1407,
      "epoch": 16.27648,
      "grad_norm": 1.203125,
      "learning_rate": 0.003,
      "loss": 2.3522,
      "router_z_loss": 0.0442,
      "step": 3190
    },
    {
      "balancing_entropy": 0.9867,
      "balancing_loss": 1.0605,
      "epoch": 16.32768,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.3689,
      "router_z_loss": 0.0557,
      "step": 3200
    },
    {
      "balancing_entropy": 1.043,
      "balancing_loss": 1.1108,
      "epoch": 16.37888,
      "grad_norm": 1.1328125,
      "learning_rate": 0.003,
      "loss": 2.3602,
      "router_z_loss": 0.1298,
      "step": 3210
    },
    {
      "balancing_entropy": 0.9911,
      "balancing_loss": 1.1195,
      "epoch": 16.43008,
      "grad_norm": 0.90625,
      "learning_rate": 0.003,
      "loss": 2.3637,
      "router_z_loss": 0.0937,
      "step": 3220
    },
    {
      "balancing_entropy": 1.0403,
      "balancing_loss": 1.0654,
      "epoch": 16.48128,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.3477,
      "router_z_loss": 0.084,
      "step": 3230
    },
    {
      "balancing_entropy": 1.0483,
      "balancing_loss": 1.1128,
      "epoch": 16.53248,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.3539,
      "router_z_loss": 0.0796,
      "step": 3240
    },
    {
      "balancing_entropy": 1.0536,
      "balancing_loss": 1.0791,
      "epoch": 16.58368,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.3591,
      "router_z_loss": 0.0557,
      "step": 3250
    },
    {
      "balancing_entropy": 1.003,
      "balancing_loss": 1.1178,
      "epoch": 16.63488,
      "grad_norm": 0.9921875,
      "learning_rate": 0.003,
      "loss": 2.3649,
      "router_z_loss": 0.0448,
      "step": 3260
    },
    {
      "balancing_entropy": 1.0046,
      "balancing_loss": 1.1006,
      "epoch": 16.68608,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.3611,
      "router_z_loss": 0.0521,
      "step": 3270
    },
    {
      "balancing_entropy": 1.0141,
      "balancing_loss": 1.0902,
      "epoch": 16.73728,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.3618,
      "router_z_loss": 0.0535,
      "step": 3280
    },
    {
      "balancing_entropy": 0.9902,
      "balancing_loss": 1.0827,
      "epoch": 16.78848,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.3545,
      "router_z_loss": 0.0521,
      "step": 3290
    },
    {
      "balancing_entropy": 1.0342,
      "balancing_loss": 1.0892,
      "epoch": 16.83968,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.3506,
      "router_z_loss": 0.0485,
      "step": 3300
    },
    {
      "balancing_entropy": 1.0373,
      "balancing_loss": 1.1182,
      "epoch": 16.89088,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.3608,
      "router_z_loss": 0.0629,
      "step": 3310
    },
    {
      "balancing_entropy": 1.0241,
      "balancing_loss": 1.1207,
      "epoch": 16.94208,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.3584,
      "router_z_loss": 0.0431,
      "step": 3320
    },
    {
      "balancing_entropy": 1.0706,
      "balancing_loss": 1.084,
      "epoch": 16.99328,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.3569,
      "router_z_loss": 0.0401,
      "step": 3330
    },
    {
      "balancing_entropy": 1.0154,
      "balancing_loss": 1.1176,
      "epoch": 17.04096,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.3457,
      "router_z_loss": 0.0513,
      "step": 3340
    },
    {
      "balancing_entropy": 1.0223,
      "balancing_loss": 1.0951,
      "epoch": 17.09216,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.3396,
      "router_z_loss": 0.0497,
      "step": 3350
    },
    {
      "balancing_entropy": 1.0352,
      "balancing_loss": 1.0953,
      "epoch": 17.14336,
      "grad_norm": 1.0859375,
      "learning_rate": 0.003,
      "loss": 2.3378,
      "router_z_loss": 0.0517,
      "step": 3360
    },
    {
      "balancing_entropy": 1.0611,
      "balancing_loss": 1.0779,
      "epoch": 17.19456,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.3503,
      "router_z_loss": 0.0439,
      "step": 3370
    },
    {
      "balancing_entropy": 1.0086,
      "balancing_loss": 1.1052,
      "epoch": 17.24576,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.3458,
      "router_z_loss": 0.0452,
      "step": 3380
    },
    {
      "balancing_entropy": 1.0328,
      "balancing_loss": 1.0718,
      "epoch": 17.29696,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.3437,
      "router_z_loss": 0.0552,
      "step": 3390
    },
    {
      "balancing_entropy": 1.0428,
      "balancing_loss": 1.079,
      "epoch": 17.34816,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.356,
      "router_z_loss": 0.058,
      "step": 3400
    },
    {
      "balancing_entropy": 0.9512,
      "balancing_loss": 1.1035,
      "epoch": 17.39936,
      "grad_norm": 0.77734375,
      "learning_rate": 0.003,
      "loss": 2.3539,
      "router_z_loss": 0.0753,
      "step": 3410
    },
    {
      "balancing_entropy": 1.008,
      "balancing_loss": 1.1258,
      "epoch": 17.45056,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.3389,
      "router_z_loss": 0.0623,
      "step": 3420
    },
    {
      "balancing_entropy": 1.0431,
      "balancing_loss": 1.1152,
      "epoch": 17.50176,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.3449,
      "router_z_loss": 0.0512,
      "step": 3430
    },
    {
      "balancing_entropy": 1.0846,
      "balancing_loss": 1.0551,
      "epoch": 17.55296,
      "grad_norm": 0.2421875,
      "learning_rate": 0.003,
      "loss": 2.3533,
      "router_z_loss": 0.0691,
      "step": 3440
    },
    {
      "balancing_entropy": 1.0761,
      "balancing_loss": 1.0856,
      "epoch": 17.60416,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.3416,
      "router_z_loss": 0.1074,
      "step": 3450
    },
    {
      "balancing_entropy": 1.0482,
      "balancing_loss": 1.0903,
      "epoch": 17.65536,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.3406,
      "router_z_loss": 0.0676,
      "step": 3460
    },
    {
      "balancing_entropy": 1.0215,
      "balancing_loss": 1.091,
      "epoch": 17.70656,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.3453,
      "router_z_loss": 0.0529,
      "step": 3470
    },
    {
      "balancing_entropy": 0.9921,
      "balancing_loss": 1.1252,
      "epoch": 17.75776,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.3457,
      "router_z_loss": 0.0548,
      "step": 3480
    },
    {
      "balancing_entropy": 0.9884,
      "balancing_loss": 1.0654,
      "epoch": 17.80896,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.355,
      "router_z_loss": 0.0614,
      "step": 3490
    },
    {
      "balancing_entropy": 0.9843,
      "balancing_loss": 1.1226,
      "epoch": 17.86016,
      "grad_norm": 1.1796875,
      "learning_rate": 0.003,
      "loss": 2.3484,
      "router_z_loss": 0.0927,
      "step": 3500
    },
    {
      "balancing_entropy": 1.0228,
      "balancing_loss": 1.0765,
      "epoch": 17.91136,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.3442,
      "router_z_loss": 0.0903,
      "step": 3510
    },
    {
      "balancing_entropy": 1.0648,
      "balancing_loss": 1.0793,
      "epoch": 17.96256,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.3466,
      "router_z_loss": 0.0656,
      "step": 3520
    },
    {
      "balancing_entropy": 0.947,
      "balancing_loss": 1.1157,
      "epoch": 18.01024,
      "grad_norm": 1.4375,
      "learning_rate": 0.003,
      "loss": 2.3538,
      "router_z_loss": 0.0801,
      "step": 3530
    },
    {
      "balancing_entropy": 1.0146,
      "balancing_loss": 1.0741,
      "epoch": 18.06144,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.3487,
      "router_z_loss": 0.0976,
      "step": 3540
    },
    {
      "balancing_entropy": 1.0187,
      "balancing_loss": 1.1294,
      "epoch": 18.11264,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.3364,
      "router_z_loss": 0.0804,
      "step": 3550
    },
    {
      "balancing_entropy": 1.0352,
      "balancing_loss": 1.0875,
      "epoch": 18.16384,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.3268,
      "router_z_loss": 0.0595,
      "step": 3560
    },
    {
      "balancing_entropy": 1.0358,
      "balancing_loss": 1.0897,
      "epoch": 18.21504,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.3243,
      "router_z_loss": 0.0608,
      "step": 3570
    },
    {
      "balancing_entropy": 1.0602,
      "balancing_loss": 1.1065,
      "epoch": 18.26624,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.3292,
      "router_z_loss": 0.0522,
      "step": 3580
    },
    {
      "balancing_entropy": 0.9802,
      "balancing_loss": 1.1189,
      "epoch": 18.31744,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.3302,
      "router_z_loss": 0.0463,
      "step": 3590
    },
    {
      "balancing_entropy": 0.9674,
      "balancing_loss": 1.1104,
      "epoch": 18.36864,
      "grad_norm": 1.2265625,
      "learning_rate": 0.003,
      "loss": 2.3297,
      "router_z_loss": 0.0928,
      "step": 3600
    },
    {
      "balancing_entropy": 0.998,
      "balancing_loss": 1.0953,
      "epoch": 18.41984,
      "grad_norm": 0.84375,
      "learning_rate": 0.003,
      "loss": 2.3332,
      "router_z_loss": 0.0945,
      "step": 3610
    },
    {
      "balancing_entropy": 1.0178,
      "balancing_loss": 1.0844,
      "epoch": 18.47104,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.338,
      "router_z_loss": 0.0609,
      "step": 3620
    },
    {
      "balancing_entropy": 1.0119,
      "balancing_loss": 1.1044,
      "epoch": 18.52224,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.3388,
      "router_z_loss": 0.0631,
      "step": 3630
    },
    {
      "balancing_entropy": 0.9985,
      "balancing_loss": 1.1041,
      "epoch": 18.57344,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.3379,
      "router_z_loss": 0.0484,
      "step": 3640
    },
    {
      "balancing_entropy": 1.0274,
      "balancing_loss": 1.0759,
      "epoch": 18.62464,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.3373,
      "router_z_loss": 0.061,
      "step": 3650
    },
    {
      "balancing_entropy": 1.0301,
      "balancing_loss": 1.1106,
      "epoch": 18.67584,
      "grad_norm": 0.80859375,
      "learning_rate": 0.003,
      "loss": 2.3334,
      "router_z_loss": 0.0876,
      "step": 3660
    },
    {
      "balancing_entropy": 1.0386,
      "balancing_loss": 1.1028,
      "epoch": 18.72704,
      "grad_norm": 0.9921875,
      "learning_rate": 0.003,
      "loss": 2.3328,
      "router_z_loss": 0.0413,
      "step": 3670
    },
    {
      "balancing_entropy": 1.0404,
      "balancing_loss": 1.1131,
      "epoch": 18.77824,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.338,
      "router_z_loss": 0.0609,
      "step": 3680
    },
    {
      "balancing_entropy": 1.0088,
      "balancing_loss": 1.1,
      "epoch": 18.829439999999998,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.3335,
      "router_z_loss": 0.0503,
      "step": 3690
    },
    {
      "balancing_entropy": 0.9662,
      "balancing_loss": 1.1399,
      "epoch": 18.88064,
      "grad_norm": 0.6875,
      "learning_rate": 0.003,
      "loss": 2.3452,
      "router_z_loss": 0.0526,
      "step": 3700
    },
    {
      "balancing_entropy": 0.9891,
      "balancing_loss": 1.0983,
      "epoch": 18.93184,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.3373,
      "router_z_loss": 0.067,
      "step": 3710
    },
    {
      "balancing_entropy": 1.0732,
      "balancing_loss": 1.0684,
      "epoch": 18.98304,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.3377,
      "router_z_loss": 0.0655,
      "step": 3720
    },
    {
      "balancing_entropy": 1.019,
      "balancing_loss": 1.087,
      "epoch": 19.03072,
      "grad_norm": 0.7109375,
      "learning_rate": 0.003,
      "loss": 2.3307,
      "router_z_loss": 0.039,
      "step": 3730
    },
    {
      "balancing_entropy": 1.015,
      "balancing_loss": 1.1249,
      "epoch": 19.08192,
      "grad_norm": 0.890625,
      "learning_rate": 0.003,
      "loss": 2.3189,
      "router_z_loss": 0.0614,
      "step": 3740
    },
    {
      "balancing_entropy": 1.0194,
      "balancing_loss": 1.1258,
      "epoch": 19.13312,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.3473,
      "router_z_loss": 0.0593,
      "step": 3750
    },
    {
      "balancing_entropy": 1.0694,
      "balancing_loss": 1.0487,
      "epoch": 19.18432,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.3326,
      "router_z_loss": 0.1236,
      "step": 3760
    },
    {
      "balancing_entropy": 1.0581,
      "balancing_loss": 1.105,
      "epoch": 19.23552,
      "grad_norm": 1.5078125,
      "learning_rate": 0.003,
      "loss": 2.3175,
      "router_z_loss": 0.148,
      "step": 3770
    },
    {
      "balancing_entropy": 0.9623,
      "balancing_loss": 1.1516,
      "epoch": 19.28672,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.3231,
      "router_z_loss": 0.0582,
      "step": 3780
    },
    {
      "balancing_entropy": 1.0277,
      "balancing_loss": 1.0792,
      "epoch": 19.33792,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.3391,
      "router_z_loss": 0.0461,
      "step": 3790
    },
    {
      "balancing_entropy": 0.9559,
      "balancing_loss": 1.1244,
      "epoch": 19.38912,
      "grad_norm": 1.140625,
      "learning_rate": 0.003,
      "loss": 2.3342,
      "router_z_loss": 0.0644,
      "step": 3800
    },
    {
      "balancing_entropy": 1.0394,
      "balancing_loss": 1.0628,
      "epoch": 19.44032,
      "grad_norm": 0.66015625,
      "learning_rate": 0.003,
      "loss": 2.3174,
      "router_z_loss": 0.1052,
      "step": 3810
    },
    {
      "balancing_entropy": 0.9799,
      "balancing_loss": 1.1481,
      "epoch": 19.49152,
      "grad_norm": 1.828125,
      "learning_rate": 0.003,
      "loss": 2.3237,
      "router_z_loss": 0.062,
      "step": 3820
    },
    {
      "balancing_entropy": 1.0131,
      "balancing_loss": 1.1006,
      "epoch": 19.54272,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.3311,
      "router_z_loss": 0.0451,
      "step": 3830
    },
    {
      "balancing_entropy": 0.9852,
      "balancing_loss": 1.1217,
      "epoch": 19.59392,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.3227,
      "router_z_loss": 0.0488,
      "step": 3840
    },
    {
      "balancing_entropy": 0.9467,
      "balancing_loss": 1.1036,
      "epoch": 19.64512,
      "grad_norm": 1.0546875,
      "learning_rate": 0.003,
      "loss": 2.324,
      "router_z_loss": 0.0505,
      "step": 3850
    },
    {
      "balancing_entropy": 0.9709,
      "balancing_loss": 1.1268,
      "epoch": 19.69632,
      "grad_norm": 1.046875,
      "learning_rate": 0.003,
      "loss": 2.357,
      "router_z_loss": 0.0569,
      "step": 3860
    },
    {
      "balancing_entropy": 0.9909,
      "balancing_loss": 1.0966,
      "epoch": 19.74752,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.3385,
      "router_z_loss": 0.0812,
      "step": 3870
    },
    {
      "balancing_entropy": 0.988,
      "balancing_loss": 1.0871,
      "epoch": 19.79872,
      "grad_norm": 0.796875,
      "learning_rate": 0.003,
      "loss": 2.3343,
      "router_z_loss": 0.1159,
      "step": 3880
    },
    {
      "balancing_entropy": 0.9813,
      "balancing_loss": 1.1034,
      "epoch": 19.84992,
      "grad_norm": 0.73828125,
      "learning_rate": 0.003,
      "loss": 2.3376,
      "router_z_loss": 0.0731,
      "step": 3890
    },
    {
      "balancing_entropy": 0.9973,
      "balancing_loss": 1.1281,
      "epoch": 19.90112,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.3492,
      "router_z_loss": 0.115,
      "step": 3900
    },
    {
      "balancing_entropy": 1.076,
      "balancing_loss": 1.0642,
      "epoch": 19.95232,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.3296,
      "router_z_loss": 0.097,
      "step": 3910
    },
    {
      "balancing_entropy": 1.0037,
      "balancing_loss": 1.1148,
      "epoch": 20.0,
      "grad_norm": 0.68359375,
      "learning_rate": 0.003,
      "loss": 2.3262,
      "router_z_loss": 0.0903,
      "step": 3920
    },
    {
      "balancing_entropy": 0.9806,
      "balancing_loss": 1.1113,
      "epoch": 20.0512,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.3089,
      "router_z_loss": 0.0482,
      "step": 3930
    },
    {
      "balancing_entropy": 0.9922,
      "balancing_loss": 1.1062,
      "epoch": 20.1024,
      "grad_norm": 1.3125,
      "learning_rate": 0.003,
      "loss": 2.3214,
      "router_z_loss": 0.1051,
      "step": 3940
    },
    {
      "balancing_entropy": 0.9845,
      "balancing_loss": 1.0857,
      "epoch": 20.1536,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.3089,
      "router_z_loss": 0.1009,
      "step": 3950
    },
    {
      "balancing_entropy": 0.92,
      "balancing_loss": 1.1608,
      "epoch": 20.2048,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.3282,
      "router_z_loss": 0.0542,
      "step": 3960
    },
    {
      "balancing_entropy": 0.9917,
      "balancing_loss": 1.0879,
      "epoch": 20.256,
      "grad_norm": 1.0234375,
      "learning_rate": 0.003,
      "loss": 2.3276,
      "router_z_loss": 0.0623,
      "step": 3970
    },
    {
      "balancing_entropy": 1.0161,
      "balancing_loss": 1.0919,
      "epoch": 20.3072,
      "grad_norm": 1.1484375,
      "learning_rate": 0.003,
      "loss": 2.3176,
      "router_z_loss": 0.095,
      "step": 3980
    },
    {
      "balancing_entropy": 0.9801,
      "balancing_loss": 1.1301,
      "epoch": 20.3584,
      "grad_norm": 0.96875,
      "learning_rate": 0.003,
      "loss": 2.3176,
      "router_z_loss": 0.0819,
      "step": 3990
    },
    {
      "balancing_entropy": 0.9796,
      "balancing_loss": 1.0714,
      "epoch": 20.4096,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.3211,
      "router_z_loss": 0.0573,
      "step": 4000
    },
    {
      "balancing_entropy": 0.994,
      "balancing_loss": 1.0704,
      "epoch": 20.4608,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.3222,
      "router_z_loss": 0.1033,
      "step": 4010
    },
    {
      "balancing_entropy": 1.0144,
      "balancing_loss": 1.1084,
      "epoch": 20.512,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.3282,
      "router_z_loss": 0.0519,
      "step": 4020
    },
    {
      "balancing_entropy": 0.9651,
      "balancing_loss": 1.1064,
      "epoch": 20.5632,
      "grad_norm": 0.984375,
      "learning_rate": 0.003,
      "loss": 2.3266,
      "router_z_loss": 0.0578,
      "step": 4030
    },
    {
      "balancing_entropy": 1.0006,
      "balancing_loss": 1.0973,
      "epoch": 20.6144,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.3216,
      "router_z_loss": 0.104,
      "step": 4040
    },
    {
      "balancing_entropy": 0.9431,
      "balancing_loss": 1.1347,
      "epoch": 20.6656,
      "grad_norm": 2.21875,
      "learning_rate": 0.003,
      "loss": 2.3185,
      "router_z_loss": 0.0726,
      "step": 4050
    },
    {
      "balancing_entropy": 0.9313,
      "balancing_loss": 1.1189,
      "epoch": 20.7168,
      "grad_norm": 1.578125,
      "learning_rate": 0.003,
      "loss": 2.3255,
      "router_z_loss": 0.0886,
      "step": 4060
    },
    {
      "balancing_entropy": 0.9636,
      "balancing_loss": 1.1192,
      "epoch": 20.768,
      "grad_norm": 0.8515625,
      "learning_rate": 0.003,
      "loss": 2.3247,
      "router_z_loss": 0.1498,
      "step": 4070
    },
    {
      "balancing_entropy": 1.0145,
      "balancing_loss": 1.1077,
      "epoch": 20.8192,
      "grad_norm": 0.93359375,
      "learning_rate": 0.003,
      "loss": 2.3155,
      "router_z_loss": 0.0511,
      "step": 4080
    },
    {
      "balancing_entropy": 0.9809,
      "balancing_loss": 1.0943,
      "epoch": 20.8704,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.3231,
      "router_z_loss": 0.0397,
      "step": 4090
    },
    {
      "balancing_entropy": 0.9702,
      "balancing_loss": 1.1199,
      "epoch": 20.9216,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.3174,
      "router_z_loss": 0.1303,
      "step": 4100
    },
    {
      "balancing_entropy": 1.0509,
      "balancing_loss": 1.0949,
      "epoch": 20.9728,
      "grad_norm": 0.7578125,
      "learning_rate": 0.003,
      "loss": 2.3065,
      "router_z_loss": 0.0783,
      "step": 4110
    },
    {
      "balancing_entropy": 1.0083,
      "balancing_loss": 1.117,
      "epoch": 21.02048,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.3131,
      "router_z_loss": 0.0365,
      "step": 4120
    },
    {
      "balancing_entropy": 1.0365,
      "balancing_loss": 1.1034,
      "epoch": 21.07168,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.3043,
      "router_z_loss": 0.0433,
      "step": 4130
    },
    {
      "balancing_entropy": 1.0347,
      "balancing_loss": 1.0772,
      "epoch": 21.12288,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.304,
      "router_z_loss": 0.0697,
      "step": 4140
    },
    {
      "balancing_entropy": 0.9986,
      "balancing_loss": 1.0839,
      "epoch": 21.17408,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.303,
      "router_z_loss": 0.073,
      "step": 4150
    },
    {
      "balancing_entropy": 1.0446,
      "balancing_loss": 1.0769,
      "epoch": 21.22528,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.301,
      "router_z_loss": 0.1243,
      "step": 4160
    },
    {
      "balancing_entropy": 1.0344,
      "balancing_loss": 1.0994,
      "epoch": 21.27648,
      "grad_norm": 1.28125,
      "learning_rate": 0.003,
      "loss": 2.2984,
      "router_z_loss": 0.0585,
      "step": 4170
    },
    {
      "balancing_entropy": 0.991,
      "balancing_loss": 1.1033,
      "epoch": 21.32768,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.2993,
      "router_z_loss": 0.0423,
      "step": 4180
    },
    {
      "balancing_entropy": 1.0169,
      "balancing_loss": 1.1242,
      "epoch": 21.37888,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.2987,
      "router_z_loss": 0.0553,
      "step": 4190
    },
    {
      "balancing_entropy": 1.0114,
      "balancing_loss": 1.1047,
      "epoch": 21.43008,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.3023,
      "router_z_loss": 0.0401,
      "step": 4200
    },
    {
      "balancing_entropy": 1.0318,
      "balancing_loss": 1.0789,
      "epoch": 21.48128,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.3104,
      "router_z_loss": 0.0617,
      "step": 4210
    },
    {
      "balancing_entropy": 1.0092,
      "balancing_loss": 1.1223,
      "epoch": 21.53248,
      "grad_norm": 1.4140625,
      "learning_rate": 0.003,
      "loss": 2.3093,
      "router_z_loss": 0.0825,
      "step": 4220
    },
    {
      "balancing_entropy": 1.0164,
      "balancing_loss": 1.0979,
      "epoch": 21.58368,
      "grad_norm": 1.34375,
      "learning_rate": 0.003,
      "loss": 2.3205,
      "router_z_loss": 0.0757,
      "step": 4230
    },
    {
      "balancing_entropy": 1.0386,
      "balancing_loss": 1.0734,
      "epoch": 21.63488,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.318,
      "router_z_loss": 0.108,
      "step": 4240
    },
    {
      "balancing_entropy": 0.9882,
      "balancing_loss": 1.1051,
      "epoch": 21.68608,
      "grad_norm": 0.921875,
      "learning_rate": 0.003,
      "loss": 2.3093,
      "router_z_loss": 0.0465,
      "step": 4250
    },
    {
      "balancing_entropy": 1.0052,
      "balancing_loss": 1.0697,
      "epoch": 21.73728,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.3104,
      "router_z_loss": 0.0694,
      "step": 4260
    },
    {
      "balancing_entropy": 0.9736,
      "balancing_loss": 1.1144,
      "epoch": 21.78848,
      "grad_norm": 1.15625,
      "learning_rate": 0.003,
      "loss": 2.314,
      "router_z_loss": 0.0864,
      "step": 4270
    },
    {
      "balancing_entropy": 1.0139,
      "balancing_loss": 1.0985,
      "epoch": 21.83968,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.3061,
      "router_z_loss": 0.0618,
      "step": 4280
    },
    {
      "balancing_entropy": 1.0323,
      "balancing_loss": 1.1025,
      "epoch": 21.89088,
      "grad_norm": 1.09375,
      "learning_rate": 0.003,
      "loss": 2.3149,
      "router_z_loss": 0.0769,
      "step": 4290
    },
    {
      "balancing_entropy": 1.0299,
      "balancing_loss": 1.1,
      "epoch": 21.94208,
      "grad_norm": 0.431640625,
      "learning_rate": 0.003,
      "loss": 2.3244,
      "router_z_loss": 0.0814,
      "step": 4300
    },
    {
      "balancing_entropy": 0.9933,
      "balancing_loss": 1.1094,
      "epoch": 21.99328,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.3168,
      "router_z_loss": 0.0887,
      "step": 4310
    },
    {
      "balancing_entropy": 0.9649,
      "balancing_loss": 1.1384,
      "epoch": 22.04096,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.2981,
      "router_z_loss": 0.0754,
      "step": 4320
    },
    {
      "balancing_entropy": 1.0121,
      "balancing_loss": 1.1211,
      "epoch": 22.09216,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.3025,
      "router_z_loss": 0.0914,
      "step": 4330
    },
    {
      "balancing_entropy": 1.0386,
      "balancing_loss": 1.1008,
      "epoch": 22.14336,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.3058,
      "router_z_loss": 0.0684,
      "step": 4340
    },
    {
      "balancing_entropy": 1.0207,
      "balancing_loss": 1.0878,
      "epoch": 22.19456,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.3017,
      "router_z_loss": 0.0768,
      "step": 4350
    },
    {
      "balancing_entropy": 1.0589,
      "balancing_loss": 1.1096,
      "epoch": 22.24576,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.2902,
      "router_z_loss": 0.079,
      "step": 4360
    },
    {
      "balancing_entropy": 1.0234,
      "balancing_loss": 1.1073,
      "epoch": 22.29696,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.2966,
      "router_z_loss": 0.0689,
      "step": 4370
    },
    {
      "balancing_entropy": 1.0521,
      "balancing_loss": 1.0981,
      "epoch": 22.34816,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.3038,
      "router_z_loss": 0.0632,
      "step": 4380
    },
    {
      "balancing_entropy": 0.997,
      "balancing_loss": 1.1125,
      "epoch": 22.39936,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.2963,
      "router_z_loss": 0.0627,
      "step": 4390
    },
    {
      "balancing_entropy": 1.0683,
      "balancing_loss": 1.0894,
      "epoch": 22.45056,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.3012,
      "router_z_loss": 0.0861,
      "step": 4400
    },
    {
      "balancing_entropy": 1.036,
      "balancing_loss": 1.1074,
      "epoch": 22.50176,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.3073,
      "router_z_loss": 0.074,
      "step": 4410
    },
    {
      "balancing_entropy": 1.0442,
      "balancing_loss": 1.1011,
      "epoch": 22.55296,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.2987,
      "router_z_loss": 0.0626,
      "step": 4420
    },
    {
      "balancing_entropy": 1.0869,
      "balancing_loss": 1.0935,
      "epoch": 22.60416,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.2967,
      "router_z_loss": 0.0607,
      "step": 4430
    },
    {
      "balancing_entropy": 1.0312,
      "balancing_loss": 1.0961,
      "epoch": 22.65536,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.2994,
      "router_z_loss": 0.0545,
      "step": 4440
    },
    {
      "balancing_entropy": 0.9888,
      "balancing_loss": 1.1291,
      "epoch": 22.70656,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.3054,
      "router_z_loss": 0.092,
      "step": 4450
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.0989,
      "epoch": 22.75776,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.305,
      "router_z_loss": 0.112,
      "step": 4460
    },
    {
      "balancing_entropy": 1.0807,
      "balancing_loss": 1.1079,
      "epoch": 22.80896,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.2933,
      "router_z_loss": 0.0689,
      "step": 4470
    },
    {
      "balancing_entropy": 1.0455,
      "balancing_loss": 1.129,
      "epoch": 22.86016,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.2949,
      "router_z_loss": 0.0749,
      "step": 4480
    },
    {
      "balancing_entropy": 0.9741,
      "balancing_loss": 1.1139,
      "epoch": 22.91136,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.3037,
      "router_z_loss": 0.0778,
      "step": 4490
    },
    {
      "balancing_entropy": 1.0449,
      "balancing_loss": 1.0985,
      "epoch": 22.96256,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.2982,
      "router_z_loss": 0.0701,
      "step": 4500
    },
    {
      "balancing_entropy": 0.9671,
      "balancing_loss": 1.1432,
      "epoch": 23.01024,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.2958,
      "router_z_loss": 0.0773,
      "step": 4510
    },
    {
      "balancing_entropy": 1.0522,
      "balancing_loss": 1.0749,
      "epoch": 23.06144,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.2931,
      "router_z_loss": 0.0479,
      "step": 4520
    },
    {
      "balancing_entropy": 1.058,
      "balancing_loss": 1.1032,
      "epoch": 23.11264,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.2799,
      "router_z_loss": 0.0531,
      "step": 4530
    },
    {
      "balancing_entropy": 1.0112,
      "balancing_loss": 1.1211,
      "epoch": 23.16384,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.2895,
      "router_z_loss": 0.0557,
      "step": 4540
    },
    {
      "balancing_entropy": 1.0503,
      "balancing_loss": 1.0983,
      "epoch": 23.21504,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.2897,
      "router_z_loss": 0.051,
      "step": 4550
    },
    {
      "balancing_entropy": 0.9587,
      "balancing_loss": 1.1447,
      "epoch": 23.26624,
      "grad_norm": 1.0078125,
      "learning_rate": 0.003,
      "loss": 2.2945,
      "router_z_loss": 0.0513,
      "step": 4560
    },
    {
      "balancing_entropy": 1.0157,
      "balancing_loss": 1.1117,
      "epoch": 23.31744,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.2826,
      "router_z_loss": 0.1023,
      "step": 4570
    },
    {
      "balancing_entropy": 1.0767,
      "balancing_loss": 1.0876,
      "epoch": 23.36864,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.289,
      "router_z_loss": 0.0497,
      "step": 4580
    },
    {
      "balancing_entropy": 1.017,
      "balancing_loss": 1.1066,
      "epoch": 23.41984,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.2897,
      "router_z_loss": 0.0466,
      "step": 4590
    },
    {
      "balancing_entropy": 1.0203,
      "balancing_loss": 1.0965,
      "epoch": 23.47104,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.2904,
      "router_z_loss": 0.0992,
      "step": 4600
    },
    {
      "balancing_entropy": 1.0458,
      "balancing_loss": 1.1226,
      "epoch": 23.52224,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.2933,
      "router_z_loss": 0.0592,
      "step": 4610
    },
    {
      "balancing_entropy": 1.0644,
      "balancing_loss": 1.0938,
      "epoch": 23.57344,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.2888,
      "router_z_loss": 0.0783,
      "step": 4620
    },
    {
      "balancing_entropy": 0.9832,
      "balancing_loss": 1.1158,
      "epoch": 23.62464,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.2939,
      "router_z_loss": 0.0818,
      "step": 4630
    },
    {
      "balancing_entropy": 1.0649,
      "balancing_loss": 1.11,
      "epoch": 23.67584,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.286,
      "router_z_loss": 0.0492,
      "step": 4640
    },
    {
      "balancing_entropy": 1.0653,
      "balancing_loss": 1.1182,
      "epoch": 23.72704,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.2813,
      "router_z_loss": 0.0603,
      "step": 4650
    },
    {
      "balancing_entropy": 1.0693,
      "balancing_loss": 1.1085,
      "epoch": 23.77824,
      "grad_norm": 0.337890625,
      "learning_rate": 0.003,
      "loss": 2.2936,
      "router_z_loss": 0.0592,
      "step": 4660
    },
    {
      "balancing_entropy": 1.0412,
      "balancing_loss": 1.1048,
      "epoch": 23.829439999999998,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.2894,
      "router_z_loss": 0.0553,
      "step": 4670
    },
    {
      "balancing_entropy": 1.0439,
      "balancing_loss": 1.0996,
      "epoch": 23.88064,
      "grad_norm": 0.2294921875,
      "learning_rate": 0.003,
      "loss": 2.2905,
      "router_z_loss": 0.0826,
      "step": 4680
    },
    {
      "balancing_entropy": 1.0721,
      "balancing_loss": 1.1193,
      "epoch": 23.93184,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.2849,
      "router_z_loss": 0.1303,
      "step": 4690
    },
    {
      "balancing_entropy": 1.0599,
      "balancing_loss": 1.1147,
      "epoch": 23.98304,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.2854,
      "router_z_loss": 0.0824,
      "step": 4700
    },
    {
      "balancing_entropy": 1.0174,
      "balancing_loss": 1.1029,
      "epoch": 24.03072,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.284,
      "router_z_loss": 0.0621,
      "step": 4710
    },
    {
      "balancing_entropy": 1.0622,
      "balancing_loss": 1.1321,
      "epoch": 24.08192,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.279,
      "router_z_loss": 0.0651,
      "step": 4720
    },
    {
      "balancing_entropy": 1.0697,
      "balancing_loss": 1.102,
      "epoch": 24.13312,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.2731,
      "router_z_loss": 0.0745,
      "step": 4730
    },
    {
      "balancing_entropy": 1.0307,
      "balancing_loss": 1.1016,
      "epoch": 24.18432,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.2747,
      "router_z_loss": 0.0626,
      "step": 4740
    },
    {
      "balancing_entropy": 1.0247,
      "balancing_loss": 1.112,
      "epoch": 24.23552,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.2801,
      "router_z_loss": 0.074,
      "step": 4750
    },
    {
      "balancing_entropy": 1.0522,
      "balancing_loss": 1.1336,
      "epoch": 24.28672,
      "grad_norm": 0.73828125,
      "learning_rate": 0.003,
      "loss": 2.2689,
      "router_z_loss": 0.0944,
      "step": 4760
    },
    {
      "balancing_entropy": 1.0666,
      "balancing_loss": 1.0862,
      "epoch": 24.33792,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.2843,
      "router_z_loss": 0.0878,
      "step": 4770
    },
    {
      "balancing_entropy": 1.0625,
      "balancing_loss": 1.1256,
      "epoch": 24.38912,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.2736,
      "router_z_loss": 0.0909,
      "step": 4780
    },
    {
      "balancing_entropy": 1.0004,
      "balancing_loss": 1.1064,
      "epoch": 24.44032,
      "grad_norm": 0.9921875,
      "learning_rate": 0.003,
      "loss": 2.2815,
      "router_z_loss": 0.0825,
      "step": 4790
    },
    {
      "balancing_entropy": 1.0582,
      "balancing_loss": 1.1133,
      "epoch": 24.49152,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.2827,
      "router_z_loss": 0.0775,
      "step": 4800
    },
    {
      "balancing_entropy": 1.0575,
      "balancing_loss": 1.1083,
      "epoch": 24.54272,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.2786,
      "router_z_loss": 0.0529,
      "step": 4810
    },
    {
      "balancing_entropy": 1.051,
      "balancing_loss": 1.1038,
      "epoch": 24.59392,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.2794,
      "router_z_loss": 0.0623,
      "step": 4820
    },
    {
      "balancing_entropy": 1.0784,
      "balancing_loss": 1.1123,
      "epoch": 24.64512,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.2812,
      "router_z_loss": 0.0692,
      "step": 4830
    },
    {
      "balancing_entropy": 1.0182,
      "balancing_loss": 1.0972,
      "epoch": 24.69632,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.2852,
      "router_z_loss": 0.0656,
      "step": 4840
    },
    {
      "balancing_entropy": 1.0105,
      "balancing_loss": 1.1418,
      "epoch": 24.74752,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.2834,
      "router_z_loss": 0.1043,
      "step": 4850
    },
    {
      "balancing_entropy": 1.0733,
      "balancing_loss": 1.1052,
      "epoch": 24.79872,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.278,
      "router_z_loss": 0.0783,
      "step": 4860
    },
    {
      "balancing_entropy": 1.0419,
      "balancing_loss": 1.1216,
      "epoch": 24.84992,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.2794,
      "router_z_loss": 0.0613,
      "step": 4870
    },
    {
      "balancing_entropy": 1.0456,
      "balancing_loss": 1.088,
      "epoch": 24.90112,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.2817,
      "router_z_loss": 0.0701,
      "step": 4880
    },
    {
      "balancing_entropy": 1.0644,
      "balancing_loss": 1.1213,
      "epoch": 24.95232,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.2767,
      "router_z_loss": 0.0718,
      "step": 4890
    },
    {
      "balancing_entropy": 0.9996,
      "balancing_loss": 1.0968,
      "epoch": 25.0,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.2842,
      "router_z_loss": 0.0697,
      "step": 4900
    },
    {
      "balancing_entropy": 1.062,
      "balancing_loss": 1.1139,
      "epoch": 25.0512,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.2632,
      "router_z_loss": 0.0678,
      "step": 4910
    },
    {
      "balancing_entropy": 1.0703,
      "balancing_loss": 1.1125,
      "epoch": 25.1024,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.2709,
      "router_z_loss": 0.0703,
      "step": 4920
    },
    {
      "balancing_entropy": 1.068,
      "balancing_loss": 1.0997,
      "epoch": 25.1536,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.2618,
      "router_z_loss": 0.0866,
      "step": 4930
    },
    {
      "balancing_entropy": 1.0453,
      "balancing_loss": 1.1311,
      "epoch": 25.2048,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.2662,
      "router_z_loss": 0.0703,
      "step": 4940
    },
    {
      "balancing_entropy": 1.0575,
      "balancing_loss": 1.1103,
      "epoch": 25.256,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.2687,
      "router_z_loss": 0.0756,
      "step": 4950
    },
    {
      "balancing_entropy": 1.0336,
      "balancing_loss": 1.1236,
      "epoch": 25.3072,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.2704,
      "router_z_loss": 0.0754,
      "step": 4960
    },
    {
      "balancing_entropy": 1.0695,
      "balancing_loss": 1.1229,
      "epoch": 25.3584,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.2712,
      "router_z_loss": 0.0758,
      "step": 4970
    },
    {
      "balancing_entropy": 1.0707,
      "balancing_loss": 1.0993,
      "epoch": 25.4096,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.2599,
      "router_z_loss": 0.0969,
      "step": 4980
    },
    {
      "balancing_entropy": 1.0772,
      "balancing_loss": 1.1077,
      "epoch": 25.4608,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.2655,
      "router_z_loss": 0.0703,
      "step": 4990
    },
    {
      "balancing_entropy": 1.0099,
      "balancing_loss": 1.1209,
      "epoch": 25.512,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.2664,
      "router_z_loss": 0.0592,
      "step": 5000
    },
    {
      "balancing_entropy": 1.0519,
      "balancing_loss": 1.1157,
      "epoch": 25.5632,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.2703,
      "router_z_loss": 0.0764,
      "step": 5010
    },
    {
      "balancing_entropy": 1.0657,
      "balancing_loss": 1.0938,
      "epoch": 25.6144,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.2714,
      "router_z_loss": 0.0514,
      "step": 5020
    },
    {
      "balancing_entropy": 1.058,
      "balancing_loss": 1.1137,
      "epoch": 25.6656,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.2666,
      "router_z_loss": 0.0519,
      "step": 5030
    },
    {
      "balancing_entropy": 1.084,
      "balancing_loss": 1.1117,
      "epoch": 25.7168,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.2764,
      "router_z_loss": 0.0599,
      "step": 5040
    },
    {
      "balancing_entropy": 1.0462,
      "balancing_loss": 1.0983,
      "epoch": 25.768,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.2645,
      "router_z_loss": 0.0594,
      "step": 5050
    },
    {
      "balancing_entropy": 1.0299,
      "balancing_loss": 1.1031,
      "epoch": 25.8192,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.282,
      "router_z_loss": 0.0594,
      "step": 5060
    },
    {
      "balancing_entropy": 1.06,
      "balancing_loss": 1.112,
      "epoch": 25.8704,
      "grad_norm": 0.310546875,
      "learning_rate": 0.003,
      "loss": 2.2785,
      "router_z_loss": 0.0837,
      "step": 5070
    },
    {
      "balancing_entropy": 1.0362,
      "balancing_loss": 1.101,
      "epoch": 25.9216,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.2733,
      "router_z_loss": 0.0764,
      "step": 5080
    },
    {
      "balancing_entropy": 1.0879,
      "balancing_loss": 1.1054,
      "epoch": 25.9728,
      "grad_norm": 0.232421875,
      "learning_rate": 0.003,
      "loss": 2.2792,
      "router_z_loss": 0.0648,
      "step": 5090
    },
    {
      "balancing_entropy": 1.0136,
      "balancing_loss": 1.1205,
      "epoch": 26.02048,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.267,
      "router_z_loss": 0.066,
      "step": 5100
    },
    {
      "balancing_entropy": 0.9935,
      "balancing_loss": 1.0999,
      "epoch": 26.07168,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.2552,
      "router_z_loss": 0.0783,
      "step": 5110
    },
    {
      "balancing_entropy": 0.9758,
      "balancing_loss": 1.1462,
      "epoch": 26.12288,
      "grad_norm": 0.8125,
      "learning_rate": 0.003,
      "loss": 2.2642,
      "router_z_loss": 0.0918,
      "step": 5120
    },
    {
      "balancing_entropy": 1.0589,
      "balancing_loss": 1.1103,
      "epoch": 26.17408,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.2603,
      "router_z_loss": 0.1041,
      "step": 5130
    },
    {
      "balancing_entropy": 1.0798,
      "balancing_loss": 1.0954,
      "epoch": 26.22528,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.2587,
      "router_z_loss": 0.0762,
      "step": 5140
    },
    {
      "balancing_entropy": 1.0407,
      "balancing_loss": 1.1139,
      "epoch": 26.27648,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.2628,
      "router_z_loss": 0.0712,
      "step": 5150
    },
    {
      "balancing_entropy": 1.0638,
      "balancing_loss": 1.0893,
      "epoch": 26.32768,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.2626,
      "router_z_loss": 0.0845,
      "step": 5160
    },
    {
      "balancing_entropy": 1.049,
      "balancing_loss": 1.1213,
      "epoch": 26.37888,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.2601,
      "router_z_loss": 0.089,
      "step": 5170
    },
    {
      "balancing_entropy": 1.0782,
      "balancing_loss": 1.1171,
      "epoch": 26.43008,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.2603,
      "router_z_loss": 0.0841,
      "step": 5180
    },
    {
      "balancing_entropy": 1.0742,
      "balancing_loss": 1.1185,
      "epoch": 26.48128,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.2593,
      "router_z_loss": 0.0634,
      "step": 5190
    },
    {
      "balancing_entropy": 1.0526,
      "balancing_loss": 1.1141,
      "epoch": 26.53248,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.2661,
      "router_z_loss": 0.0536,
      "step": 5200
    },
    {
      "balancing_entropy": 1.0757,
      "balancing_loss": 1.1057,
      "epoch": 26.58368,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.2574,
      "router_z_loss": 0.0694,
      "step": 5210
    },
    {
      "balancing_entropy": 1.0242,
      "balancing_loss": 1.1216,
      "epoch": 26.63488,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.2664,
      "router_z_loss": 0.0753,
      "step": 5220
    },
    {
      "balancing_entropy": 1.0898,
      "balancing_loss": 1.1108,
      "epoch": 26.68608,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.2561,
      "router_z_loss": 0.0769,
      "step": 5230
    },
    {
      "balancing_entropy": 1.0466,
      "balancing_loss": 1.1026,
      "epoch": 26.73728,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.2637,
      "router_z_loss": 0.0691,
      "step": 5240
    },
    {
      "balancing_entropy": 1.0533,
      "balancing_loss": 1.086,
      "epoch": 26.78848,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.267,
      "router_z_loss": 0.0621,
      "step": 5250
    },
    {
      "balancing_entropy": 1.0343,
      "balancing_loss": 1.1469,
      "epoch": 26.83968,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.2592,
      "router_z_loss": 0.0691,
      "step": 5260
    },
    {
      "balancing_entropy": 1.065,
      "balancing_loss": 1.0988,
      "epoch": 26.89088,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.2602,
      "router_z_loss": 0.089,
      "step": 5270
    },
    {
      "balancing_entropy": 1.062,
      "balancing_loss": 1.1036,
      "epoch": 26.94208,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.2738,
      "router_z_loss": 0.0746,
      "step": 5280
    },
    {
      "balancing_entropy": 1.0687,
      "balancing_loss": 1.1189,
      "epoch": 26.99328,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.2598,
      "router_z_loss": 0.058,
      "step": 5290
    },
