{
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 2.0013335833802173,
  "global_step": 3000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "acc": 0.0,
      "epoch": 0.0,
      "learning_rate": 2e-05,
      "loss": 0.1583,
      "rl_loss": 0.0773,
      "step": 1
    },
    {
      "acc": 0.0,
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 0.1955,
      "rl_loss": -0.0025,
      "step": 10
    },
    {
      "acc": 0.0,
      "epoch": 0.01,
      "learning_rate": 2e-05,
      "loss": 0.1993,
      "rl_loss": -0.0069,
      "step": 20
    },
    {
      "acc": 0.0,
      "epoch": 0.02,
      "learning_rate": 2e-05,
      "loss": 0.2039,
      "rl_loss": -0.0095,
      "step": 30
    },
    {
      "acc": 0.0,
      "epoch": 0.03,
      "learning_rate": 2e-05,
      "loss": 0.1866,
      "rl_loss": -0.0069,
      "step": 40
    },
    {
      "acc": 0.0,
      "epoch": 0.03,
      "learning_rate": 2e-05,
      "loss": 0.2016,
      "rl_loss": -0.012,
      "step": 50
    },
    {
      "acc": 0.0,
      "epoch": 0.04,
      "learning_rate": 2e-05,
      "loss": 0.1938,
      "rl_loss": -0.0042,
      "step": 60
    },
    {
      "acc": 0.0,
      "epoch": 0.05,
      "learning_rate": 2e-05,
      "loss": 0.1973,
      "rl_loss": -0.0091,
      "step": 70
    },
    {
      "acc": 0.0,
      "epoch": 0.05,
      "learning_rate": 2e-05,
      "loss": 0.1937,
      "rl_loss": -0.0073,
      "step": 80
    },
    {
      "acc": 0.0,
      "epoch": 0.06,
      "learning_rate": 2e-05,
      "loss": 0.1995,
      "rl_loss": -0.0092,
      "step": 90
    },
    {
      "acc": 0.0,
      "epoch": 0.07,
      "learning_rate": 2e-05,
      "loss": 0.1954,
      "rl_loss": -0.0138,
      "step": 100
    },
    {
      "acc": 0.0,
      "epoch": 0.07,
      "learning_rate": 2e-05,
      "loss": 0.1963,
      "rl_loss": -0.009,
      "step": 110
    },
    {
      "acc": 0.0,
      "epoch": 0.08,
      "learning_rate": 2e-05,
      "loss": 0.2045,
      "rl_loss": -0.0015,
      "step": 120
    },
    {
      "acc": 0.0,
      "epoch": 0.09,
      "learning_rate": 2e-05,
      "loss": 0.1878,
      "rl_loss": -0.0053,
      "step": 130
    },
    {
      "acc": 0.0,
      "epoch": 0.09,
      "learning_rate": 2e-05,
      "loss": 0.1941,
      "rl_loss": -0.0131,
      "step": 140
    },
    {
      "acc": 0.0,
      "epoch": 0.1,
      "learning_rate": 2e-05,
      "loss": 0.1934,
      "rl_loss": -0.0095,
      "step": 150
    },
    {
      "acc": 0.0,
      "epoch": 0.11,
      "learning_rate": 2e-05,
      "loss": 0.1897,
      "rl_loss": -0.0079,
      "step": 160
    },
    {
      "acc": 0.0,
      "epoch": 0.11,
      "learning_rate": 2e-05,
      "loss": 0.1909,
      "rl_loss": -0.0124,
      "step": 170
    },
    {
      "acc": 0.0,
      "epoch": 0.12,
      "learning_rate": 2e-05,
      "loss": 0.192,
      "rl_loss": -0.0121,
      "step": 180
    },
    {
      "acc": 0.0,
      "epoch": 0.13,
      "learning_rate": 2e-05,
      "loss": 0.1972,
      "rl_loss": 0.0008,
      "step": 190
    },
    {
      "acc": 0.0,
      "epoch": 0.13,
      "learning_rate": 2e-05,
      "loss": 0.2,
      "rl_loss": -0.0046,
      "step": 200
    },
    {
      "acc": 0.0,
      "epoch": 0.14,
      "learning_rate": 2e-05,
