{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 51.02048,
  "eval_steps": 500,
  "global_step": 10000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "balancing_entropy": 1.0585,
      "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
    }
  ],
  "logging_steps": 10,
  "max_steps": 19500,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 100,
  "save_steps": 5000,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 6.208056806237798e+17,
  "train_batch_size": 8,
  "trial_name": null,
  "trial_params": null
}