    {
      "balancing_entropy": 1.0503,
      "balancing_loss": 1.0973,
      "epoch": 27.04096,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.2601,
      "router_z_loss": 0.0598,
      "step": 5300
    },
    {
      "balancing_entropy": 1.0341,
      "balancing_loss": 1.107,
      "epoch": 27.09216,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.2525,
      "router_z_loss": 0.0573,
      "step": 5310
    },
    {
      "balancing_entropy": 1.0636,
      "balancing_loss": 1.1166,
      "epoch": 27.14336,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.2528,
      "router_z_loss": 0.0577,
      "step": 5320
    },
    {
      "balancing_entropy": 1.0274,
      "balancing_loss": 1.1319,
      "epoch": 27.19456,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.2542,
      "router_z_loss": 0.0623,
      "step": 5330
    },
    {
      "balancing_entropy": 1.052,
      "balancing_loss": 1.1113,
      "epoch": 27.24576,
      "grad_norm": 1.0234375,
      "learning_rate": 0.003,
      "loss": 2.2575,
      "router_z_loss": 0.0603,
      "step": 5340
    },
    {
      "balancing_entropy": 1.0635,
      "balancing_loss": 1.0983,
      "epoch": 27.29696,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.255,
      "router_z_loss": 0.0657,
      "step": 5350
    },
    {
      "balancing_entropy": 1.0269,
      "balancing_loss": 1.128,
      "epoch": 27.34816,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.2484,
      "router_z_loss": 0.0941,
      "step": 5360
    },
    {
      "balancing_entropy": 1.0691,
      "balancing_loss": 1.1092,
      "epoch": 27.39936,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.2518,
      "router_z_loss": 0.0799,
      "step": 5370
    },
    {
      "balancing_entropy": 1.0661,
      "balancing_loss": 1.1152,
      "epoch": 27.45056,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.2535,
      "router_z_loss": 0.0547,
      "step": 5380
    },
    {
      "balancing_entropy": 1.0512,
      "balancing_loss": 1.1246,
      "epoch": 27.50176,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.2583,
      "router_z_loss": 0.0615,
      "step": 5390
    },
    {
      "balancing_entropy": 1.0345,
      "balancing_loss": 1.1092,
      "epoch": 27.55296,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.2528,
      "router_z_loss": 0.0619,
      "step": 5400
    },
    {
      "balancing_entropy": 1.0376,
      "balancing_loss": 1.1056,
      "epoch": 27.60416,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.2579,
      "router_z_loss": 0.0608,
      "step": 5410
    },
    {
      "balancing_entropy": 1.0468,
      "balancing_loss": 1.112,
      "epoch": 27.65536,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.2551,
      "router_z_loss": 0.0635,
      "step": 5420
    },
    {
      "balancing_entropy": 1.06,
      "balancing_loss": 1.1296,
      "epoch": 27.70656,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.2599,
      "router_z_loss": 0.0636,
      "step": 5430
    },
    {
      "balancing_entropy": 1.0532,
      "balancing_loss": 1.1051,
      "epoch": 27.75776,
      "grad_norm": 0.7421875,
      "learning_rate": 0.003,
      "loss": 2.2599,
      "router_z_loss": 0.0533,
      "step": 5440
    },
    {
      "balancing_entropy": 1.0357,
      "balancing_loss": 1.1139,
      "epoch": 27.80896,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.2595,
      "router_z_loss": 0.0565,
      "step": 5450
    },
    {
      "balancing_entropy": 1.0857,
      "balancing_loss": 1.1223,
      "epoch": 27.86016,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.2522,
      "router_z_loss": 0.0746,
      "step": 5460
    },
    {
      "balancing_entropy": 1.0409,
      "balancing_loss": 1.1084,
      "epoch": 27.91136,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.261,
      "router_z_loss": 0.0578,
      "step": 5470
    },
    {
      "balancing_entropy": 1.0715,
      "balancing_loss": 1.1053,
      "epoch": 27.96256,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.257,
      "router_z_loss": 0.0521,
      "step": 5480
    },
    {
      "balancing_entropy": 1.017,
      "balancing_loss": 1.0934,
      "epoch": 28.01024,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.2645,
      "router_z_loss": 0.056,
      "step": 5490
    },
    {
      "balancing_entropy": 1.0234,
      "balancing_loss": 1.1356,
      "epoch": 28.06144,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.2519,
      "router_z_loss": 0.085,
      "step": 5500
    },
    {
      "balancing_entropy": 1.0679,
      "balancing_loss": 1.0915,
      "epoch": 28.11264,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.2495,
      "router_z_loss": 0.0779,
      "step": 5510
    },
    {
      "balancing_entropy": 1.0404,
      "balancing_loss": 1.1107,
      "epoch": 28.16384,
      "grad_norm": 0.859375,
      "learning_rate": 0.003,
      "loss": 2.2463,
      "router_z_loss": 0.0602,
      "step": 5520
    },
    {
      "balancing_entropy": 1.0338,
      "balancing_loss": 1.132,
      "epoch": 28.21504,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.2388,
      "router_z_loss": 0.0857,
      "step": 5530
    },
    {
      "balancing_entropy": 1.0831,
      "balancing_loss": 1.0998,
      "epoch": 28.26624,
      "grad_norm": 0.271484375,
      "learning_rate": 0.003,
      "loss": 2.2456,
      "router_z_loss": 0.1085,
      "step": 5540
    },
    {
      "balancing_entropy": 1.0891,
      "balancing_loss": 1.1108,
      "epoch": 28.31744,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.2479,
      "router_z_loss": 1.1948,
      "step": 5550
    },
    {
      "balancing_entropy": 1.0815,
      "balancing_loss": 1.1151,
      "epoch": 28.36864,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.2498,
      "router_z_loss": 0.1149,
      "step": 5560
    },
    {
      "balancing_entropy": 1.0447,
      "balancing_loss": 1.1265,
      "epoch": 28.41984,
      "grad_norm": 0.2890625,
      "learning_rate": 0.003,
      "loss": 2.2481,
      "router_z_loss": 0.1026,
      "step": 5570
    },
    {
      "balancing_entropy": 1.0508,
      "balancing_loss": 1.1224,
      "epoch": 28.47104,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.256,
      "router_z_loss": 0.0829,
      "step": 5580
    },
    {
      "balancing_entropy": 1.0472,
      "balancing_loss": 1.1145,
      "epoch": 28.52224,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.2526,
      "router_z_loss": 0.0682,
      "step": 5590
    },
    {
      "balancing_entropy": 1.0612,
      "balancing_loss": 1.0956,
      "epoch": 28.57344,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.2481,
      "router_z_loss": 0.0679,
      "step": 5600
    },
    {
      "balancing_entropy": 0.9237,
      "balancing_loss": 1.128,
      "epoch": 28.62464,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.245,
      "router_z_loss": 0.1209,
      "step": 5610
    },
    {
      "balancing_entropy": 0.9924,
      "balancing_loss": 1.1057,
      "epoch": 28.67584,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.2567,
      "router_z_loss": 0.151,
      "step": 5620
    },
    {
      "balancing_entropy": 0.9231,
      "balancing_loss": 1.1094,
      "epoch": 28.72704,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.2552,
      "router_z_loss": 0.1372,
      "step": 5630
    },
    {
      "balancing_entropy": 1.0584,
      "balancing_loss": 1.12,
      "epoch": 28.77824,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.2502,
      "router_z_loss": 0.1233,
      "step": 5640
    },
    {
      "balancing_entropy": 1.036,
      "balancing_loss": 1.1149,
      "epoch": 28.829439999999998,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.2517,
      "router_z_loss": 0.1208,
      "step": 5650
    },
    {
      "balancing_entropy": 0.9966,
      "balancing_loss": 1.1098,
      "epoch": 28.88064,
      "grad_norm": 0.30078125,
      "learning_rate": 0.003,
      "loss": 2.2554,
      "router_z_loss": 0.1008,
      "step": 5660
    },
    {
      "balancing_entropy": 1.0721,
      "balancing_loss": 1.0993,
      "epoch": 28.93184,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.2577,
      "router_z_loss": 0.0722,
      "step": 5670
    },
    {
      "balancing_entropy": 1.0566,
      "balancing_loss": 1.1084,
      "epoch": 28.98304,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.254,
      "router_z_loss": 0.066,
      "step": 5680
    },
    {
      "balancing_entropy": 0.9906,
      "balancing_loss": 1.1132,
      "epoch": 29.03072,
      "grad_norm": 1.296875,
      "learning_rate": 0.003,
      "loss": 2.2478,
      "router_z_loss": 0.0822,
      "step": 5690
    },
    {
      "balancing_entropy": 1.0614,
      "balancing_loss": 1.1268,
      "epoch": 29.08192,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.2419,
      "router_z_loss": 0.0952,
      "step": 5700
    },
    {
      "balancing_entropy": 1.0962,
      "balancing_loss": 1.1015,
      "epoch": 29.13312,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.234,
      "router_z_loss": 0.1338,
      "step": 5710
    },
    {
      "balancing_entropy": 1.0418,
      "balancing_loss": 1.1119,
      "epoch": 29.18432,
      "grad_norm": 0.2734375,
      "learning_rate": 0.003,
      "loss": 2.2489,
      "router_z_loss": 0.0976,
      "step": 5720
    },
    {
      "balancing_entropy": 1.0249,
      "balancing_loss": 1.109,
      "epoch": 29.23552,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.2444,
      "router_z_loss": 0.0899,
      "step": 5730
    },
    {
      "balancing_entropy": 1.0573,
      "balancing_loss": 1.121,
      "epoch": 29.28672,
      "grad_norm": 1.125,
      "learning_rate": 0.003,
      "loss": 2.2352,
      "router_z_loss": 0.1019,
      "step": 5740
    },
    {
      "balancing_entropy": 1.0079,
      "balancing_loss": 1.1114,
      "epoch": 29.33792,
      "grad_norm": 0.859375,
      "learning_rate": 0.003,
      "loss": 2.2433,
      "router_z_loss": 0.087,
      "step": 5750
    },
    {
      "balancing_entropy": 1.0769,
      "balancing_loss": 1.113,
      "epoch": 29.38912,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.2455,
      "router_z_loss": 0.0885,
      "step": 5760
    },
    {
      "balancing_entropy": 0.9391,
      "balancing_loss": 1.1108,
      "epoch": 29.44032,
      "grad_norm": 0.95703125,
      "learning_rate": 0.003,
      "loss": 2.2483,
      "router_z_loss": 0.0661,
      "step": 5770
    },
    {
      "balancing_entropy": 0.8859,
      "balancing_loss": 1.1115,
      "epoch": 29.49152,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.255,
      "router_z_loss": 0.084,
      "step": 5780
    },
    {
      "balancing_entropy": 0.9401,
      "balancing_loss": 1.1269,
      "epoch": 29.54272,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.2469,
      "router_z_loss": 0.0953,
      "step": 5790
    },
    {
      "balancing_entropy": 1.03,
      "balancing_loss": 1.1134,
      "epoch": 29.59392,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.2391,
      "router_z_loss": 0.1085,
      "step": 5800
    },
    {
      "balancing_entropy": 1.0839,
      "balancing_loss": 1.1083,
      "epoch": 29.64512,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.2458,
      "router_z_loss": 0.0835,
      "step": 5810
    },
    {
      "balancing_entropy": 1.0112,
      "balancing_loss": 1.1115,
      "epoch": 29.69632,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.2458,
      "router_z_loss": 0.0765,
      "step": 5820
    },
    {
      "balancing_entropy": 1.0902,
      "balancing_loss": 1.1104,
      "epoch": 29.74752,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.2493,
      "router_z_loss": 0.0782,
      "step": 5830
    },
    {
      "balancing_entropy": 1.089,
      "balancing_loss": 1.1087,
      "epoch": 29.79872,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.2456,
      "router_z_loss": 0.083,
      "step": 5840
    },
    {
      "balancing_entropy": 0.9726,
      "balancing_loss": 1.1133,
      "epoch": 29.84992,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.2421,
      "router_z_loss": 0.0655,
      "step": 5850
    },
    {
      "balancing_entropy": 1.0072,
      "balancing_loss": 1.1221,
      "epoch": 29.90112,
      "grad_norm": 1.6953125,
      "learning_rate": 0.003,
      "loss": 2.2452,
      "router_z_loss": 0.083,
      "step": 5860
    },
    {
      "balancing_entropy": 1.0231,
      "balancing_loss": 1.1039,
      "epoch": 29.95232,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.2558,
      "router_z_loss": 0.0742,
      "step": 5870
    },
    {
      "balancing_entropy": 1.0169,
      "balancing_loss": 1.1196,
      "epoch": 30.0,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.2488,
      "router_z_loss": 0.0786,
      "step": 5880
    },
    {
      "balancing_entropy": 1.0776,
      "balancing_loss": 1.1219,
      "epoch": 30.0512,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.227,
      "router_z_loss": 0.0716,
      "step": 5890
    },
    {
      "balancing_entropy": 1.0702,
      "balancing_loss": 1.0987,
      "epoch": 30.1024,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.2349,
      "router_z_loss": 0.0685,
      "step": 5900
    },
    {
      "balancing_entropy": 0.9096,
      "balancing_loss": 1.117,
      "epoch": 30.1536,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.2356,
      "router_z_loss": 0.0721,
      "step": 5910
    },
    {
      "balancing_entropy": 1.039,
      "balancing_loss": 1.0973,
      "epoch": 30.2048,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.241,
      "router_z_loss": 0.0707,
      "step": 5920
    },
    {
      "balancing_entropy": 1.0398,
      "balancing_loss": 1.0978,
      "epoch": 30.256,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.2368,
      "router_z_loss": 0.0799,
      "step": 5930
    },
    {
      "balancing_entropy": 1.0689,
      "balancing_loss": 1.1169,
      "epoch": 30.3072,
      "grad_norm": 0.2412109375,
      "learning_rate": 0.003,
      "loss": 2.2314,
      "router_z_loss": 0.08,
      "step": 5940
    },
    {
      "balancing_entropy": 1.0581,
      "balancing_loss": 1.1059,
      "epoch": 30.3584,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.231,
      "router_z_loss": 0.1012,
      "step": 5950
    },
    {
      "balancing_entropy": 0.9286,
      "balancing_loss": 1.1149,
      "epoch": 30.4096,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.2326,
      "router_z_loss": 0.0661,
      "step": 5960
    },
    {
      "balancing_entropy": 0.922,
      "balancing_loss": 1.1045,
      "epoch": 30.4608,
      "grad_norm": 0.68359375,
      "learning_rate": 0.003,
      "loss": 2.2368,
      "router_z_loss": 0.0834,
      "step": 5970
    },
    {
      "balancing_entropy": 0.9997,
      "balancing_loss": 1.0946,
      "epoch": 30.512,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.2366,
      "router_z_loss": 0.0851,
      "step": 5980
    },
    {
      "balancing_entropy": 0.9418,
      "balancing_loss": 1.1101,
      "epoch": 30.5632,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.2457,
      "router_z_loss": 0.0828,
      "step": 5990
    },
    {
      "balancing_entropy": 1.0274,
      "balancing_loss": 1.1082,
      "epoch": 30.6144,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.2403,
      "router_z_loss": 0.084,
      "step": 6000
    },
    {
      "balancing_entropy": 0.9224,
      "balancing_loss": 1.1165,
      "epoch": 30.6656,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.243,
      "router_z_loss": 0.0858,
      "step": 6010
    },
    {
      "balancing_entropy": 1.001,
      "balancing_loss": 1.1168,
      "epoch": 30.7168,
      "grad_norm": 0.89453125,
      "learning_rate": 0.003,
      "loss": 2.2398,
      "router_z_loss": 0.0784,
      "step": 6020
    },
    {
      "balancing_entropy": 0.9679,
      "balancing_loss": 1.1275,
      "epoch": 30.768,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.242,
      "router_z_loss": 0.0873,
      "step": 6030
    },
    {
      "balancing_entropy": 1.0209,
      "balancing_loss": 1.1233,
      "epoch": 30.8192,
      "grad_norm": 0.84765625,
      "learning_rate": 0.003,
      "loss": 2.2408,
      "router_z_loss": 0.1096,
      "step": 6040
    },
    {
      "balancing_entropy": 1.0429,
      "balancing_loss": 1.1081,
      "epoch": 30.8704,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.2453,
      "router_z_loss": 0.0842,
      "step": 6050
    },
    {
      "balancing_entropy": 1.0865,
      "balancing_loss": 1.1161,
      "epoch": 30.9216,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.242,
      "router_z_loss": 0.07,
      "step": 6060
    },
    {
      "balancing_entropy": 1.0556,
      "balancing_loss": 1.1018,
      "epoch": 30.9728,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.2478,
      "router_z_loss": 0.0588,
      "step": 6070
    },
    {
      "balancing_entropy": 0.9794,
      "balancing_loss": 1.118,
      "epoch": 31.02048,
      "grad_norm": 1.1640625,
      "learning_rate": 0.003,
      "loss": 2.2331,
      "router_z_loss": 0.0625,
      "step": 6080
    },
    {
      "balancing_entropy": 0.9941,
      "balancing_loss": 1.1143,
      "epoch": 31.07168,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.2396,
      "router_z_loss": 0.0679,
      "step": 6090
    },
    {
      "balancing_entropy": 1.0776,
      "balancing_loss": 1.1017,
      "epoch": 31.12288,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.2257,
      "router_z_loss": 0.067,
      "step": 6100
    },
    {
      "balancing_entropy": 1.0375,
      "balancing_loss": 1.1198,
      "epoch": 31.17408,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.2332,
      "router_z_loss": 0.075,
      "step": 6110
    },
    {
      "balancing_entropy": 1.0916,
      "balancing_loss": 1.1106,
      "epoch": 31.22528,
      "grad_norm": 0.28125,
      "learning_rate": 0.003,
      "loss": 2.2238,
      "router_z_loss": 0.0614,
      "step": 6120
    },
    {
      "balancing_entropy": 0.9994,
      "balancing_loss": 1.1117,
      "epoch": 31.27648,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.2243,
      "router_z_loss": 0.0551,
      "step": 6130
    },
    {
      "balancing_entropy": 0.9634,
      "balancing_loss": 1.0934,
      "epoch": 31.32768,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.2304,
      "router_z_loss": 0.0755,
      "step": 6140
    },
    {
      "balancing_entropy": 0.975,
      "balancing_loss": 1.1173,
      "epoch": 31.37888,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.2281,
      "router_z_loss": 0.089,
      "step": 6150
    },
    {
      "balancing_entropy": 0.9062,
      "balancing_loss": 1.1235,
      "epoch": 31.43008,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.2385,
      "router_z_loss": 0.098,
      "step": 6160
    },
    {
      "balancing_entropy": 1.0421,
      "balancing_loss": 1.1203,
      "epoch": 31.48128,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.2436,
      "router_z_loss": 0.0853,
      "step": 6170
    },
    {
      "balancing_entropy": 1.0242,
      "balancing_loss": 1.1085,
      "epoch": 31.53248,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.234,
      "router_z_loss": 0.0779,
      "step": 6180
    },
    {
      "balancing_entropy": 0.9857,
      "balancing_loss": 1.1306,
      "epoch": 31.58368,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.2367,
      "router_z_loss": 0.0786,
      "step": 6190
    },
    {
      "balancing_entropy": 1.0517,
      "balancing_loss": 1.1082,
      "epoch": 31.63488,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.239,
      "router_z_loss": 0.0741,
      "step": 6200
    },
    {
      "balancing_entropy": 1.0,
      "balancing_loss": 1.1084,
      "epoch": 31.68608,
      "grad_norm": 0.98828125,
      "learning_rate": 0.003,
      "loss": 2.2366,
      "router_z_loss": 0.0677,
      "step": 6210
    },
    {
      "balancing_entropy": 0.8891,
      "balancing_loss": 1.1102,
      "epoch": 31.73728,
      "grad_norm": 1.125,
      "learning_rate": 0.003,
      "loss": 2.2437,
      "router_z_loss": 0.071,
      "step": 6220
    },
    {
      "balancing_entropy": 1.0031,
      "balancing_loss": 1.1085,
      "epoch": 31.78848,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.2365,
      "router_z_loss": 0.098,
      "step": 6230
    },
    {
      "balancing_entropy": 1.0071,
      "balancing_loss": 1.1112,
      "epoch": 31.83968,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.2382,
      "router_z_loss": 0.0916,
      "step": 6240
    },
    {
      "balancing_entropy": 1.0673,
      "balancing_loss": 1.1056,
      "epoch": 31.89088,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.2367,
      "router_z_loss": 0.0896,
      "step": 6250
    },
    {
      "balancing_entropy": 1.0808,
      "balancing_loss": 1.1013,
      "epoch": 31.94208,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.2364,
      "router_z_loss": 0.0739,
      "step": 6260
    },
    {
      "balancing_entropy": 1.0145,
      "balancing_loss": 1.1207,
      "epoch": 31.99328,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.2371,
      "router_z_loss": 0.0663,
      "step": 6270
    },
    {
      "balancing_entropy": 0.9915,
      "balancing_loss": 1.1067,
      "epoch": 32.04096,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.2276,
      "router_z_loss": 0.0896,
      "step": 6280
    },
    {
      "balancing_entropy": 1.0701,
      "balancing_loss": 1.1034,
      "epoch": 32.09216,
      "grad_norm": 0.76953125,
      "learning_rate": 0.003,
      "loss": 2.2187,
      "router_z_loss": 0.0877,
      "step": 6290
    },
    {
      "balancing_entropy": 1.0733,
      "balancing_loss": 1.111,
      "epoch": 32.14336,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.221,
      "router_z_loss": 0.0912,
      "step": 6300
    },
    {
      "balancing_entropy": 0.999,
      "balancing_loss": 1.113,
      "epoch": 32.19456,
      "grad_norm": 1.03125,
      "learning_rate": 0.003,
      "loss": 2.223,
      "router_z_loss": 0.0746,
      "step": 6310
    },
    {
      "balancing_entropy": 0.9679,
      "balancing_loss": 1.1057,
      "epoch": 32.24576,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.2233,
      "router_z_loss": 0.0628,
      "step": 6320
    },
    {
      "balancing_entropy": 1.0733,
      "balancing_loss": 1.1118,
      "epoch": 32.29696,
      "grad_norm": 0.330078125,
      "learning_rate": 0.003,
      "loss": 2.2201,
      "router_z_loss": 0.0762,
      "step": 6330
    },
    {
      "balancing_entropy": 0.9746,
      "balancing_loss": 1.1163,
      "epoch": 32.34816,
      "grad_norm": 1.3046875,
      "learning_rate": 0.003,
      "loss": 2.2195,
      "router_z_loss": 0.0618,
      "step": 6340
    },
    {
      "balancing_entropy": 1.0477,
      "balancing_loss": 1.0974,
      "epoch": 32.39936,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.2265,
      "router_z_loss": 0.0654,
      "step": 6350
    },
    {
      "balancing_entropy": 1.0723,
      "balancing_loss": 1.1054,
      "epoch": 32.45056,
      "grad_norm": 0.3125,
      "learning_rate": 0.003,
      "loss": 2.2265,
      "router_z_loss": 0.0798,
      "step": 6360
    },
    {
      "balancing_entropy": 1.0743,
      "balancing_loss": 1.1119,
      "epoch": 32.50176,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.2248,
      "router_z_loss": 0.0774,
      "step": 6370
    },
    {
      "balancing_entropy": 1.0624,
      "balancing_loss": 1.1199,
      "epoch": 32.55296,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.2314,
      "router_z_loss": 0.0872,
      "step": 6380
    },
    {
      "balancing_entropy": 1.0293,
      "balancing_loss": 1.1101,
      "epoch": 32.60416,
      "grad_norm": 1.59375,
      "learning_rate": 0.003,
      "loss": 2.2326,
      "router_z_loss": 0.0589,
      "step": 6390
    },
    {
      "balancing_entropy": 0.8998,
      "balancing_loss": 1.1295,
      "epoch": 32.65536,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.2391,
      "router_z_loss": 0.0607,
      "step": 6400
    },
    {
      "balancing_entropy": 1.0614,
      "balancing_loss": 1.113,
      "epoch": 32.70656,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.2215,
      "router_z_loss": 0.0682,
      "step": 6410
    },
    {
      "balancing_entropy": 0.9089,
      "balancing_loss": 1.1088,
      "epoch": 32.75776,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.2335,
      "router_z_loss": 0.0629,
      "step": 6420
    },
    {
      "balancing_entropy": 0.9369,
      "balancing_loss": 1.1194,
      "epoch": 32.80896,
      "grad_norm": 1.140625,
      "learning_rate": 0.003,
      "loss": 2.2377,
      "router_z_loss": 0.0618,
      "step": 6430
    },
    {
      "balancing_entropy": 1.0353,
      "balancing_loss": 1.1299,
      "epoch": 32.86016,
      "grad_norm": 1.0078125,
      "learning_rate": 0.003,
      "loss": 2.2297,
      "router_z_loss": 0.0931,
      "step": 6440
    },
    {
      "balancing_entropy": 0.9435,
      "balancing_loss": 1.0929,
      "epoch": 32.91136,
      "grad_norm": 1.515625,
      "learning_rate": 0.003,
      "loss": 2.2402,
      "router_z_loss": 0.0878,
      "step": 6450
    },
    {
      "balancing_entropy": 0.979,
      "balancing_loss": 1.1145,
      "epoch": 32.96256,
      "grad_norm": 1.578125,
      "learning_rate": 0.003,
      "loss": 2.2238,
      "router_z_loss": 0.0559,
      "step": 6460
    },
    {
      "balancing_entropy": 0.8971,
      "balancing_loss": 1.1029,
      "epoch": 33.01024,
      "grad_norm": 1.2109375,
      "learning_rate": 0.003,
      "loss": 2.2295,
      "router_z_loss": 0.0636,
      "step": 6470
    },
    {
      "balancing_entropy": 1.0111,
      "balancing_loss": 1.1231,
      "epoch": 33.06144,
      "grad_norm": 1.328125,
      "learning_rate": 0.003,
      "loss": 2.2177,
      "router_z_loss": 0.0845,
      "step": 6480
    },
    {
      "balancing_entropy": 0.9803,
      "balancing_loss": 1.1043,
      "epoch": 33.11264,
      "grad_norm": 0.37890625,
      "learning_rate": 0.003,
      "loss": 2.21,
      "router_z_loss": 0.0805,
      "step": 6490
    },
    {
      "balancing_entropy": 1.0434,
      "balancing_loss": 1.1117,
      "epoch": 33.16384,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.2207,
      "router_z_loss": 0.0778,
      "step": 6500
    },
    {
      "balancing_entropy": 1.0025,
      "balancing_loss": 1.1096,
      "epoch": 33.21504,
      "grad_norm": 1.15625,
      "learning_rate": 0.003,
      "loss": 2.2135,
      "router_z_loss": 0.0584,
      "step": 6510
    },
    {
      "balancing_entropy": 0.9567,
      "balancing_loss": 1.1151,
      "epoch": 33.26624,
      "grad_norm": 0.828125,
      "learning_rate": 0.003,
      "loss": 2.219,
      "router_z_loss": 0.0769,
      "step": 6520
    },
    {
      "balancing_entropy": 0.8368,
      "balancing_loss": 1.1125,
      "epoch": 33.31744,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.2368,
      "router_z_loss": 0.0623,
      "step": 6530
    },
    {
      "balancing_entropy": 0.9755,
      "balancing_loss": 1.0944,
      "epoch": 33.36864,
      "grad_norm": 1.359375,
      "learning_rate": 0.003,
      "loss": 2.2247,
      "router_z_loss": 0.0671,
      "step": 6540
    },
    {
      "balancing_entropy": 0.8953,
      "balancing_loss": 1.1329,
      "epoch": 33.41984,
      "grad_norm": 1.3203125,
      "learning_rate": 0.003,
      "loss": 2.2261,
      "router_z_loss": 0.0797,
      "step": 6550
    },
    {
      "balancing_entropy": 0.9882,
      "balancing_loss": 1.1069,
      "epoch": 33.47104,
      "grad_norm": 1.5546875,
      "learning_rate": 0.003,
      "loss": 2.2246,
      "router_z_loss": 0.0804,
      "step": 6560
    },
    {
      "balancing_entropy": 0.979,
      "balancing_loss": 1.0976,
      "epoch": 33.52224,
      "grad_norm": 0.85546875,
      "learning_rate": 0.003,
      "loss": 2.2375,
      "router_z_loss": 0.0598,
      "step": 6570
    },
    {
      "balancing_entropy": 1.0001,
      "balancing_loss": 1.1059,
      "epoch": 33.57344,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.2293,
      "router_z_loss": 0.0703,
      "step": 6580
    },
    {
      "balancing_entropy": 0.9563,
      "balancing_loss": 1.1172,
      "epoch": 33.62464,
      "grad_norm": 1.609375,
      "learning_rate": 0.003,
      "loss": 2.2287,
      "router_z_loss": 0.0675,
      "step": 6590
    },
    {
      "balancing_entropy": 0.9513,
      "balancing_loss": 1.1058,
      "epoch": 33.67584,
      "grad_norm": 0.78515625,
      "learning_rate": 0.003,
      "loss": 2.2329,
      "router_z_loss": 0.0682,
      "step": 6600
    },
    {
      "balancing_entropy": 1.0574,
      "balancing_loss": 1.1139,
      "epoch": 33.72704,
      "grad_norm": 0.94921875,
      "learning_rate": 0.003,
      "loss": 2.2277,
      "router_z_loss": 0.0655,
      "step": 6610
    },
    {
      "balancing_entropy": 1.0063,
      "balancing_loss": 1.1019,
      "epoch": 33.77824,
      "grad_norm": 2.046875,
      "learning_rate": 0.003,
      "loss": 2.2352,
      "router_z_loss": 0.0524,
      "step": 6620
    },
    {
      "balancing_entropy": 0.9378,
      "balancing_loss": 1.0973,
      "epoch": 33.82944,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.2227,
      "router_z_loss": 0.0482,
      "step": 6630
    },
    {
      "balancing_entropy": 0.9331,
      "balancing_loss": 1.1047,
      "epoch": 33.88064,
      "grad_norm": 1.8203125,
      "learning_rate": 0.003,
      "loss": 2.2207,
      "router_z_loss": 0.0887,
      "step": 6640
    },
    {
      "balancing_entropy": 0.9197,
      "balancing_loss": 1.1298,
      "epoch": 33.93184,
      "grad_norm": 1.984375,
      "learning_rate": 0.003,
      "loss": 2.2337,
      "router_z_loss": 0.0772,
      "step": 6650
    },
    {
      "balancing_entropy": 0.885,
      "balancing_loss": 1.1065,
      "epoch": 33.98304,
      "grad_norm": 0.8828125,
      "learning_rate": 0.003,
      "loss": 2.2531,
      "router_z_loss": 0.0582,
      "step": 6660
    },
    {
      "balancing_entropy": 0.9797,
      "balancing_loss": 1.1007,
      "epoch": 34.03072,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.2432,
      "router_z_loss": 0.0899,
      "step": 6670
    },
    {
      "balancing_entropy": 0.9986,
      "balancing_loss": 1.0998,
      "epoch": 34.08192,
      "grad_norm": 1.8046875,
      "learning_rate": 0.003,
      "loss": 2.2268,
      "router_z_loss": 0.0611,
      "step": 6680
    },
    {
      "balancing_entropy": 0.8903,
      "balancing_loss": 1.1461,
      "epoch": 34.13312,
      "grad_norm": 1.2421875,
      "learning_rate": 0.003,
      "loss": 2.229,
      "router_z_loss": 0.0684,
      "step": 6690
    },
    {
      "balancing_entropy": 0.961,
      "balancing_loss": 1.092,
      "epoch": 34.18432,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.213,
      "router_z_loss": 0.0947,
      "step": 6700
    },
    {
      "balancing_entropy": 1.093,
      "balancing_loss": 1.1075,
      "epoch": 34.23552,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.2172,
      "router_z_loss": 0.0591,
      "step": 6710
    },
    {
      "balancing_entropy": 1.0922,
      "balancing_loss": 1.104,
      "epoch": 34.28672,
      "grad_norm": 0.98828125,
      "learning_rate": 0.003,
      "loss": 2.2168,
      "router_z_loss": 0.0476,
      "step": 6720
    },
    {
      "balancing_entropy": 0.8779,
      "balancing_loss": 1.1093,
      "epoch": 34.33792,
      "grad_norm": 1.640625,
      "learning_rate": 0.003,
      "loss": 2.2274,
      "router_z_loss": 0.0496,
      "step": 6730
    },
    {
      "balancing_entropy": 1.048,
      "balancing_loss": 1.0986,
      "epoch": 34.38912,
      "grad_norm": 1.5859375,
      "learning_rate": 0.003,
      "loss": 2.2141,
      "router_z_loss": 0.0533,
      "step": 6740
    },
    {
      "balancing_entropy": 0.9463,
      "balancing_loss": 1.1186,
      "epoch": 34.44032,
      "grad_norm": 2.171875,
      "learning_rate": 0.003,
      "loss": 2.2252,
      "router_z_loss": 0.0705,
      "step": 6750
    },
    {
      "balancing_entropy": 0.9026,
      "balancing_loss": 1.125,
      "epoch": 34.49152,
      "grad_norm": 1.1640625,
      "learning_rate": 0.003,
      "loss": 2.2161,
      "router_z_loss": 0.0502,
      "step": 6760
    },
    {
      "balancing_entropy": 0.9165,
      "balancing_loss": 1.1301,
      "epoch": 34.54272,
      "grad_norm": 1.765625,
      "learning_rate": 0.003,
      "loss": 2.2226,
      "router_z_loss": 0.0552,
      "step": 6770
    },
    {
      "balancing_entropy": 0.8562,
      "balancing_loss": 1.1212,
      "epoch": 34.59392,
      "grad_norm": 1.4375,
      "learning_rate": 0.003,
      "loss": 2.2281,
      "router_z_loss": 0.056,
      "step": 6780
    },
    {
      "balancing_entropy": 0.9815,
      "balancing_loss": 1.1211,
      "epoch": 34.64512,
      "grad_norm": 1.90625,
      "learning_rate": 0.003,
      "loss": 2.2247,
      "router_z_loss": 0.0946,
      "step": 6790
    },
    {
      "balancing_entropy": 1.0143,
      "balancing_loss": 1.1207,
      "epoch": 34.69632,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.2332,
      "router_z_loss": 0.0619,
      "step": 6800
    },
    {
      "balancing_entropy": 0.9652,
      "balancing_loss": 1.1075,
      "epoch": 34.74752,
      "grad_norm": 1.3515625,
      "learning_rate": 0.003,
      "loss": 2.2262,
      "router_z_loss": 0.056,
      "step": 6810
    },
    {
      "balancing_entropy": 0.9757,
      "balancing_loss": 1.1261,
      "epoch": 34.79872,
      "grad_norm": 1.578125,
      "learning_rate": 0.003,
      "loss": 2.2327,
      "router_z_loss": 0.0718,
      "step": 6820
    },
    {
      "balancing_entropy": 0.9265,
      "balancing_loss": 1.0957,
      "epoch": 34.84992,
      "grad_norm": 1.59375,
      "learning_rate": 0.003,
      "loss": 2.2305,
      "router_z_loss": 0.0474,
      "step": 6830
    },
    {
      "balancing_entropy": 0.9297,
      "balancing_loss": 1.1202,
      "epoch": 34.90112,
      "grad_norm": 2.1875,
      "learning_rate": 0.003,
      "loss": 2.2309,
      "router_z_loss": 0.0601,
      "step": 6840
    },
    {
      "balancing_entropy": 0.9484,
      "balancing_loss": 1.1124,
      "epoch": 34.95232,
      "grad_norm": 1.5390625,
      "learning_rate": 0.003,
      "loss": 2.2266,
      "router_z_loss": 0.0497,
      "step": 6850
    },
    {
      "balancing_entropy": 0.9668,
      "balancing_loss": 1.1085,
      "epoch": 35.0,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.2265,
      "router_z_loss": 0.0521,
      "step": 6860
    },
    {
      "balancing_entropy": 0.9205,
      "balancing_loss": 1.1058,
      "epoch": 35.0512,
      "grad_norm": 1.390625,
      "learning_rate": 0.003,
      "loss": 2.2378,
      "router_z_loss": 0.043,
      "step": 6870
    },
    {
      "balancing_entropy": 0.9657,
      "balancing_loss": 1.1042,
      "epoch": 35.1024,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.2335,
      "router_z_loss": 0.0589,
      "step": 6880
    },
    {
      "balancing_entropy": 0.9691,
      "balancing_loss": 1.0975,
      "epoch": 35.1536,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.2266,
      "router_z_loss": 0.0627,
      "step": 6890
    },
    {
      "balancing_entropy": 0.8553,
      "balancing_loss": 1.0912,
      "epoch": 35.2048,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.2136,
      "router_z_loss": 0.0804,
      "step": 6900
    },
    {
      "balancing_entropy": 0.9108,
      "balancing_loss": 1.1195,
      "epoch": 35.256,
      "grad_norm": 1.2421875,
      "learning_rate": 0.003,
      "loss": 2.2139,
      "router_z_loss": 0.1192,
      "step": 6910
    },
    {
      "balancing_entropy": 0.9106,
      "balancing_loss": 1.1409,
      "epoch": 35.3072,
      "grad_norm": 2.15625,
      "learning_rate": 0.003,
      "loss": 2.2085,
      "router_z_loss": 0.0729,
      "step": 6920
    },
    {
      "balancing_entropy": 0.8405,
      "balancing_loss": 1.1386,
      "epoch": 35.3584,
      "grad_norm": 1.4765625,
      "learning_rate": 0.003,
      "loss": 2.223,
      "router_z_loss": 0.052,
      "step": 6930
    },
    {
      "balancing_entropy": 0.9611,
      "balancing_loss": 1.0966,
      "epoch": 35.4096,
      "grad_norm": 1.34375,
      "learning_rate": 0.003,
      "loss": 2.2306,
      "router_z_loss": 0.0898,
      "step": 6940
    },
    {
      "balancing_entropy": 0.8963,
      "balancing_loss": 1.111,
      "epoch": 35.4608,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.2205,
      "router_z_loss": 0.0658,
      "step": 6950
    },
    {
      "balancing_entropy": 0.9254,
      "balancing_loss": 1.1179,
      "epoch": 35.512,
      "grad_norm": 1.2109375,
      "learning_rate": 0.003,
      "loss": 2.2184,
      "router_z_loss": 0.0859,
      "step": 6960
    },
    {
      "balancing_entropy": 1.0174,
      "balancing_loss": 1.1022,
      "epoch": 35.5632,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.229,
      "router_z_loss": 0.0561,
      "step": 6970
    },
    {
      "balancing_entropy": 0.9459,
      "balancing_loss": 1.1173,
      "epoch": 35.6144,
      "grad_norm": 2.171875,
      "learning_rate": 0.003,
      "loss": 2.2362,
      "router_z_loss": 0.0584,
      "step": 6980
    },
    {
      "balancing_entropy": 0.8793,
      "balancing_loss": 1.1258,
      "epoch": 35.6656,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.2253,
      "router_z_loss": 0.0692,
      "step": 6990
    },
    {
      "balancing_entropy": 0.9393,
      "balancing_loss": 1.0816,
      "epoch": 35.7168,
      "grad_norm": 0.9609375,
      "learning_rate": 0.003,
      "loss": 2.2226,
      "router_z_loss": 0.0632,
      "step": 7000
    },
    {
      "balancing_entropy": 0.8865,
      "balancing_loss": 1.1133,
      "epoch": 35.768,
      "grad_norm": 1.390625,
      "learning_rate": 0.003,
      "loss": 2.2204,
      "router_z_loss": 0.0485,
      "step": 7010
    },
    {
      "balancing_entropy": 0.9159,
      "balancing_loss": 1.085,
      "epoch": 35.8192,
      "grad_norm": 1.4921875,
      "learning_rate": 0.003,
      "loss": 2.2232,
      "router_z_loss": 0.0559,
      "step": 7020
    },
    {
      "balancing_entropy": 0.9327,
      "balancing_loss": 1.1314,
      "epoch": 35.8704,
      "grad_norm": 1.0546875,
      "learning_rate": 0.003,
      "loss": 2.2227,
      "router_z_loss": 0.0477,
      "step": 7030
    },
    {
      "balancing_entropy": 0.8975,
      "balancing_loss": 1.1154,
      "epoch": 35.9216,
      "grad_norm": 1.2734375,
      "learning_rate": 0.003,
      "loss": 2.2267,
      "router_z_loss": 0.0497,
      "step": 7040
    },
    {
      "balancing_entropy": 0.9287,
      "balancing_loss": 1.1214,
      "epoch": 35.9728,
      "grad_norm": 0.96484375,
      "learning_rate": 0.003,
      "loss": 2.2376,
      "router_z_loss": 0.0744,
      "step": 7050
    },
    {
      "balancing_entropy": 0.8691,
      "balancing_loss": 1.0798,
      "epoch": 36.02048,
      "grad_norm": 1.5390625,
      "learning_rate": 0.003,
      "loss": 2.2121,
      "router_z_loss": 0.1036,
      "step": 7060
    },
    {
      "balancing_entropy": 0.8882,
      "balancing_loss": 1.1176,
      "epoch": 36.07168,
      "grad_norm": 1.2578125,
      "learning_rate": 0.003,
      "loss": 2.2106,
      "router_z_loss": 0.0866,
      "step": 7070
    },
    {
      "balancing_entropy": 0.9759,
      "balancing_loss": 1.068,
      "epoch": 36.12288,
      "grad_norm": 1.3046875,
      "learning_rate": 0.003,
      "loss": 2.2118,
      "router_z_loss": 0.0698,
      "step": 7080
    },
    {
      "balancing_entropy": 0.9278,
      "balancing_loss": 1.0936,
      "epoch": 36.174080000000004,
      "grad_norm": 1.9453125,
      "learning_rate": 0.003,
      "loss": 2.2691,
      "router_z_loss": 0.0683,
      "step": 7090
    },
    {
      "balancing_entropy": 0.8833,
      "balancing_loss": 1.0957,
      "epoch": 36.22528,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.2387,
      "router_z_loss": 0.1179,
      "step": 7100
    },
    {
      "balancing_entropy": 0.8925,
      "balancing_loss": 1.1126,
      "epoch": 36.27648,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.2069,
      "router_z_loss": 0.0556,
      "step": 7110
    },
    {
      "balancing_entropy": 0.9273,
      "balancing_loss": 1.119,
      "epoch": 36.32768,
      "grad_norm": 1.4140625,
      "learning_rate": 0.003,
      "loss": 2.2147,
      "router_z_loss": 0.0637,
      "step": 7120
    },
    {
      "balancing_entropy": 1.024,
      "balancing_loss": 1.1156,
      "epoch": 36.37888,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.2049,
      "router_z_loss": 0.0629,
      "step": 7130
    },
    {
      "balancing_entropy": 0.89,
      "balancing_loss": 1.1079,
      "epoch": 36.43008,
      "grad_norm": 1.5703125,
      "learning_rate": 0.003,
      "loss": 2.2109,
      "router_z_loss": 0.0533,
      "step": 7140
    },
    {
      "balancing_entropy": 0.8717,
      "balancing_loss": 1.1093,
      "epoch": 36.48128,
      "grad_norm": 1.59375,
      "learning_rate": 0.003,
      "loss": 2.2119,
      "router_z_loss": 0.0608,
      "step": 7150
    },
    {
      "balancing_entropy": 0.888,
      "balancing_loss": 1.1156,
      "epoch": 36.53248,
      "grad_norm": 1.984375,
      "learning_rate": 0.003,
      "loss": 2.2118,
      "router_z_loss": 0.0807,
      "step": 7160
    },
    {
      "balancing_entropy": 0.894,
      "balancing_loss": 1.1093,
      "epoch": 36.58368,
      "grad_norm": 1.2734375,
      "learning_rate": 0.003,
      "loss": 2.2247,
      "router_z_loss": 0.0722,
      "step": 7170
    },
    {
      "balancing_entropy": 0.8587,
      "balancing_loss": 1.1314,
      "epoch": 36.63488,
      "grad_norm": 1.1640625,
      "learning_rate": 0.003,
      "loss": 2.2362,
      "router_z_loss": 0.0787,
      "step": 7180
    },
    {
      "balancing_entropy": 0.8204,
      "balancing_loss": 1.1298,
      "epoch": 36.68608,
      "grad_norm": 2.140625,
      "learning_rate": 0.003,
      "loss": 2.264,
      "router_z_loss": 0.1043,
      "step": 7190
    },
    {
      "balancing_entropy": 0.9817,
      "balancing_loss": 1.0956,
      "epoch": 36.73728,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.2526,
      "router_z_loss": 0.0729,
      "step": 7200
    },
    {
      "balancing_entropy": 1.0004,
      "balancing_loss": 1.0917,
      "epoch": 36.78848,
      "grad_norm": 1.4375,
      "learning_rate": 0.003,
      "loss": 2.2195,
      "router_z_loss": 0.0832,
      "step": 7210
    },
    {
      "balancing_entropy": 0.9207,
      "balancing_loss": 1.0941,
      "epoch": 36.83968,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.2293,
      "router_z_loss": 0.0981,
      "step": 7220
    },
    {
      "balancing_entropy": 0.8633,
      "balancing_loss": 1.1292,
      "epoch": 36.89088,
      "grad_norm": 1.4375,
      "learning_rate": 0.003,
      "loss": 2.2219,
      "router_z_loss": 0.0908,
      "step": 7230
    },
    {
      "balancing_entropy": 0.9705,
      "balancing_loss": 1.0851,
      "epoch": 36.94208,
      "grad_norm": 0.7890625,
      "learning_rate": 0.003,
      "loss": 2.2315,
      "router_z_loss": 0.085,
      "step": 7240
    },
    {
      "balancing_entropy": 0.861,
      "balancing_loss": 1.1118,
      "epoch": 36.99328,
      "grad_norm": 1.71875,
      "learning_rate": 0.003,
      "loss": 2.2358,
      "router_z_loss": 0.112,
      "step": 7250
    },
    {
      "balancing_entropy": 0.9045,
      "balancing_loss": 1.1148,
      "epoch": 37.04096,
      "grad_norm": 1.359375,
      "learning_rate": 0.003,
      "loss": 2.2054,
      "router_z_loss": 0.0725,
      "step": 7260
    },
    {
      "balancing_entropy": 0.9138,
      "balancing_loss": 1.1143,
      "epoch": 37.09216,
      "grad_norm": 1.921875,
      "learning_rate": 0.003,
      "loss": 2.1985,
      "router_z_loss": 0.0608,
      "step": 7270
    },
    {
      "balancing_entropy": 0.8401,
      "balancing_loss": 1.1295,
      "epoch": 37.14336,
      "grad_norm": 1.5390625,
      "learning_rate": 0.003,
      "loss": 2.2042,
      "router_z_loss": 0.0846,
      "step": 7280
    },
    {
      "balancing_entropy": 0.879,
      "balancing_loss": 1.1012,
      "epoch": 37.19456,
      "grad_norm": 1.5546875,
      "learning_rate": 0.003,
      "loss": 2.2136,
      "router_z_loss": 0.1195,
      "step": 7290
    },
    {
      "balancing_entropy": 0.8464,
      "balancing_loss": 1.1154,
      "epoch": 37.24576,
      "grad_norm": 1.609375,
      "learning_rate": 0.003,
      "loss": 2.2219,
      "router_z_loss": 0.0726,
      "step": 7300
    },
    {
      "balancing_entropy": 0.8835,
      "balancing_loss": 1.1203,
      "epoch": 37.29696,
      "grad_norm": 1.015625,
      "learning_rate": 0.003,
      "loss": 2.2117,
      "router_z_loss": 0.0654,
      "step": 7310
    },
    {
      "balancing_entropy": 0.9391,
      "balancing_loss": 1.0879,
      "epoch": 37.34816,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.2108,
      "router_z_loss": 0.1181,
      "step": 7320
    },
    {
      "balancing_entropy": 1.0249,
      "balancing_loss": 1.0928,
      "epoch": 37.39936,
      "grad_norm": 1.5703125,
      "learning_rate": 0.003,
      "loss": 2.215,
      "router_z_loss": 0.0722,
      "step": 7330
    },
    {
      "balancing_entropy": 0.8982,
      "balancing_loss": 1.1424,
      "epoch": 37.45056,
      "grad_norm": 1.15625,
      "learning_rate": 0.003,
      "loss": 2.2139,
      "router_z_loss": 0.0464,
      "step": 7340
    },
    {
      "balancing_entropy": 0.9106,
      "balancing_loss": 1.112,
      "epoch": 37.50176,
      "grad_norm": 1.0546875,
      "learning_rate": 0.003,