      "loss": 0.198,
      "rl_loss": -0.0104,
      "step": 210
    },
    {
      "acc": 0.0,
      "epoch": 0.15,
      "learning_rate": 2e-05,
      "loss": 0.1969,
      "rl_loss": -0.0194,
      "step": 220
    },
    {
      "acc": 0.0,
      "epoch": 0.15,
      "learning_rate": 2e-05,
      "loss": 0.1933,
      "rl_loss": -0.0105,
      "step": 230
    },
    {
      "acc": 0.0,
      "epoch": 0.16,
      "learning_rate": 2e-05,
      "loss": 0.181,
      "rl_loss": -0.0166,
      "step": 240
    },
    {
      "acc": 0.0,
      "epoch": 0.17,
      "learning_rate": 2e-05,
      "loss": 0.1853,
      "rl_loss": -0.0174,
      "step": 250
    },
    {
      "acc": 0.0,
      "epoch": 0.17,
      "learning_rate": 2e-05,
      "loss": 0.206,
      "rl_loss": -0.0192,
      "step": 260
    },
    {
      "acc": 0.0,
      "epoch": 0.18,
      "learning_rate": 2e-05,
      "loss": 0.1902,
      "rl_loss": -0.0071,
      "step": 270
    },
    {
      "acc": 0.0,
      "epoch": 0.19,
      "learning_rate": 2e-05,
      "loss": 0.2093,
      "rl_loss": -0.0172,
      "step": 280
    },
    {
      "acc": 0.0,
      "epoch": 0.19,
      "learning_rate": 2e-05,
      "loss": 0.1916,
      "rl_loss": -0.0226,
      "step": 290
    },
    {
      "acc": 0.0,
      "epoch": 0.2,
      "learning_rate": 2e-05,
      "loss": 0.1899,
      "rl_loss": -0.0175,
      "step": 300
    },
    {
      "acc": 0.0,
      "epoch": 0.21,
      "learning_rate": 2e-05,
      "loss": 0.2031,
      "rl_loss": -0.0283,
      "step": 310
    },
    {
      "acc": 0.0,
      "epoch": 0.21,
      "learning_rate": 2e-05,
      "loss": 0.2002,
      "rl_loss": -0.0079,
      "step": 320
    },
    {
      "acc": 0.0,
      "epoch": 0.22,
      "learning_rate": 2e-05,
      "loss": 0.2052,
      "rl_loss": -0.0195,
      "step": 330
    },
    {
      "acc": 0.0,
      "epoch": 0.23,
      "learning_rate": 2e-05,
      "loss": 0.207,
      "rl_loss": -0.0204,
      "step": 340
    },
    {
      "acc": 0.0,
      "epoch": 0.23,
      "learning_rate": 2e-05,
      "loss": 0.1994,
      "rl_loss": -0.0027,
      "step": 350
    },
    {
      "acc": 0.0,
      "epoch": 0.24,
      "learning_rate": 2e-05,
      "loss": 0.1969,
      "rl_loss": -0.0091,
      "step": 360
    },
    {
      "acc": 0.0,
      "epoch": 0.25,
      "learning_rate": 2e-05,
      "loss": 0.1933,
      "rl_loss": -0.0141,
      "step": 370
    },
    {
      "acc": 0.0,
      "epoch": 0.25,
      "learning_rate": 2e-05,
      "loss": 0.1981,
      "rl_loss": -0.0188,
      "step": 380
    },
    {
      "acc": 0.0,
      "epoch": 0.26,
      "learning_rate": 2e-05,
      "loss": 0.1954,
      "rl_loss": -0.0101,
      "step": 390
    },
    {
      "acc": 0.0,
      "epoch": 0.27,
      "learning_rate": 2e-05,
      "loss": 0.1946,
      "rl_loss": -0.0218,
      "step": 400
    },
    {
      "acc": 0.0,
      "epoch": 0.27,
      "learning_rate": 2e-05,
      "loss": 0.1913,
      "rl_loss": -0.0112,
      "step": 410
    },
    {
      "acc": 0.0,
      "epoch": 0.28,
      "learning_rate": 2e-05,
      "loss": 0.2008,
      "rl_loss": -0.006,
      "step": 420
    },
    {
      "acc": 0.0,
      "epoch": 0.29,
      "learning_rate": 2e-05,
      "loss": 0.1908,