      "loss": 2.2037,
      "router_z_loss": 0.0556,
      "step": 7350
    },
    {
      "balancing_entropy": 0.8865,
      "balancing_loss": 1.1,
      "epoch": 37.55296,
      "grad_norm": 1.3515625,
      "learning_rate": 0.003,
      "loss": 2.2174,
      "router_z_loss": 0.0778,
      "step": 7360
    },
    {
      "balancing_entropy": 0.8381,
      "balancing_loss": 1.1125,
      "epoch": 37.60416,
      "grad_norm": 1.265625,
      "learning_rate": 0.003,
      "loss": 2.2303,
      "router_z_loss": 0.0757,
      "step": 7370
    },
    {
      "balancing_entropy": 0.8621,
      "balancing_loss": 1.1058,
      "epoch": 37.65536,
      "grad_norm": 1.296875,
      "learning_rate": 0.003,
      "loss": 2.2306,
      "router_z_loss": 0.0753,
      "step": 7380
    },
    {
      "balancing_entropy": 0.8687,
      "balancing_loss": 1.0875,
      "epoch": 37.70656,
      "grad_norm": 1.1484375,
      "learning_rate": 0.003,
      "loss": 2.2254,
      "router_z_loss": 0.1182,
      "step": 7390
    },
    {
      "balancing_entropy": 0.879,
      "balancing_loss": 1.1181,
      "epoch": 37.75776,
      "grad_norm": 1.3671875,
      "learning_rate": 0.003,
      "loss": 2.224,
      "router_z_loss": 0.0971,
      "step": 7400
    },
    {
      "balancing_entropy": 0.9419,
      "balancing_loss": 1.0909,
      "epoch": 37.80896,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.2086,
      "router_z_loss": 0.0813,
      "step": 7410
    },
    {
      "balancing_entropy": 0.9521,
      "balancing_loss": 1.0942,
      "epoch": 37.86016,
      "grad_norm": 0.92578125,
      "learning_rate": 0.003,
      "loss": 2.2059,
      "router_z_loss": 0.0768,
      "step": 7420
    },
    {
      "balancing_entropy": 0.866,
      "balancing_loss": 1.1023,
      "epoch": 37.91136,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.2119,
      "router_z_loss": 0.0617,
      "step": 7430
    },
    {
      "balancing_entropy": 0.9592,
      "balancing_loss": 1.0824,
      "epoch": 37.96256,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.2084,
      "router_z_loss": 0.0908,
      "step": 7440
    },
    {
      "balancing_entropy": 0.8595,
      "balancing_loss": 1.1183,
      "epoch": 38.01024,
      "grad_norm": 1.6875,
      "learning_rate": 0.003,
      "loss": 2.2156,
      "router_z_loss": 0.0544,
      "step": 7450
    },
    {
      "balancing_entropy": 0.8771,
      "balancing_loss": 1.1131,
      "epoch": 38.06144,
      "grad_norm": 1.65625,
      "learning_rate": 0.003,
      "loss": 2.1999,
      "router_z_loss": 0.0761,
      "step": 7460
    },
    {
      "balancing_entropy": 0.868,
      "balancing_loss": 1.122,
      "epoch": 38.11264,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.1986,
      "router_z_loss": 0.1246,
      "step": 7470
    },
    {
      "balancing_entropy": 1.015,
      "balancing_loss": 1.119,
      "epoch": 38.16384,
      "grad_norm": 1.34375,
      "learning_rate": 0.003,
      "loss": 2.2038,
      "router_z_loss": 0.0793,
      "step": 7480
    },
    {
      "balancing_entropy": 0.9002,
      "balancing_loss": 1.1135,
      "epoch": 38.21504,
      "grad_norm": 0.796875,
      "learning_rate": 0.003,
      "loss": 2.2061,
      "router_z_loss": 0.0595,
      "step": 7490
    },
    {
      "balancing_entropy": 0.8676,
      "balancing_loss": 1.084,
      "epoch": 38.26624,
      "grad_norm": 1.1953125,
      "learning_rate": 0.003,
      "loss": 2.2,
      "router_z_loss": 0.0844,
      "step": 7500
    },
    {
      "balancing_entropy": 0.8887,
      "balancing_loss": 1.1082,
      "epoch": 38.31744,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.2032,
      "router_z_loss": 0.0827,
      "step": 7510
    },
    {
      "balancing_entropy": 1.0354,
      "balancing_loss": 1.1087,
      "epoch": 38.36864,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.2075,
      "router_z_loss": 0.0607,
      "step": 7520
    },
    {
      "balancing_entropy": 1.0325,
      "balancing_loss": 1.1045,
      "epoch": 38.41984,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.2043,
      "router_z_loss": 0.0476,
      "step": 7530
    },
    {
      "balancing_entropy": 0.8426,
      "balancing_loss": 1.1298,
      "epoch": 38.47104,
      "grad_norm": 0.84765625,
      "learning_rate": 0.003,
      "loss": 2.2017,
      "router_z_loss": 0.0553,
      "step": 7540
    },
    {
      "balancing_entropy": 0.8385,
      "balancing_loss": 1.1345,
      "epoch": 38.52224,
      "grad_norm": 1.375,
      "learning_rate": 0.003,
      "loss": 2.1996,
      "router_z_loss": 0.0817,
      "step": 7550
    },
    {
      "balancing_entropy": 0.8887,
      "balancing_loss": 1.0899,
      "epoch": 38.57344,
      "grad_norm": 0.98828125,
      "learning_rate": 0.003,
      "loss": 2.1976,
      "router_z_loss": 0.1106,
      "step": 7560
    },
    {
      "balancing_entropy": 0.9402,
      "balancing_loss": 1.0876,
      "epoch": 38.62464,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.2021,
      "router_z_loss": 0.1427,
      "step": 7570
    },
    {
      "balancing_entropy": 1.0598,
      "balancing_loss": 1.0824,
      "epoch": 38.67584,
      "grad_norm": 0.3046875,
      "learning_rate": 0.003,
      "loss": 2.1994,
      "router_z_loss": 0.0674,
      "step": 7580
    },
    {
      "balancing_entropy": 1.0382,
      "balancing_loss": 1.1174,
      "epoch": 38.72704,
      "grad_norm": 1.5390625,
      "learning_rate": 0.003,
      "loss": 2.2087,
      "router_z_loss": 0.049,
      "step": 7590
    },
    {
      "balancing_entropy": 0.8765,
      "balancing_loss": 1.0829,
      "epoch": 38.77824,
      "grad_norm": 1.21875,
      "learning_rate": 0.003,
      "loss": 2.2177,
      "router_z_loss": 0.0667,
      "step": 7600
    },
    {
      "balancing_entropy": 0.8833,
      "balancing_loss": 1.1102,
      "epoch": 38.82944,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.2044,
      "router_z_loss": 0.1057,
      "step": 7610
    },
    {
      "balancing_entropy": 0.8459,
      "balancing_loss": 1.086,
      "epoch": 38.88064,
      "grad_norm": 1.078125,
      "learning_rate": 0.003,
      "loss": 2.2183,
      "router_z_loss": 0.0981,
      "step": 7620
    },
    {
      "balancing_entropy": 0.8354,
      "balancing_loss": 1.1089,
      "epoch": 38.93184,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.2142,
      "router_z_loss": 0.0781,
      "step": 7630
    },
    {
      "balancing_entropy": 0.9046,
      "balancing_loss": 1.0868,
      "epoch": 38.98304,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.2169,
      "router_z_loss": 0.0591,
      "step": 7640
    },
    {
      "balancing_entropy": 0.8785,
      "balancing_loss": 1.0888,
      "epoch": 39.03072,
      "grad_norm": 0.69921875,
      "learning_rate": 0.003,
      "loss": 2.2012,
      "router_z_loss": 0.0702,
      "step": 7650
    },
    {
      "balancing_entropy": 0.9841,
      "balancing_loss": 1.1059,
      "epoch": 39.08192,
      "grad_norm": 1.078125,
      "learning_rate": 0.003,
      "loss": 2.1922,
      "router_z_loss": 0.0743,
      "step": 7660
    },
    {
      "balancing_entropy": 0.945,
      "balancing_loss": 1.1046,
      "epoch": 39.13312,
      "grad_norm": 1.65625,
      "learning_rate": 0.003,
      "loss": 2.1927,
      "router_z_loss": 0.0599,
      "step": 7670
    },
    {
      "balancing_entropy": 0.9044,
      "balancing_loss": 1.1074,
      "epoch": 39.18432,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.2017,
      "router_z_loss": 0.0549,
      "step": 7680
    },
    {
      "balancing_entropy": 0.8673,
      "balancing_loss": 1.0891,
      "epoch": 39.23552,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.1936,
      "router_z_loss": 0.0915,
      "step": 7690
    },
    {
      "balancing_entropy": 0.8771,
      "balancing_loss": 1.1084,
      "epoch": 39.28672,
      "grad_norm": 1.234375,
      "learning_rate": 0.003,
      "loss": 2.2097,
      "router_z_loss": 0.06,
      "step": 7700
    },
    {
      "balancing_entropy": 0.8968,
      "balancing_loss": 1.0936,
      "epoch": 39.33792,
      "grad_norm": 1.234375,
      "learning_rate": 0.003,
      "loss": 2.195,
      "router_z_loss": 0.0747,
      "step": 7710
    },
    {
      "balancing_entropy": 0.8671,
      "balancing_loss": 1.0958,
      "epoch": 39.38912,
      "grad_norm": 0.7890625,
      "learning_rate": 0.003,
      "loss": 2.1984,
      "router_z_loss": 0.0792,
      "step": 7720
    },
    {
      "balancing_entropy": 0.8089,
      "balancing_loss": 1.1111,
      "epoch": 39.44032,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.1996,
      "router_z_loss": 0.0738,
      "step": 7730
    },
    {
      "balancing_entropy": 0.8843,
      "balancing_loss": 1.1023,
      "epoch": 39.49152,
      "grad_norm": 1.3125,
      "learning_rate": 0.003,
      "loss": 2.2035,
      "router_z_loss": 0.108,
      "step": 7740
    },
    {
      "balancing_entropy": 0.8551,
      "balancing_loss": 1.1061,
      "epoch": 39.54272,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.2089,
      "router_z_loss": 0.0689,
      "step": 7750
    },
    {
      "balancing_entropy": 0.8994,
      "balancing_loss": 1.0909,
      "epoch": 39.59392,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.2061,
      "router_z_loss": 0.0791,
      "step": 7760
    },
    {
      "balancing_entropy": 0.8477,
      "balancing_loss": 1.116,
      "epoch": 39.64512,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1997,
      "router_z_loss": 0.0584,
      "step": 7770
    },
    {
      "balancing_entropy": 0.9264,
      "balancing_loss": 1.0999,
      "epoch": 39.69632,
      "grad_norm": 0.875,
      "learning_rate": 0.003,
      "loss": 2.2,
      "router_z_loss": 0.0864,
      "step": 7780
    },
    {
      "balancing_entropy": 0.8604,
      "balancing_loss": 1.1114,
      "epoch": 39.74752,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.2034,
      "router_z_loss": 0.0663,
      "step": 7790
    },
    {
      "balancing_entropy": 0.8668,
      "balancing_loss": 1.1044,
      "epoch": 39.79872,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.2043,
      "router_z_loss": 0.0722,
      "step": 7800
    },
    {
      "balancing_entropy": 0.8191,
      "balancing_loss": 1.0932,
      "epoch": 39.84992,
      "grad_norm": 0.86328125,
      "learning_rate": 0.003,
      "loss": 2.1927,
      "router_z_loss": 0.0707,
      "step": 7810
    },
    {
      "balancing_entropy": 0.8541,
      "balancing_loss": 1.1116,
      "epoch": 39.90112,
      "grad_norm": 1.6875,
      "learning_rate": 0.003,
      "loss": 2.2088,
      "router_z_loss": 0.0901,
      "step": 7820
    },
    {
      "balancing_entropy": 0.8134,
      "balancing_loss": 1.1132,
      "epoch": 39.95232,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.2016,
      "router_z_loss": 0.0607,
      "step": 7830
    },
    {
      "balancing_entropy": 0.7896,
      "balancing_loss": 1.1056,
      "epoch": 40.0,
      "grad_norm": 1.5234375,
      "learning_rate": 0.003,
      "loss": 2.2025,
      "router_z_loss": 0.079,
      "step": 7840
    },
    {
      "balancing_entropy": 0.795,
      "balancing_loss": 1.1054,
      "epoch": 40.0512,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1972,
      "router_z_loss": 0.0844,
      "step": 7850
    },
    {
      "balancing_entropy": 0.8153,
      "balancing_loss": 1.1019,
      "epoch": 40.1024,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.2054,
      "router_z_loss": 0.1159,
      "step": 7860
    },
    {
      "balancing_entropy": 0.8247,
      "balancing_loss": 1.1287,
      "epoch": 40.1536,
      "grad_norm": 2.171875,
      "learning_rate": 0.003,
      "loss": 2.2298,
      "router_z_loss": 0.1053,
      "step": 7870
    },
    {
      "balancing_entropy": 0.8065,
      "balancing_loss": 1.096,
      "epoch": 40.2048,
      "grad_norm": 1.2265625,
      "learning_rate": 0.003,
      "loss": 2.2095,
      "router_z_loss": 0.1017,
      "step": 7880
    },
    {
      "balancing_entropy": 0.8809,
      "balancing_loss": 1.0761,
      "epoch": 40.256,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.1907,
      "router_z_loss": 0.0965,
      "step": 7890
    },
    {
      "balancing_entropy": 0.9166,
      "balancing_loss": 1.102,
      "epoch": 40.3072,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.1959,
      "router_z_loss": 0.0644,
      "step": 7900
    },
    {
      "balancing_entropy": 1.0026,
      "balancing_loss": 1.0965,
      "epoch": 40.3584,
      "grad_norm": 1.1484375,
      "learning_rate": 0.003,
      "loss": 2.1921,
      "router_z_loss": 0.0649,
      "step": 7910
    },
    {
      "balancing_entropy": 0.8315,
      "balancing_loss": 1.117,
      "epoch": 40.4096,
      "grad_norm": 1.4765625,
      "learning_rate": 0.003,
      "loss": 2.19,
      "router_z_loss": 0.0772,
      "step": 7920
    },
    {
      "balancing_entropy": 0.8311,
      "balancing_loss": 1.1102,
      "epoch": 40.4608,
      "grad_norm": 0.9609375,
      "learning_rate": 0.003,
      "loss": 2.1848,
      "router_z_loss": 0.0648,
      "step": 7930
    },
    {
      "balancing_entropy": 0.8778,
      "balancing_loss": 1.101,
      "epoch": 40.512,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.1998,
      "router_z_loss": 0.057,
      "step": 7940
    },
    {
      "balancing_entropy": 0.9043,
      "balancing_loss": 1.1062,
      "epoch": 40.5632,
      "grad_norm": 1.0859375,
      "learning_rate": 0.003,
      "loss": 2.2028,
      "router_z_loss": 0.0767,
      "step": 7950
    },
    {
      "balancing_entropy": 0.9328,
      "balancing_loss": 1.1223,
      "epoch": 40.6144,
      "grad_norm": 1.5,
      "learning_rate": 0.003,
      "loss": 2.2163,
      "router_z_loss": 0.0786,
      "step": 7960
    },
    {
      "balancing_entropy": 0.8326,
      "balancing_loss": 1.104,
      "epoch": 40.6656,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.2061,
      "router_z_loss": 0.0781,
      "step": 7970
    },
    {
      "balancing_entropy": 0.8228,
      "balancing_loss": 1.1024,
      "epoch": 40.7168,
      "grad_norm": 0.984375,
      "learning_rate": 0.003,
      "loss": 2.1962,
      "router_z_loss": 0.1165,
      "step": 7980
    },
    {
      "balancing_entropy": 1.016,
      "balancing_loss": 1.1016,
      "epoch": 40.768,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.2005,
      "router_z_loss": 0.0843,
      "step": 7990
    },
    {
      "balancing_entropy": 0.996,
      "balancing_loss": 1.0955,
      "epoch": 40.8192,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.1987,
      "router_z_loss": 0.0492,
      "step": 8000
    },
    {
      "balancing_entropy": 0.8629,
      "balancing_loss": 1.108,
      "epoch": 40.8704,
      "grad_norm": 1.0859375,
      "learning_rate": 0.003,
      "loss": 2.194,
      "router_z_loss": 0.0625,
      "step": 8010
    },
    {
      "balancing_entropy": 0.9055,
      "balancing_loss": 1.1081,
      "epoch": 40.9216,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.1995,
      "router_z_loss": 0.0703,
      "step": 8020
    },
    {
      "balancing_entropy": 0.9273,
      "balancing_loss": 1.0899,
      "epoch": 40.9728,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.198,
      "router_z_loss": 0.0914,
      "step": 8030
    },
    {
      "balancing_entropy": 0.8937,
      "balancing_loss": 1.1008,
      "epoch": 41.02048,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.2,
      "router_z_loss": 0.0513,
      "step": 8040
    },
    {
      "balancing_entropy": 0.982,
      "balancing_loss": 1.0895,
      "epoch": 41.07168,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1856,
      "router_z_loss": 0.0848,
      "step": 8050
    },
    {
      "balancing_entropy": 0.9215,
      "balancing_loss": 1.1012,
      "epoch": 41.12288,
      "grad_norm": 1.234375,
      "learning_rate": 0.003,
      "loss": 2.1869,
      "router_z_loss": 0.0775,
      "step": 8060
    },
    {
      "balancing_entropy": 0.8851,
      "balancing_loss": 1.1041,
      "epoch": 41.174080000000004,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.187,
      "router_z_loss": 0.0504,
      "step": 8070
    },
    {
      "balancing_entropy": 0.827,
      "balancing_loss": 1.1137,
      "epoch": 41.22528,
      "grad_norm": 1.3203125,
      "learning_rate": 0.003,
      "loss": 2.1833,
      "router_z_loss": 0.0775,
      "step": 8080
    },
    {
      "balancing_entropy": 0.8709,
      "balancing_loss": 1.1109,
      "epoch": 41.27648,
      "grad_norm": 1.2109375,
      "learning_rate": 0.003,
      "loss": 2.191,
      "router_z_loss": 0.0703,
      "step": 8090
    },
    {
      "balancing_entropy": 0.8538,
      "balancing_loss": 1.1026,
      "epoch": 41.32768,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.1923,
      "router_z_loss": 0.0784,
      "step": 8100
    },
    {
      "balancing_entropy": 1.0136,
      "balancing_loss": 1.103,
      "epoch": 41.37888,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.1881,
      "router_z_loss": 0.0858,
      "step": 8110
    },
    {
      "balancing_entropy": 0.9086,
      "balancing_loss": 1.098,
      "epoch": 41.43008,
      "grad_norm": 1.3515625,
      "learning_rate": 0.003,
      "loss": 2.1902,
      "router_z_loss": 0.0625,
      "step": 8120
    },
    {
      "balancing_entropy": 0.8924,
      "balancing_loss": 1.1181,
      "epoch": 41.48128,
      "grad_norm": 0.3515625,
      "learning_rate": 0.003,
      "loss": 2.1816,
      "router_z_loss": 0.0734,
      "step": 8130
    },
    {
      "balancing_entropy": 1.046,
      "balancing_loss": 1.1152,
      "epoch": 41.53248,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.1843,
      "router_z_loss": 0.0549,
      "step": 8140
    },
    {
      "balancing_entropy": 0.9242,
      "balancing_loss": 1.1145,
      "epoch": 41.58368,
      "grad_norm": 1.2265625,
      "learning_rate": 0.003,
      "loss": 2.1822,
      "router_z_loss": 0.0469,
      "step": 8150
    },
    {
      "balancing_entropy": 0.8937,
      "balancing_loss": 1.0983,
      "epoch": 41.63488,
      "grad_norm": 1.0859375,
      "learning_rate": 0.003,
      "loss": 2.1823,
      "router_z_loss": 0.0797,
      "step": 8160
    },
    {
      "balancing_entropy": 0.7973,
      "balancing_loss": 1.1187,
      "epoch": 41.68608,
      "grad_norm": 1.265625,
      "learning_rate": 0.003,
      "loss": 2.1931,
      "router_z_loss": 0.0767,
      "step": 8170
    },
    {
      "balancing_entropy": 0.8701,
      "balancing_loss": 1.0882,
      "epoch": 41.73728,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.2037,
      "router_z_loss": 0.106,
      "step": 8180
    },
    {
      "balancing_entropy": 0.8496,
      "balancing_loss": 1.1115,
      "epoch": 41.78848,
      "grad_norm": 1.015625,
      "learning_rate": 0.003,
      "loss": 2.2121,
      "router_z_loss": 0.087,
      "step": 8190
    },
    {
      "balancing_entropy": 0.8155,
      "balancing_loss": 1.1076,
      "epoch": 41.83968,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.2094,
      "router_z_loss": 0.1205,
      "step": 8200
    },
    {
      "balancing_entropy": 0.8161,
      "balancing_loss": 1.1025,
      "epoch": 41.89088,
      "grad_norm": 1.140625,
      "learning_rate": 0.003,
      "loss": 2.2151,
      "router_z_loss": 0.1214,
      "step": 8210
    },
    {
      "balancing_entropy": 0.852,
      "balancing_loss": 1.0726,
      "epoch": 41.94208,
      "grad_norm": 1.1640625,
      "learning_rate": 0.003,
      "loss": 2.2047,
      "router_z_loss": 0.113,
      "step": 8220
    },
    {
      "balancing_entropy": 0.8409,
      "balancing_loss": 1.1291,
      "epoch": 41.99328,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.201,
      "router_z_loss": 0.0888,
      "step": 8230
    },
    {
      "balancing_entropy": 0.8407,
      "balancing_loss": 1.0694,
      "epoch": 42.04096,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.1924,
      "router_z_loss": 0.1051,
      "step": 8240
    },
    {
      "balancing_entropy": 0.8726,
      "balancing_loss": 1.0974,
      "epoch": 42.09216,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.1821,
      "router_z_loss": 0.0913,
      "step": 8250
    },
    {
      "balancing_entropy": 0.9444,
      "balancing_loss": 1.088,
      "epoch": 42.14336,
      "grad_norm": 1.65625,
      "learning_rate": 0.003,
      "loss": 2.18,
      "router_z_loss": 0.0797,
      "step": 8260
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.1206,
      "epoch": 42.19456,
      "grad_norm": 0.84765625,
      "learning_rate": 0.003,
      "loss": 2.1748,
      "router_z_loss": 0.0551,
      "step": 8270
    },
    {
      "balancing_entropy": 0.9994,
      "balancing_loss": 1.0895,
      "epoch": 42.24576,
      "grad_norm": 1.234375,
      "learning_rate": 0.003,
      "loss": 2.1784,
      "router_z_loss": 0.0852,
      "step": 8280
    },
    {
      "balancing_entropy": 0.8552,
      "balancing_loss": 1.1233,
      "epoch": 42.29696,
      "grad_norm": 1.046875,
      "learning_rate": 0.003,
      "loss": 2.1849,
      "router_z_loss": 0.0589,
      "step": 8290
    },
    {
      "balancing_entropy": 0.8472,
      "balancing_loss": 1.0771,
      "epoch": 42.34816,
      "grad_norm": 0.8984375,
      "learning_rate": 0.003,
      "loss": 2.1742,
      "router_z_loss": 0.1084,
      "step": 8300
    },
    {
      "balancing_entropy": 0.8002,
      "balancing_loss": 1.1362,
      "epoch": 42.39936,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.1881,
      "router_z_loss": 0.0952,
      "step": 8310
    },
    {
      "balancing_entropy": 1.007,
      "balancing_loss": 1.1105,
      "epoch": 42.45056,
      "grad_norm": 1.359375,
      "learning_rate": 0.003,
      "loss": 2.1828,
      "router_z_loss": 0.0776,
      "step": 8320
    },
    {
      "balancing_entropy": 0.9781,
      "balancing_loss": 1.0892,
      "epoch": 42.50176,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.1811,
      "router_z_loss": 0.063,
      "step": 8330
    },
    {
      "balancing_entropy": 0.9358,
      "balancing_loss": 1.105,
      "epoch": 42.55296,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1887,
      "router_z_loss": 0.0701,
      "step": 8340
    },
    {
      "balancing_entropy": 0.8805,
      "balancing_loss": 1.1142,
      "epoch": 42.60416,
      "grad_norm": 0.86328125,
      "learning_rate": 0.003,
      "loss": 2.1913,
      "router_z_loss": 0.0768,
      "step": 8350
    },
    {
      "balancing_entropy": 0.9045,
      "balancing_loss": 1.1062,
      "epoch": 42.65536,
      "grad_norm": 1.3359375,
      "learning_rate": 0.003,
      "loss": 2.1834,
      "router_z_loss": 0.0765,
      "step": 8360
    },
    {
      "balancing_entropy": 0.827,
      "balancing_loss": 1.1115,
      "epoch": 42.70656,
      "grad_norm": 1.453125,
      "learning_rate": 0.003,
      "loss": 2.1881,
      "router_z_loss": 0.0795,
      "step": 8370
    },
    {
      "balancing_entropy": 0.7947,
      "balancing_loss": 1.1167,
      "epoch": 42.75776,
      "grad_norm": 1.71875,
      "learning_rate": 0.003,
      "loss": 2.1907,
      "router_z_loss": 0.084,
      "step": 8380
    },
    {
      "balancing_entropy": 0.816,
      "balancing_loss": 1.0892,
      "epoch": 42.80896,
      "grad_norm": 1.015625,
      "learning_rate": 0.003,
      "loss": 2.2021,
      "router_z_loss": 0.0855,
      "step": 8390
    },
    {
      "balancing_entropy": 0.8457,
      "balancing_loss": 1.091,
      "epoch": 42.86016,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.2091,
      "router_z_loss": 0.085,
      "step": 8400
    },
    {
      "balancing_entropy": 0.816,
      "balancing_loss": 1.0921,
      "epoch": 42.91136,
      "grad_norm": 0.99609375,
      "learning_rate": 0.003,
      "loss": 2.1974,
      "router_z_loss": 0.0748,
      "step": 8410
    },
    {
      "balancing_entropy": 0.7915,
      "balancing_loss": 1.1196,
      "epoch": 42.96256,
      "grad_norm": 1.375,
      "learning_rate": 0.003,
      "loss": 2.1976,
      "router_z_loss": 0.0762,
      "step": 8420
    },
    {
      "balancing_entropy": 0.8698,
      "balancing_loss": 1.0916,
      "epoch": 43.01024,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1848,
      "router_z_loss": 0.0958,
      "step": 8430
    },
    {
      "balancing_entropy": 0.8955,
      "balancing_loss": 1.0978,
      "epoch": 43.06144,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.1809,
      "router_z_loss": 0.0526,
      "step": 8440
    },
    {
      "balancing_entropy": 0.9372,
      "balancing_loss": 1.1181,
      "epoch": 43.11264,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1713,
      "router_z_loss": 0.0692,
      "step": 8450
    },
    {
      "balancing_entropy": 1.0067,
      "balancing_loss": 1.108,
      "epoch": 43.16384,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1729,
      "router_z_loss": 0.0574,
      "step": 8460
    },
    {
      "balancing_entropy": 1.0277,
      "balancing_loss": 1.1053,
      "epoch": 43.21504,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.175,
      "router_z_loss": 0.0541,
      "step": 8470
    },
    {
      "balancing_entropy": 1.031,
      "balancing_loss": 1.0916,
      "epoch": 43.26624,
      "grad_norm": 1.1328125,
      "learning_rate": 0.003,
      "loss": 2.1787,
      "router_z_loss": 0.056,
      "step": 8480
    },
    {
      "balancing_entropy": 0.8862,
      "balancing_loss": 1.0948,
      "epoch": 43.31744,
      "grad_norm": 1.1953125,
      "learning_rate": 0.003,
      "loss": 2.1811,
      "router_z_loss": 0.0727,
      "step": 8490
    },
    {
      "balancing_entropy": 0.7861,
      "balancing_loss": 1.1233,
      "epoch": 43.36864,
      "grad_norm": 1.515625,
      "learning_rate": 0.003,
      "loss": 2.1854,
      "router_z_loss": 0.0759,
      "step": 8500
    },
    {
      "balancing_entropy": 0.8116,
      "balancing_loss": 1.0945,
      "epoch": 43.41984,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.1714,
      "router_z_loss": 0.0781,
      "step": 8510
    },
    {
      "balancing_entropy": 0.8223,
      "balancing_loss": 1.1052,
      "epoch": 43.47104,
      "grad_norm": 1.25,
      "learning_rate": 0.003,
      "loss": 2.1829,
      "router_z_loss": 0.0868,
      "step": 8520
    },
    {
      "balancing_entropy": 0.845,
      "balancing_loss": 1.1103,
      "epoch": 43.52224,
      "grad_norm": 1.046875,
      "learning_rate": 0.003,
      "loss": 2.1887,
      "router_z_loss": 0.0753,
      "step": 8530
    },
    {
      "balancing_entropy": 0.8091,
      "balancing_loss": 1.14,
      "epoch": 43.57344,
      "grad_norm": 1.140625,
      "learning_rate": 0.003,
      "loss": 2.1872,
      "router_z_loss": 0.0982,
      "step": 8540
    },
    {
      "balancing_entropy": 0.8834,
      "balancing_loss": 1.0912,
      "epoch": 43.62464,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.1897,
      "router_z_loss": 0.1257,
      "step": 8550
    },
    {
      "balancing_entropy": 0.935,
      "balancing_loss": 1.0994,
      "epoch": 43.67584,
      "grad_norm": 0.953125,
      "learning_rate": 0.003,
      "loss": 2.2056,
      "router_z_loss": 0.0882,
      "step": 8560
    },
    {
      "balancing_entropy": 0.8634,
      "balancing_loss": 1.1157,
      "epoch": 43.72704,
      "grad_norm": 0.93359375,
      "learning_rate": 0.003,
      "loss": 2.2005,
      "router_z_loss": 0.0699,
      "step": 8570
    },
    {
      "balancing_entropy": 0.9566,
      "balancing_loss": 1.1018,
      "epoch": 43.77824,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.1868,
      "router_z_loss": 0.1045,
      "step": 8580
    },
    {
      "balancing_entropy": 0.832,
      "balancing_loss": 1.1019,
      "epoch": 43.82944,
      "grad_norm": 0.85546875,
      "learning_rate": 0.003,
      "loss": 2.1938,
      "router_z_loss": 0.0632,
      "step": 8590
    },
    {
      "balancing_entropy": 0.8296,
      "balancing_loss": 1.1237,
      "epoch": 43.88064,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.1836,
      "router_z_loss": 0.0812,
      "step": 8600
    },
    {
      "balancing_entropy": 0.9812,
      "balancing_loss": 1.0852,
      "epoch": 43.93184,
      "grad_norm": 1.1796875,
      "learning_rate": 0.003,
      "loss": 2.1867,
      "router_z_loss": 0.0629,
      "step": 8610
    },
    {
      "balancing_entropy": 1.0168,
      "balancing_loss": 1.0925,
      "epoch": 43.98304,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.1805,
      "router_z_loss": 0.0499,
      "step": 8620
    },
    {
      "balancing_entropy": 0.9493,
      "balancing_loss": 1.1117,
      "epoch": 44.03072,
      "grad_norm": 1.1328125,
      "learning_rate": 0.003,
      "loss": 2.1783,
      "router_z_loss": 0.0554,
      "step": 8630
    },
    {
      "balancing_entropy": 0.8727,
      "balancing_loss": 1.0877,
      "epoch": 44.08192,
      "grad_norm": 0.921875,
      "learning_rate": 0.003,
      "loss": 2.1824,
      "router_z_loss": 0.0805,
      "step": 8640
    },
    {
      "balancing_entropy": 0.8234,
      "balancing_loss": 1.0997,
      "epoch": 44.13312,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.1742,
      "router_z_loss": 0.1446,
      "step": 8650
    },
    {
      "balancing_entropy": 0.8688,
      "balancing_loss": 1.1374,
      "epoch": 44.18432,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.1785,
      "router_z_loss": 0.0731,
      "step": 8660
    },
    {
      "balancing_entropy": 0.844,
      "balancing_loss": 1.103,
      "epoch": 44.23552,
      "grad_norm": 1.6484375,
      "learning_rate": 0.003,
      "loss": 2.19,
      "router_z_loss": 0.0988,
      "step": 8670
    },
    {
      "balancing_entropy": 0.9051,
      "balancing_loss": 1.0917,
      "epoch": 44.28672,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1814,
      "router_z_loss": 0.0699,
      "step": 8680
    },
    {
      "balancing_entropy": 0.8989,
      "balancing_loss": 1.0898,
      "epoch": 44.33792,
      "grad_norm": 1.1484375,
      "learning_rate": 0.003,
      "loss": 2.1783,
      "router_z_loss": 0.051,
      "step": 8690
    },
    {
      "balancing_entropy": 1.0121,
      "balancing_loss": 1.112,
      "epoch": 44.38912,
      "grad_norm": 0.267578125,
      "learning_rate": 0.003,
      "loss": 2.178,
      "router_z_loss": 0.0621,
      "step": 8700
    },
    {
      "balancing_entropy": 0.8428,
      "balancing_loss": 1.1143,
      "epoch": 44.44032,
      "grad_norm": 0.76171875,
      "learning_rate": 0.003,
      "loss": 2.1792,
      "router_z_loss": 0.0603,
      "step": 8710
    },
    {
      "balancing_entropy": 0.86,
      "balancing_loss": 1.0926,
      "epoch": 44.49152,
      "grad_norm": 0.84765625,
      "learning_rate": 0.003,
      "loss": 2.1708,
      "router_z_loss": 0.1162,
      "step": 8720
    },
    {
      "balancing_entropy": 0.8468,
      "balancing_loss": 1.102,
      "epoch": 44.54272,
      "grad_norm": 1.09375,
      "learning_rate": 0.003,
      "loss": 2.1718,
      "router_z_loss": 0.141,
      "step": 8730
    },
    {
      "balancing_entropy": 0.9079,
      "balancing_loss": 1.1195,
      "epoch": 44.59392,
      "grad_norm": 0.34765625,
      "learning_rate": 0.003,
      "loss": 2.1806,
      "router_z_loss": 0.1342,
      "step": 8740
    },
    {
      "balancing_entropy": 1.0112,
      "balancing_loss": 1.0883,
      "epoch": 44.64512,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1852,
      "router_z_loss": 0.082,
      "step": 8750
    },
    {
      "balancing_entropy": 0.8619,
      "balancing_loss": 1.1234,
      "epoch": 44.69632,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.1801,
      "router_z_loss": 0.0576,
      "step": 8760
    },
    {
      "balancing_entropy": 0.9981,
      "balancing_loss": 1.0775,
      "epoch": 44.74752,
      "grad_norm": 1.296875,
      "learning_rate": 0.003,
      "loss": 2.1863,
      "router_z_loss": 0.0748,
      "step": 8770
    },
    {
      "balancing_entropy": 0.9679,
      "balancing_loss": 1.1181,
      "epoch": 44.79872,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.1822,
      "router_z_loss": 0.0519,
      "step": 8780
    },
    {
      "balancing_entropy": 1.0162,
      "balancing_loss": 1.0899,
      "epoch": 44.84992,
      "grad_norm": 1.140625,
      "learning_rate": 0.003,
      "loss": 2.1728,
      "router_z_loss": 0.0528,
      "step": 8790
    },
    {
      "balancing_entropy": 0.9532,
      "balancing_loss": 1.0967,
      "epoch": 44.90112,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1863,
      "router_z_loss": 0.0711,
      "step": 8800
    },
    {
      "balancing_entropy": 0.8728,
      "balancing_loss": 1.092,
      "epoch": 44.95232,
      "grad_norm": 1.046875,
      "learning_rate": 0.003,
      "loss": 2.1842,
      "router_z_loss": 0.083,
      "step": 8810
    },
    {
      "balancing_entropy": 0.7997,
      "balancing_loss": 1.0964,
      "epoch": 45.0,
      "grad_norm": 1.015625,
      "learning_rate": 0.003,
      "loss": 2.1885,
      "router_z_loss": 0.063,
      "step": 8820
    },
    {
      "balancing_entropy": 0.8474,
      "balancing_loss": 1.1145,
      "epoch": 45.0512,
      "grad_norm": 1.03125,
      "learning_rate": 0.003,
      "loss": 2.1774,
      "router_z_loss": 0.0961,
      "step": 8830
    },
    {
      "balancing_entropy": 0.8301,
      "balancing_loss": 1.1006,
      "epoch": 45.1024,
      "grad_norm": 1.15625,
      "learning_rate": 0.003,
      "loss": 2.1885,
      "router_z_loss": 0.0714,
      "step": 8840
    },
    {
      "balancing_entropy": 0.8954,
      "balancing_loss": 1.0885,
      "epoch": 45.1536,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1751,
      "router_z_loss": 0.0883,
      "step": 8850
    },
    {
      "balancing_entropy": 1.0927,
      "balancing_loss": 1.0845,
      "epoch": 45.2048,
      "grad_norm": 0.25390625,
      "learning_rate": 0.003,
      "loss": 2.1689,
      "router_z_loss": 0.0633,
      "step": 8860
    },
    {
      "balancing_entropy": 0.9439,
      "balancing_loss": 1.1088,
      "epoch": 45.256,
      "grad_norm": 0.90234375,
      "learning_rate": 0.003,
      "loss": 2.1634,
      "router_z_loss": 0.0473,
      "step": 8870
    },
    {
      "balancing_entropy": 0.8557,
      "balancing_loss": 1.1535,
      "epoch": 45.3072,
      "grad_norm": 1.671875,
      "learning_rate": 0.003,
      "loss": 2.1762,
      "router_z_loss": 0.1907,
      "step": 8880
    },
    {
      "balancing_entropy": 1.0295,
      "balancing_loss": 1.0992,
      "epoch": 45.3584,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.1771,
      "router_z_loss": 0.0787,
      "step": 8890
    },
    {
      "balancing_entropy": 1.0492,
      "balancing_loss": 1.1052,
      "epoch": 45.4096,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.1665,
      "router_z_loss": 0.0649,
      "step": 8900
    },
    {
      "balancing_entropy": 0.9267,
      "balancing_loss": 1.0885,
      "epoch": 45.4608,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1781,
      "router_z_loss": 0.0638,
      "step": 8910
    },
    {
      "balancing_entropy": 0.8105,
      "balancing_loss": 1.1075,
      "epoch": 45.512,
      "grad_norm": 1.40625,
      "learning_rate": 0.003,
      "loss": 2.1693,
      "router_z_loss": 0.0716,
      "step": 8920
    },
    {
      "balancing_entropy": 0.8436,
      "balancing_loss": 1.1071,
      "epoch": 45.5632,
      "grad_norm": 1.09375,
      "learning_rate": 0.003,
      "loss": 2.1726,
      "router_z_loss": 0.1261,
      "step": 8930
    },
    {
      "balancing_entropy": 0.8667,
      "balancing_loss": 1.1219,
      "epoch": 45.6144,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1756,
      "router_z_loss": 0.0987,
      "step": 8940
    },
    {
      "balancing_entropy": 0.8477,
      "balancing_loss": 1.1256,
      "epoch": 45.6656,
      "grad_norm": 1.6484375,
      "learning_rate": 0.003,
      "loss": 2.1912,
      "router_z_loss": 0.0829,
      "step": 8950
    },
    {
      "balancing_entropy": 0.8423,
      "balancing_loss": 1.1021,
      "epoch": 45.7168,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.1695,
      "router_z_loss": 0.1267,
      "step": 8960
    },
    {
      "balancing_entropy": 0.9484,
      "balancing_loss": 1.0913,
      "epoch": 45.768,
      "grad_norm": 0.62109375,
      "learning_rate": 0.003,
      "loss": 2.1816,
      "router_z_loss": 0.1536,
      "step": 8970
    },
    {
      "balancing_entropy": 0.952,
      "balancing_loss": 1.1098,
      "epoch": 45.8192,
      "grad_norm": 1.078125,
      "learning_rate": 0.003,
      "loss": 2.1795,
      "router_z_loss": 0.0805,
      "step": 8980
    },
    {
      "balancing_entropy": 0.9081,
      "balancing_loss": 1.1036,
      "epoch": 45.8704,
      "grad_norm": 1.1875,
      "learning_rate": 0.003,
      "loss": 2.1812,
      "router_z_loss": 0.0771,
      "step": 8990
    },
    {
      "balancing_entropy": 0.9535,
      "balancing_loss": 1.1114,
      "epoch": 45.9216,
      "grad_norm": 0.890625,
      "learning_rate": 0.003,
      "loss": 2.1768,
      "router_z_loss": 0.055,
      "step": 9000
    },
    {
      "balancing_entropy": 0.968,
      "balancing_loss": 1.0938,
      "epoch": 45.9728,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1818,
      "router_z_loss": 0.046,
      "step": 9010
    },
    {
      "balancing_entropy": 0.9299,
      "balancing_loss": 1.1004,
      "epoch": 46.02048,
      "grad_norm": 1.2578125,
      "learning_rate": 0.003,
      "loss": 2.1704,
      "router_z_loss": 0.056,
      "step": 9020
    },
    {
      "balancing_entropy": 0.8111,
      "balancing_loss": 1.1076,
      "epoch": 46.07168,
      "grad_norm": 0.8828125,
      "learning_rate": 0.003,
      "loss": 2.1681,
      "router_z_loss": 0.0524,
      "step": 9030
    },
    {
      "balancing_entropy": 0.8366,
      "balancing_loss": 1.1103,
      "epoch": 46.12288,
      "grad_norm": 1.125,
      "learning_rate": 0.003,
      "loss": 2.1583,
      "router_z_loss": 0.0998,
      "step": 9040
    },
    {
      "balancing_entropy": 0.8308,
      "balancing_loss": 1.0995,
      "epoch": 46.174080000000004,
      "grad_norm": 0.8515625,
      "learning_rate": 0.003,
      "loss": 2.1616,
      "router_z_loss": 0.1117,
      "step": 9050
    },
    {
      "balancing_entropy": 0.8177,
      "balancing_loss": 1.1176,
      "epoch": 46.22528,
      "grad_norm": 0.77734375,
      "learning_rate": 0.003,
      "loss": 2.1753,
      "router_z_loss": 0.1001,
      "step": 9060
    },
    {
      "balancing_entropy": 0.8623,
      "balancing_loss": 1.1072,
      "epoch": 46.27648,
      "grad_norm": 0.6875,
      "learning_rate": 0.003,
      "loss": 2.1703,
      "router_z_loss": 0.0828,
      "step": 9070
    },
    {
      "balancing_entropy": 0.8238,
      "balancing_loss": 1.1071,
      "epoch": 46.32768,
      "grad_norm": 1.1328125,
      "learning_rate": 0.003,
      "loss": 2.1606,
      "router_z_loss": 0.084,
      "step": 9080
    },
    {
      "balancing_entropy": 0.8169,
      "balancing_loss": 1.1011,
      "epoch": 46.37888,
      "grad_norm": 0.90625,
      "learning_rate": 0.003,
      "loss": 2.1697,
      "router_z_loss": 0.0927,
      "step": 9090
    },
    {
      "balancing_entropy": 0.8832,
      "balancing_loss": 1.1206,
      "epoch": 46.43008,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.1727,
      "router_z_loss": 0.1094,
      "step": 9100
    },
    {
      "balancing_entropy": 0.8707,
      "balancing_loss": 1.1046,
      "epoch": 46.48128,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.1802,
      "router_z_loss": 0.088,
      "step": 9110
    },
    {
      "balancing_entropy": 0.811,
      "balancing_loss": 1.1292,
      "epoch": 46.53248,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.1674,
      "router_z_loss": 0.0964,
      "step": 9120
    },
    {
      "balancing_entropy": 0.9634,
      "balancing_loss": 1.1036,
      "epoch": 46.58368,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1698,
      "router_z_loss": 0.1049,
      "step": 9130
    },
    {
      "balancing_entropy": 1.0804,
      "balancing_loss": 1.1009,
      "epoch": 46.63488,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.1713,
      "router_z_loss": 0.051,
      "step": 9140
    },
    {
      "balancing_entropy": 1.0292,
      "balancing_loss": 1.1075,
      "epoch": 46.68608,
      "grad_norm": 0.236328125,
      "learning_rate": 0.003,
      "loss": 2.166,
      "router_z_loss": 0.0516,
      "step": 9150
    },
    {
      "balancing_entropy": 0.9162,
      "balancing_loss": 1.1065,
      "epoch": 46.73728,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.1731,
      "router_z_loss": 0.0538,
      "step": 9160
    },
    {
      "balancing_entropy": 0.8775,
      "balancing_loss": 1.1098,
      "epoch": 46.78848,
      "grad_norm": 0.92578125,
      "learning_rate": 0.003,
      "loss": 2.1782,
      "router_z_loss": 0.081,
      "step": 9170
    },
    {
      "balancing_entropy": 0.788,
      "balancing_loss": 1.1122,
      "epoch": 46.83968,
      "grad_norm": 0.89453125,
      "learning_rate": 0.003,
      "loss": 2.1808,
      "router_z_loss": 0.0862,
      "step": 9180
    },
    {
      "balancing_entropy": 0.8475,
      "balancing_loss": 1.0997,
      "epoch": 46.89088,
      "grad_norm": 0.7890625,
      "learning_rate": 0.003,
      "loss": 2.1894,
      "router_z_loss": 0.0992,
      "step": 9190
    },
    {
      "balancing_entropy": 0.8997,
      "balancing_loss": 1.0728,
      "epoch": 46.94208,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1856,
      "router_z_loss": 0.1077,
      "step": 9200
    },
    {
      "balancing_entropy": 1.0577,
      "balancing_loss": 1.0943,
      "epoch": 46.99328,
      "grad_norm": 0.291015625,
      "learning_rate": 0.003,
      "loss": 2.1804,
      "router_z_loss": 0.0736,
      "step": 9210
    },
    {
      "balancing_entropy": 1.0275,
      "balancing_loss": 1.0883,
      "epoch": 47.04096,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.1677,
      "router_z_loss": 0.0571,
      "step": 9220
    },
    {
      "balancing_entropy": 0.8863,
      "balancing_loss": 1.1111,
      "epoch": 47.09216,
      "grad_norm": 1.25,
      "learning_rate": 0.003,
      "loss": 2.1598,
      "router_z_loss": 0.0761,
      "step": 9230
    },
    {
      "balancing_entropy": 0.8933,
      "balancing_loss": 1.0971,
      "epoch": 47.14336,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.1614,
      "router_z_loss": 0.0596,
      "step": 9240
    },
    {
      "balancing_entropy": 0.8644,
      "balancing_loss": 1.0926,
      "epoch": 47.19456,
      "grad_norm": 0.8671875,
      "learning_rate": 0.003,
      "loss": 2.1623,
      "router_z_loss": 0.0791,
      "step": 9250
    },
    {
      "balancing_entropy": 0.8337,
      "balancing_loss": 1.0914,
      "epoch": 47.24576,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1654,
      "router_z_loss": 0.0636,
      "step": 9260
    },
    {
      "balancing_entropy": 0.9841,
      "balancing_loss": 1.0983,
      "epoch": 47.29696,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1676,
      "router_z_loss": 0.0778,
      "step": 9270
    },
    {
      "balancing_entropy": 0.9545,
      "balancing_loss": 1.1324,
      "epoch": 47.34816,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.1665,
      "router_z_loss": 0.0689,
      "step": 9280
    },
    {
      "balancing_entropy": 0.9481,
      "balancing_loss": 1.0766,
      "epoch": 47.39936,
      "grad_norm": 0.7109375,
      "learning_rate": 0.003,
      "loss": 2.161,
      "router_z_loss": 0.0864,
      "step": 9290
    },
    {
      "balancing_entropy": 0.8821,
      "balancing_loss": 1.1094,
      "epoch": 47.45056,
      "grad_norm": 0.68359375,
      "learning_rate": 0.003,
      "loss": 2.1663,
      "router_z_loss": 0.0595,
      "step": 9300
    },
    {
      "balancing_entropy": 1.0664,
      "balancing_loss": 1.0959,
      "epoch": 47.50176,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.1685,
      "router_z_loss": 0.0826,
      "step": 9310
    },
    {
      "balancing_entropy": 1.0918,
      "balancing_loss": 1.1026,
      "epoch": 47.55296,
      "grad_norm": 0.27734375,
      "learning_rate": 0.003,
      "loss": 2.1649,
      "router_z_loss": 0.0571,
      "step": 9320
    },
    {
      "balancing_entropy": 0.9146,
      "balancing_loss": 1.09,
      "epoch": 47.60416,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.1657,
      "router_z_loss": 0.0547,
      "step": 9330
    },
    {
      "balancing_entropy": 0.8005,
      "balancing_loss": 1.1427,
      "epoch": 47.65536,
      "grad_norm": 0.8046875,
      "learning_rate": 0.003,
      "loss": 2.1661,
      "router_z_loss": 0.0818,
      "step": 9340
    },
    {
      "balancing_entropy": 0.9592,
      "balancing_loss": 1.1073,
      "epoch": 47.70656,
      "grad_norm": 0.82421875,
      "learning_rate": 0.003,
      "loss": 2.1653,
      "router_z_loss": 0.0956,
      "step": 9350
    },
    {
      "balancing_entropy": 0.9615,
      "balancing_loss": 1.0965,
      "epoch": 47.75776,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.1751,
      "router_z_loss": 0.0727,
      "step": 9360
    },
    {
      "balancing_entropy": 0.9002,
      "balancing_loss": 1.0997,
      "epoch": 47.80896,
      "grad_norm": 1.015625,
      "learning_rate": 0.003,
      "loss": 2.1785,
      "router_z_loss": 0.0619,
      "step": 9370
    },
    {
      "balancing_entropy": 0.8309,
      "balancing_loss": 1.0939,
      "epoch": 47.86016,
      "grad_norm": 1.03125,
      "learning_rate": 0.003,
      "loss": 2.1724,
      "router_z_loss": 0.0718,
      "step": 9380
    },
    {
      "balancing_entropy": 0.8879,
      "balancing_loss": 1.1088,