      "rl_loss": -0.0199,
      "step": 430
    },
    {
      "acc": 0.0,
      "epoch": 0.29,
      "learning_rate": 2e-05,
      "loss": 0.192,
      "rl_loss": -0.0201,
      "step": 440
    },
    {
      "acc": 0.0,
      "epoch": 0.3,
      "learning_rate": 2e-05,
      "loss": 0.2064,
      "rl_loss": -0.0144,
      "step": 450
    },
    {
      "acc": 0.0,
      "epoch": 0.31,
      "learning_rate": 2e-05,
      "loss": 0.1998,
      "rl_loss": -0.0295,
      "step": 460
    },
    {
      "acc": 0.0,
      "epoch": 0.31,
      "learning_rate": 2e-05,
      "loss": 0.1996,
      "rl_loss": -0.0151,
      "step": 470
    },
    {
      "acc": 0.0,
      "epoch": 0.32,
      "learning_rate": 2e-05,
      "loss": 0.2179,
      "rl_loss": -0.0062,
      "step": 480
    },
    {
      "acc": 0.0,
      "epoch": 0.33,
      "learning_rate": 2e-05,
      "loss": 0.189,
      "rl_loss": -0.014,
      "step": 490
    },
    {
      "acc": 0.0,
      "epoch": 0.33,
      "learning_rate": 2e-05,
      "loss": 0.2153,
      "rl_loss": -0.0176,
      "step": 500
    },
    {
      "acc": 0.0,
      "epoch": 0.34,
      "learning_rate": 2e-05,
      "loss": 0.1954,
      "rl_loss": -0.0051,
      "step": 510
    },
    {
      "acc": 0.0,
      "epoch": 0.35,
      "learning_rate": 2e-05,
      "loss": 0.1877,
      "rl_loss": -0.0055,
      "step": 520
    },
    {
      "acc": 0.0,
      "epoch": 0.35,
      "learning_rate": 2e-05,
      "loss": 0.1952,
      "rl_loss": -0.0128,
      "step": 530
    },
    {
      "acc": 0.0,
      "epoch": 0.36,
      "learning_rate": 2e-05,
      "loss": 0.1946,
      "rl_loss": -0.0079,
      "step": 540
    },
    {
      "acc": 0.0,
      "epoch": 0.37,
      "learning_rate": 2e-05,
      "loss": 0.2009,
      "rl_loss": -0.0594,
      "step": 550
    },
    {
      "acc": 0.0,
      "epoch": 0.37,
      "learning_rate": 2e-05,
      "loss": 0.2025,
      "rl_loss": -0.0103,
      "step": 560
    },
    {
      "acc": 0.0,
      "epoch": 0.38,
      "learning_rate": 2e-05,
      "loss": 0.1932,
      "rl_loss": -0.0185,
      "step": 570
    },
    {
      "acc": 0.0,
      "epoch": 0.39,
      "learning_rate": 2e-05,
      "loss": 0.207,
      "rl_loss": -0.0175,
      "step": 580
    },
    {
      "acc": 0.0,
      "epoch": 0.39,
      "learning_rate": 2e-05,
      "loss": 0.1869,
      "rl_loss": -0.0559,
      "step": 590
    },
    {
      "acc": 0.0,
      "epoch": 0.4,
      "learning_rate": 2e-05,
      "loss": 0.2019,
      "rl_loss": -0.009,
      "step": 600
    },
    {
      "acc": 0.0,
      "epoch": 0.41,
      "learning_rate": 2e-05,
      "loss": 0.2078,
      "rl_loss": -0.0115,
      "step": 610
    },
    {
      "acc": 0.0,
      "epoch": 0.41,
      "learning_rate": 2e-05,
      "loss": 0.1918,
      "rl_loss": -0.0094,
      "step": 620
    },
    {
      "acc": 0.0,
      "epoch": 0.42,
      "learning_rate": 2e-05,
      "loss": 0.2049,
      "rl_loss": -0.0436,
      "step": 630
    },
    {
      "acc": 0.0,
      "epoch": 0.43,
      "learning_rate": 2e-05,
      "loss": 0.1891,
      "rl_loss": -0.0128,
      "step": 640
    },
    {
      "acc": 0.0,
      "epoch": 0.43,
      "learning_rate": 2e-05,
      "loss": 0.1971,
      "rl_loss": -0.0107,
      "step": 650