      "epoch": 47.91136,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.1728,
      "router_z_loss": 0.082,
      "step": 9390
    },
    {
      "balancing_entropy": 1.0801,
      "balancing_loss": 1.0971,
      "epoch": 47.96256,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1683,
      "router_z_loss": 0.0536,
      "step": 9400
    },
    {
      "balancing_entropy": 0.8075,
      "balancing_loss": 1.1292,
      "epoch": 48.01024,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1682,
      "router_z_loss": 0.0723,
      "step": 9410
    },
    {
      "balancing_entropy": 0.874,
      "balancing_loss": 1.1022,
      "epoch": 48.06144,
      "grad_norm": 2.03125,
      "learning_rate": 0.003,
      "loss": 2.157,
      "router_z_loss": 0.0852,
      "step": 9420
    },
    {
      "balancing_entropy": 0.8853,
      "balancing_loss": 1.1002,
      "epoch": 48.11264,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.1525,
      "router_z_loss": 0.0712,
      "step": 9430
    },
    {
      "balancing_entropy": 0.9081,
      "balancing_loss": 1.0822,
      "epoch": 48.16384,
      "grad_norm": 0.8671875,
      "learning_rate": 0.003,
      "loss": 2.1572,
      "router_z_loss": 0.0705,
      "step": 9440
    },
    {
      "balancing_entropy": 0.8179,
      "balancing_loss": 1.0965,
      "epoch": 48.21504,
      "grad_norm": 1.09375,
      "learning_rate": 0.003,
      "loss": 2.1613,
      "router_z_loss": 0.0835,
      "step": 9450
    },
    {
      "balancing_entropy": 0.8275,
      "balancing_loss": 1.1106,
      "epoch": 48.26624,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.1614,
      "router_z_loss": 0.1005,
      "step": 9460
    },
    {
      "balancing_entropy": 0.9282,
      "balancing_loss": 1.1088,
      "epoch": 48.31744,
      "grad_norm": 1.40625,
      "learning_rate": 0.003,
      "loss": 2.1628,
      "router_z_loss": 0.1014,
      "step": 9470
    },
    {
      "balancing_entropy": 0.8332,
      "balancing_loss": 1.0902,
      "epoch": 48.36864,
      "grad_norm": 0.8359375,
      "learning_rate": 0.003,
      "loss": 2.1761,
      "router_z_loss": 0.0814,
      "step": 9480
    },
    {
      "balancing_entropy": 0.8105,
      "balancing_loss": 1.1234,
      "epoch": 48.41984,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.1711,
      "router_z_loss": 0.1254,
      "step": 9490
    },
    {
      "balancing_entropy": 0.8744,
      "balancing_loss": 1.1087,
      "epoch": 48.47104,
      "grad_norm": 0.7890625,
      "learning_rate": 0.003,
      "loss": 2.171,
      "router_z_loss": 0.0877,
      "step": 9500
    },
    {
      "balancing_entropy": 0.8674,
      "balancing_loss": 1.093,
      "epoch": 48.52224,
      "grad_norm": 1.046875,
      "learning_rate": 0.003,
      "loss": 2.1795,
      "router_z_loss": 0.11,
      "step": 9510
    },
    {
      "balancing_entropy": 0.8568,
      "balancing_loss": 1.0945,
      "epoch": 48.57344,
      "grad_norm": 0.953125,
      "learning_rate": 0.003,
      "loss": 2.1731,
      "router_z_loss": 0.0729,
      "step": 9520
    },
    {
      "balancing_entropy": 0.8322,
      "balancing_loss": 1.0879,
      "epoch": 48.62464,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1578,
      "router_z_loss": 0.0938,
      "step": 9530
    },
    {
      "balancing_entropy": 0.8214,
      "balancing_loss": 1.116,
      "epoch": 48.67584,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1674,
      "router_z_loss": 0.0719,
      "step": 9540
    },
    {
      "balancing_entropy": 0.7958,
      "balancing_loss": 1.1067,
      "epoch": 48.72704,
      "grad_norm": 0.90625,
      "learning_rate": 0.003,
      "loss": 2.1675,
      "router_z_loss": 0.095,
      "step": 9550
    },
    {
      "balancing_entropy": 0.8552,
      "balancing_loss": 1.0812,
      "epoch": 48.77824,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.1714,
      "router_z_loss": 0.1209,
      "step": 9560
    },
    {
      "balancing_entropy": 0.8645,
      "balancing_loss": 1.1158,
      "epoch": 48.82944,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1689,
      "router_z_loss": 0.0857,
      "step": 9570
    },
    {
      "balancing_entropy": 0.878,
      "balancing_loss": 1.0929,
      "epoch": 48.88064,
      "grad_norm": 1.0078125,
      "learning_rate": 0.003,
      "loss": 2.1798,
      "router_z_loss": 0.0849,
      "step": 9580
    },
    {
      "balancing_entropy": 0.8119,
      "balancing_loss": 1.115,
      "epoch": 48.93184,
      "grad_norm": 1.1171875,
      "learning_rate": 0.003,
      "loss": 2.1773,
      "router_z_loss": 0.0764,
      "step": 9590
    },
    {
      "balancing_entropy": 0.7832,
      "balancing_loss": 1.1085,
      "epoch": 48.98304,
      "grad_norm": 1.203125,
      "learning_rate": 0.003,
      "loss": 2.1742,
      "router_z_loss": 0.0892,
      "step": 9600
    },
    {
      "balancing_entropy": 0.8995,
      "balancing_loss": 1.0795,
      "epoch": 49.03072,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.1714,
      "router_z_loss": 0.0805,
      "step": 9610
    },
    {
      "balancing_entropy": 0.851,
      "balancing_loss": 1.0971,
      "epoch": 49.08192,
      "grad_norm": 0.97265625,
      "learning_rate": 0.003,
      "loss": 2.1649,
      "router_z_loss": 0.0626,
      "step": 9620
    },
    {
      "balancing_entropy": 0.9511,
      "balancing_loss": 1.0939,
      "epoch": 49.13312,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.1538,
      "router_z_loss": 0.0736,
      "step": 9630
    },
    {
      "balancing_entropy": 0.8582,
      "balancing_loss": 1.1216,
      "epoch": 49.18432,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1557,
      "router_z_loss": 0.0654,
      "step": 9640
    },
    {
      "balancing_entropy": 0.9318,
      "balancing_loss": 1.0939,
      "epoch": 49.23552,
      "grad_norm": 0.66015625,
      "learning_rate": 0.003,
      "loss": 2.1472,
      "router_z_loss": 0.0794,
      "step": 9650
    },
    {
      "balancing_entropy": 0.8376,
      "balancing_loss": 1.0997,
      "epoch": 49.28672,
      "grad_norm": 1.265625,
      "learning_rate": 0.003,
      "loss": 2.1548,
      "router_z_loss": 0.0632,
      "step": 9660
    },
    {
      "balancing_entropy": 0.8428,
      "balancing_loss": 1.1003,
      "epoch": 49.33792,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1619,
      "router_z_loss": 0.0699,
      "step": 9670
    },
    {
      "balancing_entropy": 0.8626,
      "balancing_loss": 1.098,
      "epoch": 49.38912,
      "grad_norm": 1.2265625,
      "learning_rate": 0.003,
      "loss": 2.1614,
      "router_z_loss": 0.0696,
      "step": 9680
    },
    {
      "balancing_entropy": 1.0263,
      "balancing_loss": 1.1029,
      "epoch": 49.44032,
      "grad_norm": 0.298828125,
      "learning_rate": 0.003,
      "loss": 2.159,
      "router_z_loss": 0.0596,
      "step": 9690
    },
    {
      "balancing_entropy": 1.0916,
      "balancing_loss": 1.0894,
      "epoch": 49.49152,
      "grad_norm": 0.26171875,
      "learning_rate": 0.003,
      "loss": 2.1555,
      "router_z_loss": 0.0467,
      "step": 9700
    },
    {
      "balancing_entropy": 0.9875,
      "balancing_loss": 1.1152,
      "epoch": 49.54272,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1596,
      "router_z_loss": 0.0593,
      "step": 9710
    },
    {
      "balancing_entropy": 0.8452,
      "balancing_loss": 1.1052,
      "epoch": 49.59392,
      "grad_norm": 0.95703125,
      "learning_rate": 0.003,
      "loss": 2.1609,
      "router_z_loss": 0.0773,
      "step": 9720
    },
    {
      "balancing_entropy": 0.8529,
      "balancing_loss": 1.1041,
      "epoch": 49.64512,
      "grad_norm": 1.1328125,
      "learning_rate": 0.003,
      "loss": 2.1569,
      "router_z_loss": 0.1565,
      "step": 9730
    },
    {
      "balancing_entropy": 0.8735,
      "balancing_loss": 1.1105,
      "epoch": 49.69632,
      "grad_norm": 1.03125,
      "learning_rate": 0.003,
      "loss": 2.1672,
      "router_z_loss": 0.1008,
      "step": 9740
    },
    {
      "balancing_entropy": 1.0064,
      "balancing_loss": 1.0892,
      "epoch": 49.74752,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.1661,
      "router_z_loss": 0.0752,
      "step": 9750
    },
    {
      "balancing_entropy": 0.8651,
      "balancing_loss": 1.0958,
      "epoch": 49.79872,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.1608,
      "router_z_loss": 0.0943,
      "step": 9760
    },
    {
      "balancing_entropy": 0.862,
      "balancing_loss": 1.1277,
      "epoch": 49.84992,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1633,
      "router_z_loss": 0.0644,
      "step": 9770
    },
    {
      "balancing_entropy": 0.8463,
      "balancing_loss": 1.0793,
      "epoch": 49.90112,
      "grad_norm": 0.9609375,
      "learning_rate": 0.003,
      "loss": 2.1671,
      "router_z_loss": 0.1369,
      "step": 9780
    },
    {
      "balancing_entropy": 0.8243,
      "balancing_loss": 1.1142,
      "epoch": 49.95232,
      "grad_norm": 0.98046875,
      "learning_rate": 0.003,
      "loss": 2.1848,
      "router_z_loss": 0.1311,
      "step": 9790
    },
    {
      "balancing_entropy": 0.8846,
      "balancing_loss": 1.0981,
      "epoch": 50.0,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.172,
      "router_z_loss": 0.1524,
      "step": 9800
    },
    {
      "balancing_entropy": 0.9082,
      "balancing_loss": 1.1001,
      "epoch": 50.0512,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.1515,
      "router_z_loss": 0.0693,
      "step": 9810
    },
    {
      "balancing_entropy": 0.9334,
      "balancing_loss": 1.1037,
      "epoch": 50.1024,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1489,
      "router_z_loss": 0.0785,
      "step": 9820
    },
    {
      "balancing_entropy": 1.0334,
      "balancing_loss": 1.1026,
      "epoch": 50.1536,
      "grad_norm": 1.1171875,
      "learning_rate": 0.003,
      "loss": 2.1502,
      "router_z_loss": 0.0623,
      "step": 9830
    },
    {
      "balancing_entropy": 0.8177,
      "balancing_loss": 1.1246,
      "epoch": 50.2048,
      "grad_norm": 1.5078125,
      "learning_rate": 0.003,
      "loss": 2.1536,
      "router_z_loss": 0.058,
      "step": 9840
    },
    {
      "balancing_entropy": 0.7926,
      "balancing_loss": 1.109,
      "epoch": 50.256,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.1539,
      "router_z_loss": 0.1565,
      "step": 9850
    },
    {
      "balancing_entropy": 0.8218,
      "balancing_loss": 1.0992,
      "epoch": 50.3072,
      "grad_norm": 1.0078125,
      "learning_rate": 0.003,
      "loss": 2.1488,
      "router_z_loss": 0.1226,
      "step": 9860
    },
    {
      "balancing_entropy": 0.833,
      "balancing_loss": 1.104,
      "epoch": 50.3584,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.1636,
      "router_z_loss": 0.1044,
      "step": 9870
    },
    {
      "balancing_entropy": 0.9838,
      "balancing_loss": 1.0897,
      "epoch": 50.4096,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.162,
      "router_z_loss": 0.0729,
      "step": 9880
    },
    {
      "balancing_entropy": 0.8914,
      "balancing_loss": 1.0919,
      "epoch": 50.4608,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.16,
      "router_z_loss": 0.0601,
      "step": 9890
    },
    {
      "balancing_entropy": 0.8371,
      "balancing_loss": 1.1069,
      "epoch": 50.512,
      "grad_norm": 0.80078125,
      "learning_rate": 0.003,
      "loss": 2.157,
      "router_z_loss": 0.0669,
      "step": 9900
    },
    {
      "balancing_entropy": 0.8384,
      "balancing_loss": 1.1127,
      "epoch": 50.5632,
      "grad_norm": 0.765625,
      "learning_rate": 0.003,
      "loss": 2.1603,
      "router_z_loss": 0.0977,
      "step": 9910
    },
    {
      "balancing_entropy": 0.8528,
      "balancing_loss": 1.0998,
      "epoch": 50.6144,
      "grad_norm": 0.9453125,
      "learning_rate": 0.003,
      "loss": 2.166,
      "router_z_loss": 0.0953,
      "step": 9920
    },
    {
      "balancing_entropy": 0.8584,
      "balancing_loss": 1.0966,
      "epoch": 50.6656,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.166,
      "router_z_loss": 0.0757,
      "step": 9930
    },
    {
      "balancing_entropy": 0.9737,
      "balancing_loss": 1.1288,
      "epoch": 50.7168,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.1705,
      "router_z_loss": 0.0639,
      "step": 9940
    },
    {
      "balancing_entropy": 0.8865,
      "balancing_loss": 1.0854,
      "epoch": 50.768,
      "grad_norm": 0.88671875,
      "learning_rate": 0.003,
      "loss": 2.1637,
      "router_z_loss": 0.084,
      "step": 9950
    },
    {
      "balancing_entropy": 0.8138,
      "balancing_loss": 1.0928,
      "epoch": 50.8192,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.1593,
      "router_z_loss": 0.0808,
      "step": 9960
    },
    {
      "balancing_entropy": 0.8241,
      "balancing_loss": 1.1026,
      "epoch": 50.8704,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.1712,
      "router_z_loss": 0.0837,
      "step": 9970
    },
    {
      "balancing_entropy": 0.8399,
      "balancing_loss": 1.0916,
      "epoch": 50.9216,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1689,
      "router_z_loss": 0.1108,
      "step": 9980
    },
    {
      "balancing_entropy": 0.8367,
      "balancing_loss": 1.1186,
      "epoch": 50.9728,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.1663,
      "router_z_loss": 0.0725,
      "step": 9990
    },
    {
      "balancing_entropy": 0.7805,
      "balancing_loss": 1.1028,
      "epoch": 51.02048,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.1735,
      "router_z_loss": 0.1047,
      "step": 10000
    },
    {
      "balancing_entropy": 0.8526,
      "balancing_loss": 1.1094,
      "epoch": 51.07168,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.161,
      "router_z_loss": 0.1093,
      "step": 10010
    },
    {
      "balancing_entropy": 0.8294,
      "balancing_loss": 1.1072,
      "epoch": 51.12288,
      "grad_norm": 2.203125,
      "learning_rate": 0.003,
      "loss": 2.1845,
      "router_z_loss": 0.1213,
      "step": 10020
    },
    {
      "balancing_entropy": 0.7949,
      "balancing_loss": 1.1287,
      "epoch": 51.174080000000004,
      "grad_norm": 2.4375,
      "learning_rate": 0.003,
      "loss": 2.166,
      "router_z_loss": 0.1383,
      "step": 10030
    },
    {
      "balancing_entropy": 0.8195,
      "balancing_loss": 1.1156,
      "epoch": 51.22528,
      "grad_norm": 1.2734375,
      "learning_rate": 0.003,
      "loss": 2.1558,
      "router_z_loss": 0.1415,
      "step": 10040
    },
    {
      "balancing_entropy": 0.8586,
      "balancing_loss": 1.0809,
      "epoch": 51.27648,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.1559,
      "router_z_loss": 0.0873,
      "step": 10050
    },
    {
      "balancing_entropy": 0.9087,
      "balancing_loss": 1.1204,
      "epoch": 51.32768,
      "grad_norm": 0.82421875,
      "learning_rate": 0.003,
      "loss": 2.1536,
      "router_z_loss": 0.0862,
      "step": 10060
    },
    {
      "balancing_entropy": 0.7995,
      "balancing_loss": 1.1247,
      "epoch": 51.37888,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.165,
      "router_z_loss": 0.1222,
      "step": 10070
    },
    {
      "balancing_entropy": 0.9465,
      "balancing_loss": 1.0504,
      "epoch": 51.43008,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.1595,
      "router_z_loss": 0.1276,
      "step": 10080
    },
    {
      "balancing_entropy": 0.8809,
      "balancing_loss": 1.1615,
      "epoch": 51.48128,
      "grad_norm": 1.3125,
      "learning_rate": 0.003,
      "loss": 2.1679,
      "router_z_loss": 0.1039,
      "step": 10090
    },
    {
      "balancing_entropy": 0.8939,
      "balancing_loss": 1.1025,
      "epoch": 51.53248,
      "grad_norm": 0.9453125,
      "learning_rate": 0.003,
      "loss": 2.1568,
      "router_z_loss": 0.137,
      "step": 10100
    },
    {
      "balancing_entropy": 0.9524,
      "balancing_loss": 1.1175,
      "epoch": 51.58368,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.175,
      "router_z_loss": 0.0751,
      "step": 10110
    },
    {
      "balancing_entropy": 1.0004,
      "balancing_loss": 1.0758,
      "epoch": 51.63488,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.1645,
      "router_z_loss": 0.0842,
      "step": 10120
    },
    {
      "balancing_entropy": 0.9375,
      "balancing_loss": 1.0663,
      "epoch": 51.68608,
      "grad_norm": 0.97265625,
      "learning_rate": 0.003,
      "loss": 2.1642,
      "router_z_loss": 0.0809,
      "step": 10130
    },
    {
      "balancing_entropy": 0.8378,
      "balancing_loss": 1.1193,
      "epoch": 51.73728,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.1617,
      "router_z_loss": 0.0777,
      "step": 10140
    },
    {
      "balancing_entropy": 0.9812,
      "balancing_loss": 1.114,
      "epoch": 51.78848,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.1657,
      "router_z_loss": 0.0799,
      "step": 10150
    },
    {
      "balancing_entropy": 0.9566,
      "balancing_loss": 1.0725,
      "epoch": 51.83968,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.1646,
      "router_z_loss": 0.0694,
      "step": 10160
    },
    {
      "balancing_entropy": 0.8254,
      "balancing_loss": 1.162,
      "epoch": 51.89088,
      "grad_norm": 1.078125,
      "learning_rate": 0.003,
      "loss": 2.1619,
      "router_z_loss": 0.0707,
      "step": 10170
    },
    {
      "balancing_entropy": 0.88,
      "balancing_loss": 1.0937,
      "epoch": 51.94208,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.1617,
      "router_z_loss": 0.0749,
      "step": 10180
    },
    {
      "balancing_entropy": 0.7759,
      "balancing_loss": 1.177,
      "epoch": 51.99328,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.1702,
      "router_z_loss": 0.0711,
      "step": 10190
    },
    {
      "balancing_entropy": 0.8372,
      "balancing_loss": 1.0767,
      "epoch": 52.04096,
      "grad_norm": 0.671875,
      "learning_rate": 0.003,
      "loss": 2.1644,
      "router_z_loss": 0.1468,
      "step": 10200
    },
    {
      "balancing_entropy": 0.9957,
      "balancing_loss": 1.0798,
      "epoch": 52.09216,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.1571,
      "router_z_loss": 0.1214,
      "step": 10210
    },
    {
      "balancing_entropy": 0.8894,
      "balancing_loss": 1.0877,
      "epoch": 52.14336,
      "grad_norm": 0.90234375,
      "learning_rate": 0.003,
      "loss": 2.1486,
      "router_z_loss": 0.1207,
      "step": 10220
    },
    {
      "balancing_entropy": 0.9072,
      "balancing_loss": 1.1204,
      "epoch": 52.19456,
      "grad_norm": 0.93359375,
      "learning_rate": 0.003,
      "loss": 2.1506,
      "router_z_loss": 0.0965,
      "step": 10230
    },
    {
      "balancing_entropy": 0.8725,
      "balancing_loss": 1.133,
      "epoch": 52.24576,
      "grad_norm": 0.87890625,
      "learning_rate": 0.003,
      "loss": 2.1518,
      "router_z_loss": 0.1118,
      "step": 10240
    },
    {
      "balancing_entropy": 0.875,
      "balancing_loss": 1.0607,
      "epoch": 52.29696,
      "grad_norm": 0.90234375,
      "learning_rate": 0.003,
      "loss": 2.1632,
      "router_z_loss": 0.1223,
      "step": 10250
    },
    {
      "balancing_entropy": 0.8346,
      "balancing_loss": 1.1421,
      "epoch": 52.34816,
      "grad_norm": 1.2265625,
      "learning_rate": 0.003,
      "loss": 2.1584,
      "router_z_loss": 0.1018,
      "step": 10260
    },
    {
      "balancing_entropy": 0.8129,
      "balancing_loss": 1.1165,
      "epoch": 52.39936,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1579,
      "router_z_loss": 0.115,
      "step": 10270
    },
    {
      "balancing_entropy": 0.8056,
      "balancing_loss": 1.1149,
      "epoch": 52.45056,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.1558,
      "router_z_loss": 0.1133,
      "step": 10280
    },
    {
      "balancing_entropy": 0.8064,
      "balancing_loss": 1.085,
      "epoch": 52.50176,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.1749,
      "router_z_loss": 0.1488,
      "step": 10290
    },
    {
      "balancing_entropy": 0.9062,
      "balancing_loss": 1.0927,
      "epoch": 52.55296,
      "grad_norm": 1.3515625,
      "learning_rate": 0.003,
      "loss": 2.1714,
      "router_z_loss": 0.1192,
      "step": 10300
    },
    {
      "balancing_entropy": 0.8317,
      "balancing_loss": 1.1256,
      "epoch": 52.60416,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.1563,
      "router_z_loss": 0.0946,
      "step": 10310
    },
    {
      "balancing_entropy": 0.8428,
      "balancing_loss": 1.0929,
      "epoch": 52.65536,
      "grad_norm": 1.09375,
      "learning_rate": 0.003,
      "loss": 2.153,
      "router_z_loss": 0.1391,
      "step": 10320
    },
    {
      "balancing_entropy": 0.8232,
      "balancing_loss": 1.1253,
      "epoch": 52.70656,
      "grad_norm": 1.1875,
      "learning_rate": 0.003,
      "loss": 2.1588,
      "router_z_loss": 0.1762,
      "step": 10330
    },
    {
      "balancing_entropy": 0.8767,
      "balancing_loss": 1.0713,
      "epoch": 52.75776,
      "grad_norm": 1.296875,
      "learning_rate": 0.003,
      "loss": 2.1663,
      "router_z_loss": 0.1016,
      "step": 10340
    },
    {
      "balancing_entropy": 0.8742,
      "balancing_loss": 1.1182,
      "epoch": 52.80896,
      "grad_norm": 1.0625,
      "learning_rate": 0.003,
      "loss": 2.1582,
      "router_z_loss": 0.1127,
      "step": 10350
    },
    {
      "balancing_entropy": 0.8385,
      "balancing_loss": 1.0806,
      "epoch": 52.86016,
      "grad_norm": 0.91796875,
      "learning_rate": 0.003,
      "loss": 2.153,
      "router_z_loss": 0.1106,
      "step": 10360
    },
    {
      "balancing_entropy": 0.8261,
      "balancing_loss": 1.1458,
      "epoch": 52.91136,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.1564,
      "router_z_loss": 0.1413,
      "step": 10370
    },
    {
      "balancing_entropy": 0.857,
      "balancing_loss": 1.0661,
      "epoch": 52.96256,
      "grad_norm": 1.0390625,
      "learning_rate": 0.003,
      "loss": 2.1674,
      "router_z_loss": 0.1617,
      "step": 10380
    },
    {
      "balancing_entropy": 0.6948,
      "balancing_loss": 1.1792,
      "epoch": 53.01024,
      "grad_norm": 0.90234375,
      "learning_rate": 0.003,
      "loss": 2.1737,
      "router_z_loss": 0.1282,
      "step": 10390
    },
    {
      "balancing_entropy": 0.8248,
      "balancing_loss": 1.0884,
      "epoch": 53.06144,
      "grad_norm": 0.984375,
      "learning_rate": 0.003,
      "loss": 2.1624,
      "router_z_loss": 0.1563,
      "step": 10400
    },
    {
      "balancing_entropy": 0.7833,
      "balancing_loss": 1.1047,
      "epoch": 53.11264,
      "grad_norm": 0.87109375,
      "learning_rate": 0.003,
      "loss": 2.1568,
      "router_z_loss": 0.2073,
      "step": 10410
    },
    {
      "balancing_entropy": 0.8034,
      "balancing_loss": 1.1016,
      "epoch": 53.16384,
      "grad_norm": 0.9453125,
      "learning_rate": 0.003,
      "loss": 2.1641,
      "router_z_loss": 0.1581,
      "step": 10420
    },
    {
      "balancing_entropy": 0.8341,
      "balancing_loss": 1.1516,
      "epoch": 53.21504,
      "grad_norm": 0.69140625,
      "learning_rate": 0.003,
      "loss": 2.1543,
      "router_z_loss": 0.1771,
      "step": 10430
    },
    {
      "balancing_entropy": 0.8444,
      "balancing_loss": 1.0666,
      "epoch": 53.26624,
      "grad_norm": 0.79296875,
      "learning_rate": 0.003,
      "loss": 2.1548,
      "router_z_loss": 0.1817,
      "step": 10440
    },
    {
      "balancing_entropy": 0.7953,
      "balancing_loss": 1.1262,
      "epoch": 53.31744,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.1641,
      "router_z_loss": 0.185,
      "step": 10450
    },
    {
      "balancing_entropy": 0.9109,
      "balancing_loss": 1.1061,
      "epoch": 53.36864,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.1594,
      "router_z_loss": 0.1751,
      "step": 10460
    },
    {
      "balancing_entropy": 0.806,
      "balancing_loss": 1.09,
      "epoch": 53.41984,
      "grad_norm": 0.8828125,
      "learning_rate": 0.003,
      "loss": 2.1516,
      "router_z_loss": 0.192,
      "step": 10470
    },
    {
      "balancing_entropy": 0.753,
      "balancing_loss": 1.1292,
      "epoch": 53.47104,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.1497,
      "router_z_loss": 0.2006,
      "step": 10480
    },
    {
      "balancing_entropy": 0.8474,
      "balancing_loss": 1.0716,
      "epoch": 53.52224,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1517,
      "router_z_loss": 0.1828,
      "step": 10490
    },
    {
      "balancing_entropy": 0.8043,
      "balancing_loss": 1.1205,
      "epoch": 53.57344,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.1581,
      "router_z_loss": 0.151,
      "step": 10500
    },
    {
      "balancing_entropy": 0.8688,
      "balancing_loss": 1.1115,
      "epoch": 53.62464,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1586,
      "router_z_loss": 0.1435,
      "step": 10510
    },
    {
      "balancing_entropy": 0.9263,
      "balancing_loss": 1.0791,
      "epoch": 53.67584,
      "grad_norm": 0.91015625,
      "learning_rate": 0.003,
      "loss": 2.1598,
      "router_z_loss": 0.1621,
      "step": 10520
    },
    {
      "balancing_entropy": 0.8058,
      "balancing_loss": 1.1217,
      "epoch": 53.72704,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.1687,
      "router_z_loss": 0.1584,
      "step": 10530
    },
    {
      "balancing_entropy": 0.9017,
      "balancing_loss": 1.0725,
      "epoch": 53.77824,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1674,
      "router_z_loss": 0.1914,
      "step": 10540
    },
    {
      "balancing_entropy": 0.7243,
      "balancing_loss": 1.1807,
      "epoch": 53.82944,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.1677,
      "router_z_loss": 0.1456,
      "step": 10550
    },
    {
      "balancing_entropy": 0.8124,
      "balancing_loss": 1.0955,
      "epoch": 53.88064,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1561,
      "router_z_loss": 0.2612,
      "step": 10560
    },
    {
      "balancing_entropy": 0.8514,
      "balancing_loss": 1.0793,
      "epoch": 53.93184,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1553,
      "router_z_loss": 0.2212,
      "step": 10570
    },
    {
      "balancing_entropy": 0.7331,
      "balancing_loss": 1.1535,
      "epoch": 53.98304,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.1544,
      "router_z_loss": 0.216,
      "step": 10580
    },
    {
      "balancing_entropy": 0.8193,
      "balancing_loss": 1.0677,
      "epoch": 54.03072,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1466,
      "router_z_loss": 0.2115,
      "step": 10590
    },
    {
      "balancing_entropy": 0.8226,
      "balancing_loss": 1.0849,
      "epoch": 54.08192,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.1426,
      "router_z_loss": 0.177,
      "step": 10600
    },
    {
      "balancing_entropy": 0.7627,
      "balancing_loss": 1.1242,
      "epoch": 54.13312,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.1384,
      "router_z_loss": 0.2111,
      "step": 10610
    },
    {
      "balancing_entropy": 0.858,
      "balancing_loss": 1.0713,
      "epoch": 54.18432,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1479,
      "router_z_loss": 0.2088,
      "step": 10620
    },
    {
      "balancing_entropy": 0.795,
      "balancing_loss": 1.1042,
      "epoch": 54.23552,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.1445,
      "router_z_loss": 0.2261,
      "step": 10630
    },
    {
      "balancing_entropy": 0.7975,
      "balancing_loss": 1.1268,
      "epoch": 54.28672,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1585,
      "router_z_loss": 0.249,
      "step": 10640
    },
    {
      "balancing_entropy": 0.8273,
      "balancing_loss": 1.0773,
      "epoch": 54.33792,
      "grad_norm": 0.44140625,
      "learning_rate": 0.003,
      "loss": 2.1544,
      "router_z_loss": 0.2313,
      "step": 10650
    },
    {
      "balancing_entropy": 0.8121,
      "balancing_loss": 1.1553,
      "epoch": 54.38912,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.1549,
      "router_z_loss": 0.2785,
      "step": 10660
    },
    {
      "balancing_entropy": 0.8332,
      "balancing_loss": 1.0762,
      "epoch": 54.44032,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.1573,
      "router_z_loss": 0.2797,
      "step": 10670
    },
    {
      "balancing_entropy": 0.8637,
      "balancing_loss": 1.1673,
      "epoch": 54.49152,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1571,
      "router_z_loss": 0.2637,
      "step": 10680
    },
    {
      "balancing_entropy": 0.8349,
      "balancing_loss": 1.0609,
      "epoch": 54.54272,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1656,
      "router_z_loss": 0.2361,
      "step": 10690
    },
    {
      "balancing_entropy": 0.8008,
      "balancing_loss": 1.0996,
      "epoch": 54.59392,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.1656,
      "router_z_loss": 0.1757,
      "step": 10700
    },
    {
      "balancing_entropy": 0.8475,
      "balancing_loss": 1.0998,
      "epoch": 54.64512,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.1569,
      "router_z_loss": 0.1669,
      "step": 10710
    },
    {
      "balancing_entropy": 0.8944,
      "balancing_loss": 1.1351,
      "epoch": 54.69632,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1471,
      "router_z_loss": 0.1651,
      "step": 10720
    },
    {
      "balancing_entropy": 0.8915,
      "balancing_loss": 1.055,
      "epoch": 54.74752,
      "grad_norm": 0.70703125,
      "learning_rate": 0.003,
      "loss": 2.1478,
      "router_z_loss": 0.1802,
      "step": 10730
    },
    {
      "balancing_entropy": 0.8128,
      "balancing_loss": 1.0998,
      "epoch": 54.79872,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.1537,
      "router_z_loss": 0.1877,
      "step": 10740
    },
    {
      "balancing_entropy": 0.7846,
      "balancing_loss": 1.1452,
      "epoch": 54.84992,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.1642,
      "router_z_loss": 0.2075,
      "step": 10750
    },
    {
      "balancing_entropy": 0.902,
      "balancing_loss": 1.0594,
      "epoch": 54.90112,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.1614,
      "router_z_loss": 0.2011,
      "step": 10760
    },
    {
      "balancing_entropy": 0.8781,
      "balancing_loss": 1.0919,
      "epoch": 54.95232,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.1612,
      "router_z_loss": 0.1496,
      "step": 10770
    },
    {
      "balancing_entropy": 0.7976,
      "balancing_loss": 1.0948,
      "epoch": 55.0,
      "grad_norm": 0.74609375,
      "learning_rate": 0.003,
      "loss": 2.1529,
      "router_z_loss": 0.146,
      "step": 10780
    },
    {
      "balancing_entropy": 0.814,
      "balancing_loss": 1.1003,
      "epoch": 55.0512,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1341,
      "router_z_loss": 0.1668,
      "step": 10790
    },
    {
      "balancing_entropy": 0.9215,
      "balancing_loss": 1.0967,
      "epoch": 55.1024,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.1497,
      "router_z_loss": 0.1586,
      "step": 10800
    },
    {
      "balancing_entropy": 0.7971,
      "balancing_loss": 1.0905,
      "epoch": 55.1536,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1372,
      "router_z_loss": 0.1884,
      "step": 10810
    },
    {
      "balancing_entropy": 0.8281,
      "balancing_loss": 1.1029,
      "epoch": 55.2048,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.1356,
      "router_z_loss": 0.1891,
      "step": 10820
    },
    {
      "balancing_entropy": 0.8269,
      "balancing_loss": 1.0818,
      "epoch": 55.256,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1422,
      "router_z_loss": 0.3302,
      "step": 10830
    },
    {
      "balancing_entropy": 0.8097,
      "balancing_loss": 1.1347,
      "epoch": 55.3072,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.1383,
      "router_z_loss": 0.1812,
      "step": 10840
    },
    {
      "balancing_entropy": 0.8585,
      "balancing_loss": 1.0716,
      "epoch": 55.3584,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1458,
      "router_z_loss": 0.18,
      "step": 10850
    },
    {
      "balancing_entropy": 0.839,
      "balancing_loss": 1.1107,
      "epoch": 55.4096,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1468,
      "router_z_loss": 0.1949,
      "step": 10860
    },
    {
      "balancing_entropy": 0.8566,
      "balancing_loss": 1.0757,
      "epoch": 55.4608,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.1486,
      "router_z_loss": 0.1594,
      "step": 10870
    },
    {
      "balancing_entropy": 0.7935,
      "balancing_loss": 1.1148,
      "epoch": 55.512,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.1418,
      "router_z_loss": 0.1431,
      "step": 10880
    },
    {
      "balancing_entropy": 0.8239,
      "balancing_loss": 1.0679,
      "epoch": 55.5632,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1447,
      "router_z_loss": 0.1615,
      "step": 10890
    },
    {
      "balancing_entropy": 0.8454,
      "balancing_loss": 1.1104,
      "epoch": 55.6144,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.1421,
      "router_z_loss": 0.1516,
      "step": 10900
    },
    {
      "balancing_entropy": 0.7768,
      "balancing_loss": 1.102,
      "epoch": 55.6656,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.1461,
      "router_z_loss": 0.1752,
      "step": 10910
    },
    {
      "balancing_entropy": 0.7223,
      "balancing_loss": 1.1361,
      "epoch": 55.7168,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1575,
      "router_z_loss": 0.1623,
      "step": 10920
    },
    {
      "balancing_entropy": 0.8878,
      "balancing_loss": 1.072,
      "epoch": 55.768,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1579,
      "router_z_loss": 0.2058,
      "step": 10930
    },
    {
      "balancing_entropy": 0.8442,
      "balancing_loss": 1.0989,
      "epoch": 55.8192,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.1617,
      "router_z_loss": 0.1437,
      "step": 10940
    },
    {
      "balancing_entropy": 0.8414,
      "balancing_loss": 1.0797,
      "epoch": 55.8704,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.1531,
      "router_z_loss": 0.1593,
      "step": 10950
    },
    {
      "balancing_entropy": 0.8525,
      "balancing_loss": 1.1053,
      "epoch": 55.9216,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.1479,
      "router_z_loss": 0.1459,
      "step": 10960
    },
    {
      "balancing_entropy": 0.8154,
      "balancing_loss": 1.1099,
      "epoch": 55.9728,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1504,
      "router_z_loss": 0.1553,
      "step": 10970
    },
    {
      "balancing_entropy": 0.8006,
      "balancing_loss": 1.1035,
      "epoch": 56.02048,
      "grad_norm": 1.078125,
      "learning_rate": 0.003,
      "loss": 2.1422,
      "router_z_loss": 0.1521,
      "step": 10980
    },
    {
      "balancing_entropy": 0.8314,
      "balancing_loss": 1.064,
      "epoch": 56.07168,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.142,
      "router_z_loss": 0.1794,
      "step": 10990
    },
    {
      "balancing_entropy": 0.746,
      "balancing_loss": 1.1713,
      "epoch": 56.12288,
      "grad_norm": 0.62109375,
      "learning_rate": 0.003,
      "loss": 2.1392,
      "router_z_loss": 0.3513,
      "step": 11000
    },
    {
      "balancing_entropy": 0.7898,
      "balancing_loss": 1.0983,
      "epoch": 56.174080000000004,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.1351,
      "router_z_loss": 0.1377,
      "step": 11010
    },
    {
      "balancing_entropy": 0.7744,
      "balancing_loss": 1.1001,
      "epoch": 56.22528,
      "grad_norm": 1.0,
      "learning_rate": 0.003,
      "loss": 2.1386,
      "router_z_loss": 0.1442,
      "step": 11020
    },
    {
      "balancing_entropy": 0.8269,
      "balancing_loss": 1.0768,
      "epoch": 56.27648,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1336,
      "router_z_loss": 0.1578,
      "step": 11030
    },
    {
      "balancing_entropy": 0.8496,
      "balancing_loss": 1.1332,
      "epoch": 56.32768,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.1347,
      "router_z_loss": 0.1423,
      "step": 11040
    },
    {
      "balancing_entropy": 0.8346,
      "balancing_loss": 1.0762,
      "epoch": 56.37888,
      "grad_norm": 0.6875,
      "learning_rate": 0.003,
      "loss": 2.142,
      "router_z_loss": 0.135,
      "step": 11050
    },
    {
      "balancing_entropy": 0.7854,
      "balancing_loss": 1.0962,
      "epoch": 56.43008,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1418,
      "router_z_loss": 0.1432,
      "step": 11060
    },
    {
      "balancing_entropy": 0.7321,
      "balancing_loss": 1.1744,
      "epoch": 56.48128,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.145,
      "router_z_loss": 0.1466,
      "step": 11070
    },
    {
      "balancing_entropy": 0.8058,
      "balancing_loss": 1.0794,
      "epoch": 56.53248,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1525,
      "router_z_loss": 0.1667,
      "step": 11080
    },
    {
      "balancing_entropy": 0.8304,
      "balancing_loss": 1.063,
      "epoch": 56.58368,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.1444,
      "router_z_loss": 0.1207,
      "step": 11090
    },
    {
      "balancing_entropy": 0.7547,
      "balancing_loss": 1.1445,
      "epoch": 56.63488,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.1546,
      "router_z_loss": 0.1448,
      "step": 11100
    },
    {
      "balancing_entropy": 0.8294,
      "balancing_loss": 1.0822,
      "epoch": 56.68608,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1476,
      "router_z_loss": 0.1269,
      "step": 11110
    },
    {
      "balancing_entropy": 0.781,
      "balancing_loss": 1.1083,
      "epoch": 56.73728,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.1439,
      "router_z_loss": 0.1235,
      "step": 11120
    },
    {
      "balancing_entropy": 0.7994,
      "balancing_loss": 1.105,
      "epoch": 56.78848,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.1456,
      "router_z_loss": 0.128,
      "step": 11130
    },
    {
      "balancing_entropy": 0.7807,
      "balancing_loss": 1.0979,
      "epoch": 56.83968,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.1399,
      "router_z_loss": 0.1524,
      "step": 11140
    },
    {
      "balancing_entropy": 0.7831,
      "balancing_loss": 1.08,
      "epoch": 56.89088,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1433,
      "router_z_loss": 0.2079,
      "step": 11150
    },
    {
      "balancing_entropy": 0.7521,
      "balancing_loss": 1.1136,
      "epoch": 56.94208,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.1528,
      "router_z_loss": 0.1709,
      "step": 11160
    },
    {
      "balancing_entropy": 0.7742,
      "balancing_loss": 1.1241,
      "epoch": 56.99328,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.1671,
      "router_z_loss": 0.1762,
      "step": 11170
    },
    {
      "balancing_entropy": 0.7958,
      "balancing_loss": 1.0832,
      "epoch": 57.04096,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.1425,
      "router_z_loss": 0.1851,
      "step": 11180
    },
    {
      "balancing_entropy": 0.8068,
      "balancing_loss": 1.0946,
      "epoch": 57.09216,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.1272,
      "router_z_loss": 0.1502,
      "step": 11190
    },
    {
      "balancing_entropy": 0.7338,
      "balancing_loss": 1.1224,
      "epoch": 57.14336,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1268,
      "router_z_loss": 0.1663,
      "step": 11200
    },
    {
      "balancing_entropy": 0.8375,
      "balancing_loss": 1.0697,
      "epoch": 57.19456,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.1392,
      "router_z_loss": 0.1445,
      "step": 11210
    },
    {
      "balancing_entropy": 0.7956,
      "balancing_loss": 1.0931,
      "epoch": 57.24576,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.1396,
      "router_z_loss": 0.114,
      "step": 11220
    },
    {
      "balancing_entropy": 0.8136,
      "balancing_loss": 1.1087,
      "epoch": 57.29696,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1342,
      "router_z_loss": 0.1528,
      "step": 11230
    },
    {
      "balancing_entropy": 0.7707,
      "balancing_loss": 1.1023,
      "epoch": 57.34816,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1329,
      "router_z_loss": 0.1628,
      "step": 11240
    },
    {
      "balancing_entropy": 0.746,
      "balancing_loss": 1.1,
      "epoch": 57.39936,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.1398,
      "router_z_loss": 0.1765,
      "step": 11250
    },
    {
      "balancing_entropy": 0.7385,
      "balancing_loss": 1.1346,
      "epoch": 57.45056,
      "grad_norm": 0.8984375,
      "learning_rate": 0.003,
      "loss": 2.1493,
      "router_z_loss": 0.1983,
      "step": 11260
    },
    {
      "balancing_entropy": 0.8222,
      "balancing_loss": 1.0612,
      "epoch": 57.50176,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1555,
      "router_z_loss": 0.1871,
      "step": 11270
    },
    {
      "balancing_entropy": 0.8094,
      "balancing_loss": 1.0651,
      "epoch": 57.55296,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1524,
      "router_z_loss": 0.2467,
      "step": 11280
    },
    {
      "balancing_entropy": 0.823,
      "balancing_loss": 1.0912,
      "epoch": 57.60416,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.1493,
      "router_z_loss": 0.1336,
      "step": 11290
    },
    {
      "balancing_entropy": 0.741,
      "balancing_loss": 1.141,
      "epoch": 57.65536,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1615,
      "router_z_loss": 0.1211,
      "step": 11300
    },
    {
      "balancing_entropy": 0.7943,
      "balancing_loss": 1.0691,
      "epoch": 57.70656,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1471,
      "router_z_loss": 0.1796,
      "step": 11310
    },
    {
      "balancing_entropy": 0.741,
      "balancing_loss": 1.1141,
      "epoch": 57.75776,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.154,
      "router_z_loss": 0.1825,
      "step": 11320
    },
    {
      "balancing_entropy": 0.8339,
      "balancing_loss": 1.0646,
      "epoch": 57.80896,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1538,
      "router_z_loss": 0.1389,
      "step": 11330
    },
    {
      "balancing_entropy": 0.8071,
      "balancing_loss": 1.102,
      "epoch": 57.86016,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.1395,
      "router_z_loss": 0.1141,
      "step": 11340
    },
    {
      "balancing_entropy": 0.8051,
      "balancing_loss": 1.1365,
      "epoch": 57.91136,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1446,
      "router_z_loss": 0.123,
      "step": 11350
    },
    {
      "balancing_entropy": 0.8403,
      "balancing_loss": 1.07,
      "epoch": 57.96256,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.1384,
      "router_z_loss": 0.1044,
      "step": 11360
    },
    {
      "balancing_entropy": 0.7295,
      "balancing_loss": 1.1137,
      "epoch": 58.01024,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.1408,
      "router_z_loss": 0.125,
      "step": 11370
    },
    {
      "balancing_entropy": 0.7238,
      "balancing_loss": 1.1215,
      "epoch": 58.06144,
      "grad_norm": 0.69921875,
      "learning_rate": 0.003,
      "loss": 2.1307,
      "router_z_loss": 0.1379,
      "step": 11380
    },
    {
      "balancing_entropy": 0.7807,
      "balancing_loss": 1.1092,
      "epoch": 58.11264,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1272,
      "router_z_loss": 0.1907,
      "step": 11390
    },
    {
      "balancing_entropy": 0.7439,
      "balancing_loss": 1.1105,
      "epoch": 58.16384,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1263,
      "router_z_loss": 0.1754,
      "step": 11400
    },
    {
      "balancing_entropy": 0.8049,
      "balancing_loss": 1.0823,
      "epoch": 58.21504,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.1343,
      "router_z_loss": 0.3226,
      "step": 11410
    },
    {
      "balancing_entropy": 0.7976,
      "balancing_loss": 1.0761,
      "epoch": 58.26624,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.1326,
      "router_z_loss": 0.177,
      "step": 11420
    },
    {
      "balancing_entropy": 0.8211,
      "balancing_loss": 1.1417,
      "epoch": 58.31744,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.1422,
      "router_z_loss": 0.1403,
      "step": 11430
    },
    {
      "balancing_entropy": 0.7953,
      "balancing_loss": 1.0744,
      "epoch": 58.36864,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1503,
      "router_z_loss": 0.1566,
      "step": 11440
    },
    {
      "balancing_entropy": 0.7991,
      "balancing_loss": 1.1071,
      "epoch": 58.41984,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1465,
      "router_z_loss": 0.1525,
      "step": 11450
    },
    {
      "balancing_entropy": 0.8952,
      "balancing_loss": 1.1023,
      "epoch": 58.47104,
      "grad_norm": 0.306640625,
      "learning_rate": 0.003,
      "loss": 2.1389,
      "router_z_loss": 0.132,
      "step": 11460
    },
    {
      "balancing_entropy": 0.875,
      "balancing_loss": 1.0724,