    },
    {
      "acc": 0.0,
      "epoch": 0.44,
      "learning_rate": 2e-05,
      "loss": 0.2326,
      "rl_loss": -0.0503,
      "step": 660
    },
    {
      "acc": 0.0,
      "epoch": 0.45,
      "learning_rate": 2e-05,
      "loss": 0.2079,
      "rl_loss": -0.0093,
      "step": 670
    },
    {
      "acc": 0.0,
      "epoch": 0.45,
      "learning_rate": 2e-05,
      "loss": 0.1986,
      "rl_loss": -0.0099,
      "step": 680
    },
    {
      "acc": 0.0,
      "epoch": 0.46,
      "learning_rate": 2e-05,
      "loss": 0.2051,
      "rl_loss": -0.0126,
      "step": 690
    },
    {
      "acc": 0.0,
      "epoch": 0.47,
      "learning_rate": 2e-05,
      "loss": 0.2038,
      "rl_loss": -0.0255,
      "step": 700
    },
    {
      "acc": 0.0,
      "epoch": 0.47,
      "learning_rate": 2e-05,
      "loss": 0.1982,
      "rl_loss": -0.1333,
      "step": 710
    },
    {
      "acc": 0.0,
      "epoch": 0.48,
      "learning_rate": 2e-05,
      "loss": 0.1991,
      "rl_loss": -0.0114,
      "step": 720
    },
    {
      "acc": 0.0,
      "epoch": 0.49,
      "learning_rate": 2e-05,
      "loss": 0.1953,
      "rl_loss": -0.0084,
      "step": 730
    },
    {
      "acc": 0.0,
      "epoch": 0.49,
      "learning_rate": 2e-05,
      "loss": 0.1884,
      "rl_loss": -0.0047,
      "step": 740
    },
    {
      "acc": 0.0,
      "epoch": 0.5,
      "learning_rate": 2e-05,
      "loss": 0.2078,
      "rl_loss": -0.0588,
      "step": 750
    },
    {
      "acc": 0.0,
      "epoch": 0.51,
      "learning_rate": 2e-05,
      "loss": 0.2105,
      "rl_loss": -0.0238,
      "step": 760
    },
    {
      "acc": 0.0,
      "epoch": 0.51,
      "learning_rate": 2e-05,
      "loss": 0.1908,
      "rl_loss": -0.0122,
      "step": 770
    },
    {
      "acc": 0.0,
      "epoch": 0.52,
      "learning_rate": 2e-05,
      "loss": 0.1997,
      "rl_loss": -0.039,
      "step": 780
    },
    {
      "acc": 0.0,
      "epoch": 0.53,
      "learning_rate": 2e-05,
      "loss": 0.1942,
      "rl_loss": -0.0666,
      "step": 790
    },
    {
      "acc": 0.0,
      "epoch": 0.53,
      "learning_rate": 2e-05,
      "loss": 0.2019,
      "rl_loss": -0.0694,
      "step": 800
    },
    {
      "acc": 0.0,
      "epoch": 0.54,
      "learning_rate": 2e-05,
      "loss": 0.2407,
      "rl_loss": -0.0509,
      "step": 810
    },
    {
      "acc": 0.0,
      "epoch": 0.55,
      "learning_rate": 2e-05,
      "loss": 0.2672,
      "rl_loss": 0.0018,
      "step": 820
    },
    {
      "acc": 0.0,
      "epoch": 0.55,
      "learning_rate": 2e-05,
      "loss": 0.2007,
      "rl_loss": -0.0075,
      "step": 830
    },
    {
      "acc": 0.0,
      "epoch": 0.56,
      "learning_rate": 2e-05,
      "loss": 0.2021,
      "rl_loss": -0.072,
      "step": 840
    },
    {
      "acc": 0.0,
      "epoch": 0.57,
      "learning_rate": 2e-05,
      "loss": 0.2401,
      "rl_loss": -0.0257,
      "step": 850
    },
    {
      "acc": 0.0,
      "epoch": 0.57,
      "learning_rate": 2e-05,
      "loss": 0.2168,
      "rl_loss": -0.0371,
      "step": 860
    },
    {
      "acc": 0.0,
      "epoch": 0.58,
      "learning_rate": 2e-05,
      "loss": 0.2008,
      "rl_loss": -0.0549,
      "step": 870
    },
    {
      "acc": 0.0,