      "epoch": 58.52224,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.1392,
      "router_z_loss": 0.1272,
      "step": 11470
    },
    {
      "balancing_entropy": 0.7872,
      "balancing_loss": 1.107,
      "epoch": 58.57344,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.1295,
      "router_z_loss": 0.1216,
      "step": 11480
    },
    {
      "balancing_entropy": 0.8278,
      "balancing_loss": 1.0939,
      "epoch": 58.62464,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.136,
      "router_z_loss": 0.1901,
      "step": 11490
    },
    {
      "balancing_entropy": 0.77,
      "balancing_loss": 1.1096,
      "epoch": 58.67584,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.1448,
      "router_z_loss": 0.1566,
      "step": 11500
    },
    {
      "balancing_entropy": 0.7829,
      "balancing_loss": 1.0992,
      "epoch": 58.72704,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1497,
      "router_z_loss": 0.1614,
      "step": 11510
    },
    {
      "balancing_entropy": 0.8249,
      "balancing_loss": 1.0754,
      "epoch": 58.77824,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1491,
      "router_z_loss": 0.1809,
      "step": 11520
    },
    {
      "balancing_entropy": 0.8412,
      "balancing_loss": 1.1284,
      "epoch": 58.82944,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.1556,
      "router_z_loss": 0.126,
      "step": 11530
    },
    {
      "balancing_entropy": 0.8112,
      "balancing_loss": 1.0753,
      "epoch": 58.88064,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1448,
      "router_z_loss": 0.1335,
      "step": 11540
    },
    {
      "balancing_entropy": 0.7684,
      "balancing_loss": 1.1103,
      "epoch": 58.93184,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1412,
      "router_z_loss": 0.1685,
      "step": 11550
    },
    {
      "balancing_entropy": 0.7655,
      "balancing_loss": 1.1102,
      "epoch": 58.98304,
      "grad_norm": 0.98046875,
      "learning_rate": 0.003,
      "loss": 2.1433,
      "router_z_loss": 0.1302,
      "step": 11560
    },
    {
      "balancing_entropy": 0.8067,
      "balancing_loss": 1.076,
      "epoch": 59.03072,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1305,
      "router_z_loss": 0.1717,
      "step": 11570
    },
    {
      "balancing_entropy": 0.7927,
      "balancing_loss": 1.1028,
      "epoch": 59.08192,
      "grad_norm": 0.60546875,
      "learning_rate": 0.003,
      "loss": 2.1372,
      "router_z_loss": 0.1098,
      "step": 11580
    },
    {
      "balancing_entropy": 0.8379,
      "balancing_loss": 1.0784,
      "epoch": 59.13312,
      "grad_norm": 0.671875,
      "learning_rate": 0.003,
      "loss": 2.1307,
      "router_z_loss": 0.1594,
      "step": 11590
    },
    {
      "balancing_entropy": 0.7207,
      "balancing_loss": 1.1481,
      "epoch": 59.18432,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.1357,
      "router_z_loss": 0.1633,
      "step": 11600
    },
    {
      "balancing_entropy": 0.8061,
      "balancing_loss": 1.0673,
      "epoch": 59.23552,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.1501,
      "router_z_loss": 0.1588,
      "step": 11610
    },
    {
      "balancing_entropy": 0.7504,
      "balancing_loss": 1.1257,
      "epoch": 59.28672,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1392,
      "router_z_loss": 0.1536,
      "step": 11620
    },
    {
      "balancing_entropy": 0.8268,
      "balancing_loss": 1.1033,
      "epoch": 59.33792,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.1351,
      "router_z_loss": 0.106,
      "step": 11630
    },
    {
      "balancing_entropy": 0.8315,
      "balancing_loss": 1.0767,
      "epoch": 59.38912,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.1279,
      "router_z_loss": 0.1177,
      "step": 11640
    },
    {
      "balancing_entropy": 0.8167,
      "balancing_loss": 1.1149,
      "epoch": 59.44032,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.1284,
      "router_z_loss": 0.1387,
      "step": 11650
    },
    {
      "balancing_entropy": 0.8283,
      "balancing_loss": 1.0983,
      "epoch": 59.49152,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.1323,
      "router_z_loss": 0.0944,
      "step": 11660
    },
    {
      "balancing_entropy": 0.7664,
      "balancing_loss": 1.1094,
      "epoch": 59.54272,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1356,
      "router_z_loss": 0.1254,
      "step": 11670
    },
    {
      "balancing_entropy": 0.8004,
      "balancing_loss": 1.0737,
      "epoch": 59.59392,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.1344,
      "router_z_loss": 0.1773,
      "step": 11680
    },
    {
      "balancing_entropy": 0.757,
      "balancing_loss": 1.1204,
      "epoch": 59.64512,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1421,
      "router_z_loss": 0.15,
      "step": 11690
    },
    {
      "balancing_entropy": 0.7622,
      "balancing_loss": 1.1069,
      "epoch": 59.69632,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1576,
      "router_z_loss": 0.1483,
      "step": 11700
    },
    {
      "balancing_entropy": 0.7808,
      "balancing_loss": 1.084,
      "epoch": 59.74752,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1483,
      "router_z_loss": 0.1338,
      "step": 11710
    },
    {
      "balancing_entropy": 0.8292,
      "balancing_loss": 1.1218,
      "epoch": 59.79872,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.1403,
      "router_z_loss": 0.1524,
      "step": 11720
    },
    {
      "balancing_entropy": 0.7979,
      "balancing_loss": 1.0748,
      "epoch": 59.84992,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.1442,
      "router_z_loss": 0.1857,
      "step": 11730
    },
    {
      "balancing_entropy": 0.8145,
      "balancing_loss": 1.108,
      "epoch": 59.90112,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1603,
      "router_z_loss": 0.1646,
      "step": 11740
    },
    {
      "balancing_entropy": 0.8151,
      "balancing_loss": 1.0698,
      "epoch": 59.95232,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1434,
      "router_z_loss": 0.1552,
      "step": 11750
    },
    {
      "balancing_entropy": 0.8121,
      "balancing_loss": 1.1238,
      "epoch": 60.0,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.1401,
      "router_z_loss": 0.1359,
      "step": 11760
    },
    {
      "balancing_entropy": 0.7799,
      "balancing_loss": 1.0908,
      "epoch": 60.0512,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.1193,
      "router_z_loss": 0.1596,
      "step": 11770
    },
    {
      "balancing_entropy": 0.8573,
      "balancing_loss": 1.0977,
      "epoch": 60.1024,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.1203,
      "router_z_loss": 0.1146,
      "step": 11780
    },
    {
      "balancing_entropy": 0.866,
      "balancing_loss": 1.1075,
      "epoch": 60.1536,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1257,
      "router_z_loss": 0.091,
      "step": 11790
    },
    {
      "balancing_entropy": 0.797,
      "balancing_loss": 1.0783,
      "epoch": 60.2048,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.125,
      "router_z_loss": 0.1498,
      "step": 11800
    },
    {
      "balancing_entropy": 0.7731,
      "balancing_loss": 1.1108,
      "epoch": 60.256,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.1256,
      "router_z_loss": 0.1369,
      "step": 11810
    },
    {
      "balancing_entropy": 0.7944,
      "balancing_loss": 1.0861,
      "epoch": 60.3072,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.1306,
      "router_z_loss": 0.1268,
      "step": 11820
    },
    {
      "balancing_entropy": 0.8591,
      "balancing_loss": 1.1144,
      "epoch": 60.3584,
      "grad_norm": 0.44140625,
      "learning_rate": 0.003,
      "loss": 2.1337,
      "router_z_loss": 0.1256,
      "step": 11830
    },
    {
      "balancing_entropy": 0.8552,
      "balancing_loss": 1.1163,
      "epoch": 60.4096,
      "grad_norm": 0.62109375,
      "learning_rate": 0.003,
      "loss": 2.1307,
      "router_z_loss": 0.1035,
      "step": 11840
    },
    {
      "balancing_entropy": 0.8312,
      "balancing_loss": 1.1069,
      "epoch": 60.4608,
      "grad_norm": 0.34375,
      "learning_rate": 0.003,
      "loss": 2.132,
      "router_z_loss": 0.1325,
      "step": 11850
    },
    {
      "balancing_entropy": 0.8307,
      "balancing_loss": 1.071,
      "epoch": 60.512,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1321,
      "router_z_loss": 0.093,
      "step": 11860
    },
    {
      "balancing_entropy": 0.7843,
      "balancing_loss": 1.1241,
      "epoch": 60.5632,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.1308,
      "router_z_loss": 0.1096,
      "step": 11870
    },
    {
      "balancing_entropy": 0.781,
      "balancing_loss": 1.0988,
      "epoch": 60.6144,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.1265,
      "router_z_loss": 0.1735,
      "step": 11880
    },
    {
      "balancing_entropy": 0.8251,
      "balancing_loss": 1.0809,
      "epoch": 60.6656,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1301,
      "router_z_loss": 0.1363,
      "step": 11890
    },
    {
      "balancing_entropy": 0.7819,
      "balancing_loss": 1.1178,
      "epoch": 60.7168,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.1306,
      "router_z_loss": 0.2279,
      "step": 11900
    },
    {
      "balancing_entropy": 0.7502,
      "balancing_loss": 1.1284,
      "epoch": 60.768,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1303,
      "router_z_loss": 0.1696,
      "step": 11910
    },
    {
      "balancing_entropy": 0.8349,
      "balancing_loss": 1.0769,
      "epoch": 60.8192,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1412,
      "router_z_loss": 0.1404,
      "step": 11920
    },
    {
      "balancing_entropy": 0.7261,
      "balancing_loss": 1.1296,
      "epoch": 60.8704,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.1398,
      "router_z_loss": 0.1591,
      "step": 11930
    },
    {
      "balancing_entropy": 0.8065,
      "balancing_loss": 1.0872,
      "epoch": 60.9216,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1364,
      "router_z_loss": 0.1459,
      "step": 11940
    },
    {
      "balancing_entropy": 0.8341,
      "balancing_loss": 1.0855,
      "epoch": 60.9728,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.1354,
      "router_z_loss": 0.1114,
      "step": 11950
    },
    {
      "balancing_entropy": 0.7901,
      "balancing_loss": 1.1337,
      "epoch": 61.02048,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.1354,
      "router_z_loss": 0.1382,
      "step": 11960
    },
    {
      "balancing_entropy": 0.7812,
      "balancing_loss": 1.0901,
      "epoch": 61.07168,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.12,
      "router_z_loss": 0.1516,
      "step": 11970
    },
    {
      "balancing_entropy": 0.6942,
      "balancing_loss": 1.1592,
      "epoch": 61.12288,
      "grad_norm": 0.80859375,
      "learning_rate": 0.003,
      "loss": 2.1172,
      "router_z_loss": 0.1844,
      "step": 11980
    },
    {
      "balancing_entropy": 0.7899,
      "balancing_loss": 1.0882,
      "epoch": 61.174080000000004,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.1266,
      "router_z_loss": 0.1669,
      "step": 11990
    },
    {
      "balancing_entropy": 0.8945,
      "balancing_loss": 1.2095,
      "epoch": 61.22528,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.1294,
      "router_z_loss": 0.5637,
      "step": 12000
    },
    {
      "balancing_entropy": 0.9354,
      "balancing_loss": 1.1505,
      "epoch": 61.27648,
      "grad_norm": 0.390625,
      "learning_rate": 0.003,
      "loss": 2.1239,
      "router_z_loss": 0.4785,
      "step": 12010
    },
    {
      "balancing_entropy": 0.9063,
      "balancing_loss": 1.1219,
      "epoch": 61.32768,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1239,
      "router_z_loss": 0.4214,
      "step": 12020
    },
    {
      "balancing_entropy": 0.9663,
      "balancing_loss": 1.1494,
      "epoch": 61.37888,
      "grad_norm": 0.326171875,
      "learning_rate": 0.003,
      "loss": 2.1346,
      "router_z_loss": 0.457,
      "step": 12030
    },
    {
      "balancing_entropy": 0.8892,
      "balancing_loss": 1.1015,
      "epoch": 61.43008,
      "grad_norm": 0.625,
      "learning_rate": 0.003,
      "loss": 2.1219,
      "router_z_loss": 0.4376,
      "step": 12040
    },
    {
      "balancing_entropy": 0.867,
      "balancing_loss": 1.0407,
      "epoch": 61.48128,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.126,
      "router_z_loss": 0.5526,
      "step": 12050
    },
    {
      "balancing_entropy": 0.8274,
      "balancing_loss": 1.049,
      "epoch": 61.53248,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.1397,
      "router_z_loss": 0.577,
      "step": 12060
    },
    {
      "balancing_entropy": 0.8414,
      "balancing_loss": 1.0447,
      "epoch": 61.58368,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.1376,
      "router_z_loss": 0.6027,
      "step": 12070
    },
    {
      "balancing_entropy": 0.8898,
      "balancing_loss": 1.0557,
      "epoch": 61.63488,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1476,
      "router_z_loss": 0.5877,
      "step": 12080
    },
    {
      "balancing_entropy": 0.8192,
      "balancing_loss": 1.0866,
      "epoch": 61.68608,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1624,
      "router_z_loss": 0.5837,
      "step": 12090
    },
    {
      "balancing_entropy": 0.8549,
      "balancing_loss": 1.0874,
      "epoch": 61.73728,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1434,
      "router_z_loss": 0.5779,
      "step": 12100
    },
    {
      "balancing_entropy": 0.857,
      "balancing_loss": 1.0953,
      "epoch": 61.78848,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.146,
      "router_z_loss": 0.5782,
      "step": 12110
    },
    {
      "balancing_entropy": 0.8956,
      "balancing_loss": 1.1414,
      "epoch": 61.83968,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.151,
      "router_z_loss": 0.4744,
      "step": 12120
    },
    {
      "balancing_entropy": 0.8658,
      "balancing_loss": 1.1032,
      "epoch": 61.89088,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.139,
      "router_z_loss": 0.2209,
      "step": 12130
    },
    {
      "balancing_entropy": 0.8123,
      "balancing_loss": 1.1081,
      "epoch": 61.94208,
      "grad_norm": 0.765625,
      "learning_rate": 0.003,
      "loss": 2.1337,
      "router_z_loss": 0.1843,
      "step": 12140
    },
    {
      "balancing_entropy": 0.7398,
      "balancing_loss": 1.1115,
      "epoch": 61.99328,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.1491,
      "router_z_loss": 0.1914,
      "step": 12150
    },
    {
      "balancing_entropy": 0.7399,
      "balancing_loss": 1.0869,
      "epoch": 62.04096,
      "grad_norm": 1.109375,
      "learning_rate": 0.003,
      "loss": 2.1339,
      "router_z_loss": 0.1284,
      "step": 12160
    },
    {
      "balancing_entropy": 0.7923,
      "balancing_loss": 1.087,
      "epoch": 62.09216,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.1547,
      "router_z_loss": 0.2009,
      "step": 12170
    },
    {
      "balancing_entropy": 0.8093,
      "balancing_loss": 1.0532,
      "epoch": 62.14336,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.131,
      "router_z_loss": 0.2127,
      "step": 12180
    },
    {
      "balancing_entropy": 0.8353,
      "balancing_loss": 1.149,
      "epoch": 62.19456,
      "grad_norm": 1.1953125,
      "learning_rate": 0.003,
      "loss": 2.131,
      "router_z_loss": 0.1667,
      "step": 12190
    },
    {
      "balancing_entropy": 0.8044,
      "balancing_loss": 1.0958,
      "epoch": 62.24576,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.1288,
      "router_z_loss": 0.1148,
      "step": 12200
    },
    {
      "balancing_entropy": 0.8028,
      "balancing_loss": 1.0978,
      "epoch": 62.29696,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.1245,
      "router_z_loss": 0.1169,
      "step": 12210
    },
    {
      "balancing_entropy": 0.7632,
      "balancing_loss": 1.1201,
      "epoch": 62.34816,
      "grad_norm": 1.328125,
      "learning_rate": 0.003,
      "loss": 2.1206,
      "router_z_loss": 0.1198,
      "step": 12220
    },
    {
      "balancing_entropy": 0.787,
      "balancing_loss": 1.083,
      "epoch": 62.39936,
      "grad_norm": 0.6875,
      "learning_rate": 0.003,
      "loss": 2.1249,
      "router_z_loss": 0.1059,
      "step": 12230
    },
    {
      "balancing_entropy": 0.8377,
      "balancing_loss": 1.0947,
      "epoch": 62.45056,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1339,
      "router_z_loss": 0.1207,
      "step": 12240
    },
    {
      "balancing_entropy": 0.784,
      "balancing_loss": 1.1169,
      "epoch": 62.50176,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.142,
      "router_z_loss": 0.125,
      "step": 12250
    },
    {
      "balancing_entropy": 0.812,
      "balancing_loss": 1.0925,
      "epoch": 62.55296,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1449,
      "router_z_loss": 0.1107,
      "step": 12260
    },
    {
      "balancing_entropy": 0.7604,
      "balancing_loss": 1.1206,
      "epoch": 62.60416,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.1408,
      "router_z_loss": 0.1343,
      "step": 12270
    },
    {
      "balancing_entropy": 0.8027,
      "balancing_loss": 1.0893,
      "epoch": 62.65536,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1339,
      "router_z_loss": 0.125,
      "step": 12280
    },
    {
      "balancing_entropy": 0.7918,
      "balancing_loss": 1.0937,
      "epoch": 62.70656,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.1324,
      "router_z_loss": 0.1042,
      "step": 12290
    },
    {
      "balancing_entropy": 0.7459,
      "balancing_loss": 1.1305,
      "epoch": 62.75776,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.1365,
      "router_z_loss": 0.1359,
      "step": 12300
    },
    {
      "balancing_entropy": 0.7932,
      "balancing_loss": 1.1024,
      "epoch": 62.80896,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.1241,
      "router_z_loss": 0.2438,
      "step": 12310
    },
    {
      "balancing_entropy": 0.7717,
      "balancing_loss": 1.1529,
      "epoch": 62.86016,
      "grad_norm": 0.94140625,
      "learning_rate": 0.003,
      "loss": 2.1296,
      "router_z_loss": 0.172,
      "step": 12320
    },
    {
      "balancing_entropy": 0.8085,
      "balancing_loss": 1.081,
      "epoch": 62.91136,
      "grad_norm": 0.7265625,
      "learning_rate": 0.003,
      "loss": 2.134,
      "router_z_loss": 0.1406,
      "step": 12330
    },
    {
      "balancing_entropy": 0.7861,
      "balancing_loss": 1.0915,
      "epoch": 62.96256,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.1336,
      "router_z_loss": 0.1102,
      "step": 12340
    },
    {
      "balancing_entropy": 0.7165,
      "balancing_loss": 1.1257,
      "epoch": 63.01024,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.1307,
      "router_z_loss": 0.1116,
      "step": 12350
    },
    {
      "balancing_entropy": 0.7973,
      "balancing_loss": 1.0819,
      "epoch": 63.06144,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.1202,
      "router_z_loss": 0.1419,
      "step": 12360
    },
    {
      "balancing_entropy": 0.7915,
      "balancing_loss": 1.0943,
      "epoch": 63.11264,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.1274,
      "router_z_loss": 0.15,
      "step": 12370
    },
    {
      "balancing_entropy": 0.7712,
      "balancing_loss": 1.0988,
      "epoch": 63.16384,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1366,
      "router_z_loss": 0.1539,
      "step": 12380
    },
    {
      "balancing_entropy": 0.7207,
      "balancing_loss": 1.1609,
      "epoch": 63.21504,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1329,
      "router_z_loss": 0.1443,
      "step": 12390
    },
    {
      "balancing_entropy": 0.8396,
      "balancing_loss": 1.0762,
      "epoch": 63.26624,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.1319,
      "router_z_loss": 0.1907,
      "step": 12400
    },
    {
      "balancing_entropy": 0.7921,
      "balancing_loss": 1.0865,
      "epoch": 63.31744,
      "grad_norm": 0.431640625,
      "learning_rate": 0.003,
      "loss": 2.1226,
      "router_z_loss": 0.1883,
      "step": 12410
    },
    {
      "balancing_entropy": 0.7812,
      "balancing_loss": 1.1347,
      "epoch": 63.36864,
      "grad_norm": 0.67578125,
      "learning_rate": 0.003,
      "loss": 2.1195,
      "router_z_loss": 0.14,
      "step": 12420
    },
    {
      "balancing_entropy": 0.8457,
      "balancing_loss": 1.0861,
      "epoch": 63.41984,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.1342,
      "router_z_loss": 0.1465,
      "step": 12430
    },
    {
      "balancing_entropy": 0.7789,
      "balancing_loss": 1.0927,
      "epoch": 63.47104,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.1219,
      "router_z_loss": 0.1054,
      "step": 12440
    },
    {
      "balancing_entropy": 0.729,
      "balancing_loss": 1.1501,
      "epoch": 63.52224,
      "grad_norm": 0.9296875,
      "learning_rate": 0.003,
      "loss": 2.1345,
      "router_z_loss": 0.1625,
      "step": 12450
    },
    {
      "balancing_entropy": 0.8039,
      "balancing_loss": 1.0965,
      "epoch": 63.57344,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.1276,
      "router_z_loss": 0.0976,
      "step": 12460
    },
    {
      "balancing_entropy": 0.8475,
      "balancing_loss": 1.085,
      "epoch": 63.62464,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.1255,
      "router_z_loss": 0.0831,
      "step": 12470
    },
    {
      "balancing_entropy": 0.8775,
      "balancing_loss": 1.1083,
      "epoch": 63.67584,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.129,
      "router_z_loss": 0.0774,
      "step": 12480
    },
    {
      "balancing_entropy": 0.8237,
      "balancing_loss": 1.1217,
      "epoch": 63.72704,
      "grad_norm": 0.671875,
      "learning_rate": 0.003,
      "loss": 2.1304,
      "router_z_loss": 0.0699,
      "step": 12490
    },
    {
      "balancing_entropy": 0.824,
      "balancing_loss": 1.0962,
      "epoch": 63.77824,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1242,
      "router_z_loss": 0.0683,
      "step": 12500
    },
    {
      "balancing_entropy": 0.8081,
      "balancing_loss": 1.0818,
      "epoch": 63.82944,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.1248,
      "router_z_loss": 0.1269,
      "step": 12510
    },
    {
      "balancing_entropy": 0.8791,
      "balancing_loss": 1.1225,
      "epoch": 63.88064,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1287,
      "router_z_loss": 0.0942,
      "step": 12520
    },
    {
      "balancing_entropy": 0.7929,
      "balancing_loss": 1.094,
      "epoch": 63.93184,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1302,
      "router_z_loss": 0.106,
      "step": 12530
    },
    {
      "balancing_entropy": 0.7889,
      "balancing_loss": 1.0869,
      "epoch": 63.98304,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.1284,
      "router_z_loss": 0.1145,
      "step": 12540
    },
    {
      "balancing_entropy": 0.7635,
      "balancing_loss": 1.0917,
      "epoch": 64.03072,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.125,
      "router_z_loss": 0.1765,
      "step": 12550
    },
    {
      "balancing_entropy": 0.8104,
      "balancing_loss": 1.127,
      "epoch": 64.08192,
      "grad_norm": 0.49609375,
      "learning_rate": 0.003,
      "loss": 2.124,
      "router_z_loss": 0.1235,
      "step": 12560
    },
    {
      "balancing_entropy": 0.7712,
      "balancing_loss": 1.1223,
      "epoch": 64.13312,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1229,
      "router_z_loss": 0.098,
      "step": 12570
    },
    {
      "balancing_entropy": 0.7995,
      "balancing_loss": 1.0801,
      "epoch": 64.18432,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1166,
      "router_z_loss": 0.1598,
      "step": 12580
    },
    {
      "balancing_entropy": 0.8097,
      "balancing_loss": 1.082,
      "epoch": 64.23552,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.1137,
      "router_z_loss": 0.2422,
      "step": 12590
    },
    {
      "balancing_entropy": 0.6762,
      "balancing_loss": 1.171,
      "epoch": 64.28672,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.1237,
      "router_z_loss": 0.1457,
      "step": 12600
    },
    {
      "balancing_entropy": 0.8014,
      "balancing_loss": 1.0956,
      "epoch": 64.33792,
      "grad_norm": 0.99609375,
      "learning_rate": 0.003,
      "loss": 2.1286,
      "router_z_loss": 0.1232,
      "step": 12610
    },
    {
      "balancing_entropy": 0.7945,
      "balancing_loss": 1.0876,
      "epoch": 64.38912,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1175,
      "router_z_loss": 0.1072,
      "step": 12620
    },
    {
      "balancing_entropy": 0.7431,
      "balancing_loss": 1.1485,
      "epoch": 64.44032,
      "grad_norm": 0.9921875,
      "learning_rate": 0.003,
      "loss": 2.1195,
      "router_z_loss": 0.1547,
      "step": 12630
    },
    {
      "balancing_entropy": 0.7958,
      "balancing_loss": 1.0858,
      "epoch": 64.49152,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.1242,
      "router_z_loss": 0.1582,
      "step": 12640
    },
    {
      "balancing_entropy": 0.8049,
      "balancing_loss": 1.082,
      "epoch": 64.54272,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.1277,
      "router_z_loss": 0.1423,
      "step": 12650
    },
    {
      "balancing_entropy": 0.7932,
      "balancing_loss": 1.0912,
      "epoch": 64.59392,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1351,
      "router_z_loss": 0.0974,
      "step": 12660
    },
    {
      "balancing_entropy": 0.8217,
      "balancing_loss": 1.1223,
      "epoch": 64.64512,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1242,
      "router_z_loss": 0.0936,
      "step": 12670
    },
    {
      "balancing_entropy": 0.8492,
      "balancing_loss": 1.0849,
      "epoch": 64.69632,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1207,
      "router_z_loss": 0.09,
      "step": 12680
    },
    {
      "balancing_entropy": 0.7906,
      "balancing_loss": 1.0927,
      "epoch": 64.74752,
      "grad_norm": 0.7578125,
      "learning_rate": 0.003,
      "loss": 2.1169,
      "router_z_loss": 0.069,
      "step": 12690
    },
    {
      "balancing_entropy": 0.7638,
      "balancing_loss": 1.1192,
      "epoch": 64.79872,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1271,
      "router_z_loss": 0.1087,
      "step": 12700
    },
    {
      "balancing_entropy": 0.7602,
      "balancing_loss": 1.1016,
      "epoch": 64.84992,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1281,
      "router_z_loss": 0.1392,
      "step": 12710
    },
    {
      "balancing_entropy": 0.7936,
      "balancing_loss": 1.0757,
      "epoch": 64.90112,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1287,
      "router_z_loss": 0.1748,
      "step": 12720
    },
    {
      "balancing_entropy": 0.7268,
      "balancing_loss": 1.1233,
      "epoch": 64.95232,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.137,
      "router_z_loss": 0.1911,
      "step": 12730
    },
    {
      "balancing_entropy": 0.7596,
      "balancing_loss": 1.1072,
      "epoch": 65.0,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.1315,
      "router_z_loss": 0.2178,
      "step": 12740
    },
    {
      "balancing_entropy": 0.7735,
      "balancing_loss": 1.1293,
      "epoch": 65.0512,
      "grad_norm": 0.8046875,
      "learning_rate": 0.003,
      "loss": 2.1339,
      "router_z_loss": 0.1808,
      "step": 12750
    },
    {
      "balancing_entropy": 0.8148,
      "balancing_loss": 1.1277,
      "epoch": 65.1024,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1317,
      "router_z_loss": 0.1275,
      "step": 12760
    },
    {
      "balancing_entropy": 0.8123,
      "balancing_loss": 1.0762,
      "epoch": 65.1536,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.1183,
      "router_z_loss": 0.1451,
      "step": 12770
    },
    {
      "balancing_entropy": 0.7634,
      "balancing_loss": 1.1237,
      "epoch": 65.2048,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1059,
      "router_z_loss": 0.1257,
      "step": 12780
    },
    {
      "balancing_entropy": 0.7872,
      "balancing_loss": 1.0964,
      "epoch": 65.256,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.1102,
      "router_z_loss": 0.1366,
      "step": 12790
    },
    {
      "balancing_entropy": 0.8046,
      "balancing_loss": 1.0769,
      "epoch": 65.3072,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.118,
      "router_z_loss": 0.1431,
      "step": 12800
    },
    {
      "balancing_entropy": 0.7555,
      "balancing_loss": 1.1022,
      "epoch": 65.3584,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1192,
      "router_z_loss": 0.1364,
      "step": 12810
    },
    {
      "balancing_entropy": 0.8013,
      "balancing_loss": 1.0763,
      "epoch": 65.4096,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1225,
      "router_z_loss": 0.1302,
      "step": 12820
    },
    {
      "balancing_entropy": 0.7728,
      "balancing_loss": 1.1369,
      "epoch": 65.4608,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1251,
      "router_z_loss": 0.1581,
      "step": 12830
    },
    {
      "balancing_entropy": 0.7718,
      "balancing_loss": 1.1032,
      "epoch": 65.512,
      "grad_norm": 0.375,
      "learning_rate": 0.003,
      "loss": 2.1254,
      "router_z_loss": 0.1418,
      "step": 12840
    },
    {
      "balancing_entropy": 0.7325,
      "balancing_loss": 1.119,
      "epoch": 65.5632,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.1341,
      "router_z_loss": 0.1482,
      "step": 12850
    },
    {
      "balancing_entropy": 0.7815,
      "balancing_loss": 1.0791,
      "epoch": 65.6144,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.128,
      "router_z_loss": 0.2231,
      "step": 12860
    },
    {
      "balancing_entropy": 0.6913,
      "balancing_loss": 1.1504,
      "epoch": 65.6656,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.136,
      "router_z_loss": 0.2412,
      "step": 12870
    },
    {
      "balancing_entropy": 0.8159,
      "balancing_loss": 1.1258,
      "epoch": 65.7168,
      "grad_norm": 0.7890625,
      "learning_rate": 0.003,
      "loss": 2.1579,
      "router_z_loss": 0.4114,
      "step": 12880
    },
    {
      "balancing_entropy": 0.7323,
      "balancing_loss": 1.1609,
      "epoch": 65.768,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.1337,
      "router_z_loss": 0.136,
      "step": 12890
    },
    {
      "balancing_entropy": 0.7944,
      "balancing_loss": 1.08,
      "epoch": 65.8192,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1278,
      "router_z_loss": 0.2138,
      "step": 12900
    },
    {
      "balancing_entropy": 0.8127,
      "balancing_loss": 1.0678,
      "epoch": 65.8704,
      "grad_norm": 0.76171875,
      "learning_rate": 0.003,
      "loss": 2.1352,
      "router_z_loss": 0.1807,
      "step": 12910
    },
    {
      "balancing_entropy": 0.8055,
      "balancing_loss": 1.0914,
      "epoch": 65.9216,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1353,
      "router_z_loss": 0.0991,
      "step": 12920
    },
    {
      "balancing_entropy": 0.6786,
      "balancing_loss": 1.1766,
      "epoch": 65.9728,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.1292,
      "router_z_loss": 0.1721,
      "step": 12930
    },
    {
      "balancing_entropy": 0.7123,
      "balancing_loss": 1.1096,
      "epoch": 66.02048,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.1245,
      "router_z_loss": 0.128,
      "step": 12940
    },
    {
      "balancing_entropy": 0.7882,
      "balancing_loss": 1.068,
      "epoch": 66.07168,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.1225,
      "router_z_loss": 0.1592,
      "step": 12950
    },
    {
      "balancing_entropy": 0.7662,
      "balancing_loss": 1.1209,
      "epoch": 66.12288,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.1199,
      "router_z_loss": 0.1547,
      "step": 12960
    },
    {
      "balancing_entropy": 0.7748,
      "balancing_loss": 1.0836,
      "epoch": 66.17408,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1236,
      "router_z_loss": 0.1701,
      "step": 12970
    },
    {
      "balancing_entropy": 0.7861,
      "balancing_loss": 1.1413,
      "epoch": 66.22528,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.119,
      "router_z_loss": 0.1881,
      "step": 12980
    },
    {
      "balancing_entropy": 0.7468,
      "balancing_loss": 1.1394,
      "epoch": 66.27648,
      "grad_norm": 0.365234375,
      "learning_rate": 0.003,
      "loss": 2.1166,
      "router_z_loss": 0.0916,
      "step": 12990
    },
    {
      "balancing_entropy": 0.7831,
      "balancing_loss": 1.0707,
      "epoch": 66.32768,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.1198,
      "router_z_loss": 0.1098,
      "step": 13000
    },
    {
      "balancing_entropy": 0.7987,
      "balancing_loss": 1.0664,
      "epoch": 66.37888,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.1234,
      "router_z_loss": 0.1544,
      "step": 13010
    },
    {
      "balancing_entropy": 0.7467,
      "balancing_loss": 1.1066,
      "epoch": 66.43008,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1264,
      "router_z_loss": 0.1378,
      "step": 13020
    },
    {
      "balancing_entropy": 0.7431,
      "balancing_loss": 1.1041,
      "epoch": 66.48128,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1326,
      "router_z_loss": 0.1526,
      "step": 13030
    },
    {
      "balancing_entropy": 0.7939,
      "balancing_loss": 1.0731,
      "epoch": 66.53248,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.1253,
      "router_z_loss": 0.106,
      "step": 13040
    },
    {
      "balancing_entropy": 0.788,
      "balancing_loss": 1.1183,
      "epoch": 66.58368,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1176,
      "router_z_loss": 0.0923,
      "step": 13050
    },
    {
      "balancing_entropy": 0.7938,
      "balancing_loss": 1.0759,
      "epoch": 66.63488,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1142,
      "router_z_loss": 0.1607,
      "step": 13060
    },
    {
      "balancing_entropy": 0.7736,
      "balancing_loss": 1.0841,
      "epoch": 66.68608,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1159,
      "router_z_loss": 0.1638,
      "step": 13070
    },
    {
      "balancing_entropy": 0.7194,
      "balancing_loss": 1.1591,
      "epoch": 66.73728,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.1311,
      "router_z_loss": 0.1256,
      "step": 13080
    },
    {
      "balancing_entropy": 0.7932,
      "balancing_loss": 1.0706,
      "epoch": 66.78848,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.1223,
      "router_z_loss": 0.14,
      "step": 13090
    },
    {
      "balancing_entropy": 0.7585,
      "balancing_loss": 1.1313,
      "epoch": 66.83968,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1236,
      "router_z_loss": 0.1442,
      "step": 13100
    },
    {
      "balancing_entropy": 0.7819,
      "balancing_loss": 1.0825,
      "epoch": 66.89088,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1304,
      "router_z_loss": 0.107,
      "step": 13110
    },
    {
      "balancing_entropy": 0.7736,
      "balancing_loss": 1.1354,
      "epoch": 66.94208,
      "grad_norm": 0.921875,
      "learning_rate": 0.003,
      "loss": 2.1324,
      "router_z_loss": 0.2692,
      "step": 13120
    },
    {
      "balancing_entropy": 0.7838,
      "balancing_loss": 1.0854,
      "epoch": 66.99328,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.1247,
      "router_z_loss": 0.0781,
      "step": 13130
    },
    {
      "balancing_entropy": 0.7572,
      "balancing_loss": 1.1531,
      "epoch": 67.04096,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.1206,
      "router_z_loss": 0.1435,
      "step": 13140
    },
    {
      "balancing_entropy": 0.7804,
      "balancing_loss": 1.0813,
      "epoch": 67.09216,
      "grad_norm": 0.68359375,
      "learning_rate": 0.003,
      "loss": 2.1226,
      "router_z_loss": 0.089,
      "step": 13150
    },
    {
      "balancing_entropy": 0.7787,
      "balancing_loss": 1.0875,
      "epoch": 67.14336,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1187,
      "router_z_loss": 0.1047,
      "step": 13160
    },
    {
      "balancing_entropy": 0.735,
      "balancing_loss": 1.1155,
      "epoch": 67.19456,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.1083,
      "router_z_loss": 0.1362,
      "step": 13170
    },
    {
      "balancing_entropy": 0.7807,
      "balancing_loss": 1.0824,
      "epoch": 67.24576,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.107,
      "router_z_loss": 0.2209,
      "step": 13180
    },
    {
      "balancing_entropy": 0.7155,
      "balancing_loss": 1.1261,
      "epoch": 67.29696,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1186,
      "router_z_loss": 0.1456,
      "step": 13190
    },
    {
      "balancing_entropy": 0.7673,
      "balancing_loss": 1.0883,
      "epoch": 67.34816000000001,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.122,
      "router_z_loss": 0.1883,
      "step": 13200
    },
    {
      "balancing_entropy": 0.7793,
      "balancing_loss": 1.082,
      "epoch": 67.39936,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.1372,
      "router_z_loss": 0.127,
      "step": 13210
    },
    {
      "balancing_entropy": 0.7147,
      "balancing_loss": 1.1265,
      "epoch": 67.45056,
      "grad_norm": 0.76171875,
      "learning_rate": 0.003,
      "loss": 2.1198,
      "router_z_loss": 0.143,
      "step": 13220
    },
    {
      "balancing_entropy": 0.7877,
      "balancing_loss": 1.0791,
      "epoch": 67.50176,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.1267,
      "router_z_loss": 0.1487,
      "step": 13230
    },
    {
      "balancing_entropy": 0.7687,
      "balancing_loss": 1.0889,
      "epoch": 67.55296,
      "grad_norm": 0.48828125,
      "learning_rate": 0.003,
      "loss": 2.1216,
      "router_z_loss": 0.1193,
      "step": 13240
    },
    {
      "balancing_entropy": 0.7962,
      "balancing_loss": 1.1131,
      "epoch": 67.60416,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.1305,
      "router_z_loss": 0.1313,
      "step": 13250
    },
    {
      "balancing_entropy": 0.7714,
      "balancing_loss": 1.0913,
      "epoch": 67.65536,
      "grad_norm": 0.77734375,
      "learning_rate": 0.003,
      "loss": 2.1301,
      "router_z_loss": 0.102,
      "step": 13260
    },
    {
      "balancing_entropy": 0.7744,
      "balancing_loss": 1.0846,
      "epoch": 67.70656,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1206,
      "router_z_loss": 0.115,
      "step": 13270
    },
    {
      "balancing_entropy": 0.6602,
      "balancing_loss": 1.167,
      "epoch": 67.75776,
      "grad_norm": 0.765625,
      "learning_rate": 0.003,
      "loss": 2.1408,
      "router_z_loss": 0.1315,
      "step": 13280
    },
    {
      "balancing_entropy": 0.7873,
      "balancing_loss": 1.0836,
      "epoch": 67.80896,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.1296,
      "router_z_loss": 0.1278,
      "step": 13290
    },
    {
      "balancing_entropy": 0.7512,
      "balancing_loss": 1.1045,
      "epoch": 67.86016,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.1258,
      "router_z_loss": 0.1249,
      "step": 13300
    },
    {
      "balancing_entropy": 0.7763,
      "balancing_loss": 1.0961,
      "epoch": 67.91136,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.1185,
      "router_z_loss": 0.1069,
      "step": 13310
    },
    {
      "balancing_entropy": 0.7728,
      "balancing_loss": 1.0894,
      "epoch": 67.96256,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.1232,
      "router_z_loss": 0.1912,
      "step": 13320
    },
    {
      "balancing_entropy": 0.7129,
      "balancing_loss": 1.1253,
      "epoch": 68.01024,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1254,
      "router_z_loss": 0.1131,
      "step": 13330
    },
    {
      "balancing_entropy": 0.7638,
      "balancing_loss": 1.0957,
      "epoch": 68.06144,
      "grad_norm": 0.57421875,
      "learning_rate": 0.003,
      "loss": 2.1104,
      "router_z_loss": 0.1149,
      "step": 13340
    },
    {
      "balancing_entropy": 0.7615,
      "balancing_loss": 1.0943,
      "epoch": 68.11264,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.1041,
      "router_z_loss": 0.1301,
      "step": 13350
    },
    {
      "balancing_entropy": 0.7092,
      "balancing_loss": 1.1307,
      "epoch": 68.16384,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.1185,
      "router_z_loss": 0.1228,
      "step": 13360
    },
    {
      "balancing_entropy": 0.7663,
      "balancing_loss": 1.0954,
      "epoch": 68.21504,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.121,
      "router_z_loss": 0.1847,
      "step": 13370
    },
    {
      "balancing_entropy": 0.7772,
      "balancing_loss": 1.0926,
      "epoch": 68.26624,
      "grad_norm": 0.85546875,
      "learning_rate": 0.003,
      "loss": 2.1321,
      "router_z_loss": 0.1325,
      "step": 13380
    },
    {
      "balancing_entropy": 0.7759,
      "balancing_loss": 1.0929,
      "epoch": 68.31744,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1146,
      "router_z_loss": 0.0921,
      "step": 13390
    },
    {
      "balancing_entropy": 0.719,
      "balancing_loss": 1.1239,
      "epoch": 68.36864,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1092,
      "router_z_loss": 0.1472,
      "step": 13400
    },
    {
      "balancing_entropy": 0.7968,
      "balancing_loss": 1.0867,
      "epoch": 68.41984,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.1072,
      "router_z_loss": 0.0981,
      "step": 13410
    },
    {
      "balancing_entropy": 0.779,
      "balancing_loss": 1.0937,
      "epoch": 68.47104,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.1088,
      "router_z_loss": 0.0793,
      "step": 13420
    },
    {
      "balancing_entropy": 0.7794,
      "balancing_loss": 1.1224,
      "epoch": 68.52224,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.1177,
      "router_z_loss": 0.0928,
      "step": 13430
    },
    {
      "balancing_entropy": 0.7692,
      "balancing_loss": 1.0939,
      "epoch": 68.57344,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1099,
      "router_z_loss": 0.1198,
      "step": 13440
    },
    {
      "balancing_entropy": 0.678,
      "balancing_loss": 1.1618,
      "epoch": 68.62464,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.1326,
      "router_z_loss": 0.2226,
      "step": 13450
    },
    {
      "balancing_entropy": 0.7831,
      "balancing_loss": 1.0919,
      "epoch": 68.67584,
      "grad_norm": 0.7421875,
      "learning_rate": 0.003,
      "loss": 2.131,
      "router_z_loss": 0.1277,
      "step": 13460
    },
    {
      "balancing_entropy": 0.779,
      "balancing_loss": 1.0799,
      "epoch": 68.72704,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.1229,
      "router_z_loss": 0.1136,
      "step": 13470
    },
    {
      "balancing_entropy": 0.7765,
      "balancing_loss": 1.1504,
      "epoch": 68.77824,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.1296,
      "router_z_loss": 0.1617,
      "step": 13480
    },
    {
      "balancing_entropy": 0.7908,
      "balancing_loss": 1.0726,
      "epoch": 68.82944,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.1445,
      "router_z_loss": 0.1197,
      "step": 13490
    },
    {
      "balancing_entropy": 0.758,
      "balancing_loss": 1.1495,
      "epoch": 68.88064,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.141,
      "router_z_loss": 0.1507,
      "step": 13500
    },
    {
      "balancing_entropy": 0.7871,
      "balancing_loss": 1.0719,
      "epoch": 68.93184,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.119,
      "router_z_loss": 0.2216,
      "step": 13510
    },
    {
      "balancing_entropy": 0.783,
      "balancing_loss": 1.1053,
      "epoch": 68.98304,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1209,
      "router_z_loss": 0.1149,
      "step": 13520
    },
    {
      "balancing_entropy": 0.7407,
      "balancing_loss": 1.081,
      "epoch": 69.03072,
      "grad_norm": 0.6328125,
      "learning_rate": 0.003,
      "loss": 2.1136,
      "router_z_loss": 0.159,
      "step": 13530
    },
    {
      "balancing_entropy": 0.7083,
      "balancing_loss": 1.1155,
      "epoch": 69.08192,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.1123,
      "router_z_loss": 0.1593,
      "step": 13540
    },
    {
      "balancing_entropy": 0.7641,
      "balancing_loss": 1.0791,
      "epoch": 69.13312,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.1088,
      "router_z_loss": 0.1586,
      "step": 13550
    },
    {
      "balancing_entropy": 0.7857,
      "balancing_loss": 1.1054,
      "epoch": 69.18432,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.11,
      "router_z_loss": 0.1365,
      "step": 13560
    },
    {
      "balancing_entropy": 0.7711,
      "balancing_loss": 1.0784,
      "epoch": 69.23552,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.1115,
      "router_z_loss": 0.1518,
      "step": 13570
    },
    {
      "balancing_entropy": 0.7877,
      "balancing_loss": 1.1134,