      "epoch": 0.59,
      "learning_rate": 2e-05,
      "loss": 0.1992,
      "rl_loss": -0.0534,
      "step": 880
    },
    {
      "acc": 0.0,
      "epoch": 0.59,
      "learning_rate": 2e-05,
      "loss": 0.2226,
      "rl_loss": -0.0824,
      "step": 890
    },
    {
      "acc": 0.0,
      "epoch": 0.6,
      "learning_rate": 2e-05,
      "loss": 0.2067,
      "rl_loss": -0.0092,
      "step": 900
    },
    {
      "acc": 0.0,
      "epoch": 0.61,
      "learning_rate": 2e-05,
      "loss": 0.2192,
      "rl_loss": -0.0334,
      "step": 910
    },
    {
      "acc": 0.0,
      "epoch": 0.61,
      "learning_rate": 2e-05,
      "loss": 0.2127,
      "rl_loss": -0.0402,
      "step": 920
    },
    {
      "acc": 0.0,
      "epoch": 0.62,
      "learning_rate": 2e-05,
      "loss": 0.2399,
      "rl_loss": -0.0901,
      "step": 930
    },
    {
      "acc": 0.0,
      "epoch": 0.63,
      "learning_rate": 2e-05,
      "loss": 0.2302,
      "rl_loss": -0.0274,
      "step": 940
    },
    {
      "acc": 0.0,
      "epoch": 0.63,
      "learning_rate": 2e-05,
      "loss": 0.2229,
      "rl_loss": -0.0317,
      "step": 950
    },
    {
      "acc": 0.0,
      "epoch": 0.64,
      "learning_rate": 2e-05,
      "loss": 0.1961,
      "rl_loss": 0.0009,
      "step": 960
    },
    {
      "acc": 0.0,
      "epoch": 0.65,
      "learning_rate": 2e-05,
      "loss": 0.1909,
      "rl_loss": -0.015,
      "step": 970
    },
    {
      "acc": 0.0,
      "epoch": 0.65,
      "learning_rate": 2e-05,
      "loss": 0.2071,
      "rl_loss": -0.0675,
      "step": 980
    },
    {
      "acc": 0.0,
      "epoch": 0.66,
      "learning_rate": 2e-05,
      "loss": 0.2073,
      "rl_loss": -0.0286,
      "step": 990
    },
    {
      "acc": 0.0,
      "epoch": 0.67,
      "learning_rate": 2e-05,
      "loss": 0.2251,
      "rl_loss": 0.0071,
      "step": 1000
    },
    {
      "acc": 0.0,
      "epoch": 0.67,
      "learning_rate": 2e-05,
      "loss": 0.1874,
      "rl_loss": -0.0176,
      "step": 1010
    },
    {
      "acc": 0.0,
      "epoch": 0.68,
      "learning_rate": 2e-05,
      "loss": 0.2342,
      "rl_loss": -0.0224,
      "step": 1020
    },
    {
      "acc": 0.0,
      "epoch": 0.69,
      "learning_rate": 2e-05,
      "loss": 0.2244,
      "rl_loss": -0.0204,
      "step": 1030
    },
    {
      "acc": 0.0,
      "epoch": 0.69,
      "learning_rate": 2e-05,
      "loss": 0.2059,
      "rl_loss": -0.0206,
      "step": 1040
    },
    {
      "acc": 0.0,
      "epoch": 0.7,
      "learning_rate": 2e-05,
      "loss": 0.1919,
      "rl_loss": -0.0619,
      "step": 1050
    },
    {
      "acc": 0.0,
      "epoch": 0.71,
      "learning_rate": 2e-05,
      "loss": 0.2103,
      "rl_loss": -0.0141,
      "step": 1060
    },
    {
      "acc": 0.0,
      "epoch": 0.71,
      "learning_rate": 2e-05,
      "loss": 0.2018,
      "rl_loss": -0.0241,
      "step": 1070
    },
    {
      "acc": 0.0,
      "epoch": 0.72,
      "learning_rate": 2e-05,
      "loss": 0.2551,
      "rl_loss": -0.0541,
      "step": 1080
    },
    {
      "acc": 0.0,
      "epoch": 0.73,
      "learning_rate": 2e-05,
      "loss": 0.201,
      "rl_loss": -0.0098,
      "step": 1090
    },
    {
      "acc": 0.0,
      "epoch": 0.73,
      "learning_rate": 2e-05,
      "loss": 0.2008,
      "rl_loss": -0.0105,
      "step": 1100
    },
    {
      "acc": 0.0,
      "epoch": 0.74,
      "learning_rate": 2e-05,
      "loss": 0.2028,
      "rl_loss": -0.05,
      "step": 1110
    },
    {
      "acc": 0.0,
      "epoch": 0.75,
      "learning_rate": 2e-05,
      "loss": 0.1966,
      "rl_loss": -0.006,
      "step": 1120
    },
    {
      "acc": 0.0,
      "epoch": 0.75,
      "learning_rate": 2e-05,
      "loss": 0.2089,
      "rl_loss": -0.0066,
      "step": 1130
    },
    {
      "acc": 0.0,
      "epoch": 0.76,
      "learning_rate": 2e-05,
      "loss": 0.1951,
      "rl_loss": -0.0268,
      "step": 1140
    },
    {
      "acc": 0.0,
      "epoch": 0.77,
      "learning_rate": 2e-05,
      "loss": 0.1885,
      "rl_loss": -0.0478,
      "step": 1150
    },
    {
      "acc": 0.0,
      "epoch": 0.77,
      "learning_rate": 2e-05,
      "loss": 0.2258,
      "rl_loss": -0.0383,
      "step": 1160
    },
    {
      "acc": 0.0,
      "epoch": 0.78,
      "learning_rate": 2e-05,
      "loss": 0.2002,
      "rl_loss": -0.0201,
      "step": 1170
    },
    {
      "acc": 0.0,
      "epoch": 0.79,
      "learning_rate": 2e-05,
      "loss": 0.2319,
      "rl_loss": -0.1138,
      "step": 1180
    },
    {
      "acc": 0.0,
      "epoch": 0.79,
      "learning_rate": 2e-05,
      "loss": 0.2828,
      "rl_loss": -0.0367,
      "step": 1190
    },
    {
      "acc": 0.0,
      "epoch": 0.8,
      "learning_rate": 2e-05,
      "loss": 0.2212,
      "rl_loss": -0.1415,
      "step": 1200
    },
    {
      "acc": 0.0,
      "epoch": 0.81,
      "learning_rate": 2e-05,
      "loss": 0.2619,
      "rl_loss": -0.0687,
      "step": 1210
    },
    {
      "acc": 0.0,
      "epoch": 0.81,
      "learning_rate": 2e-05,
      "loss": 0.2034,
      "rl_loss": -0.0475,
      "step": 1220
    },
    {
      "acc": 0.0,
      "epoch": 0.82,
      "learning_rate": 2e-05,
      "loss": 0.2583,
      "rl_loss": -0.2057,
      "step": 1230
    },
    {
      "acc": 0.0,
      "epoch": 0.83,
      "learning_rate": 2e-05,
      "loss": 0.2258,
      "rl_loss": -0.0758,
      "step": 1240
    },
    {
      "acc": 0.0,
      "epoch": 0.83,
      "learning_rate": 2e-05,
      "loss": 0.2015,
      "rl_loss": -0.2222,
      "step": 1250
    },
    {
      "acc": 0.0,
      "epoch": 0.84,
      "learning_rate": 2e-05,
      "loss": 0.2088,
      "rl_loss": -0.0407,
      "step": 1260
    },
    {
      "acc": 0.0,
      "epoch": 0.85,
      "learning_rate": 2e-05,
      "loss": 0.1986,
      "rl_loss": -0.0952,
      "step": 1270
    },
    {
      "acc": 0.0,
      "epoch": 0.85,
      "learning_rate": 2e-05,
      "loss": 0.2644,
      "rl_loss": -0.0654,
      "step": 1280
    },
    {
      "acc": 0.0,
      "epoch": 0.86,
      "learning_rate": 2e-05,
      "loss": 0.2119,
      "rl_loss": -0.099,
      "step": 1290
    },
    {
      "acc": 0.0,
      "epoch": 0.87,
      "learning_rate": 2e-05,
      "loss": 0.2162,
      "rl_loss": 0.003,
      "step": 1300
    },
    {
      "acc": 0.0,
      "epoch": 0.87,
      "learning_rate": 2e-05,
      "loss": 0.2345,
      "rl_loss": -0.1278,
      "step": 1310
    },
    {
      "acc": 0.0,
      "epoch": 0.88,