      "epoch": 69.28672,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1178,
      "router_z_loss": 0.1147,
      "step": 13580
    },
    {
      "balancing_entropy": 0.7717,
      "balancing_loss": 1.0822,
      "epoch": 69.33792,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1145,
      "router_z_loss": 0.1494,
      "step": 13590
    },
    {
      "balancing_entropy": 0.6599,
      "balancing_loss": 1.1547,
      "epoch": 69.38912,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.1154,
      "router_z_loss": 0.1546,
      "step": 13600
    },
    {
      "balancing_entropy": 0.7652,
      "balancing_loss": 1.0844,
      "epoch": 69.44032,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.1141,
      "router_z_loss": 0.1611,
      "step": 13610
    },
    {
      "balancing_entropy": 0.7755,
      "balancing_loss": 1.0864,
      "epoch": 69.49152,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.127,
      "router_z_loss": 0.1602,
      "step": 13620
    },
    {
      "balancing_entropy": 0.7939,
      "balancing_loss": 1.0936,
      "epoch": 69.54272,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1291,
      "router_z_loss": 0.1179,
      "step": 13630
    },
    {
      "balancing_entropy": 0.7894,
      "balancing_loss": 1.0885,
      "epoch": 69.59392,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.112,
      "router_z_loss": 0.089,
      "step": 13640
    },
    {
      "balancing_entropy": 0.6834,
      "balancing_loss": 1.1392,
      "epoch": 69.64512,
      "grad_norm": 0.8046875,
      "learning_rate": 0.003,
      "loss": 2.113,
      "router_z_loss": 0.1229,
      "step": 13650
    },
    {
      "balancing_entropy": 0.7826,
      "balancing_loss": 1.0974,
      "epoch": 69.69632,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1166,
      "router_z_loss": 0.1239,
      "step": 13660
    },
    {
      "balancing_entropy": 0.6992,
      "balancing_loss": 1.138,
      "epoch": 69.74752,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1155,
      "router_z_loss": 0.1064,
      "step": 13670
    },
    {
      "balancing_entropy": 0.7583,
      "balancing_loss": 1.0952,
      "epoch": 69.79872,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1142,
      "router_z_loss": 0.1158,
      "step": 13680
    },
    {
      "balancing_entropy": 0.7133,
      "balancing_loss": 1.1298,
      "epoch": 69.84992,
      "grad_norm": 0.73828125,
      "learning_rate": 0.003,
      "loss": 2.1223,
      "router_z_loss": 0.1725,
      "step": 13690
    },
    {
      "balancing_entropy": 0.7565,
      "balancing_loss": 1.1088,
      "epoch": 69.90112,
      "grad_norm": 0.55078125,
      "learning_rate": 0.003,
      "loss": 2.1225,
      "router_z_loss": 0.2619,
      "step": 13700
    },
    {
      "balancing_entropy": 0.7465,
      "balancing_loss": 1.1728,
      "epoch": 69.95232,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.1218,
      "router_z_loss": 0.1693,
      "step": 13710
    },
    {
      "balancing_entropy": 0.7643,
      "balancing_loss": 1.0947,
      "epoch": 70.0,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.119,
      "router_z_loss": 0.1126,
      "step": 13720
    },
    {
      "balancing_entropy": 0.7671,
      "balancing_loss": 1.0947,
      "epoch": 70.0512,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1032,
      "router_z_loss": 0.1113,
      "step": 13730
    },
    {
      "balancing_entropy": 0.7764,
      "balancing_loss": 1.0827,
      "epoch": 70.1024,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1011,
      "router_z_loss": 0.107,
      "step": 13740
    },
    {
      "balancing_entropy": 0.7541,
      "balancing_loss": 1.0949,
      "epoch": 70.1536,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1102,
      "router_z_loss": 0.1366,
      "step": 13750
    },
    {
      "balancing_entropy": 0.6952,
      "balancing_loss": 1.1319,
      "epoch": 70.2048,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1325,
      "router_z_loss": 0.1055,
      "step": 13760
    },
    {
      "balancing_entropy": 0.7643,
      "balancing_loss": 1.0876,
      "epoch": 70.256,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.1256,
      "router_z_loss": 0.1365,
      "step": 13770
    },
    {
      "balancing_entropy": 0.7562,
      "balancing_loss": 1.0942,
      "epoch": 70.3072,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1096,
      "router_z_loss": 0.1339,
      "step": 13780
    },
    {
      "balancing_entropy": 0.7063,
      "balancing_loss": 1.129,
      "epoch": 70.3584,
      "grad_norm": 0.83203125,
      "learning_rate": 0.003,
      "loss": 2.1054,
      "router_z_loss": 0.2049,
      "step": 13790
    },
    {
      "balancing_entropy": 0.7689,
      "balancing_loss": 1.084,
      "epoch": 70.4096,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.1226,
      "router_z_loss": 0.1785,
      "step": 13800
    },
    {
      "balancing_entropy": 0.7768,
      "balancing_loss": 1.0862,
      "epoch": 70.4608,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1159,
      "router_z_loss": 0.1655,
      "step": 13810
    },
    {
      "balancing_entropy": 0.7644,
      "balancing_loss": 1.094,
      "epoch": 70.512,
      "grad_norm": 0.44140625,
      "learning_rate": 0.003,
      "loss": 2.1161,
      "router_z_loss": 0.1351,
      "step": 13820
    },
    {
      "balancing_entropy": 0.6672,
      "balancing_loss": 1.1617,
      "epoch": 70.5632,
      "grad_norm": 0.81640625,
      "learning_rate": 0.003,
      "loss": 2.1472,
      "router_z_loss": 0.1269,
      "step": 13830
    },
    {
      "balancing_entropy": 0.758,
      "balancing_loss": 1.0961,
      "epoch": 70.6144,
      "grad_norm": 0.73828125,
      "learning_rate": 0.003,
      "loss": 2.1342,
      "router_z_loss": 0.0794,
      "step": 13840
    },
    {
      "balancing_entropy": 0.7739,
      "balancing_loss": 1.0839,
      "epoch": 70.6656,
      "grad_norm": 0.486328125,
      "learning_rate": 0.003,
      "loss": 2.1134,
      "router_z_loss": 0.1126,
      "step": 13850
    },
    {
      "balancing_entropy": 0.767,
      "balancing_loss": 1.0949,
      "epoch": 70.7168,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1147,
      "router_z_loss": 0.0738,
      "step": 13860
    },
    {
      "balancing_entropy": 0.6993,
      "balancing_loss": 1.1286,
      "epoch": 70.768,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1135,
      "router_z_loss": 0.1171,
      "step": 13870
    },
    {
      "balancing_entropy": 0.7575,
      "balancing_loss": 1.0947,
      "epoch": 70.8192,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1119,
      "router_z_loss": 0.1331,
      "step": 13880
    },
    {
      "balancing_entropy": 0.7769,
      "balancing_loss": 1.0869,
      "epoch": 70.8704,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1197,
      "router_z_loss": 0.1086,
      "step": 13890
    },
    {
      "balancing_entropy": 0.7543,
      "balancing_loss": 1.0955,
      "epoch": 70.9216,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.1172,
      "router_z_loss": 0.0992,
      "step": 13900
    },
    {
      "balancing_entropy": 0.7607,
      "balancing_loss": 1.0891,
      "epoch": 70.9728,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.1141,
      "router_z_loss": 0.1561,
      "step": 13910
    },
    {
      "balancing_entropy": 0.6898,
      "balancing_loss": 1.1342,
      "epoch": 71.02048,
      "grad_norm": 0.609375,
      "learning_rate": 0.003,
      "loss": 2.1149,
      "router_z_loss": 0.1464,
      "step": 13920
    },
    {
      "balancing_entropy": 0.7832,
      "balancing_loss": 1.0851,
      "epoch": 71.07168,
      "grad_norm": 0.86328125,
      "learning_rate": 0.003,
      "loss": 2.1167,
      "router_z_loss": 0.1182,
      "step": 13930
    },
    {
      "balancing_entropy": 0.7143,
      "balancing_loss": 1.1283,
      "epoch": 71.12288,
      "grad_norm": 0.921875,
      "learning_rate": 0.003,
      "loss": 2.124,
      "router_z_loss": 0.0939,
      "step": 13940
    },
    {
      "balancing_entropy": 0.7768,
      "balancing_loss": 1.0831,
      "epoch": 71.17408,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.1031,
      "router_z_loss": 0.1352,
      "step": 13950
    },
    {
      "balancing_entropy": 0.7674,
      "balancing_loss": 1.0842,
      "epoch": 71.22528,
      "grad_norm": 0.466796875,
      "learning_rate": 0.003,
      "loss": 2.0992,
      "router_z_loss": 0.1616,
      "step": 13960
    },
    {
      "balancing_entropy": 0.7036,
      "balancing_loss": 1.1286,
      "epoch": 71.27648,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1107,
      "router_z_loss": 0.1185,
      "step": 13970
    },
    {
      "balancing_entropy": 0.7963,
      "balancing_loss": 1.0832,
      "epoch": 71.32768,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1106,
      "router_z_loss": 0.1076,
      "step": 13980
    },
    {
      "balancing_entropy": 0.793,
      "balancing_loss": 1.0875,
      "epoch": 71.37888,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.1055,
      "router_z_loss": 0.0788,
      "step": 13990
    },
    {
      "balancing_entropy": 0.7324,
      "balancing_loss": 1.1213,
      "epoch": 71.43008,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.1175,
      "router_z_loss": 0.0795,
      "step": 14000
    },
    {
      "balancing_entropy": 0.769,
      "balancing_loss": 1.0819,
      "epoch": 71.48128,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1056,
      "router_z_loss": 0.13,
      "step": 14010
    },
    {
      "balancing_entropy": 0.7692,
      "balancing_loss": 1.0871,
      "epoch": 71.53248,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.1099,
      "router_z_loss": 0.107,
      "step": 14020
    },
    {
      "balancing_entropy": 0.7127,
      "balancing_loss": 1.1255,
      "epoch": 71.58368,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.11,
      "router_z_loss": 0.0966,
      "step": 14030
    },
    {
      "balancing_entropy": 0.7804,
      "balancing_loss": 1.0858,
      "epoch": 71.63488,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1144,
      "router_z_loss": 0.1029,
      "step": 14040
    },
    {
      "balancing_entropy": 0.6973,
      "balancing_loss": 1.1668,
      "epoch": 71.68608,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1168,
      "router_z_loss": 0.1207,
      "step": 14050
    },
    {
      "balancing_entropy": 0.8135,
      "balancing_loss": 1.0848,
      "epoch": 71.73728,
      "grad_norm": 0.76171875,
      "learning_rate": 0.003,
      "loss": 2.1199,
      "router_z_loss": 0.115,
      "step": 14060
    },
    {
      "balancing_entropy": 0.796,
      "balancing_loss": 1.0874,
      "epoch": 71.78848,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.1112,
      "router_z_loss": 0.0918,
      "step": 14070
    },
    {
      "balancing_entropy": 0.7928,
      "balancing_loss": 1.1182,
      "epoch": 71.83968,
      "grad_norm": 0.6640625,
      "learning_rate": 0.003,
      "loss": 2.114,
      "router_z_loss": 0.0776,
      "step": 14080
    },
    {
      "balancing_entropy": 0.774,
      "balancing_loss": 1.0811,
      "epoch": 71.89088,
      "grad_norm": 0.56640625,
      "learning_rate": 0.003,
      "loss": 2.1087,
      "router_z_loss": 0.1108,
      "step": 14090
    },
    {
      "balancing_entropy": 0.7586,
      "balancing_loss": 1.0873,
      "epoch": 71.94208,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.116,
      "router_z_loss": 0.116,
      "step": 14100
    },
    {
      "balancing_entropy": 0.6695,
      "balancing_loss": 1.1999,
      "epoch": 71.99328,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.1402,
      "router_z_loss": 0.2841,
      "step": 14110
    },
    {
      "balancing_entropy": 0.8121,
      "balancing_loss": 1.1063,
      "epoch": 72.04096,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.1158,
      "router_z_loss": 16.9742,
      "step": 14120
    },
    {
      "balancing_entropy": 0.8637,
      "balancing_loss": 1.1871,
      "epoch": 72.09216,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.1082,
      "router_z_loss": 0.2995,
      "step": 14130
    },
    {
      "balancing_entropy": 0.8568,
      "balancing_loss": 1.173,
      "epoch": 72.14336,
      "grad_norm": 0.7265625,
      "learning_rate": 0.003,
      "loss": 2.1011,
      "router_z_loss": 0.1989,
      "step": 14140
    },
    {
      "balancing_entropy": 0.8856,
      "balancing_loss": 1.1448,
      "epoch": 72.19456,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1053,
      "router_z_loss": 0.1338,
      "step": 14150
    },
    {
      "balancing_entropy": 0.8846,
      "balancing_loss": 1.1213,
      "epoch": 72.24576,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.0948,
      "router_z_loss": 0.1354,
      "step": 14160
    },
    {
      "balancing_entropy": 0.8742,
      "balancing_loss": 1.1121,
      "epoch": 72.29696,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1153,
      "router_z_loss": 0.1004,
      "step": 14170
    },
    {
      "balancing_entropy": 0.8752,
      "balancing_loss": 1.1186,
      "epoch": 72.34816000000001,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.108,
      "router_z_loss": 0.0859,
      "step": 14180
    },
    {
      "balancing_entropy": 0.8687,
      "balancing_loss": 1.1291,
      "epoch": 72.39936,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.1067,
      "router_z_loss": 0.0889,
      "step": 14190
    },
    {
      "balancing_entropy": 0.8624,
      "balancing_loss": 1.135,
      "epoch": 72.45056,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.1083,
      "router_z_loss": 0.1103,
      "step": 14200
    },
    {
      "balancing_entropy": 0.9049,
      "balancing_loss": 1.102,
      "epoch": 72.50176,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.125,
      "router_z_loss": 0.0867,
      "step": 14210
    },
    {
      "balancing_entropy": 0.8595,
      "balancing_loss": 1.1293,
      "epoch": 72.55296,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.1218,
      "router_z_loss": 0.108,
      "step": 14220
    },
    {
      "balancing_entropy": 0.8884,
      "balancing_loss": 1.1067,
      "epoch": 72.60416,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.1049,
      "router_z_loss": 0.118,
      "step": 14230
    },
    {
      "balancing_entropy": 0.8283,
      "balancing_loss": 1.1506,
      "epoch": 72.65536,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.1098,
      "router_z_loss": 0.1068,
      "step": 14240
    },
    {
      "balancing_entropy": 0.8767,
      "balancing_loss": 1.1111,
      "epoch": 72.70656,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.1271,
      "router_z_loss": 0.096,
      "step": 14250
    },
    {
      "balancing_entropy": 0.882,
      "balancing_loss": 1.1295,
      "epoch": 72.75776,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.1192,
      "router_z_loss": 0.0814,
      "step": 14260
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.106,
      "epoch": 72.80896,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.117,
      "router_z_loss": 0.099,
      "step": 14270
    },
    {
      "balancing_entropy": 0.89,
      "balancing_loss": 1.1186,
      "epoch": 72.86016,
      "grad_norm": 0.3203125,
      "learning_rate": 0.003,
      "loss": 2.1133,
      "router_z_loss": 0.1073,
      "step": 14280
    },
    {
      "balancing_entropy": 0.8591,
      "balancing_loss": 1.14,
      "epoch": 72.91136,
      "grad_norm": 0.486328125,
      "learning_rate": 0.003,
      "loss": 2.1156,
      "router_z_loss": 0.1385,
      "step": 14290
    },
    {
      "balancing_entropy": 0.8999,
      "balancing_loss": 1.1177,
      "epoch": 72.96256,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1211,
      "router_z_loss": 0.0952,
      "step": 14300
    },
    {
      "balancing_entropy": 0.8681,
      "balancing_loss": 1.1206,
      "epoch": 73.01024,
      "grad_norm": 0.408203125,
      "learning_rate": 0.003,
      "loss": 2.1093,
      "router_z_loss": 0.0873,
      "step": 14310
    },
    {
      "balancing_entropy": 0.8654,
      "balancing_loss": 1.1409,
      "epoch": 73.06144,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.1008,
      "router_z_loss": 0.1169,
      "step": 14320
    },
    {
      "balancing_entropy": 0.8948,
      "balancing_loss": 1.117,
      "epoch": 73.11264,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.1019,
      "router_z_loss": 0.1996,
      "step": 14330
    },
    {
      "balancing_entropy": 0.8851,
      "balancing_loss": 1.1262,
      "epoch": 73.16384,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.1034,
      "router_z_loss": 0.1042,
      "step": 14340
    },
    {
      "balancing_entropy": 0.889,
      "balancing_loss": 1.1286,
      "epoch": 73.21504,
      "grad_norm": 0.37109375,
      "learning_rate": 0.003,
      "loss": 2.0963,
      "router_z_loss": 0.1087,
      "step": 14350
    },
    {
      "balancing_entropy": 0.8952,
      "balancing_loss": 1.1179,
      "epoch": 73.26624,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.1013,
      "router_z_loss": 0.1187,
      "step": 14360
    },
    {
      "balancing_entropy": 0.8841,
      "balancing_loss": 1.1121,
      "epoch": 73.31744,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.1051,
      "router_z_loss": 0.1023,
      "step": 14370
    },
    {
      "balancing_entropy": 0.8938,
      "balancing_loss": 1.1152,
      "epoch": 73.36864,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.0946,
      "step": 14380
    },
    {
      "balancing_entropy": 0.8716,
      "balancing_loss": 1.1327,
      "epoch": 73.41984,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.1105,
      "router_z_loss": 0.1097,
      "step": 14390
    },
    {
      "balancing_entropy": 0.9268,
      "balancing_loss": 1.1085,
      "epoch": 73.47104,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.1199,
      "router_z_loss": 0.0904,
      "step": 14400
    },
    {
      "balancing_entropy": 0.8798,
      "balancing_loss": 1.1209,
      "epoch": 73.52224,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.1195,
      "router_z_loss": 0.12,
      "step": 14410
    },
    {
      "balancing_entropy": 0.8967,
      "balancing_loss": 1.1158,
      "epoch": 73.57344,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.1067,
      "router_z_loss": 0.0928,
      "step": 14420
    },
    {
      "balancing_entropy": 0.8881,
      "balancing_loss": 1.1115,
      "epoch": 73.62464,
      "grad_norm": 0.359375,
      "learning_rate": 0.003,
      "loss": 2.1046,
      "router_z_loss": 0.1066,
      "step": 14430
    },
    {
      "balancing_entropy": 0.9059,
      "balancing_loss": 1.114,
      "epoch": 73.67584,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.1062,
      "router_z_loss": 0.0888,
      "step": 14440
    },
    {
      "balancing_entropy": 0.9448,
      "balancing_loss": 1.1052,
      "epoch": 73.72704,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.1163,
      "router_z_loss": 0.0831,
      "step": 14450
    },
    {
      "balancing_entropy": 0.8972,
      "balancing_loss": 1.1181,
      "epoch": 73.77824,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.1203,
      "router_z_loss": 0.09,
      "step": 14460
    },
    {
      "balancing_entropy": 0.8987,
      "balancing_loss": 1.1262,
      "epoch": 73.82944,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.1143,
      "router_z_loss": 0.081,
      "step": 14470
    },
    {
      "balancing_entropy": 0.919,
      "balancing_loss": 1.1107,
      "epoch": 73.88064,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1117,
      "router_z_loss": 0.0874,
      "step": 14480
    },
    {
      "balancing_entropy": 0.8804,
      "balancing_loss": 1.138,
      "epoch": 73.93184,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1168,
      "router_z_loss": 0.0608,
      "step": 14490
    },
    {
      "balancing_entropy": 0.915,
      "balancing_loss": 1.11,
      "epoch": 73.98304,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.1129,
      "router_z_loss": 0.0637,
      "step": 14500
    },
    {
      "balancing_entropy": 0.8885,
      "balancing_loss": 1.1113,
      "epoch": 74.03072,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.1032,
      "router_z_loss": 0.0789,
      "step": 14510
    },
    {
      "balancing_entropy": 0.8829,
      "balancing_loss": 1.1303,
      "epoch": 74.08192,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.0957,
      "router_z_loss": 0.0813,
      "step": 14520
    },
    {
      "balancing_entropy": 0.8975,
      "balancing_loss": 1.113,
      "epoch": 74.13312,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1031,
      "router_z_loss": 0.0773,
      "step": 14530
    },
    {
      "balancing_entropy": 0.873,
      "balancing_loss": 1.1282,
      "epoch": 74.18432,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.0956,
      "router_z_loss": 0.0735,
      "step": 14540
    },
    {
      "balancing_entropy": 0.9058,
      "balancing_loss": 1.1031,
      "epoch": 74.23552,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.1059,
      "router_z_loss": 0.0868,
      "step": 14550
    },
    {
      "balancing_entropy": 0.8791,
      "balancing_loss": 1.1183,
      "epoch": 74.28672,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.1075,
      "router_z_loss": 0.1154,
      "step": 14560
    },
    {
      "balancing_entropy": 0.8635,
      "balancing_loss": 1.1288,
      "epoch": 74.33792,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1088,
      "router_z_loss": 0.1308,
      "step": 14570
    },
    {
      "balancing_entropy": 0.8695,
      "balancing_loss": 1.1338,
      "epoch": 74.38912,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1168,
      "router_z_loss": 0.0665,
      "step": 14580
    },
    {
      "balancing_entropy": 0.909,
      "balancing_loss": 1.1221,
      "epoch": 74.44032,
      "grad_norm": 0.31640625,
      "learning_rate": 0.003,
      "loss": 2.1226,
      "router_z_loss": 0.1137,
      "step": 14590
    },
    {
      "balancing_entropy": 0.9059,
      "balancing_loss": 1.1099,
      "epoch": 74.49152,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1099,
      "router_z_loss": 0.0786,
      "step": 14600
    },
    {
      "balancing_entropy": 0.8768,
      "balancing_loss": 1.1447,
      "epoch": 74.54272,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.108,
      "router_z_loss": 0.136,
      "step": 14610
    },
    {
      "balancing_entropy": 0.884,
      "balancing_loss": 1.1307,
      "epoch": 74.59392,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.108,
      "router_z_loss": 0.2253,
      "step": 14620
    },
    {
      "balancing_entropy": 0.8716,
      "balancing_loss": 1.1191,
      "epoch": 74.64512,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.111,
      "router_z_loss": 0.0859,
      "step": 14630
    },
    {
      "balancing_entropy": 0.8737,
      "balancing_loss": 1.1154,
      "epoch": 74.69632,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.1113,
      "router_z_loss": 0.0894,
      "step": 14640
    },
    {
      "balancing_entropy": 0.914,
      "balancing_loss": 1.0922,
      "epoch": 74.74752,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.1031,
      "router_z_loss": 0.1083,
      "step": 14650
    },
    {
      "balancing_entropy": 0.8843,
      "balancing_loss": 1.1238,
      "epoch": 74.79872,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.1119,
      "router_z_loss": 0.0995,
      "step": 14660
    },
    {
      "balancing_entropy": 0.8825,
      "balancing_loss": 1.1167,
      "epoch": 74.84992,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1094,
      "router_z_loss": 0.0777,
      "step": 14670
    },
    {
      "balancing_entropy": 0.8927,
      "balancing_loss": 1.1118,
      "epoch": 74.90112,
      "grad_norm": 0.5546875,
      "learning_rate": 0.003,
      "loss": 2.1138,
      "router_z_loss": 0.0765,
      "step": 14680
    },
    {
      "balancing_entropy": 0.8891,
      "balancing_loss": 1.1242,
      "epoch": 74.95232,
      "grad_norm": 0.67578125,
      "learning_rate": 0.003,
      "loss": 2.1182,
      "router_z_loss": 0.0608,
      "step": 14690
    },
    {
      "balancing_entropy": 0.8737,
      "balancing_loss": 1.1109,
      "epoch": 75.0,
      "grad_norm": 0.41015625,
      "learning_rate": 0.003,
      "loss": 2.1148,
      "router_z_loss": 0.0729,
      "step": 14700
    },
    {
      "balancing_entropy": 0.8868,
      "balancing_loss": 1.1202,
      "epoch": 75.0512,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.1022,
      "router_z_loss": 0.0792,
      "step": 14710
    },
    {
      "balancing_entropy": 0.9158,
      "balancing_loss": 1.1167,
      "epoch": 75.1024,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.0987,
      "router_z_loss": 0.0891,
      "step": 14720
    },
    {
      "balancing_entropy": 0.9067,
      "balancing_loss": 1.1113,
      "epoch": 75.1536,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.0959,
      "router_z_loss": 0.0661,
      "step": 14730
    },
    {
      "balancing_entropy": 0.9008,
      "balancing_loss": 1.1179,
      "epoch": 75.2048,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.0914,
      "router_z_loss": 0.0664,
      "step": 14740
    },
    {
      "balancing_entropy": 0.87,
      "balancing_loss": 1.1287,
      "epoch": 75.256,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.0999,
      "router_z_loss": 0.0854,
      "step": 14750
    },
    {
      "balancing_entropy": 0.8946,
      "balancing_loss": 1.1174,
      "epoch": 75.3072,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.1188,
      "router_z_loss": 0.0727,
      "step": 14760
    },
    {
      "balancing_entropy": 0.875,
      "balancing_loss": 1.1296,
      "epoch": 75.3584,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.1051,
      "router_z_loss": 0.0913,
      "step": 14770
    },
    {
      "balancing_entropy": 0.8912,
      "balancing_loss": 1.1315,
      "epoch": 75.4096,
      "grad_norm": 0.4765625,
      "learning_rate": 0.003,
      "loss": 2.109,
      "router_z_loss": 0.0631,
      "step": 14780
    },
    {
      "balancing_entropy": 0.8985,
      "balancing_loss": 1.1035,
      "epoch": 75.4608,
      "grad_norm": 0.400390625,
      "learning_rate": 0.003,
      "loss": 2.1047,
      "router_z_loss": 0.09,
      "step": 14790
    },
    {
      "balancing_entropy": 0.8917,
      "balancing_loss": 1.1188,
      "epoch": 75.512,
      "grad_norm": 0.44140625,
      "learning_rate": 0.003,
      "loss": 2.1013,
      "router_z_loss": 0.0683,
      "step": 14800
    },
    {
      "balancing_entropy": 0.8986,
      "balancing_loss": 1.1163,
      "epoch": 75.5632,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.1145,
      "router_z_loss": 0.0591,
      "step": 14810
    },
    {
      "balancing_entropy": 0.8718,
      "balancing_loss": 1.1356,
      "epoch": 75.6144,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.104,
      "router_z_loss": 0.0665,
      "step": 14820
    },
    {
      "balancing_entropy": 0.913,
      "balancing_loss": 1.1061,
      "epoch": 75.6656,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.1012,
      "router_z_loss": 0.0687,
      "step": 14830
    },
    {
      "balancing_entropy": 0.8915,
      "balancing_loss": 1.1138,
      "epoch": 75.7168,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.0816,
      "step": 14840
    },
    {
      "balancing_entropy": 0.948,
      "balancing_loss": 1.1146,
      "epoch": 75.768,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.104,
      "router_z_loss": 0.0456,
      "step": 14850
    },
    {
      "balancing_entropy": 0.9345,
      "balancing_loss": 1.1216,
      "epoch": 75.8192,
      "grad_norm": 0.396484375,
      "learning_rate": 0.003,
      "loss": 2.1051,
      "router_z_loss": 0.0547,
      "step": 14860
    },
    {
      "balancing_entropy": 0.877,
      "balancing_loss": 1.1306,
      "epoch": 75.8704,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.1095,
      "router_z_loss": 0.0443,
      "step": 14870
    },
    {
      "balancing_entropy": 0.8959,
      "balancing_loss": 1.1173,
      "epoch": 75.9216,
      "grad_norm": 0.578125,
      "learning_rate": 0.003,
      "loss": 2.1059,
      "router_z_loss": 0.0759,
      "step": 14880
    },
    {
      "balancing_entropy": 0.9391,
      "balancing_loss": 1.1218,
      "epoch": 75.9728,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.1002,
      "router_z_loss": 0.0754,
      "step": 14890
    },
    {
      "balancing_entropy": 0.9325,
      "balancing_loss": 1.1169,
      "epoch": 76.02048,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1017,
      "router_z_loss": 0.0603,
      "step": 14900
    },
    {
      "balancing_entropy": 0.9357,
      "balancing_loss": 1.1103,
      "epoch": 76.07168,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.0967,
      "router_z_loss": 0.0586,
      "step": 14910
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.0968,
      "epoch": 76.12288,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.0952,
      "router_z_loss": 0.0912,
      "step": 14920
    },
    {
      "balancing_entropy": 0.872,
      "balancing_loss": 1.1269,
      "epoch": 76.17408,
      "grad_norm": 0.384765625,
      "learning_rate": 0.003,
      "loss": 2.0908,
      "router_z_loss": 0.0849,
      "step": 14930
    },
    {
      "balancing_entropy": 0.8991,
      "balancing_loss": 1.1138,
      "epoch": 76.22528,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.0924,
      "router_z_loss": 0.0688,
      "step": 14940
    },
    {
      "balancing_entropy": 0.8944,
      "balancing_loss": 1.1169,
      "epoch": 76.27648,
      "grad_norm": 0.44921875,
      "learning_rate": 0.003,
      "loss": 2.0967,
      "router_z_loss": 0.0521,
      "step": 14950
    },
    {
      "balancing_entropy": 0.8935,
      "balancing_loss": 1.1152,
      "epoch": 76.32768,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.0969,
      "router_z_loss": 0.0704,
      "step": 14960
    },
    {
      "balancing_entropy": 0.9102,
      "balancing_loss": 1.1033,
      "epoch": 76.37888,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.0979,
      "router_z_loss": 0.0846,
      "step": 14970
    },
    {
      "balancing_entropy": 0.8725,
      "balancing_loss": 1.1412,
      "epoch": 76.43008,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.0985,
      "router_z_loss": 0.099,
      "step": 14980
    },
    {
      "balancing_entropy": 0.9433,
      "balancing_loss": 1.1096,
      "epoch": 76.48128,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.0986,
      "router_z_loss": 0.0602,
      "step": 14990
    },
    {
      "balancing_entropy": 0.9517,
      "balancing_loss": 1.115,
      "epoch": 76.53248,
      "grad_norm": 0.287109375,
      "learning_rate": 0.003,
      "loss": 2.097,
      "router_z_loss": 0.0629,
      "step": 15000
    },
    {
      "balancing_entropy": 0.8788,
      "balancing_loss": 1.1324,
      "epoch": 76.58368,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.1028,
      "router_z_loss": 0.0553,
      "step": 15010
    },
    {
      "balancing_entropy": 0.9127,
      "balancing_loss": 1.1164,
      "epoch": 76.63488,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.1045,
      "router_z_loss": 0.0629,
      "step": 15020
    },
    {
      "balancing_entropy": 0.9211,
      "balancing_loss": 1.1239,
      "epoch": 76.68608,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.1042,
      "router_z_loss": 0.0486,
      "step": 15030
    },
    {
      "balancing_entropy": 0.9207,
      "balancing_loss": 1.1204,
      "epoch": 76.73728,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.1064,
      "router_z_loss": 0.0477,
      "step": 15040
    },
    {
      "balancing_entropy": 0.9222,
      "balancing_loss": 1.1031,
      "epoch": 76.78848,
      "grad_norm": 0.45703125,
      "learning_rate": 0.003,
      "loss": 2.1044,
      "router_z_loss": 0.0592,
      "step": 15050
    },
    {
      "balancing_entropy": 0.8446,
      "balancing_loss": 1.1471,
      "epoch": 76.83968,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.1024,
      "router_z_loss": 0.0775,
      "step": 15060
    },
    {
      "balancing_entropy": 0.8723,
      "balancing_loss": 1.1305,
      "epoch": 76.89088,
      "grad_norm": 0.421875,
      "learning_rate": 0.003,
      "loss": 2.1013,
      "router_z_loss": 0.0711,
      "step": 15070
    },
    {
      "balancing_entropy": 0.8915,
      "balancing_loss": 1.1155,
      "epoch": 76.94208,
      "grad_norm": 0.42578125,
      "learning_rate": 0.003,
      "loss": 2.1116,
      "router_z_loss": 0.0848,
      "step": 15080
    },
    {
      "balancing_entropy": 0.896,
      "balancing_loss": 1.1169,
      "epoch": 76.99328,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.1059,
      "router_z_loss": 0.0956,
      "step": 15090
    },
    {
      "balancing_entropy": 0.8617,
      "balancing_loss": 1.1216,
      "epoch": 77.04096,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.0916,
      "router_z_loss": 0.1354,
      "step": 15100
    },
    {
      "balancing_entropy": 0.907,
      "balancing_loss": 1.1044,
      "epoch": 77.09216,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.0936,
      "router_z_loss": 0.0732,
      "step": 15110
    },
    {
      "balancing_entropy": 0.902,
      "balancing_loss": 1.1015,
      "epoch": 77.14336,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.0918,
      "router_z_loss": 0.0902,
      "step": 15120
    },
    {
      "balancing_entropy": 0.8794,
      "balancing_loss": 1.1228,
      "epoch": 77.19456,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.094,
      "router_z_loss": 0.0759,
      "step": 15130
    },
    {
      "balancing_entropy": 0.883,
      "balancing_loss": 1.1285,
      "epoch": 77.24576,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.0948,
      "router_z_loss": 0.0845,
      "step": 15140
    },
    {
      "balancing_entropy": 0.9001,
      "balancing_loss": 1.1253,
      "epoch": 77.29696,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.1058,
      "router_z_loss": 0.0525,
      "step": 15150
    },
    {
      "balancing_entropy": 0.8946,
      "balancing_loss": 1.1226,
      "epoch": 77.34816000000001,
      "grad_norm": 0.486328125,
      "learning_rate": 0.003,
      "loss": 2.1052,
      "router_z_loss": 0.0518,
      "step": 15160
    },
    {
      "balancing_entropy": 0.893,
      "balancing_loss": 1.1177,
      "epoch": 77.39936,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.0919,
      "router_z_loss": 0.0762,
      "step": 15170
    },
    {
      "balancing_entropy": 0.8758,
      "balancing_loss": 1.1353,
      "epoch": 77.45056,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.0916,
      "router_z_loss": 0.0706,
      "step": 15180
    },
    {
      "balancing_entropy": 0.9424,
      "balancing_loss": 1.1109,
      "epoch": 77.50176,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.0948,
      "router_z_loss": 0.0589,
      "step": 15190
    },
    {
      "balancing_entropy": 0.894,
      "balancing_loss": 1.1294,
      "epoch": 77.55296,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.0978,
      "router_z_loss": 0.0732,
      "step": 15200
    },
    {
      "balancing_entropy": 0.9534,
      "balancing_loss": 1.1119,
      "epoch": 77.60416,
      "grad_norm": 0.72265625,
      "learning_rate": 0.003,
      "loss": 2.1053,
      "router_z_loss": 0.045,
      "step": 15210
    },
    {
      "balancing_entropy": 0.8629,
      "balancing_loss": 1.1347,
      "epoch": 77.65536,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.1004,
      "router_z_loss": 0.0746,
      "step": 15220
    },
    {
      "balancing_entropy": 0.8764,
      "balancing_loss": 1.1212,
      "epoch": 77.70656,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.0965,
      "router_z_loss": 0.0984,
      "step": 15230
    },
    {
      "balancing_entropy": 0.8929,
      "balancing_loss": 1.1029,
      "epoch": 77.75776,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.1,
      "router_z_loss": 0.0894,
      "step": 15240
    },
    {
      "balancing_entropy": 0.9149,
      "balancing_loss": 1.1046,
      "epoch": 77.80896,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.1056,
      "router_z_loss": 0.0803,
      "step": 15250
    },
    {
      "balancing_entropy": 0.8834,
      "balancing_loss": 1.1202,
      "epoch": 77.86016,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1038,
      "router_z_loss": 0.1343,
      "step": 15260
    },
    {
      "balancing_entropy": 0.9115,
      "balancing_loss": 1.1128,
      "epoch": 77.91136,
      "grad_norm": 0.40234375,
      "learning_rate": 0.003,
      "loss": 2.1082,
      "router_z_loss": 0.0668,
      "step": 15270
    },
    {
      "balancing_entropy": 0.8841,
      "balancing_loss": 1.131,
      "epoch": 77.96256,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.1083,
      "router_z_loss": 0.0629,
      "step": 15280
    },
    {
      "balancing_entropy": 0.885,
      "balancing_loss": 1.1038,
      "epoch": 78.01024,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.1086,
      "router_z_loss": 0.0624,
      "step": 15290
    },
    {
      "balancing_entropy": 0.8661,
      "balancing_loss": 1.1356,
      "epoch": 78.06144,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.0817,
      "router_z_loss": 0.0726,
      "step": 15300
    },
    {
      "balancing_entropy": 0.9156,
      "balancing_loss": 1.126,
      "epoch": 78.11264,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.0865,
      "router_z_loss": 0.065,
      "step": 15310
    },
    {
      "balancing_entropy": 0.8988,
      "balancing_loss": 1.1241,
      "epoch": 78.16384,
      "grad_norm": 0.3359375,
      "learning_rate": 0.003,
      "loss": 2.0894,
      "router_z_loss": 0.0728,
      "step": 15320
    },
    {
      "balancing_entropy": 0.9189,
      "balancing_loss": 1.1097,
      "epoch": 78.21504,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.09,
      "router_z_loss": 0.0588,
      "step": 15330
    },
    {
      "balancing_entropy": 0.8631,
      "balancing_loss": 1.1295,
      "epoch": 78.26624,
      "grad_norm": 0.62109375,
      "learning_rate": 0.003,
      "loss": 2.1021,
      "router_z_loss": 0.0536,
      "step": 15340
    },
    {
      "balancing_entropy": 0.9163,
      "balancing_loss": 1.1157,
      "epoch": 78.31744,
      "grad_norm": 0.61328125,
      "learning_rate": 0.003,
      "loss": 2.1034,
      "router_z_loss": 0.0717,
      "step": 15350
    },
    {
      "balancing_entropy": 0.932,
      "balancing_loss": 1.0977,
      "epoch": 78.36864,
      "grad_norm": 0.33984375,
      "learning_rate": 0.003,
      "loss": 2.1212,
      "router_z_loss": 0.061,
      "step": 15360
    },
    {
      "balancing_entropy": 0.8822,
      "balancing_loss": 1.1367,
      "epoch": 78.41984,
      "grad_norm": 0.353515625,
      "learning_rate": 0.003,
      "loss": 2.1085,
      "router_z_loss": 0.0573,
      "step": 15370
    },
    {
      "balancing_entropy": 0.9,
      "balancing_loss": 1.1217,
      "epoch": 78.47104,
      "grad_norm": 0.380859375,
      "learning_rate": 0.003,
      "loss": 2.1022,
      "router_z_loss": 0.0544,
      "step": 15380
    },
    {
      "balancing_entropy": 0.8951,
      "balancing_loss": 1.1033,
      "epoch": 78.52224,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.0914,
      "router_z_loss": 0.059,
      "step": 15390
    },
    {
      "balancing_entropy": 0.8649,
      "balancing_loss": 1.1338,
      "epoch": 78.57344,
      "grad_norm": 0.345703125,
      "learning_rate": 0.003,
      "loss": 2.0988,
      "router_z_loss": 0.0701,
      "step": 15400
    },
    {
      "balancing_entropy": 0.9312,
      "balancing_loss": 1.1255,
      "epoch": 78.62464,
      "grad_norm": 0.474609375,
      "learning_rate": 0.003,
      "loss": 2.093,
      "router_z_loss": 0.0689,
      "step": 15410
    },
    {
      "balancing_entropy": 0.9085,
      "balancing_loss": 1.1143,
      "epoch": 78.67584,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.0962,
      "router_z_loss": 0.0519,
      "step": 15420
    },
    {
      "balancing_entropy": 0.9559,
      "balancing_loss": 1.1042,
      "epoch": 78.72704,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.0945,
      "router_z_loss": 0.0488,
      "step": 15430
    },
    {
      "balancing_entropy": 0.924,
      "balancing_loss": 1.1223,
      "epoch": 78.77824,
      "grad_norm": 0.373046875,
      "learning_rate": 0.003,
      "loss": 2.1063,
      "router_z_loss": 0.0594,
      "step": 15440
    },
    {
      "balancing_entropy": 0.9554,
      "balancing_loss": 1.1089,
      "epoch": 78.82944,
      "grad_norm": 0.490234375,
      "learning_rate": 0.003,
      "loss": 2.106,
      "router_z_loss": 0.0632,
      "step": 15450
    },
    {
      "balancing_entropy": 0.9888,
      "balancing_loss": 1.067,
      "epoch": 78.88064,
      "grad_norm": 0.478515625,
      "learning_rate": 0.003,
      "loss": 2.1104,
      "router_z_loss": 0.0916,
      "step": 15460
    },
    {
      "balancing_entropy": 0.9382,
      "balancing_loss": 1.0974,
      "epoch": 78.93184,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.1076,
      "router_z_loss": 0.0839,
      "step": 15470
    },
    {
      "balancing_entropy": 0.9532,
      "balancing_loss": 1.0833,
      "epoch": 78.98304,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.1085,
      "router_z_loss": 0.0704,
      "step": 15480
    },
    {
      "balancing_entropy": 0.9116,
      "balancing_loss": 1.0941,
      "epoch": 79.03072,
      "grad_norm": 0.58984375,
      "learning_rate": 0.003,
      "loss": 2.1099,
      "router_z_loss": 0.0763,
      "step": 15490
    },
    {
      "balancing_entropy": 0.9328,
      "balancing_loss": 1.1079,
      "epoch": 79.08192,
      "grad_norm": 0.423828125,
      "learning_rate": 0.003,
      "loss": 2.0952,
      "router_z_loss": 0.0808,
      "step": 15500
    },
    {
      "balancing_entropy": 0.9231,
      "balancing_loss": 1.0826,
      "epoch": 79.13312,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.0981,
      "router_z_loss": 0.1478,
      "step": 15510
    },
    {
      "balancing_entropy": 0.9392,
      "balancing_loss": 1.0742,
      "epoch": 79.18432,
      "grad_norm": 0.59375,
      "learning_rate": 0.003,
      "loss": 2.0915,
      "router_z_loss": 0.1078,
      "step": 15520
    },
    {
      "balancing_entropy": 0.8999,
      "balancing_loss": 1.0807,
      "epoch": 79.23552,
      "grad_norm": 0.5859375,
      "learning_rate": 0.003,
      "loss": 2.092,
      "router_z_loss": 0.1148,
      "step": 15530
    },
    {
      "balancing_entropy": 0.9093,
      "balancing_loss": 1.0837,
      "epoch": 79.28672,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.087,
      "router_z_loss": 0.0871,
      "step": 15540
    },
    {
      "balancing_entropy": 0.8957,
      "balancing_loss": 1.103,
      "epoch": 79.33792,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.1044,
      "router_z_loss": 0.124,
      "step": 15550
    },
    {
      "balancing_entropy": 0.9169,
      "balancing_loss": 1.0806,
      "epoch": 79.38912,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.0983,
      "router_z_loss": 0.1496,
      "step": 15560
    },
    {
      "balancing_entropy": 0.9505,
      "balancing_loss": 1.0924,
      "epoch": 79.44032,
      "grad_norm": 0.39453125,
      "learning_rate": 0.003,
      "loss": 2.1028,
      "router_z_loss": 0.0796,
      "step": 15570
    },
    {
      "balancing_entropy": 0.9129,
      "balancing_loss": 1.0986,
      "epoch": 79.49152,
      "grad_norm": 0.5703125,
      "learning_rate": 0.003,
      "loss": 2.1005,
      "router_z_loss": 0.1103,
      "step": 15580
    },
    {
      "balancing_entropy": 0.9399,
      "balancing_loss": 1.094,
      "epoch": 79.54272,
      "grad_norm": 0.66796875,
      "learning_rate": 0.003,
      "loss": 2.1026,
      "router_z_loss": 0.1102,
      "step": 15590
    },
    {
      "balancing_entropy": 0.9724,
      "balancing_loss": 1.0843,
      "epoch": 79.59392,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.1012,
      "router_z_loss": 0.0952,
      "step": 15600
    },
    {
      "balancing_entropy": 0.8898,
      "balancing_loss": 1.1098,
      "epoch": 79.64512,
      "grad_norm": 0.703125,
      "learning_rate": 0.003,
      "loss": 2.1124,
      "router_z_loss": 0.1309,
      "step": 15610
    },
    {
      "balancing_entropy": 0.9503,
      "balancing_loss": 1.1016,
      "epoch": 79.69632,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.1116,
      "router_z_loss": 0.3309,
      "step": 15620
    },
    {
      "balancing_entropy": 0.9483,
      "balancing_loss": 1.0705,
      "epoch": 79.74752,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1025,
      "router_z_loss": 0.1785,
      "step": 15630
    },
    {
      "balancing_entropy": 0.8894,
      "balancing_loss": 1.0976,
      "epoch": 79.79872,
      "grad_norm": 0.439453125,
      "learning_rate": 0.003,
      "loss": 2.1031,
      "router_z_loss": 0.1064,
      "step": 15640
    },
    {
      "balancing_entropy": 0.9185,
      "balancing_loss": 1.1009,
      "epoch": 79.84992,
      "grad_norm": 0.3984375,
      "learning_rate": 0.003,
      "loss": 2.0997,
      "router_z_loss": 0.1459,
      "step": 15650
    },
    {
      "balancing_entropy": 0.8866,
      "balancing_loss": 1.1141,
      "epoch": 79.90112,
      "grad_norm": 0.77734375,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.093,
      "step": 15660
    },
    {
      "balancing_entropy": 0.8958,
      "balancing_loss": 1.0951,