      "learning_rate": 2e-05,
      "loss": 0.2147,
      "rl_loss": 0.0178,
      "step": 1320
    },
    {
      "acc": 0.0,
      "epoch": 0.89,
      "learning_rate": 2e-05,
      "loss": 0.3013,
      "rl_loss": -0.0425,
      "step": 1330
    },
    {
      "acc": 0.0,
      "epoch": 0.89,
      "learning_rate": 2e-05,
      "loss": 0.2049,
      "rl_loss": -0.0238,
      "step": 1340
    },
    {
      "acc": 0.0,
      "epoch": 0.9,
      "learning_rate": 2e-05,
      "loss": 0.2489,
      "rl_loss": -0.0539,
      "step": 1350
    },
    {
      "acc": 0.0,
      "epoch": 0.91,
      "learning_rate": 2e-05,
      "loss": 0.2009,
      "rl_loss": -0.0724,
      "step": 1360
    },
    {
      "acc": 0.0,
      "epoch": 0.91,
      "learning_rate": 2e-05,
      "loss": 0.2411,
      "rl_loss": -0.1009,
      "step": 1370
    },
    {
      "acc": 0.0,
      "epoch": 0.92,
      "learning_rate": 2e-05,
      "loss": 0.2153,
      "rl_loss": -0.059,
      "step": 1380
    },
    {
      "acc": 0.0,
      "epoch": 0.93,
      "learning_rate": 2e-05,
      "loss": 0.2806,
      "rl_loss": -0.0582,
      "step": 1390
    },
    {
      "acc": 0.0,
      "epoch": 0.93,
      "learning_rate": 2e-05,
      "loss": 0.211,
      "rl_loss": -0.0547,
      "step": 1400
    },
    {
      "acc": 0.0,
      "epoch": 0.94,
      "learning_rate": 2e-05,
      "loss": 0.2064,
      "rl_loss": -0.0155,
      "step": 1410
    },
    {
      "acc": 0.0,
      "epoch": 0.95,
      "learning_rate": 2e-05,
      "loss": 0.2573,
      "rl_loss": -0.206,
      "step": 1420
    },
    {
      "acc": 0.0,
      "epoch": 0.95,
      "learning_rate": 2e-05,
      "loss": 0.3427,
      "rl_loss": -0.4052,
      "step": 1430
    },
    {
      "acc": 0.0,
      "epoch": 0.96,
      "learning_rate": 2e-05,
      "loss": 0.2292,
      "rl_loss": -0.0413,
      "step": 1440
    },
    {
      "acc": 0.0,
      "epoch": 0.97,
      "learning_rate": 2e-05,
      "loss": 0.2241,
      "rl_loss": -0.0907,
      "step": 1450
    },
    {
      "acc": 0.0,
      "epoch": 0.97,
      "learning_rate": 2e-05,
      "loss": 0.2004,
      "rl_loss": -0.2518,
      "step": 1460
    },
    {
      "acc": 0.0,
      "epoch": 0.98,
      "learning_rate": 2e-05,
      "loss": 0.2882,
      "rl_loss": -0.1383,
      "step": 1470
    },
    {
      "acc": 0.0,
      "epoch": 0.99,
      "learning_rate": 2e-05,
      "loss": 0.4526,
      "rl_loss": -0.3262,
      "step": 1480
    },
    {
      "acc": 0.0,
      "epoch": 0.99,
      "learning_rate": 2e-05,
      "loss": 0.2733,
      "rl_loss": -0.078,
      "step": 1490
    },
    {
      "acc": 0.0,
      "epoch": 1.0,
      "learning_rate": 2e-05,
      "loss": 0.24,
      "rl_loss": -0.0143,
      "step": 1500
    },
    {
      "acc": 0.0,
      "epoch": 1.01,
      "learning_rate": 2e-05,
      "loss": 0.2716,
      "rl_loss": -0.1058,
      "step": 1510
    },
    {
      "acc": 0.0,
      "epoch": 1.01,
      "learning_rate": 2e-05,
      "loss": 0.3491,
      "rl_loss": -0.2102,
      "step": 1520
    },
    {
      "acc": 0.0,
      "epoch": 1.02,
      "learning_rate": 2e-05,
      "loss": 0.2348,
      "rl_loss": -0.1394,
      "step": 1530
    },
    {
      "acc": 0.0,
      "epoch": 1.03,
      "learning_rate": 2e-05,