      "epoch": 79.95232,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.1187,
      "router_z_loss": 0.1695,
      "step": 15670
    },
    {
      "balancing_entropy": 0.884,
      "balancing_loss": 1.1018,
      "epoch": 80.0,
      "grad_norm": 0.41796875,
      "learning_rate": 0.003,
      "loss": 2.1136,
      "router_z_loss": 0.1569,
      "step": 15680
    },
    {
      "balancing_entropy": 0.9479,
      "balancing_loss": 1.0869,
      "epoch": 80.0512,
      "grad_norm": 0.71875,
      "learning_rate": 0.003,
      "loss": 2.098,
      "router_z_loss": 0.1027,
      "step": 15690
    },
    {
      "balancing_entropy": 0.8734,
      "balancing_loss": 1.1066,
      "epoch": 80.1024,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.0897,
      "router_z_loss": 0.0903,
      "step": 15700
    },
    {
      "balancing_entropy": 0.912,
      "balancing_loss": 1.0924,
      "epoch": 80.1536,
      "grad_norm": 0.3828125,
      "learning_rate": 0.003,
      "loss": 2.0905,
      "router_z_loss": 0.0749,
      "step": 15710
    },
    {
      "balancing_entropy": 0.9106,
      "balancing_loss": 1.0879,
      "epoch": 80.2048,
      "grad_norm": 0.494140625,
      "learning_rate": 0.003,
      "loss": 2.0919,
      "router_z_loss": 0.1192,
      "step": 15720
    },
    {
      "balancing_entropy": 0.9285,
      "balancing_loss": 1.0961,
      "epoch": 80.256,
      "grad_norm": 0.369140625,
      "learning_rate": 0.003,
      "loss": 2.0928,
      "router_z_loss": 0.1026,
      "step": 15730
    },
    {
      "balancing_entropy": 0.9376,
      "balancing_loss": 1.0893,
      "epoch": 80.3072,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.0907,
      "router_z_loss": 0.0753,
      "step": 15740
    },
    {
      "balancing_entropy": 0.8618,
      "balancing_loss": 1.1263,
      "epoch": 80.3584,
      "grad_norm": 0.74609375,
      "learning_rate": 0.003,
      "loss": 2.1003,
      "router_z_loss": 0.0768,
      "step": 15750
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.0705,
      "epoch": 80.4096,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.0997,
      "router_z_loss": 0.1349,
      "step": 15760
    },
    {
      "balancing_entropy": 0.9144,
      "balancing_loss": 1.0992,
      "epoch": 80.4608,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.0954,
      "router_z_loss": 0.0794,
      "step": 15770
    },
    {
      "balancing_entropy": 0.8932,
      "balancing_loss": 1.0932,
      "epoch": 80.512,
      "grad_norm": 0.416015625,
      "learning_rate": 0.003,
      "loss": 2.1034,
      "router_z_loss": 0.0913,
      "step": 15780
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.1043,
      "epoch": 80.5632,
      "grad_norm": 0.4609375,
      "learning_rate": 0.003,
      "loss": 2.1036,
      "router_z_loss": 0.0697,
      "step": 15790
    },
    {
      "balancing_entropy": 0.9234,
      "balancing_loss": 1.0895,
      "epoch": 80.6144,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1181,
      "router_z_loss": 0.1176,
      "step": 15800
    },
    {
      "balancing_entropy": 0.9139,
      "balancing_loss": 1.0829,
      "epoch": 80.6656,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.1052,
      "router_z_loss": 0.1125,
      "step": 15810
    },
    {
      "balancing_entropy": 0.8665,
      "balancing_loss": 1.14,
      "epoch": 80.7168,
      "grad_norm": 0.7265625,
      "learning_rate": 0.003,
      "loss": 2.1096,
      "router_z_loss": 0.1001,
      "step": 15820
    },
    {
      "balancing_entropy": 0.8915,
      "balancing_loss": 1.1095,
      "epoch": 80.768,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.1029,
      "router_z_loss": 0.0769,
      "step": 15830
    },
    {
      "balancing_entropy": 0.9552,
      "balancing_loss": 1.0845,
      "epoch": 80.8192,
      "grad_norm": 0.48046875,
      "learning_rate": 0.003,
      "loss": 2.1136,
      "router_z_loss": 0.0966,
      "step": 15840
    },
    {
      "balancing_entropy": 0.9475,
      "balancing_loss": 1.0944,
      "epoch": 80.8704,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.107,
      "router_z_loss": 0.1073,
      "step": 15850
    },
    {
      "balancing_entropy": 0.9126,
      "balancing_loss": 1.108,
      "epoch": 80.9216,
      "grad_norm": 0.52734375,
      "learning_rate": 0.003,
      "loss": 2.1167,
      "router_z_loss": 0.0717,
      "step": 15860
    },
    {
      "balancing_entropy": 0.9228,
      "balancing_loss": 1.0716,
      "epoch": 80.9728,
      "grad_norm": 0.328125,
      "learning_rate": 0.003,
      "loss": 2.1002,
      "router_z_loss": 0.1169,
      "step": 15870
    },
    {
      "balancing_entropy": 0.8394,
      "balancing_loss": 1.103,
      "epoch": 81.02048,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.0919,
      "router_z_loss": 0.1,
      "step": 15880
    },
    {
      "balancing_entropy": 0.9116,
      "balancing_loss": 1.0824,
      "epoch": 81.07168,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.0926,
      "router_z_loss": 0.0977,
      "step": 15890
    },
    {
      "balancing_entropy": 0.893,
      "balancing_loss": 1.1552,
      "epoch": 81.12288,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1003,
      "router_z_loss": 0.1896,
      "step": 15900
    },
    {
      "balancing_entropy": 0.9387,
      "balancing_loss": 1.0775,
      "epoch": 81.17408,
      "grad_norm": 0.515625,
      "learning_rate": 0.003,
      "loss": 2.0883,
      "router_z_loss": 0.1124,
      "step": 15910
    },
    {
      "balancing_entropy": 0.8395,
      "balancing_loss": 1.1391,
      "epoch": 81.22528,
      "grad_norm": 0.75390625,
      "learning_rate": 0.003,
      "loss": 2.0976,
      "router_z_loss": 0.1007,
      "step": 15920
    },
    {
      "balancing_entropy": 0.9338,
      "balancing_loss": 1.072,
      "epoch": 81.27648,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.1002,
      "router_z_loss": 0.0922,
      "step": 15930
    },
    {
      "balancing_entropy": 0.9114,
      "balancing_loss": 1.1245,
      "epoch": 81.32768,
      "grad_norm": 0.5234375,
      "learning_rate": 0.003,
      "loss": 2.0922,
      "router_z_loss": 0.0659,
      "step": 15940
    },
    {
      "balancing_entropy": 0.9549,
      "balancing_loss": 1.0811,
      "epoch": 81.37888,
      "grad_norm": 0.470703125,
      "learning_rate": 0.003,
      "loss": 2.0979,
      "router_z_loss": 0.0695,
      "step": 15950
    },
    {
      "balancing_entropy": 0.9463,
      "balancing_loss": 1.0832,
      "epoch": 81.43008,
      "grad_norm": 0.462890625,
      "learning_rate": 0.003,
      "loss": 2.0971,
      "router_z_loss": 0.0583,
      "step": 15960
    },
    {
      "balancing_entropy": 0.88,
      "balancing_loss": 1.1253,
      "epoch": 81.48128,
      "grad_norm": 0.8984375,
      "learning_rate": 0.003,
      "loss": 2.1016,
      "router_z_loss": 0.1395,
      "step": 15970
    },
    {
      "balancing_entropy": 0.9506,
      "balancing_loss": 1.077,
      "epoch": 81.53248,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.1027,
      "router_z_loss": 0.1338,
      "step": 15980
    },
    {
      "balancing_entropy": 0.9407,
      "balancing_loss": 1.0805,
      "epoch": 81.58368,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.0966,
      "router_z_loss": 0.1048,
      "step": 15990
    },
    {
      "balancing_entropy": 0.931,
      "balancing_loss": 1.0986,
      "epoch": 81.63488,
      "grad_norm": 0.63671875,
      "learning_rate": 0.003,
      "loss": 2.0928,
      "router_z_loss": 0.0697,
      "step": 16000
    },
    {
      "balancing_entropy": 0.9657,
      "balancing_loss": 1.0789,
      "epoch": 81.68608,
      "grad_norm": 0.8203125,
      "learning_rate": 0.003,
      "loss": 2.1049,
      "router_z_loss": 0.0726,
      "step": 16010
    },
    {
      "balancing_entropy": 0.9025,
      "balancing_loss": 1.1192,
      "epoch": 81.73728,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.1021,
      "router_z_loss": 0.0985,
      "step": 16020
    },
    {
      "balancing_entropy": 0.9758,
      "balancing_loss": 1.0945,
      "epoch": 81.78848,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.0964,
      "router_z_loss": 0.1409,
      "step": 16030
    },
    {
      "balancing_entropy": 0.9602,
      "balancing_loss": 1.0907,
      "epoch": 81.83968,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1007,
      "router_z_loss": 0.1063,
      "step": 16040
    },
    {
      "balancing_entropy": 0.9164,
      "balancing_loss": 1.0991,
      "epoch": 81.89088,
      "grad_norm": 0.828125,
      "learning_rate": 0.003,
      "loss": 2.0967,
      "router_z_loss": 0.0733,
      "step": 16050
    },
    {
      "balancing_entropy": 0.9086,
      "balancing_loss": 1.1001,
      "epoch": 81.94208,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.0671,
      "step": 16060
    },
    {
      "balancing_entropy": 0.9483,
      "balancing_loss": 1.1195,
      "epoch": 81.99328,
      "grad_norm": 0.4140625,
      "learning_rate": 0.003,
      "loss": 2.1124,
      "router_z_loss": 0.0833,
      "step": 16070
    },
    {
      "balancing_entropy": 0.8952,
      "balancing_loss": 1.0935,
      "epoch": 82.04096,
      "grad_norm": 0.427734375,
      "learning_rate": 0.003,
      "loss": 2.1006,
      "router_z_loss": 0.0834,
      "step": 16080
    },
    {
      "balancing_entropy": 0.9551,
      "balancing_loss": 1.0903,
      "epoch": 82.09216,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.0913,
      "router_z_loss": 0.0899,
      "step": 16090
    },
    {
      "balancing_entropy": 0.9433,
      "balancing_loss": 1.0883,
      "epoch": 82.14336,
      "grad_norm": 0.4375,
      "learning_rate": 0.003,
      "loss": 2.0889,
      "router_z_loss": 0.0838,
      "step": 16100
    },
    {
      "balancing_entropy": 0.8911,
      "balancing_loss": 1.0985,
      "epoch": 82.19456,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.0843,
      "router_z_loss": 0.0708,
      "step": 16110
    },
    {
      "balancing_entropy": 0.8494,
      "balancing_loss": 1.1284,
      "epoch": 82.24576,
      "grad_norm": 0.6953125,
      "learning_rate": 0.003,
      "loss": 2.0912,
      "router_z_loss": 0.0831,
      "step": 16120
    },
    {
      "balancing_entropy": 0.9055,
      "balancing_loss": 1.0762,
      "epoch": 82.29696,
      "grad_norm": 0.46484375,
      "learning_rate": 0.003,
      "loss": 2.0863,
      "router_z_loss": 0.0715,
      "step": 16130
    },
    {
      "balancing_entropy": 0.8808,
      "balancing_loss": 1.1069,
      "epoch": 82.34816000000001,
      "grad_norm": 0.458984375,
      "learning_rate": 0.003,
      "loss": 2.0857,
      "router_z_loss": 0.102,
      "step": 16140
    },
    {
      "balancing_entropy": 0.8876,
      "balancing_loss": 1.1067,
      "epoch": 82.39936,
      "grad_norm": 0.8515625,
      "learning_rate": 0.003,
      "loss": 2.0952,
      "router_z_loss": 0.0821,
      "step": 16150
    },
    {
      "balancing_entropy": 0.9229,
      "balancing_loss": 1.1084,
      "epoch": 82.45056,
      "grad_norm": 0.53515625,
      "learning_rate": 0.003,
      "loss": 2.1006,
      "router_z_loss": 0.0954,
      "step": 16160
    },
    {
      "balancing_entropy": 0.9382,
      "balancing_loss": 1.0811,
      "epoch": 82.50176,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.0965,
      "router_z_loss": 0.0889,
      "step": 16170
    },
    {
      "balancing_entropy": 0.9021,
      "balancing_loss": 1.0868,
      "epoch": 82.55296,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.0998,
      "router_z_loss": 0.0668,
      "step": 16180
    },
    {
      "balancing_entropy": 0.8652,
      "balancing_loss": 1.1136,
      "epoch": 82.60416,
      "grad_norm": 1.0234375,
      "learning_rate": 0.003,
      "loss": 2.0982,
      "router_z_loss": 0.0766,
      "step": 16190
    },
    {
      "balancing_entropy": 0.9885,
      "balancing_loss": 1.0769,
      "epoch": 82.65536,
      "grad_norm": 0.443359375,
      "learning_rate": 0.003,
      "loss": 2.0981,
      "router_z_loss": 0.0869,
      "step": 16200
    },
    {
      "balancing_entropy": 0.9934,
      "balancing_loss": 1.1109,
      "epoch": 82.70656,
      "grad_norm": 0.296875,
      "learning_rate": 0.003,
      "loss": 2.0938,
      "router_z_loss": 0.0907,
      "step": 16210
    },
    {
      "balancing_entropy": 0.9394,
      "balancing_loss": 1.0866,
      "epoch": 82.75776,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.0909,
      "router_z_loss": 0.0767,
      "step": 16220
    },
    {
      "balancing_entropy": 0.8631,
      "balancing_loss": 1.1047,
      "epoch": 82.80896,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.0987,
      "router_z_loss": 0.093,
      "step": 16230
    },
    {
      "balancing_entropy": 0.8862,
      "balancing_loss": 1.1184,
      "epoch": 82.86016,
      "grad_norm": 0.78125,
      "learning_rate": 0.003,
      "loss": 2.1102,
      "router_z_loss": 0.1474,
      "step": 16240
    },
    {
      "balancing_entropy": 0.9334,
      "balancing_loss": 1.0864,
      "epoch": 82.91136,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.1017,
      "router_z_loss": 0.1557,
      "step": 16250
    },
    {
      "balancing_entropy": 0.863,
      "balancing_loss": 1.1116,
      "epoch": 82.96256,
      "grad_norm": 0.74609375,
      "learning_rate": 0.003,
      "loss": 2.105,
      "router_z_loss": 0.0847,
      "step": 16260
    },
    {
      "balancing_entropy": 0.8979,
      "balancing_loss": 1.0745,
      "epoch": 83.01024,
      "grad_norm": 0.388671875,
      "learning_rate": 0.003,
      "loss": 2.0983,
      "router_z_loss": 0.092,
      "step": 16270
    },
    {
      "balancing_entropy": 0.9868,
      "balancing_loss": 1.112,
      "epoch": 83.06144,
      "grad_norm": 0.2373046875,
      "learning_rate": 0.003,
      "loss": 2.0892,
      "router_z_loss": 0.0655,
      "step": 16280
    },
    {
      "balancing_entropy": 0.9246,
      "balancing_loss": 1.089,
      "epoch": 83.11264,
      "grad_norm": 0.84375,
      "learning_rate": 0.003,
      "loss": 2.0904,
      "router_z_loss": 0.1011,
      "step": 16290
    },
    {
      "balancing_entropy": 0.8743,
      "balancing_loss": 1.133,
      "epoch": 83.16384,
      "grad_norm": 0.546875,
      "learning_rate": 0.003,
      "loss": 2.0981,
      "router_z_loss": 0.0853,
      "step": 16300
    },
    {
      "balancing_entropy": 0.9527,
      "balancing_loss": 1.0774,
      "epoch": 83.21504,
      "grad_norm": 0.5390625,
      "learning_rate": 0.003,
      "loss": 2.0951,
      "router_z_loss": 0.1174,
      "step": 16310
    },
    {
      "balancing_entropy": 0.9125,
      "balancing_loss": 1.1101,
      "epoch": 83.26624,
      "grad_norm": 0.65234375,
      "learning_rate": 0.003,
      "loss": 2.1094,
      "router_z_loss": 0.1177,
      "step": 16320
    },
    {
      "balancing_entropy": 0.9562,
      "balancing_loss": 1.0742,
      "epoch": 83.31744,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.0989,
      "router_z_loss": 0.1778,
      "step": 16330
    },
    {
      "balancing_entropy": 0.8893,
      "balancing_loss": 1.1091,
      "epoch": 83.36864,
      "grad_norm": 1.0078125,
      "learning_rate": 0.003,
      "loss": 2.0948,
      "router_z_loss": 0.1145,
      "step": 16340
    },
    {
      "balancing_entropy": 0.9106,
      "balancing_loss": 1.0835,
      "epoch": 83.41984,
      "grad_norm": 0.51171875,
      "learning_rate": 0.003,
      "loss": 2.0901,
      "router_z_loss": 0.1734,
      "step": 16350
    },
    {
      "balancing_entropy": 0.915,
      "balancing_loss": 1.1091,
      "epoch": 83.47104,
      "grad_norm": 0.4453125,
      "learning_rate": 0.003,
      "loss": 2.0977,
      "router_z_loss": 0.1299,
      "step": 16360
    },
    {
      "balancing_entropy": 0.9426,
      "balancing_loss": 1.0663,
      "epoch": 83.52224,
      "grad_norm": 0.7265625,
      "learning_rate": 0.003,
      "loss": 2.0986,
      "router_z_loss": 0.1442,
      "step": 16370
    },
    {
      "balancing_entropy": 0.8985,
      "balancing_loss": 1.1008,
      "epoch": 83.57344,
      "grad_norm": 0.54296875,
      "learning_rate": 0.003,
      "loss": 2.1031,
      "router_z_loss": 0.0868,
      "step": 16380
    },
    {
      "balancing_entropy": 0.9442,
      "balancing_loss": 1.1281,
      "epoch": 83.62464,
      "grad_norm": 0.59765625,
      "learning_rate": 0.003,
      "loss": 2.0954,
      "router_z_loss": 0.0857,
      "step": 16390
    },
    {
      "balancing_entropy": 0.8572,
      "balancing_loss": 1.1124,
      "epoch": 83.67584,
      "grad_norm": 0.35546875,
      "learning_rate": 0.003,
      "loss": 2.0997,
      "router_z_loss": 0.0945,
      "step": 16400
    },
    {
      "balancing_entropy": 0.9473,
      "balancing_loss": 1.0592,
      "epoch": 83.72704,
      "grad_norm": 0.33203125,
      "learning_rate": 0.003,
      "loss": 2.101,
      "router_z_loss": 0.1368,
      "step": 16410
    },
    {
      "balancing_entropy": 1.0025,
      "balancing_loss": 1.0926,
      "epoch": 83.77824,
      "grad_norm": 0.28515625,
      "learning_rate": 0.003,
      "loss": 2.1059,
      "router_z_loss": 0.0655,
      "step": 16420
    },
    {
      "balancing_entropy": 0.968,
      "balancing_loss": 1.1579,
      "epoch": 83.82944,
      "grad_norm": 0.36328125,
      "learning_rate": 0.003,
      "loss": 2.0988,
      "router_z_loss": 0.1595,
      "step": 16430
    },
    {
      "balancing_entropy": 0.8734,
      "balancing_loss": 1.1539,
      "epoch": 83.88064,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.108,
      "router_z_loss": 0.1269,
      "step": 16440
    },
    {
      "balancing_entropy": 0.9462,
      "balancing_loss": 1.0516,
      "epoch": 83.93184,
      "grad_norm": 0.62890625,
      "learning_rate": 0.003,
      "loss": 2.1131,
      "router_z_loss": 0.1546,
      "step": 16450
    },
    {
      "balancing_entropy": 0.9313,
      "balancing_loss": 1.1498,
      "epoch": 83.98304,
      "grad_norm": 0.435546875,
      "learning_rate": 0.003,
      "loss": 2.1058,
      "router_z_loss": 0.1488,
      "step": 16460
    },
    {
      "balancing_entropy": 0.8968,
      "balancing_loss": 1.0435,
      "epoch": 84.03072,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.1019,
      "router_z_loss": 0.121,
      "step": 16470
    },
    {
      "balancing_entropy": 1.0612,
      "balancing_loss": 1.1834,
      "epoch": 84.08192,
      "grad_norm": 0.314453125,
      "learning_rate": 0.003,
      "loss": 2.0935,
      "router_z_loss": 0.1766,
      "step": 16480
    },
    {
      "balancing_entropy": 1.0356,
      "balancing_loss": 1.0847,
      "epoch": 84.13312,
      "grad_norm": 0.58203125,
      "learning_rate": 0.003,
      "loss": 2.0885,
      "router_z_loss": 0.1578,
      "step": 16490
    },
    {
      "balancing_entropy": 0.8976,
      "balancing_loss": 1.0926,
      "epoch": 84.18432,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.0896,
      "router_z_loss": 0.1849,
      "step": 16500
    },
    {
      "balancing_entropy": 0.9196,
      "balancing_loss": 1.1013,
      "epoch": 84.23552,
      "grad_norm": 0.392578125,
      "learning_rate": 0.003,
      "loss": 2.0898,
      "router_z_loss": 0.1599,
      "step": 16510
    },
    {
      "balancing_entropy": 1.001,
      "balancing_loss": 1.166,
      "epoch": 84.28672,
      "grad_norm": 0.361328125,
      "learning_rate": 0.003,
      "loss": 2.0976,
      "router_z_loss": 0.176,
      "step": 16520
    },
    {
      "balancing_entropy": 1.0022,
      "balancing_loss": 1.0682,
      "epoch": 84.33792,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.097,
      "router_z_loss": 0.3348,
      "step": 16530
    },
    {
      "balancing_entropy": 0.9989,
      "balancing_loss": 1.1009,
      "epoch": 84.38912,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.0927,
      "router_z_loss": 0.3637,
      "step": 16540
    },
    {
      "balancing_entropy": 0.996,
      "balancing_loss": 1.048,
      "epoch": 84.44032,
      "grad_norm": 0.484375,
      "learning_rate": 0.003,
      "loss": 2.0934,
      "router_z_loss": 0.2777,
      "step": 16550
    },
    {
      "balancing_entropy": 1.0136,
      "balancing_loss": 1.1274,
      "epoch": 84.49152,
      "grad_norm": 0.6796875,
      "learning_rate": 0.003,
      "loss": 2.0994,
      "router_z_loss": 0.1576,
      "step": 16560
    },
    {
      "balancing_entropy": 1.0509,
      "balancing_loss": 1.0906,
      "epoch": 84.54272,
      "grad_norm": 0.30859375,
      "learning_rate": 0.003,
      "loss": 2.097,
      "router_z_loss": 0.1366,
      "step": 16570
    },
    {
      "balancing_entropy": 1.0598,
      "balancing_loss": 1.133,
      "epoch": 84.59392,
      "grad_norm": 0.259765625,
      "learning_rate": 0.003,
      "loss": 2.0974,
      "router_z_loss": 0.146,
      "step": 16580
    },
    {
      "balancing_entropy": 1.0667,
      "balancing_loss": 1.065,
      "epoch": 84.64512,
      "grad_norm": 0.275390625,
      "learning_rate": 0.003,
      "loss": 2.0992,
      "router_z_loss": 0.2178,
      "step": 16590
    },
    {
      "balancing_entropy": 0.9435,
      "balancing_loss": 1.0655,
      "epoch": 84.69632,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.0946,
      "router_z_loss": 0.3011,
      "step": 16600
    },
    {
      "balancing_entropy": 0.8263,
      "balancing_loss": 1.1606,
      "epoch": 84.74752,
      "grad_norm": 0.9140625,
      "learning_rate": 0.003,
      "loss": 2.0983,
      "router_z_loss": 0.1581,
      "step": 16610
    },
    {
      "balancing_entropy": 0.9055,
      "balancing_loss": 1.0742,
      "epoch": 84.79872,
      "grad_norm": 0.5,
      "learning_rate": 0.003,
      "loss": 2.1114,
      "router_z_loss": 0.112,
      "step": 16620
    },
    {
      "balancing_entropy": 0.8268,
      "balancing_loss": 1.1449,
      "epoch": 84.84992,
      "grad_norm": 0.90234375,
      "learning_rate": 0.003,
      "loss": 2.1116,
      "router_z_loss": 0.1621,
      "step": 16630
    },
    {
      "balancing_entropy": 0.892,
      "balancing_loss": 1.1052,
      "epoch": 84.90112,
      "grad_norm": 1.3203125,
      "learning_rate": 0.003,
      "loss": 2.107,
      "router_z_loss": 0.2888,
      "step": 16640
    },
    {
      "balancing_entropy": 0.9875,
      "balancing_loss": 1.0969,
      "epoch": 84.95232,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.1646,
      "step": 16650
    },
    {
      "balancing_entropy": 0.9148,
      "balancing_loss": 1.1946,
      "epoch": 85.0,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.1001,
      "router_z_loss": 0.238,
      "step": 16660
    },
    {
      "balancing_entropy": 1.0454,
      "balancing_loss": 1.0931,
      "epoch": 85.0512,
      "grad_norm": 0.341796875,
      "learning_rate": 0.003,
      "loss": 2.0901,
      "router_z_loss": 0.1113,
      "step": 16670
    },
    {
      "balancing_entropy": 0.9838,
      "balancing_loss": 1.0999,
      "epoch": 85.1024,
      "grad_norm": 0.357421875,
      "learning_rate": 0.003,
      "loss": 2.0862,
      "router_z_loss": 0.1074,
      "step": 16680
    },
    {
      "balancing_entropy": 0.952,
      "balancing_loss": 1.1158,
      "epoch": 85.1536,
      "grad_norm": 0.482421875,
      "learning_rate": 0.003,
      "loss": 2.0874,
      "router_z_loss": 0.2914,
      "step": 16690
    },
    {
      "balancing_entropy": 0.9977,
      "balancing_loss": 1.092,
      "epoch": 85.2048,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.088,
      "router_z_loss": 0.1782,
      "step": 16700
    },
    {
      "balancing_entropy": 0.9778,
      "balancing_loss": 1.1062,
      "epoch": 85.256,
      "grad_norm": 0.412109375,
      "learning_rate": 0.003,
      "loss": 2.0879,
      "router_z_loss": 0.2029,
      "step": 16710
    },
    {
      "balancing_entropy": 0.9528,
      "balancing_loss": 1.1127,
      "epoch": 85.3072,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.0872,
      "router_z_loss": 0.1535,
      "step": 16720
    },
    {
      "balancing_entropy": 0.8894,
      "balancing_loss": 1.1261,
      "epoch": 85.3584,
      "grad_norm": 0.322265625,
      "learning_rate": 0.003,
      "loss": 2.0954,
      "router_z_loss": 0.0933,
      "step": 16730
    },
    {
      "balancing_entropy": 1.0493,
      "balancing_loss": 1.0998,
      "epoch": 85.4096,
      "grad_norm": 0.2578125,
      "learning_rate": 0.003,
      "loss": 2.0924,
      "router_z_loss": 0.1031,
      "step": 16740
    },
    {
      "balancing_entropy": 1.0451,
      "balancing_loss": 1.0964,
      "epoch": 85.4608,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.0953,
      "router_z_loss": 0.0971,
      "step": 16750
    },
    {
      "balancing_entropy": 1.0201,
      "balancing_loss": 1.0613,
      "epoch": 85.512,
      "grad_norm": 0.349609375,
      "learning_rate": 0.003,
      "loss": 2.0982,
      "router_z_loss": 0.0859,
      "step": 16760
    },
    {
      "balancing_entropy": 0.9339,
      "balancing_loss": 1.1094,
      "epoch": 85.5632,
      "grad_norm": 0.6015625,
      "learning_rate": 0.003,
      "loss": 2.103,
      "router_z_loss": 0.082,
      "step": 16770
    },
    {
      "balancing_entropy": 0.9065,
      "balancing_loss": 1.1707,
      "epoch": 85.6144,
      "grad_norm": 0.498046875,
      "learning_rate": 0.003,
      "loss": 2.1,
      "router_z_loss": 0.122,
      "step": 16780
    },
    {
      "balancing_entropy": 0.9402,
      "balancing_loss": 1.0536,
      "epoch": 85.6656,
      "grad_norm": 0.7421875,
      "learning_rate": 0.003,
      "loss": 2.097,
      "router_z_loss": 0.0767,
      "step": 16790
    },
    {
      "balancing_entropy": 0.9276,
      "balancing_loss": 1.113,
      "epoch": 85.7168,
      "grad_norm": 0.953125,
      "learning_rate": 0.003,
      "loss": 2.0901,
      "router_z_loss": 0.0981,
      "step": 16800
    },
    {
      "balancing_entropy": 0.9428,
      "balancing_loss": 1.1089,
      "epoch": 85.768,
      "grad_norm": 0.84375,
      "learning_rate": 0.003,
      "loss": 2.1073,
      "router_z_loss": 0.1085,
      "step": 16810
    },
    {
      "balancing_entropy": 1.0497,
      "balancing_loss": 1.1541,
      "epoch": 85.8192,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.0984,
      "router_z_loss": 0.1611,
      "step": 16820
    },
    {
      "balancing_entropy": 1.0672,
      "balancing_loss": 1.0611,
      "epoch": 85.8704,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.0981,
      "router_z_loss": 0.1662,
      "step": 16830
    },
    {
      "balancing_entropy": 1.0036,
      "balancing_loss": 1.0658,
      "epoch": 85.9216,
      "grad_norm": 0.40625,
      "learning_rate": 0.003,
      "loss": 2.103,
      "router_z_loss": 0.2468,
      "step": 16840
    },
    {
      "balancing_entropy": 0.9919,
      "balancing_loss": 1.1476,
      "epoch": 85.9728,
      "grad_norm": 0.51953125,
      "learning_rate": 0.003,
      "loss": 2.1102,
      "router_z_loss": 0.152,
      "step": 16850
    },
    {
      "balancing_entropy": 0.9129,
      "balancing_loss": 1.076,
      "epoch": 86.02048,
      "grad_norm": 0.5625,
      "learning_rate": 0.003,
      "loss": 2.0812,
      "router_z_loss": 0.143,
      "step": 16860
    },
    {
      "balancing_entropy": 0.878,
      "balancing_loss": 1.1419,
      "epoch": 86.07168,
      "grad_norm": 0.55859375,
      "learning_rate": 0.003,
      "loss": 2.0857,
      "router_z_loss": 0.1032,
      "step": 16870
    },
    {
      "balancing_entropy": 0.9367,
      "balancing_loss": 1.0816,
      "epoch": 86.12288,
      "grad_norm": 0.85546875,
      "learning_rate": 0.003,
      "loss": 2.0882,
      "router_z_loss": 0.1657,
      "step": 16880
    },
    {
      "balancing_entropy": 0.936,
      "balancing_loss": 1.0693,
      "epoch": 86.17408,
      "grad_norm": 0.451171875,
      "learning_rate": 0.003,
      "loss": 2.0912,
      "router_z_loss": 0.1589,
      "step": 16890
    },
    {
      "balancing_entropy": 0.838,
      "balancing_loss": 1.1243,
      "epoch": 86.22528,
      "grad_norm": 1.28125,
      "learning_rate": 0.003,
      "loss": 2.0871,
      "router_z_loss": 0.16,
      "step": 16900
    },
    {
      "balancing_entropy": 0.8984,
      "balancing_loss": 1.0817,
      "epoch": 86.27648,
      "grad_norm": 1.125,
      "learning_rate": 0.003,
      "loss": 2.086,
      "router_z_loss": 0.1581,
      "step": 16910
    },
    {
      "balancing_entropy": 0.8945,
      "balancing_loss": 1.1225,
      "epoch": 86.32768,
      "grad_norm": 0.46875,
      "learning_rate": 0.003,
      "loss": 2.0916,
      "router_z_loss": 0.0884,
      "step": 16920
    },
    {
      "balancing_entropy": 0.8619,
      "balancing_loss": 1.0834,
      "epoch": 86.37888,
      "grad_norm": 1.171875,
      "learning_rate": 0.003,
      "loss": 2.0932,
      "router_z_loss": 0.1891,
      "step": 16930
    },
    {
      "balancing_entropy": 0.9703,
      "balancing_loss": 1.11,
      "epoch": 86.43008,
      "grad_norm": 0.38671875,
      "learning_rate": 0.003,
      "loss": 2.0887,
      "router_z_loss": 0.1274,
      "step": 16940
    },
    {
      "balancing_entropy": 0.9261,
      "balancing_loss": 1.0856,
      "epoch": 86.48128,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.093,
      "router_z_loss": 0.1469,
      "step": 16950
    },
    {
      "balancing_entropy": 0.8326,
      "balancing_loss": 1.1444,
      "epoch": 86.53248,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.0951,
      "router_z_loss": 0.1056,
      "step": 16960
    },
    {
      "balancing_entropy": 0.9233,
      "balancing_loss": 1.0689,
      "epoch": 86.58368,
      "grad_norm": 1.0703125,
      "learning_rate": 0.003,
      "loss": 2.0894,
      "router_z_loss": 0.1549,
      "step": 16970
    },
    {
      "balancing_entropy": 0.846,
      "balancing_loss": 1.1182,
      "epoch": 86.63488,
      "grad_norm": 0.80078125,
      "learning_rate": 0.003,
      "loss": 2.0957,
      "router_z_loss": 0.1131,
      "step": 16980
    },
    {
      "balancing_entropy": 0.9485,
      "balancing_loss": 1.0666,
      "epoch": 86.68608,
      "grad_norm": 0.376953125,
      "learning_rate": 0.003,
      "loss": 2.1002,
      "router_z_loss": 0.1075,
      "step": 16990
    },
    {
      "balancing_entropy": 0.9363,
      "balancing_loss": 1.1221,
      "epoch": 86.73728,
      "grad_norm": 0.83984375,
      "learning_rate": 0.003,
      "loss": 2.0975,
      "router_z_loss": 0.0834,
      "step": 17000
    },
    {
      "balancing_entropy": 0.9264,
      "balancing_loss": 1.0584,
      "epoch": 86.78848,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.0926,
      "router_z_loss": 0.1406,
      "step": 17010
    },
    {
      "balancing_entropy": 0.8605,
      "balancing_loss": 1.1622,
      "epoch": 86.83968,
      "grad_norm": 0.6484375,
      "learning_rate": 0.003,
      "loss": 2.0973,
      "router_z_loss": 0.1369,
      "step": 17020
    },
    {
      "balancing_entropy": 0.9532,
      "balancing_loss": 1.0862,
      "epoch": 86.89088,
      "grad_norm": 0.9375,
      "learning_rate": 0.003,
      "loss": 2.0983,
      "router_z_loss": 0.1451,
      "step": 17030
    },
    {
      "balancing_entropy": 0.921,
      "balancing_loss": 1.1022,
      "epoch": 86.94208,
      "grad_norm": 0.50390625,
      "learning_rate": 0.003,
      "loss": 2.0956,
      "router_z_loss": 0.1195,
      "step": 17040
    },
    {
      "balancing_entropy": 0.9043,
      "balancing_loss": 1.1388,
      "epoch": 86.99328,
      "grad_norm": 0.9765625,
      "learning_rate": 0.003,
      "loss": 2.1069,
      "router_z_loss": 0.0923,
      "step": 17050
    },
    {
      "balancing_entropy": 0.837,
      "balancing_loss": 1.1193,
      "epoch": 87.04096,
      "grad_norm": 0.47265625,
      "learning_rate": 0.003,
      "loss": 2.1066,
      "router_z_loss": 0.091,
      "step": 17060
    },
    {
      "balancing_entropy": 0.9204,
      "balancing_loss": 1.0755,
      "epoch": 87.09216,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.0852,
      "router_z_loss": 0.1431,
      "step": 17070
    },
    {
      "balancing_entropy": 0.9262,
      "balancing_loss": 1.1083,
      "epoch": 87.14336,
      "grad_norm": 0.98046875,
      "learning_rate": 0.003,
      "loss": 2.0943,
      "router_z_loss": 0.0847,
      "step": 17080
    },
    {
      "balancing_entropy": 0.9697,
      "balancing_loss": 1.0844,
      "epoch": 87.19456,
      "grad_norm": 0.4296875,
      "learning_rate": 0.003,
      "loss": 2.0974,
      "router_z_loss": 0.0924,
      "step": 17090
    },
    {
      "balancing_entropy": 0.8852,
      "balancing_loss": 1.1424,
      "epoch": 87.24576,
      "grad_norm": 0.71484375,
      "learning_rate": 0.003,
      "loss": 2.0913,
      "router_z_loss": 0.1083,
      "step": 17100
    },
    {
      "balancing_entropy": 0.9221,
      "balancing_loss": 1.0753,
      "epoch": 87.29696,
      "grad_norm": 1.0546875,
      "learning_rate": 0.003,
      "loss": 2.0839,
      "router_z_loss": 0.1275,
      "step": 17110
    },
    {
      "balancing_entropy": 0.9046,
      "balancing_loss": 1.1055,
      "epoch": 87.34816000000001,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.0832,
      "router_z_loss": 0.0912,
      "step": 17120
    },
    {
      "balancing_entropy": 0.9678,
      "balancing_loss": 1.127,
      "epoch": 87.39936,
      "grad_norm": 0.43359375,
      "learning_rate": 0.003,
      "loss": 2.085,
      "router_z_loss": 0.0664,
      "step": 17130
    },
    {
      "balancing_entropy": 0.932,
      "balancing_loss": 1.1143,
      "epoch": 87.45056,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.0842,
      "router_z_loss": 0.0631,
      "step": 17140
    },
    {
      "balancing_entropy": 0.9742,
      "balancing_loss": 1.1043,
      "epoch": 87.50176,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.0868,
      "router_z_loss": 0.0678,
      "step": 17150
    },
    {
      "balancing_entropy": 0.9384,
      "balancing_loss": 1.082,
      "epoch": 87.55296,
      "grad_norm": 0.76953125,
      "learning_rate": 0.003,
      "loss": 2.0894,
      "router_z_loss": 0.0839,
      "step": 17160
    },
    {
      "balancing_entropy": 1.0059,
      "balancing_loss": 1.1027,
      "epoch": 87.60416,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.0924,
      "router_z_loss": 0.0875,
      "step": 17170
    },
    {
      "balancing_entropy": 0.8867,
      "balancing_loss": 1.1238,
      "epoch": 87.65536,
      "grad_norm": 0.447265625,
      "learning_rate": 0.003,
      "loss": 2.0949,
      "router_z_loss": 0.0739,
      "step": 17180
    },
    {
      "balancing_entropy": 0.9714,
      "balancing_loss": 1.0589,
      "epoch": 87.70656,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.0981,
      "router_z_loss": 0.0889,
      "step": 17190
    },
    {
      "balancing_entropy": 0.9569,
      "balancing_loss": 1.1235,
      "epoch": 87.75776,
      "grad_norm": 0.32421875,
      "learning_rate": 0.003,
      "loss": 2.096,
      "router_z_loss": 0.0864,
      "step": 17200
    },
    {
      "balancing_entropy": 1.0278,
      "balancing_loss": 1.0954,
      "epoch": 87.80896,
      "grad_norm": 0.29296875,
      "learning_rate": 0.003,
      "loss": 2.0906,
      "router_z_loss": 0.0994,
      "step": 17210
    },
    {
      "balancing_entropy": 0.9506,
      "balancing_loss": 1.1042,
      "epoch": 87.86016,
      "grad_norm": 0.404296875,
      "learning_rate": 0.003,
      "loss": 2.0928,
      "router_z_loss": 0.0774,
      "step": 17220
    },
    {
      "balancing_entropy": 0.9096,
      "balancing_loss": 1.1532,
      "epoch": 87.91136,
      "grad_norm": 0.4921875,
      "learning_rate": 0.003,
      "loss": 2.101,
      "router_z_loss": 0.0883,
      "step": 17230
    },
    {
      "balancing_entropy": 0.9737,
      "balancing_loss": 1.0902,
      "epoch": 87.96256,
      "grad_norm": 0.453125,
      "learning_rate": 0.003,
      "loss": 2.1114,
      "router_z_loss": 0.0645,
      "step": 17240
    },
    {
      "balancing_entropy": 0.8935,
      "balancing_loss": 1.0864,
      "epoch": 88.01024,
      "grad_norm": 0.640625,
      "learning_rate": 0.003,
      "loss": 2.1038,
      "router_z_loss": 0.0806,
      "step": 17250
    },
    {
      "balancing_entropy": 0.8854,
      "balancing_loss": 1.0878,
      "epoch": 88.06144,
      "grad_norm": 1.1015625,
      "learning_rate": 0.003,
      "loss": 2.0893,
      "router_z_loss": 0.0821,
      "step": 17260
    },
    {
      "balancing_entropy": 0.9394,
      "balancing_loss": 1.0728,
      "epoch": 88.11264,
      "grad_norm": 0.53125,
      "learning_rate": 0.003,
      "loss": 2.0968,
      "router_z_loss": 0.0971,
      "step": 17270
    },
    {
      "balancing_entropy": 0.8651,
      "balancing_loss": 1.1424,
      "epoch": 88.16384,
      "grad_norm": 0.734375,
      "learning_rate": 0.003,
      "loss": 2.085,
      "router_z_loss": 0.1039,
      "step": 17280
    },
    {
      "balancing_entropy": 0.8736,
      "balancing_loss": 1.0879,
      "epoch": 88.21504,
      "grad_norm": 0.76171875,
      "learning_rate": 0.003,
      "loss": 2.0831,
      "router_z_loss": 0.1394,
      "step": 17290
    },
    {
      "balancing_entropy": 0.893,
      "balancing_loss": 1.0712,
      "epoch": 88.26624,
      "grad_norm": 0.98828125,
      "learning_rate": 0.003,
      "loss": 2.0793,
      "router_z_loss": 0.1425,
      "step": 17300
    },
    {
      "balancing_entropy": 0.9222,
      "balancing_loss": 1.1058,
      "epoch": 88.31744,
      "grad_norm": 0.5078125,
      "learning_rate": 0.003,
      "loss": 2.0886,
      "router_z_loss": 0.0856,
      "step": 17310
    },
    {
      "balancing_entropy": 0.9283,
      "balancing_loss": 1.0981,
      "epoch": 88.36864,
      "grad_norm": 0.8828125,
      "learning_rate": 0.003,
      "loss": 2.0869,
      "router_z_loss": 0.0907,
      "step": 17320
    },
    {
      "balancing_entropy": 0.9734,
      "balancing_loss": 1.1135,
      "epoch": 88.41984,
      "grad_norm": 0.65625,
      "learning_rate": 0.003,
      "loss": 2.0844,
      "router_z_loss": 0.0857,
      "step": 17330
    },
    {
      "balancing_entropy": 0.9966,
      "balancing_loss": 1.0922,
      "epoch": 88.47104,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.0879,
      "router_z_loss": 0.0825,
      "step": 17340
    },
    {
      "balancing_entropy": 0.9205,
      "balancing_loss": 1.0898,
      "epoch": 88.52224,
      "grad_norm": 0.80078125,
      "learning_rate": 0.003,
      "loss": 2.0887,
      "router_z_loss": 0.067,
      "step": 17350
    },
    {
      "balancing_entropy": 0.841,
      "balancing_loss": 1.1133,
      "epoch": 88.57344,
      "grad_norm": 0.8828125,
      "learning_rate": 0.003,
      "loss": 2.0994,
      "router_z_loss": 0.0753,
      "step": 17360
    },
    {
      "balancing_entropy": 0.8796,
      "balancing_loss": 1.1126,
      "epoch": 88.62464,
      "grad_norm": 0.318359375,
      "learning_rate": 0.003,
      "loss": 2.0935,
      "router_z_loss": 0.0775,
      "step": 17370
    },
    {
      "balancing_entropy": 0.8625,
      "balancing_loss": 1.0878,
      "epoch": 88.67584,
      "grad_norm": 0.73046875,
      "learning_rate": 0.003,
      "loss": 2.0934,
      "router_z_loss": 0.0703,
      "step": 17380
    },
    {
      "balancing_entropy": 0.8892,
      "balancing_loss": 1.0946,
      "epoch": 88.72704,
      "grad_norm": 0.6171875,
      "learning_rate": 0.003,
      "loss": 2.0922,
      "router_z_loss": 0.0996,
      "step": 17390
    },
    {
      "balancing_entropy": 0.9143,
      "balancing_loss": 1.1055,
      "epoch": 88.77824,
      "grad_norm": 0.455078125,
      "learning_rate": 0.003,
      "loss": 2.098,
      "router_z_loss": 0.0748,
      "step": 17400
    },
    {
      "balancing_entropy": 0.9166,
      "balancing_loss": 1.1245,
      "epoch": 88.82944,
      "grad_norm": 0.263671875,
      "learning_rate": 0.003,
      "loss": 2.0982,
      "router_z_loss": 0.1011,
      "step": 17410
    },
    {
      "balancing_entropy": 1.036,
      "balancing_loss": 1.103,
      "epoch": 88.88064,
      "grad_norm": 0.283203125,
      "learning_rate": 0.003,
      "loss": 2.0907,
      "router_z_loss": 0.0607,
      "step": 17420
    },
    {
      "balancing_entropy": 0.9626,
      "balancing_loss": 1.1095,
      "epoch": 88.93184,
      "grad_norm": 0.333984375,
      "learning_rate": 0.003,
      "loss": 2.0918,
      "router_z_loss": 0.0534,
      "step": 17430
    },
    {
      "balancing_entropy": 0.9466,
      "balancing_loss": 1.0653,
      "epoch": 88.98304,
      "grad_norm": 0.77734375,
      "learning_rate": 0.003,
      "loss": 2.0943,
      "router_z_loss": 0.0722,
      "step": 17440
    },
    {
      "balancing_entropy": 0.8641,
      "balancing_loss": 1.1249,
      "epoch": 89.03072,
      "grad_norm": 0.78515625,
      "learning_rate": 0.003,
      "loss": 2.097,
      "router_z_loss": 0.071,
      "step": 17450
    },
    {
      "balancing_entropy": 0.8757,
      "balancing_loss": 1.1164,
      "epoch": 89.08192,
      "grad_norm": 0.92578125,
      "learning_rate": 0.003,
      "loss": 2.0828,
      "router_z_loss": 0.0773,
      "step": 17460
    },
    {
      "balancing_entropy": 0.927,
      "balancing_loss": 1.0737,
      "epoch": 89.13312,
      "grad_norm": 0.9140625,
      "learning_rate": 0.003,
      "loss": 2.0781,
      "router_z_loss": 0.0926,
      "step": 17470
    },
    {
      "balancing_entropy": 0.9362,
      "balancing_loss": 1.072,
      "epoch": 89.18432,
      "grad_norm": 0.302734375,
      "learning_rate": 0.003,
      "loss": 2.0725,
      "router_z_loss": 0.075,
      "step": 17480
    },
    {
      "balancing_entropy": 0.9396,
      "balancing_loss": 1.1091,
      "epoch": 89.23552,
      "grad_norm": 0.3671875,
      "learning_rate": 0.003,
      "loss": 2.0797,
      "router_z_loss": 0.155,
      "step": 17490
    },
    {
      "balancing_entropy": 0.8904,
      "balancing_loss": 1.1245,
      "epoch": 89.28672,
      "grad_norm": 0.64453125,
      "learning_rate": 0.003,
      "loss": 2.0819,
      "router_z_loss": 0.1161,
      "step": 17500
    },
    {
      "balancing_entropy": 0.9445,
      "balancing_loss": 1.0853,
      "epoch": 89.33792,
      "grad_norm": 0.419921875,
      "learning_rate": 0.0029850000000000002,
      "loss": 2.0863,
      "router_z_loss": 0.0934,
      "step": 17510
    },
    {
      "balancing_entropy": 0.8886,
      "balancing_loss": 1.128,
      "epoch": 89.38912,
      "grad_norm": 0.68359375,
      "learning_rate": 0.00297,
      "loss": 2.0856,
      "router_z_loss": 0.0677,
      "step": 17520
    },
    {
      "balancing_entropy": 0.8678,
      "balancing_loss": 1.082,
      "epoch": 89.44032,
      "grad_norm": 0.65625,
      "learning_rate": 0.002955,
      "loss": 2.0869,
      "router_z_loss": 0.083,
      "step": 17530
    },
    {
      "balancing_entropy": 0.8573,
      "balancing_loss": 1.134,
      "epoch": 89.49152,
      "grad_norm": 0.765625,
      "learning_rate": 0.00294,
      "loss": 2.0844,
      "router_z_loss": 0.1058,
      "step": 17540
    },
    {
      "balancing_entropy": 0.9038,
      "balancing_loss": 1.0731,
      "epoch": 89.54272,
      "grad_norm": 0.54296875,
      "learning_rate": 0.002925,
      "loss": 2.0847,
      "router_z_loss": 0.1261,
      "step": 17550
    },
    {
      "balancing_entropy": 0.8298,
      "balancing_loss": 1.1281,
      "epoch": 89.59392,
      "grad_norm": 0.8828125,
      "learning_rate": 0.00291,
      "loss": 2.0847,
      "router_z_loss": 0.0861,
      "step": 17560
    },
    {
      "balancing_entropy": 0.9058,
      "balancing_loss": 1.0987,
      "epoch": 89.64512,
      "grad_norm": 0.408203125,
      "learning_rate": 0.002895,
      "loss": 2.094,
      "router_z_loss": 0.0867,
      "step": 17570
    },
    {
      "balancing_entropy": 0.93,
      "balancing_loss": 1.1111,
      "epoch": 89.69632,
      "grad_norm": 0.478515625,
      "learning_rate": 0.0028799999999999997,
      "loss": 2.0849,
      "router_z_loss": 0.0959,
      "step": 17580
    },
    {
      "balancing_entropy": 0.9182,
      "balancing_loss": 1.0805,
      "epoch": 89.74752,
      "grad_norm": 0.357421875,
      "learning_rate": 0.002865,
      "loss": 2.0863,
      "router_z_loss": 0.1037,
      "step": 17590
    },
    {
      "balancing_entropy": 0.9584,
      "balancing_loss": 1.1287,
      "epoch": 89.79872,
      "grad_norm": 1.359375,
      "learning_rate": 0.00285,
      "loss": 2.0935,
      "router_z_loss": 0.2021,
      "step": 17600
    },
    {
      "balancing_entropy": 1.0024,
      "balancing_loss": 1.102,
      "epoch": 89.84992,
      "grad_norm": 0.2734375,
      "learning_rate": 0.002835,
      "loss": 2.0829,
      "router_z_loss": 0.1824,
      "step": 17610
    },
    {
      "balancing_entropy": 0.8794,
      "balancing_loss": 1.097,
      "epoch": 89.90112,
      "grad_norm": 0.5625,
      "learning_rate": 0.00282,
      "loss": 2.0868,
      "router_z_loss": 0.0938,
      "step": 17620
    },
    {
      "balancing_entropy": 0.9116,
      "balancing_loss": 1.0971,
      "epoch": 89.95232,
      "grad_norm": 0.75,
      "learning_rate": 0.002805,
      "loss": 2.0842,
      "router_z_loss": 0.095,
      "step": 17630
    },
    {
      "balancing_entropy": 0.8651,
      "balancing_loss": 1.1254,
      "epoch": 90.0,
      "grad_norm": 0.703125,
      "learning_rate": 0.00279,
      "loss": 2.0921,
      "router_z_loss": 0.0649,
      "step": 17640
    },
    {
      "balancing_entropy": 0.881,
      "balancing_loss": 1.0821,
      "epoch": 90.0512,
      "grad_norm": 0.6796875,
      "learning_rate": 0.002775,
      "loss": 2.0798,
      "router_z_loss": 0.1015,
      "step": 17650
    },
    {
      "balancing_entropy": 0.8601,
      "balancing_loss": 1.1138,
      "epoch": 90.1024,
      "grad_norm": 0.375,
      "learning_rate": 0.0027600000000000003,
      "loss": 2.0585,
      "router_z_loss": 0.117,
      "step": 17660
    },
    {
      "balancing_entropy": 0.8299,
      "balancing_loss": 1.1316,
      "epoch": 90.1536,
      "grad_norm": 0.8203125,
      "learning_rate": 0.002745,
      "loss": 2.0756,
      "router_z_loss": 0.08,
      "step": 17670
    },
    {
      "balancing_entropy": 0.8742,
      "balancing_loss": 1.0803,
      "epoch": 90.2048,
      "grad_norm": 0.375,
      "learning_rate": 0.0027300000000000002,
      "loss": 2.0668,
      "router_z_loss": 0.1093,
      "step": 17680
    },
    {
      "balancing_entropy": 0.8511,
      "balancing_loss": 1.1049,
      "epoch": 90.256,
      "grad_norm": 0.65234375,
      "learning_rate": 0.002715,
      "loss": 2.0747,
      "router_z_loss": 0.09,
      "step": 17690
    },
    {
      "balancing_entropy": 0.9061,
      "balancing_loss": 1.0695,
      "epoch": 90.3072,
      "grad_norm": 0.5625,
      "learning_rate": 0.0027,
      "loss": 2.0716,
      "router_z_loss": 0.1034,
      "step": 17700
    },
    {
      "balancing_entropy": 0.8709,
      "balancing_loss": 1.1313,
      "epoch": 90.3584,
      "grad_norm": 0.671875,
      "learning_rate": 0.0026850000000000003,
      "loss": 2.0754,
      "router_z_loss": 0.1173,
      "step": 17710
    },
    {
      "balancing_entropy": 0.8952,
      "balancing_loss": 1.0997,
      "epoch": 90.4096,
      "grad_norm": 0.953125,
      "learning_rate": 0.00267,
      "loss": 2.0747,
      "router_z_loss": 0.1239,
      "step": 17720
    },
    {
      "balancing_entropy": 0.882,
      "balancing_loss": 1.1126,
      "epoch": 90.4608,
      "grad_norm": 0.5234375,
      "learning_rate": 0.0026550000000000002,
      "loss": 2.0732,
      "router_z_loss": 0.0961,
      "step": 17730
    },
    {
      "balancing_entropy": 0.9072,
      "balancing_loss": 1.1193,
      "epoch": 90.512,
      "grad_norm": 0.86328125,
      "learning_rate": 0.00264,
      "loss": 2.0755,
      "router_z_loss": 0.1527,
      "step": 17740
    },
    {
      "balancing_entropy": 0.9428,
      "balancing_loss": 1.0936,
      "epoch": 90.5632,
      "grad_norm": 1.1640625,
      "learning_rate": 0.002625,
      "loss": 2.0706,
      "router_z_loss": 0.1072,
      "step": 17750
    },
    {
      "balancing_entropy": 0.9251,
      "balancing_loss": 1.1114,
      "epoch": 90.6144,
      "grad_norm": 1.125,
      "learning_rate": 0.00261,
      "loss": 2.0666,
      "router_z_loss": 0.0857,
      "step": 17760
    },
    {
      "balancing_entropy": 0.9358,
      "balancing_loss": 1.0921,
      "epoch": 90.6656,
      "grad_norm": 0.80859375,
      "learning_rate": 0.002595,
      "loss": 2.0744,
      "router_z_loss": 0.0654,
      "step": 17770
    },
    {
      "balancing_entropy": 0.9056,
      "balancing_loss": 1.0926,
      "epoch": 90.7168,
      "grad_norm": 0.76953125,
      "learning_rate": 0.00258,
      "loss": 2.0745,
      "router_z_loss": 0.0831,
      "step": 17780
    },
    {
      "balancing_entropy": 0.9216,
      "balancing_loss": 1.0934,
      "epoch": 90.768,
      "grad_norm": 1.09375,
      "learning_rate": 0.002565,
      "loss": 2.068,
      "router_z_loss": 0.1106,
      "step": 17790
    },
    {
      "balancing_entropy": 0.8978,
      "balancing_loss": 1.1298,
      "epoch": 90.8192,
      "grad_norm": 0.69140625,
      "learning_rate": 0.00255,
      "loss": 2.0656,
      "router_z_loss": 0.0723,
      "step": 17800
    },
    {
      "balancing_entropy": 0.9076,
      "balancing_loss": 1.0857,
      "epoch": 90.8704,
      "grad_norm": 0.625,
      "learning_rate": 0.002535,
      "loss": 2.0692,
      "router_z_loss": 0.1684,
      "step": 17810
    },
    {
      "balancing_entropy": 0.9028,
      "balancing_loss": 1.1141,
      "epoch": 90.9216,
      "grad_norm": 0.45703125,
      "learning_rate": 0.00252,
      "loss": 2.0667,
      "router_z_loss": 0.11,
      "step": 17820
    },
    {
      "balancing_entropy": 0.8828,
      "balancing_loss": 1.0826,
      "epoch": 90.9728,
      "grad_norm": 0.84375,
      "learning_rate": 0.002505,
      "loss": 2.0701,
      "router_z_loss": 0.1077,
      "step": 17830
    },
    {
      "balancing_entropy": 0.8963,
      "balancing_loss": 1.1009,
      "epoch": 91.02048,
      "grad_norm": 0.357421875,
      "learning_rate": 0.00249,
      "loss": 2.0639,
      "router_z_loss": 0.0637,
      "step": 17840
    },
    {
      "balancing_entropy": 0.8589,
      "balancing_loss": 1.0947,
      "epoch": 91.07168,
      "grad_norm": 0.453125,
      "learning_rate": 0.0024749999999999998,
      "loss": 2.0545,
      "router_z_loss": 0.0644,
      "step": 17850
    },
    {
      "balancing_entropy": 0.8776,
      "balancing_loss": 1.1023,
      "epoch": 91.12288,
      "grad_norm": 0.466796875,
      "learning_rate": 0.0024600000000000004,
      "loss": 2.0504,
      "router_z_loss": 0.1265,
      "step": 17860
    },
    {
      "balancing_entropy": 0.8975,
      "balancing_loss": 1.1115,
      "epoch": 91.17408,
      "grad_norm": 0.494140625,
      "learning_rate": 0.0024449999999999997,
      "loss": 2.0572,
      "router_z_loss": 0.1065,
      "step": 17870
    },
    {
      "balancing_entropy": 0.8536,
      "balancing_loss": 1.1104,
      "epoch": 91.22528,
      "grad_norm": 0.51953125,
      "learning_rate": 0.0024300000000000003,
      "loss": 2.0538,
      "router_z_loss": 0.0953,
      "step": 17880
    },
    {
      "balancing_entropy": 0.9451,
      "balancing_loss": 1.1121,
      "epoch": 91.27648,
      "grad_norm": 0.451171875,
      "learning_rate": 0.002415,
      "loss": 2.0574,
      "router_z_loss": 0.0726,
      "step": 17890
    },
    {
      "balancing_entropy": 0.9203,
      "balancing_loss": 1.1213,
      "epoch": 91.32768,
      "grad_norm": 0.80078125,
      "learning_rate": 0.0024000000000000002,
      "loss": 2.0492,
      "router_z_loss": 0.0776,
      "step": 17900
    },
    {
      "balancing_entropy": 0.9125,
      "balancing_loss": 1.0903,
      "epoch": 91.37888,
      "grad_norm": 0.5390625,
      "learning_rate": 0.002385,
      "loss": 2.0521,
      "router_z_loss": 0.0635,
      "step": 17910
    },
    {
      "balancing_entropy": 0.9751,
      "balancing_loss": 1.1034,
      "epoch": 91.43008,
      "grad_norm": 0.255859375,
      "learning_rate": 0.00237,
      "loss": 2.0551,
      "router_z_loss": 0.0904,
      "step": 17920
    },
    {
      "balancing_entropy": 0.9957,
      "balancing_loss": 1.0919,
      "epoch": 91.48128,
      "grad_norm": 0.38671875,
      "learning_rate": 0.0023550000000000003,
      "loss": 2.0519,
      "router_z_loss": 0.058,
      "step": 17930
    },
    {
      "balancing_entropy": 0.8813,
      "balancing_loss": 1.1271,
      "epoch": 91.53248,
      "grad_norm": 0.64453125,
      "learning_rate": 0.00234,
      "loss": 2.04,
      "router_z_loss": 0.0697,
      "step": 17940
    },
    {
      "balancing_entropy": 0.9522,
      "balancing_loss": 1.101,
      "epoch": 91.58368,
      "grad_norm": 0.55859375,
      "learning_rate": 0.0023250000000000002,
      "loss": 2.0499,
      "router_z_loss": 0.089,
      "step": 17950
    },
    {
      "balancing_entropy": 0.868,
      "balancing_loss": 1.1064,
      "epoch": 91.63488,
      "grad_norm": 0.984375,
      "learning_rate": 0.00231,
      "loss": 2.0467,
      "router_z_loss": 0.0643,
      "step": 17960
    },
    {
      "balancing_entropy": 0.8813,
      "balancing_loss": 1.1024,
      "epoch": 91.68608,
      "grad_norm": 0.50390625,
      "learning_rate": 0.002295,
      "loss": 2.0574,
      "router_z_loss": 0.0735,
      "step": 17970
    },
    {
      "balancing_entropy": 0.933,
      "balancing_loss": 1.0899,
      "epoch": 91.73728,
      "grad_norm": 0.5703125,
      "learning_rate": 0.00228,
      "loss": 2.0589,
      "router_z_loss": 0.0806,
      "step": 17980
    },
    {
      "balancing_entropy": 0.913,
      "balancing_loss": 1.1464,
      "epoch": 91.78848,
      "grad_norm": 0.388671875,
      "learning_rate": 0.002265,
      "loss": 2.0554,
      "router_z_loss": 0.0703,
      "step": 17990
    },
    {
      "balancing_entropy": 0.9001,
      "balancing_loss": 1.0823,
      "epoch": 91.83968,
      "grad_norm": 0.77734375,
      "learning_rate": 0.0022500000000000003,
      "loss": 2.0486,
      "router_z_loss": 0.0813,
      "step": 18000
    },
    {
      "balancing_entropy": 0.8964,
      "balancing_loss": 1.1028,
      "epoch": 91.89088,
      "grad_norm": 0.39453125,
      "learning_rate": 0.002235,
      "loss": 2.0503,
      "router_z_loss": 0.0867,
      "step": 18010
    },
    {
      "balancing_entropy": 0.8432,
      "balancing_loss": 1.1153,
      "epoch": 91.94208,
      "grad_norm": 0.703125,
      "learning_rate": 0.00222,
      "loss": 2.0542,
      "router_z_loss": 0.0785,
      "step": 18020
    },
    {
      "balancing_entropy": 0.9116,
      "balancing_loss": 1.0616,
      "epoch": 91.99328,
      "grad_norm": 0.52734375,
      "learning_rate": 0.002205,
      "loss": 2.0565,
      "router_z_loss": 0.078,
      "step": 18030
    },
    {
      "balancing_entropy": 0.8402,
      "balancing_loss": 1.1455,
      "epoch": 92.04096,
      "grad_norm": 0.8984375,
      "learning_rate": 0.00219,
      "loss": 2.0386,
      "router_z_loss": 0.0944,
      "step": 18040
    },
    {
      "balancing_entropy": 0.904,
      "balancing_loss": 1.0988,
      "epoch": 92.09216,
      "grad_norm": 0.69140625,
      "learning_rate": 0.002175,
      "loss": 2.0354,
      "router_z_loss": 0.115,
      "step": 18050
    },
    {
      "balancing_entropy": 0.8767,
      "balancing_loss": 1.1457,
      "epoch": 92.14336,
      "grad_norm": 0.5234375,
      "learning_rate": 0.00216,
      "loss": 2.0331,
      "router_z_loss": 0.0866,
      "step": 18060
    },
    {
      "balancing_entropy": 0.9755,
      "balancing_loss": 1.0943,
      "epoch": 92.19456,
      "grad_norm": 0.2578125,
      "learning_rate": 0.002145,
      "loss": 2.0315,
      "router_z_loss": 0.101,
      "step": 18070
    },
    {
      "balancing_entropy": 0.9845,
      "balancing_loss": 1.0839,
      "epoch": 92.24576,
      "grad_norm": 0.3046875,
      "learning_rate": 0.00213,
      "loss": 2.0253,
      "router_z_loss": 0.1177,
      "step": 18080
    },
    {
      "balancing_entropy": 0.9464,
      "balancing_loss": 1.1196,
      "epoch": 92.29696,
      "grad_norm": 0.5859375,
      "learning_rate": 0.002115,
      "loss": 2.0339,
      "router_z_loss": 0.0624,
      "step": 18090
    },
    {
      "balancing_entropy": 0.8754,
      "balancing_loss": 1.0922,
      "epoch": 92.34816000000001,
      "grad_norm": 0.6484375,
      "learning_rate": 0.0021,
      "loss": 2.0281,
      "router_z_loss": 0.0855,
      "step": 18100
    },
    {
      "balancing_entropy": 0.8463,
      "balancing_loss": 1.1085,
      "epoch": 92.39936,
      "grad_norm": 0.4921875,
      "learning_rate": 0.002085,
      "loss": 2.0361,
      "router_z_loss": 0.0833,
      "step": 18110
    },
    {
      "balancing_entropy": 0.8837,
      "balancing_loss": 1.115,
      "epoch": 92.45056,
      "grad_norm": 0.99609375,
      "learning_rate": 0.00207,
      "loss": 2.0351,
      "router_z_loss": 0.0706,
      "step": 18120
    },
    {
      "balancing_entropy": 0.9135,
      "balancing_loss": 1.1027,
      "epoch": 92.50176,
      "grad_norm": 0.640625,
      "learning_rate": 0.0020550000000000004,
      "loss": 2.0373,
      "router_z_loss": 0.128,
      "step": 18130
    },
    {
      "balancing_entropy": 0.9378,
      "balancing_loss": 1.1309,
      "epoch": 92.55296,
      "grad_norm": 0.40234375,
      "learning_rate": 0.0020399999999999997,
      "loss": 2.0414,
      "router_z_loss": 0.0819,
      "step": 18140
    },
    {
      "balancing_entropy": 0.9821,
      "balancing_loss": 1.0594,
      "epoch": 92.60416,
      "grad_norm": 0.466796875,
      "learning_rate": 0.0020250000000000003,
      "loss": 2.0368,
      "router_z_loss": 0.0851,
      "step": 18150
    },
    {
      "balancing_entropy": 0.88,
      "balancing_loss": 1.157,
      "epoch": 92.65536,
      "grad_norm": 0.35546875,
      "learning_rate": 0.0020099999999999996,
      "loss": 2.0415,
      "router_z_loss": 0.0741,
      "step": 18160
    },
    {
      "balancing_entropy": 0.9096,
      "balancing_loss": 1.1098,
      "epoch": 92.70656,
      "grad_norm": 0.359375,
      "learning_rate": 0.0019950000000000002,
      "loss": 2.0273,
      "router_z_loss": 0.0826,
      "step": 18170
    },
    {
      "balancing_entropy": 0.9612,
      "balancing_loss": 1.1659,
      "epoch": 92.75776,
      "grad_norm": 0.4296875,
      "learning_rate": 0.00198,
      "loss": 2.0388,
      "router_z_loss": 0.0719,
      "step": 18180
    },
    {
      "balancing_entropy": 1.0656,
      "balancing_loss": 1.1159,
      "epoch": 92.80896,
      "grad_norm": 0.287109375,
      "learning_rate": 0.001965,
      "loss": 2.0419,
      "router_z_loss": 0.0648,
      "step": 18190
    },
    {
      "balancing_entropy": 1.0311,
      "balancing_loss": 1.1102,
      "epoch": 92.86016,
      "grad_norm": 0.390625,
      "learning_rate": 0.0019500000000000001,
      "loss": 2.0497,
      "router_z_loss": 0.0528,
      "step": 18200
    },
    {
      "balancing_entropy": 1.0087,
      "balancing_loss": 1.0759,
      "epoch": 92.91136,
      "grad_norm": 0.330078125,
      "learning_rate": 0.001935,
      "loss": 2.0451,
      "router_z_loss": 0.0634,
      "step": 18210
    },
    {
      "balancing_entropy": 0.9888,
      "balancing_loss": 1.0868,
      "epoch": 92.96256,
      "grad_norm": 0.419921875,
      "learning_rate": 0.00192,
      "loss": 2.0369,
      "router_z_loss": 0.0787,
      "step": 18220
    },
    {
      "balancing_entropy": 0.8732,
      "balancing_loss": 1.118,
      "epoch": 93.01024,
      "grad_norm": 0.42578125,
      "learning_rate": 0.001905,
      "loss": 2.0261,
      "router_z_loss": 0.0726,
      "step": 18230
    },
    {
      "balancing_entropy": 0.9604,
      "balancing_loss": 1.0513,
      "epoch": 93.06144,
      "grad_norm": 0.6015625,
      "learning_rate": 0.00189,
      "loss": 2.0189,
      "router_z_loss": 0.1097,
      "step": 18240
    },
    {
      "balancing_entropy": 0.9417,
      "balancing_loss": 1.1337,
      "epoch": 93.11264,
      "grad_norm": 0.515625,
      "learning_rate": 0.001875,
      "loss": 2.0143,
      "router_z_loss": 0.066,
      "step": 18250
    },
    {
      "balancing_entropy": 0.8983,
      "balancing_loss": 1.072,
      "epoch": 93.16384,
      "grad_norm": 0.73828125,
      "learning_rate": 0.00186,
      "loss": 2.0121,
      "router_z_loss": 0.0562,
      "step": 18260
    },
    {
      "balancing_entropy": 0.9264,
      "balancing_loss": 1.1381,
      "epoch": 93.21504,
      "grad_norm": 0.578125,
      "learning_rate": 0.001845,
      "loss": 2.0154,
      "router_z_loss": 0.0671,
      "step": 18270
    },
    {
      "balancing_entropy": 0.9216,
      "balancing_loss": 1.0815,
      "epoch": 93.26624,
      "grad_norm": 0.640625,
      "learning_rate": 0.00183,
      "loss": 2.0173,
      "router_z_loss": 0.0513,
      "step": 18280
    },
    {
      "balancing_entropy": 0.9401,
      "balancing_loss": 1.1316,
      "epoch": 93.31744,
      "grad_norm": 0.458984375,
      "learning_rate": 0.001815,
      "loss": 2.0179,
      "router_z_loss": 0.0714,
      "step": 18290
    },
    {
      "balancing_entropy": 1.0102,
      "balancing_loss": 1.1089,
      "epoch": 93.36864,
      "grad_norm": 0.546875,
      "learning_rate": 0.0018,
      "loss": 2.0106,
      "router_z_loss": 0.0551,
      "step": 18300
    },
    {
      "balancing_entropy": 0.8639,
      "balancing_loss": 1.0949,
      "epoch": 93.41984,
      "grad_norm": 0.5546875,
      "learning_rate": 0.001785,
      "loss": 2.0085,
      "router_z_loss": 0.0566,
      "step": 18310
    },
    {
      "balancing_entropy": 0.9214,
      "balancing_loss": 1.1529,
      "epoch": 93.47104,
      "grad_norm": 0.380859375,
      "learning_rate": 0.0017700000000000003,
      "loss": 2.0165,
      "router_z_loss": 0.0787,
      "step": 18320
    },
    {
      "balancing_entropy": 0.9607,
      "balancing_loss": 1.1158,
      "epoch": 93.52224,
      "grad_norm": 0.271484375,
      "learning_rate": 0.0017549999999999998,
      "loss": 2.0157,
      "router_z_loss": 0.0479,
      "step": 18330
    },
    {
      "balancing_entropy": 1.0098,
      "balancing_loss": 1.062,
      "epoch": 93.57344,
      "grad_norm": 0.59375,
      "learning_rate": 0.0017400000000000002,
      "loss": 2.0216,
      "router_z_loss": 0.0553,
      "step": 18340
    },
    {
      "balancing_entropy": 0.9745,
      "balancing_loss": 1.1623,
      "epoch": 93.62464,
      "grad_norm": 0.5234375,
      "learning_rate": 0.001725,
      "loss": 2.0147,
      "router_z_loss": 0.0487,
      "step": 18350
    },
    {
      "balancing_entropy": 0.9938,
      "balancing_loss": 1.0567,
      "epoch": 93.67584,
      "grad_norm": 0.72265625,
      "learning_rate": 0.0017100000000000001,
      "loss": 2.0124,
      "router_z_loss": 0.0639,
      "step": 18360
    },
    {
      "balancing_entropy": 0.8782,
      "balancing_loss": 1.1126,
      "epoch": 93.72704,
      "grad_norm": 0.9296875,
      "learning_rate": 0.001695,
      "loss": 2.0155,
      "router_z_loss": 0.0524,
      "step": 18370
    },
    {
      "balancing_entropy": 0.9142,
      "balancing_loss": 1.0617,
      "epoch": 93.77824,
      "grad_norm": 0.62890625,
      "learning_rate": 0.0016800000000000003,
      "loss": 2.0109,
      "router_z_loss": 0.0685,
      "step": 18380
    },
    {
      "balancing_entropy": 0.9579,
      "balancing_loss": 1.1486,
      "epoch": 93.82944,
      "grad_norm": 0.4609375,
      "learning_rate": 0.0016649999999999998,
      "loss": 2.0211,
      "router_z_loss": 0.0996,
      "step": 18390
    },
    {
      "balancing_entropy": 0.8678,
      "balancing_loss": 1.106,
      "epoch": 93.88064,
      "grad_norm": 0.6875,
      "learning_rate": 0.0016500000000000002,
      "loss": 2.0145,
      "router_z_loss": 0.0881,
      "step": 18400
    },
    {
      "balancing_entropy": 0.9255,
      "balancing_loss": 1.064,
      "epoch": 93.93184,
      "grad_norm": 0.5859375,
      "learning_rate": 0.0016349999999999997,
      "loss": 2.0127,
      "router_z_loss": 0.0589,
      "step": 18410
    },
    {
      "balancing_entropy": 0.92,
      "balancing_loss": 1.119,
      "epoch": 93.98304,
      "grad_norm": 0.33203125,
      "learning_rate": 0.0016200000000000001,
      "loss": 2.0147,
      "router_z_loss": 0.0482,
      "step": 18420
    },
    {
      "balancing_entropy": 0.8958,
      "balancing_loss": 1.1337,
      "epoch": 94.03072,
      "grad_norm": 0.2392578125,
      "learning_rate": 0.0016049999999999999,
      "loss": 2.0079,
      "router_z_loss": 0.0591,
      "step": 18430
    },
    {
      "balancing_entropy": 0.9833,
      "balancing_loss": 1.084,
      "epoch": 94.08192,
      "grad_norm": 0.4375,
      "learning_rate": 0.00159,
      "loss": 2.0026,
      "router_z_loss": 0.0509,
      "step": 18440
    },
    {
      "balancing_entropy": 0.9677,
      "balancing_loss": 1.1438,
      "epoch": 94.13312,
      "grad_norm": 0.318359375,
      "learning_rate": 0.001575,
      "loss": 2.0044,
      "router_z_loss": 0.0536,
      "step": 18450
    },
    {
      "balancing_entropy": 0.8873,
      "balancing_loss": 1.0678,
      "epoch": 94.18432,
      "grad_norm": 0.470703125,
      "learning_rate": 0.0015600000000000002,
      "loss": 2.0007,
      "router_z_loss": 0.0564,
      "step": 18460
    },
    {
      "balancing_entropy": 0.9376,
      "balancing_loss": 1.0589,
      "epoch": 94.23552,
      "grad_norm": 0.734375,
      "learning_rate": 0.0015450000000000001,
      "loss": 1.9984,
      "router_z_loss": 0.043,
      "step": 18470
    },
    {
      "balancing_entropy": 0.8548,
      "balancing_loss": 1.1779,
      "epoch": 94.28672,
      "grad_norm": 0.734375,
      "learning_rate": 0.0015300000000000001,
      "loss": 1.9968,
      "router_z_loss": 0.0552,
      "step": 18480
    },
    {
      "balancing_entropy": 0.9123,
      "balancing_loss": 1.0748,
      "epoch": 94.33792,
      "grad_norm": 0.76953125,
      "learning_rate": 0.001515,
      "loss": 2.0038,
      "router_z_loss": 0.0465,
      "step": 18490
    },
    {
      "balancing_entropy": 0.871,
      "balancing_loss": 1.1615,
      "epoch": 94.38912,
      "grad_norm": 0.3984375,
      "learning_rate": 0.0015,
      "loss": 2.0008,
      "router_z_loss": 0.1011,
      "step": 18500
    },
    {
      "balancing_entropy": 0.949,
      "balancing_loss": 1.1187,
      "epoch": 94.44032,
      "grad_norm": 0.2890625,
      "learning_rate": 0.001485,
      "loss": 1.9957,
      "router_z_loss": 0.054,
      "step": 18510
    },
    {
      "balancing_entropy": 0.9927,
      "balancing_loss": 1.0989,
      "epoch": 94.49152,
      "grad_norm": 0.49609375,
      "learning_rate": 0.00147,
      "loss": 2.0055,
      "router_z_loss": 0.0644,
      "step": 18520
    },
    {
      "balancing_entropy": 0.9265,
      "balancing_loss": 1.1094,
      "epoch": 94.54272,
      "grad_norm": 0.353515625,
      "learning_rate": 0.001455,
      "loss": 2.0043,
      "router_z_loss": 0.0693,
      "step": 18530
    },
    {
      "balancing_entropy": 0.9772,
      "balancing_loss": 1.0649,
      "epoch": 94.59392,
      "grad_norm": 0.498046875,
      "learning_rate": 0.0014399999999999999,
      "loss": 1.9992,
      "router_z_loss": 0.0634,
      "step": 18540
    },
    {
      "balancing_entropy": 0.9803,
      "balancing_loss": 1.1293,
      "epoch": 94.64512,
      "grad_norm": 0.32421875,
      "learning_rate": 0.001425,
      "loss": 1.9994,
      "router_z_loss": 0.0568,
      "step": 18550
    },
    {
      "balancing_entropy": 0.9493,
      "balancing_loss": 1.126,
      "epoch": 94.69632,
      "grad_norm": 0.4921875,
      "learning_rate": 0.00141,
      "loss": 2.0029,
      "router_z_loss": 0.0577,
      "step": 18560
    },
    {
      "balancing_entropy": 0.8737,
      "balancing_loss": 1.097,
      "epoch": 94.74752,
      "grad_norm": 0.83203125,
      "learning_rate": 0.001395,
      "loss": 1.9958,
      "router_z_loss": 0.0843,
      "step": 18570
    },
    {
      "balancing_entropy": 0.9398,
      "balancing_loss": 1.0775,
      "epoch": 94.79872,
      "grad_norm": 0.515625,
      "learning_rate": 0.00138,
      "loss": 1.9925,
      "router_z_loss": 0.052,
      "step": 18580
    },
    {
      "balancing_entropy": 0.9513,
      "balancing_loss": 1.1583,
      "epoch": 94.84992,
      "grad_norm": 0.4296875,
      "learning_rate": 0.001365,
      "loss": 2.0005,
      "router_z_loss": 0.0586,
      "step": 18590
    },
    {
      "balancing_entropy": 0.9472,
      "balancing_loss": 1.0559,
      "epoch": 94.90112,
      "grad_norm": 0.416015625,
      "learning_rate": 0.0013499999999999999,
      "loss": 1.9955,
      "router_z_loss": 0.0611,
      "step": 18600
    },
    {
      "balancing_entropy": 0.959,
      "balancing_loss": 1.0749,
      "epoch": 94.95232,
      "grad_norm": 0.86328125,
      "learning_rate": 0.0013349999999999998,
      "loss": 1.9882,
      "router_z_loss": 0.0533,
      "step": 18610
    },
    {
      "balancing_entropy": 0.9161,
      "balancing_loss": 1.1235,
      "epoch": 95.0,
      "grad_norm": 0.30859375,
      "learning_rate": 0.0013199999999999998,
      "loss": 1.9923,
      "router_z_loss": 0.0605,
      "step": 18620
    },
    {
      "balancing_entropy": 0.8829,
      "balancing_loss": 1.085,
      "epoch": 95.0512,
      "grad_norm": 0.83984375,
      "learning_rate": 0.0013050000000000002,
      "loss": 1.9894,
      "router_z_loss": 0.0906,
      "step": 18630
    },
    {
      "balancing_entropy": 0.9636,
      "balancing_loss": 1.1406,
      "epoch": 95.1024,
      "grad_norm": 0.52734375,
      "learning_rate": 0.0012900000000000001,
      "loss": 1.9898,
      "router_z_loss": 0.0615,
      "step": 18640
    },
    {
      "balancing_entropy": 0.9775,
      "balancing_loss": 1.0573,
      "epoch": 95.1536,
      "grad_norm": 0.34375,
      "learning_rate": 0.001275,
      "loss": 1.9837,
      "router_z_loss": 0.0752,
      "step": 18650
    },
    {
      "balancing_entropy": 0.9976,
      "balancing_loss": 1.0828,
      "epoch": 95.2048,
      "grad_norm": 0.39453125,
      "learning_rate": 0.00126,
      "loss": 1.9813,
      "router_z_loss": 0.0739,
      "step": 18660
    },
    {
      "balancing_entropy": 0.958,
      "balancing_loss": 1.1188,
      "epoch": 95.256,
      "grad_norm": 0.4453125,
      "learning_rate": 0.0012450000000000002,
      "loss": 1.984,
      "router_z_loss": 0.0654,
      "step": 18670
    },
    {
      "balancing_entropy": 0.9005,
      "balancing_loss": 1.0654,
      "epoch": 95.3072,
      "grad_norm": 0.52734375,
      "learning_rate": 0.0012300000000000002,
      "loss": 1.9825,
      "router_z_loss": 0.1097,
      "step": 18680
    },
    {
      "balancing_entropy": 0.9318,
      "balancing_loss": 1.1387,
      "epoch": 95.3584,
      "grad_norm": 0.45703125,
      "learning_rate": 0.0012150000000000002,
      "loss": 1.9823,
      "router_z_loss": 0.0694,
      "step": 18690
    },
    {
      "balancing_entropy": 0.9461,
      "balancing_loss": 1.0771,
      "epoch": 95.4096,
      "grad_norm": 0.6171875,
      "learning_rate": 0.0012000000000000001,
      "loss": 1.9894,
      "router_z_loss": 0.0925,
      "step": 18700
    },
    {
      "balancing_entropy": 0.9761,
      "balancing_loss": 1.0405,
      "epoch": 95.4608,
      "grad_norm": 0.2041015625,
      "learning_rate": 0.001185,
      "loss": 1.9817,
      "router_z_loss": 0.0838,
      "step": 18710
    },
    {
      "balancing_entropy": 0.8709,
      "balancing_loss": 1.131,
      "epoch": 95.512,
      "grad_norm": 0.478515625,
      "learning_rate": 0.00117,
      "loss": 1.99,
      "router_z_loss": 0.0581,
      "step": 18720
    },
    {
      "balancing_entropy": 0.9368,
      "balancing_loss": 1.0608,
      "epoch": 95.5632,
      "grad_norm": 0.169921875,
      "learning_rate": 0.001155,
      "loss": 1.9824,
      "router_z_loss": 0.0756,
      "step": 18730
    },
    {
      "balancing_entropy": 0.9641,
      "balancing_loss": 1.1629,
      "epoch": 95.6144,
      "grad_norm": 0.240234375,
      "learning_rate": 0.00114,
      "loss": 1.9792,
      "router_z_loss": 0.0586,
      "step": 18740
    },
    {
      "balancing_entropy": 1.0779,
      "balancing_loss": 1.0564,
      "epoch": 95.6656,
      "grad_norm": 0.19921875,
      "learning_rate": 0.0011250000000000001,
      "loss": 1.9854,
      "router_z_loss": 0.0688,
      "step": 18750
    },
    {
      "balancing_entropy": 1.0624,
      "balancing_loss": 1.0708,
      "epoch": 95.7168,
      "grad_norm": 0.66015625,
      "learning_rate": 0.00111,
      "loss": 1.982,
      "router_z_loss": 0.0587,
      "step": 18760
    },
    {
      "balancing_entropy": 0.9672,
      "balancing_loss": 1.1603,
      "epoch": 95.768,
      "grad_norm": 0.7109375,
      "learning_rate": 0.001095,
      "loss": 1.9762,
      "router_z_loss": 0.0632,
      "step": 18770
    },
    {
      "balancing_entropy": 0.9837,
      "balancing_loss": 1.0799,
      "epoch": 95.8192,
      "grad_norm": 0.65234375,
      "learning_rate": 0.00108,
      "loss": 1.9817,
      "router_z_loss": 0.0707,
      "step": 18780
    },
    {
      "balancing_entropy": 0.9341,
      "balancing_loss": 1.0851,
      "epoch": 95.8704,
      "grad_norm": 0.546875,
      "learning_rate": 0.001065,
      "loss": 1.9844,
      "router_z_loss": 0.0557,
      "step": 18790
    },
    {
      "balancing_entropy": 0.8325,
      "balancing_loss": 1.1338,
      "epoch": 95.9216,
      "grad_norm": 0.55859375,
      "learning_rate": 0.00105,
      "loss": 1.9812,
      "router_z_loss": 0.0556,
      "step": 18800
    },
    {
      "balancing_entropy": 0.9085,
      "balancing_loss": 1.0623,
      "epoch": 95.9728,
      "grad_norm": 0.453125,
      "learning_rate": 0.001035,
      "loss": 1.9771,
      "router_z_loss": 0.0713,
      "step": 18810
    },
    {
      "balancing_entropy": 0.9181,
      "balancing_loss": 1.0603,
      "epoch": 96.02048,
      "grad_norm": 1.0234375,
      "learning_rate": 0.0010199999999999999,
      "loss": 1.9699,
      "router_z_loss": 0.0558,
      "step": 18820
    },
    {
      "balancing_entropy": 1.0042,
      "balancing_loss": 1.1637,
      "epoch": 96.07168,
      "grad_norm": 0.2197265625,
      "learning_rate": 0.0010049999999999998,
      "loss": 1.9666,
      "router_z_loss": 0.0655,
      "step": 18830
    },
    {
      "balancing_entropy": 1.0499,
      "balancing_loss": 1.0698,
      "epoch": 96.12288,
      "grad_norm": 0.91796875,
      "learning_rate": 0.00099,
      "loss": 1.9691,
      "router_z_loss": 0.0842,
      "step": 18840
    },
    {
      "balancing_entropy": 0.9398,
      "balancing_loss": 1.0615,
      "epoch": 96.17408,
      "grad_norm": 0.91015625,
      "learning_rate": 0.0009749999999999998,
      "loss": 1.9665,
      "router_z_loss": 0.0686,
      "step": 18850
    },
    {
      "balancing_entropy": 0.8008,
      "balancing_loss": 1.1388,
      "epoch": 96.22528,
      "grad_norm": 0.462890625,
      "learning_rate": 0.0009599999999999999,
      "loss": 1.968,
      "router_z_loss": 0.0607,
      "step": 18860
    },
    {
      "balancing_entropy": 0.9471,
      "balancing_loss": 1.0553,
      "epoch": 96.27648,
      "grad_norm": 0.51171875,
      "learning_rate": 0.0009449999999999999,
      "loss": 1.9641,
      "router_z_loss": 0.0784,
      "step": 18870
    },
    {
      "balancing_entropy": 0.9506,
      "balancing_loss": 1.0811,
      "epoch": 96.32768,
      "grad_norm": 0.412109375,
      "learning_rate": 0.0009300000000000002,
      "loss": 1.9685,
      "router_z_loss": 0.0603,
      "step": 18880
    },
    {
      "balancing_entropy": 0.9324,
      "balancing_loss": 1.1758,
      "epoch": 96.37888,
      "grad_norm": 0.353515625,
      "learning_rate": 0.0009150000000000001,
      "loss": 1.9681,
      "router_z_loss": 0.0614,
      "step": 18890
    },
    {
      "balancing_entropy": 0.9587,
      "balancing_loss": 1.0604,
      "epoch": 96.43008,
      "grad_norm": 0.796875,
      "learning_rate": 0.0009000000000000002,
      "loss": 1.958,
      "router_z_loss": 0.0781,
      "step": 18900
    },
    {
      "balancing_entropy": 0.9213,
      "balancing_loss": 1.0597,
      "epoch": 96.48128,
      "grad_norm": 0.75390625,
      "learning_rate": 0.0008850000000000002,
      "loss": 1.9639,
      "router_z_loss": 0.0526,
      "step": 18910
    },
    {
      "balancing_entropy": 0.9406,
      "balancing_loss": 1.196,
      "epoch": 96.53248,
      "grad_norm": 0.326171875,
      "learning_rate": 0.0008700000000000001,
      "loss": 1.9719,
      "router_z_loss": 0.0542,
      "step": 18920
    },
    {
      "balancing_entropy": 1.0642,
      "balancing_loss": 1.064,
      "epoch": 96.58368,
      "grad_norm": 0.16015625,
      "learning_rate": 0.0008550000000000001,
      "loss": 1.9639,
      "router_z_loss": 0.077,
      "step": 18930
    },
    {
      "balancing_entropy": 1.0211,
      "balancing_loss": 1.0715,
      "epoch": 96.63488,
      "grad_norm": 0.18359375,
      "learning_rate": 0.0008400000000000001,
      "loss": 1.9715,
      "router_z_loss": 0.0638,
      "step": 18940
    },
    {
      "balancing_entropy": 1.0409,
      "balancing_loss": 1.0599,
      "epoch": 96.68608,
      "grad_norm": 0.1962890625,
      "learning_rate": 0.0008250000000000001,
      "loss": 1.9616,
      "router_z_loss": 0.0508,
      "step": 18950
    },
    {
      "balancing_entropy": 1.0139,
      "balancing_loss": 1.1982,
      "epoch": 96.73728,
      "grad_norm": 0.1708984375,
      "learning_rate": 0.0008100000000000001,
      "loss": 1.9659,
      "router_z_loss": 0.0449,
      "step": 18960
    },
    {
      "balancing_entropy": 1.0785,
      "balancing_loss": 1.0766,
      "epoch": 96.78848,
      "grad_norm": 0.146484375,
      "learning_rate": 0.000795,
      "loss": 1.9654,
      "router_z_loss": 0.0756,
      "step": 18970
    },
    {
      "balancing_entropy": 1.0478,
      "balancing_loss": 1.0651,
      "epoch": 96.83968,
      "grad_norm": 0.3125,
      "learning_rate": 0.0007800000000000001,
      "loss": 1.9607,
      "router_z_loss": 0.0635,
      "step": 18980
    },
    {
      "balancing_entropy": 0.9445,
      "balancing_loss": 1.0605,
      "epoch": 96.89088,
      "grad_norm": 0.412109375,
      "learning_rate": 0.0007650000000000001,
      "loss": 1.9626,
      "router_z_loss": 0.0518,
      "step": 18990
    },
    {
      "balancing_entropy": 0.9001,
      "balancing_loss": 1.2046,
      "epoch": 96.94208,
      "grad_norm": 0.4609375,
      "learning_rate": 0.00075,
      "loss": 1.9625,
      "router_z_loss": 0.0553,
      "step": 19000
    },
    {
      "balancing_entropy": 1.0537,
      "balancing_loss": 1.1035,
      "epoch": 96.99328,
      "grad_norm": 0.1708984375,
      "learning_rate": 0.000735,
      "loss": 1.9631,
      "router_z_loss": 0.0617,
      "step": 19010
    },
    {
      "balancing_entropy": 1.0513,
      "balancing_loss": 1.058,
      "epoch": 97.04096,
      "grad_norm": 0.1572265625,
      "learning_rate": 0.0007199999999999999,
      "loss": 1.955,
      "router_z_loss": 0.0739,
      "step": 19020
    },
    {
      "balancing_entropy": 1.0886,
      "balancing_loss": 1.0644,
      "epoch": 97.09216,
      "grad_norm": 0.3203125,
      "learning_rate": 0.000705,
      "loss": 1.9548,
      "router_z_loss": 0.058,
      "step": 19030
    },
    {
      "balancing_entropy": 1.0392,
      "balancing_loss": 1.071,
      "epoch": 97.14336,
      "grad_norm": 1.0546875,
      "learning_rate": 0.00069,
      "loss": 1.9548,
      "router_z_loss": 0.0559,
      "step": 19040
    },
    {
      "balancing_entropy": 0.9544,
      "balancing_loss": 1.1058,
      "epoch": 97.19456,
      "grad_norm": 0.8203125,
      "learning_rate": 0.0006749999999999999,
      "loss": 1.9554,
      "router_z_loss": 0.0483,
      "step": 19050
    },
    {
      "balancing_entropy": 0.9075,
      "balancing_loss": 1.2336,
      "epoch": 97.24576,
      "grad_norm": 0.40234375,
      "learning_rate": 0.0006599999999999999,
      "loss": 1.9513,
      "router_z_loss": 0.0533,
      "step": 19060
    },
    {
      "balancing_entropy": 0.9066,
      "balancing_loss": 1.0606,
      "epoch": 97.29696,
      "grad_norm": 0.80859375,
      "learning_rate": 0.000645,
      "loss": 1.9468,
      "router_z_loss": 0.0666,
      "step": 19070
    },
    {
      "balancing_entropy": 0.9242,
      "balancing_loss": 1.0519,
      "epoch": 97.34816000000001,
      "grad_norm": 0.90234375,
      "learning_rate": 0.0006299999999999999,
      "loss": 1.9545,
      "router_z_loss": 0.0543,
      "step": 19080
    },
    {
      "balancing_entropy": 0.9194,
      "balancing_loss": 1.0521,
      "epoch": 97.39936,
      "grad_norm": 0.65625,
      "learning_rate": 0.0006149999999999999,
      "loss": 1.9539,
      "router_z_loss": 0.0525,
      "step": 19090
    },
    {
      "balancing_entropy": 0.9051,
      "balancing_loss": 1.0707,
      "epoch": 97.45056,
      "grad_norm": 0.92578125,
      "learning_rate": 0.0005999999999999998,
      "loss": 1.9527,
      "router_z_loss": 0.0496,
      "step": 19100
    },
    {
      "balancing_entropy": 0.949,
      "balancing_loss": 1.0646,
      "epoch": 97.50176,
      "grad_norm": 0.50390625,
      "learning_rate": 0.0005849999999999999,
      "loss": 1.9507,
      "router_z_loss": 0.0451,
      "step": 19110
    },
    {
      "balancing_entropy": 0.9416,
      "balancing_loss": 1.0572,
      "epoch": 97.55296,
      "grad_norm": 0.8515625,
      "learning_rate": 0.0005699999999999999,
      "loss": 1.9522,
      "router_z_loss": 0.0402,
      "step": 19120
    },
    {
      "balancing_entropy": 1.0005,
      "balancing_loss": 1.2646,
      "epoch": 97.60416,
      "grad_norm": 0.2001953125,
      "learning_rate": 0.0005550000000000002,
      "loss": 1.9581,
      "router_z_loss": 0.0528,
      "step": 19130
    },
    {
      "balancing_entropy": 1.0249,
      "balancing_loss": 1.2028,
      "epoch": 97.65536,
      "grad_norm": 0.294921875,
      "learning_rate": 0.0005400000000000001,
      "loss": 1.9569,
      "router_z_loss": 0.0747,
      "step": 19140
    },
    {
      "balancing_entropy": 1.0692,
      "balancing_loss": 1.1458,
      "epoch": 97.70656,
      "grad_norm": 0.1884765625,
      "learning_rate": 0.0005250000000000002,
      "loss": 1.959,
      "router_z_loss": 0.0892,
      "step": 19150
    },
    {
      "balancing_entropy": 1.0558,
      "balancing_loss": 1.1088,
      "epoch": 97.75776,
      "grad_norm": 0.1923828125,
      "learning_rate": 0.0005100000000000001,
      "loss": 1.9555,
      "router_z_loss": 0.0769,
      "step": 19160
    },
    {
      "balancing_entropy": 1.0933,
      "balancing_loss": 1.1198,
      "epoch": 97.80896,
      "grad_norm": 0.1484375,
      "learning_rate": 0.0004950000000000001,
      "loss": 1.9585,
      "router_z_loss": 0.0642,
      "step": 19170
    },
    {
      "balancing_entropy": 1.0968,
      "balancing_loss": 1.1155,
      "epoch": 97.86016,
      "grad_norm": 0.1640625,
      "learning_rate": 0.0004800000000000001,
      "loss": 1.9549,
      "router_z_loss": 0.058,
      "step": 19180
    },
    {
      "balancing_entropy": 1.0935,
      "balancing_loss": 1.116,
      "epoch": 97.91136,
      "grad_norm": 0.1689453125,
      "learning_rate": 0.0004650000000000001,
      "loss": 1.9549,
      "router_z_loss": 0.0569,
      "step": 19190
    },
    {
      "balancing_entropy": 1.0937,
      "balancing_loss": 1.1173,
      "epoch": 97.96256,
      "grad_norm": 0.15625,
      "learning_rate": 0.0004500000000000001,
      "loss": 1.9573,
      "router_z_loss": 0.0501,
      "step": 19200
    },
    {
      "balancing_entropy": 1.0581,
      "balancing_loss": 1.1182,
      "epoch": 98.01024,
      "grad_norm": 0.142578125,
      "learning_rate": 0.00043500000000000006,
      "loss": 1.9552,
      "router_z_loss": 0.0518,
      "step": 19210
    },
    {
      "balancing_entropy": 1.0981,
      "balancing_loss": 1.1219,
      "epoch": 98.06144,
      "grad_norm": 0.1484375,
      "learning_rate": 0.00042000000000000007,
      "loss": 1.9468,
      "router_z_loss": 0.0519,
      "step": 19220
    },
    {
      "balancing_entropy": 1.0948,
      "balancing_loss": 1.1211,
      "epoch": 98.11264,
      "grad_norm": 0.2275390625,
      "learning_rate": 0.00040500000000000003,
      "loss": 1.9555,
      "router_z_loss": 0.0531,
      "step": 19230
    },
    {
      "balancing_entropy": 1.0936,
      "balancing_loss": 1.1046,
      "epoch": 98.16384,
      "grad_norm": 0.1650390625,
      "learning_rate": 0.00039000000000000005,
      "loss": 1.952,
      "router_z_loss": 0.054,
      "step": 19240
    },
    {
      "balancing_entropy": 1.0931,
      "balancing_loss": 1.12,
      "epoch": 98.21504,
      "grad_norm": 0.173828125,
      "learning_rate": 0.000375,
      "loss": 1.9529,
      "router_z_loss": 0.0567,
      "step": 19250
    },
    {
      "balancing_entropy": 1.0876,
      "balancing_loss": 1.1219,
      "epoch": 98.26624,
      "grad_norm": 0.298828125,
      "learning_rate": 0.00035999999999999997,
      "loss": 1.9476,
      "router_z_loss": 0.0515,
      "step": 19260
    },
    {
      "balancing_entropy": 1.0848,
      "balancing_loss": 1.1152,
      "epoch": 98.31744,
      "grad_norm": 0.1826171875,
      "learning_rate": 0.000345,
      "loss": 1.9455,
      "router_z_loss": 0.0473,
      "step": 19270
    },
    {
      "balancing_entropy": 1.0909,
      "balancing_loss": 1.1225,
      "epoch": 98.36864,
      "grad_norm": 0.146484375,
      "learning_rate": 0.00032999999999999994,
      "loss": 1.9554,
      "router_z_loss": 0.0473,
      "step": 19280
    },
    {
      "balancing_entropy": 1.0979,
      "balancing_loss": 1.1097,
      "epoch": 98.41984,
      "grad_norm": 0.1328125,
      "learning_rate": 0.00031499999999999996,
      "loss": 1.9499,
      "router_z_loss": 0.0476,
      "step": 19290
    },
    {
      "balancing_entropy": 1.0972,
      "balancing_loss": 1.1206,
      "epoch": 98.47104,
      "grad_norm": 0.12451171875,
      "learning_rate": 0.0002999999999999999,
      "loss": 1.9488,
      "router_z_loss": 0.0454,
      "step": 19300
    },
    {
      "balancing_entropy": 1.0976,
      "balancing_loss": 1.1132,
      "epoch": 98.52224,
      "grad_norm": 0.154296875,
      "learning_rate": 0.00028499999999999993,
      "loss": 1.9486,
      "router_z_loss": 0.045,
      "step": 19310
    },
    {
      "balancing_entropy": 1.098,
      "balancing_loss": 1.1119,
      "epoch": 98.57344,
      "grad_norm": 0.1416015625,
      "learning_rate": 0.0002699999999999999,
      "loss": 1.949,
      "router_z_loss": 0.0506,
      "step": 19320
    },
    {
      "balancing_entropy": 1.0971,
      "balancing_loss": 1.1208,
      "epoch": 98.62464,
      "grad_norm": 0.14453125,
      "learning_rate": 0.0002549999999999999,
      "loss": 1.9438,
      "router_z_loss": 0.0491,
      "step": 19330
    },
    {
      "balancing_entropy": 1.0977,
      "balancing_loss": 1.1113,
      "epoch": 98.67584,
      "grad_norm": 0.12890625,
      "learning_rate": 0.0002399999999999999,
      "loss": 1.9535,
      "router_z_loss": 0.0425,
      "step": 19340
    },
    {
      "balancing_entropy": 1.0947,
      "balancing_loss": 1.1086,
      "epoch": 98.72704,
      "grad_norm": 0.1416015625,
      "learning_rate": 0.00022499999999999986,
      "loss": 1.9485,
      "router_z_loss": 0.0446,
      "step": 19350
    },
    {
      "balancing_entropy": 1.0946,
      "balancing_loss": 1.1136,
      "epoch": 98.77824,
      "grad_norm": 0.150390625,
      "learning_rate": 0.00020999999999999985,
      "loss": 1.9482,
      "router_z_loss": 0.0453,
      "step": 19360
    },
    {
      "balancing_entropy": 1.0973,
      "balancing_loss": 1.1123,
      "epoch": 98.82944,
      "grad_norm": 0.1328125,
      "learning_rate": 0.00019499999999999983,
      "loss": 1.9478,
      "router_z_loss": 0.0509,
      "step": 19370
    },
    {
      "balancing_entropy": 1.0972,
      "balancing_loss": 1.1169,
      "epoch": 98.88064,
      "grad_norm": 0.162109375,
      "learning_rate": 0.00018000000000000017,
      "loss": 1.9425,
      "router_z_loss": 0.0529,
      "step": 19380
    },
    {
      "balancing_entropy": 1.0945,
      "balancing_loss": 1.1155,
      "epoch": 98.93184,
      "grad_norm": 0.12451171875,
      "learning_rate": 0.00016500000000000016,
      "loss": 1.9452,
      "router_z_loss": 0.0512,
      "step": 19390
    },
    {
      "balancing_entropy": 1.0976,
      "balancing_loss": 1.1269,
      "epoch": 98.98304,
      "grad_norm": 0.138671875,
      "learning_rate": 0.00015000000000000015,
      "loss": 1.9452,
      "router_z_loss": 0.0484,
      "step": 19400
    },
    {
      "balancing_entropy": 1.0561,
      "balancing_loss": 1.1062,
      "epoch": 99.03072,
      "grad_norm": 0.1494140625,
      "learning_rate": 0.0001350000000000001,
      "loss": 1.9462,
      "router_z_loss": 0.0462,
      "step": 19410
    },
    {
      "balancing_entropy": 1.0984,
      "balancing_loss": 1.1152,
      "epoch": 99.08192,
      "grad_norm": 0.1259765625,
      "learning_rate": 0.00012000000000000011,
      "loss": 1.9448,
      "router_z_loss": 0.0483,
      "step": 19420
    },
    {
      "balancing_entropy": 1.0985,
      "balancing_loss": 1.1145,
      "epoch": 99.13312,
      "grad_norm": 0.12353515625,
      "learning_rate": 0.0001050000000000001,
      "loss": 1.952,
      "router_z_loss": 0.0471,
      "step": 19430
    },
    {
      "balancing_entropy": 1.0985,
      "balancing_loss": 1.1098,
      "epoch": 99.18432,
      "grad_norm": 0.11328125,
      "learning_rate": 9.000000000000009e-05,
      "loss": 1.9518,
      "router_z_loss": 0.0458,
      "step": 19440
    },
    {
      "balancing_entropy": 1.0986,
      "balancing_loss": 1.1167,
      "epoch": 99.23552,
      "grad_norm": 0.11279296875,
      "learning_rate": 7.500000000000007e-05,
      "loss": 1.9404,
      "router_z_loss": 0.0455,
      "step": 19450
    },
    {
      "balancing_entropy": 1.0984,
      "balancing_loss": 1.1164,
      "epoch": 99.28672,
      "grad_norm": 0.115234375,
      "learning_rate": 6.0000000000000056e-05,
      "loss": 1.945,
      "router_z_loss": 0.0461,
      "step": 19460
    },
    {
      "balancing_entropy": 1.0984,
      "balancing_loss": 1.1146,
      "epoch": 99.33792,
      "grad_norm": 0.1279296875,
      "learning_rate": 4.5000000000000043e-05,
      "loss": 1.9358,
      "router_z_loss": 0.0453,
      "step": 19470
    },
    {
      "balancing_entropy": 1.0985,
      "balancing_loss": 1.1138,
      "epoch": 99.38912,
      "grad_norm": 0.1298828125,
      "learning_rate": 3.0000000000000028e-05,
      "loss": 1.9443,
      "router_z_loss": 0.0452,
      "step": 19480
    },
    {
      "balancing_entropy": 1.0985,
      "balancing_loss": 1.1168,
      "epoch": 99.44032,
      "grad_norm": 0.11767578125,
      "learning_rate": 1.5000000000000014e-05,
      "loss": 1.9425,
      "router_z_loss": 0.0451,
      "step": 19490
    },
    {
      "balancing_entropy": 1.0985,
      "balancing_loss": 1.1155,
      "epoch": 99.49152,
      "grad_norm": 0.11376953125,
      "learning_rate": 0.0,
      "loss": 1.9413,
      "router_z_loss": 0.0455,
      "step": 19500
    }
  ],
  "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": true
      },
      "attributes": {}
    }
  },
  "total_flos": 1.2105903509707162e+18,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}