      "loss": 0.2861,
      "rl_loss": -0.1298,
      "step": 1540
    },
    {
      "acc": 0.0,
      "epoch": 1.03,
      "learning_rate": 2e-05,
      "loss": 0.2077,
      "rl_loss": -0.1497,
      "step": 1550
    },
    {
      "acc": 0.0,
      "epoch": 1.04,
      "learning_rate": 2e-05,
      "loss": 0.2019,
      "rl_loss": -0.2238,
      "step": 1560
    },
    {
      "acc": 0.0,
      "epoch": 1.05,
      "learning_rate": 2e-05,
      "loss": 0.2567,
      "rl_loss": -0.1693,
      "step": 1570
    },
    {
      "acc": 0.0,
      "epoch": 1.05,
      "learning_rate": 2e-05,
      "loss": 0.2814,
      "rl_loss": -0.1605,
      "step": 1580
    },
    {
      "acc": 0.0,
      "epoch": 1.06,
      "learning_rate": 2e-05,
      "loss": 0.2215,
      "rl_loss": -0.1134,
      "step": 1590
    },
    {
      "acc": 0.0,
      "epoch": 1.07,
      "learning_rate": 2e-05,
      "loss": 0.2175,
      "rl_loss": -0.3902,
      "step": 1600
    },
    {
      "acc": 0.0,
      "epoch": 1.07,
      "learning_rate": 2e-05,
      "loss": 0.2196,
      "rl_loss": -0.301,
      "step": 1610
    },
    {
      "acc": 0.0,
      "epoch": 1.08,
      "learning_rate": 2e-05,
      "loss": 0.292,
      "rl_loss": -0.213,
      "step": 1620
    },
    {
      "acc": 0.0,
      "epoch": 1.09,
      "learning_rate": 2e-05,
      "loss": 0.2572,
      "rl_loss": -0.1575,
      "step": 1630
    },
    {
      "acc": 0.0,
      "epoch": 1.09,
      "learning_rate": 2e-05,
      "loss": 0.2362,
      "rl_loss": -0.1837,
      "step": 1640
    },
    {
      "acc": 0.0,
      "epoch": 1.1,
      "learning_rate": 2e-05,
      "loss": 0.4009,
      "rl_loss": -0.2471,
      "step": 1650
    },
    {
      "acc": 0.0,
      "epoch": 1.11,
      "learning_rate": 2e-05,
      "loss": 0.2692,
      "rl_loss": -0.1602,
      "step": 1660
    },
    {
      "acc": 0.0,
      "epoch": 1.11,
      "learning_rate": 2e-05,
      "loss": 0.3368,
      "rl_loss": -0.3032,
      "step": 1670
    },
    {
      "acc": 0.0,
      "epoch": 1.12,
      "learning_rate": 2e-05,
      "loss": 0.2074,
      "rl_loss": -0.1879,
      "step": 1680
    },
    {
      "acc": 0.0,
      "epoch": 1.13,
      "learning_rate": 2e-05,
      "loss": 0.2025,
      "rl_loss": -0.2248,
      "step": 1690
    },
    {
      "acc": 0.0,
      "epoch": 1.13,
      "learning_rate": 2e-05,
      "loss": 0.2245,
      "rl_loss": -0.1585,
      "step": 1700
    },
    {
      "acc": 0.0,
      "epoch": 1.14,
      "learning_rate": 2e-05,
      "loss": 0.2777,
      "rl_loss": -0.2912,
      "step": 1710
    },
    {
      "acc": 0.0,
      "epoch": 1.15,
      "learning_rate": 2e-05,
      "loss": 0.2496,
      "rl_loss": -0.3379,
      "step": 1720
    },
    {
      "acc": 0.0,
      "epoch": 1.15,
      "learning_rate": 2e-05,
      "loss": 0.2507,
      "rl_loss": -0.1577,
      "step": 1730
    },
    {
      "acc": 0.0,
      "epoch": 1.16,
      "learning_rate": 2e-05,
      "loss": 0.2275,
      "rl_loss": -0.2425,
      "step": 1740
    },
    {
      "acc": 0.0,
      "epoch": 1.17,
      "learning_rate": 2e-05,
      "loss": 0.2019,
      "rl_loss": -0.3212,
      "step": 1750
    },
    {
      "acc": 0.0,
      "epoch": 1.17,
      "learning_rate": 2e-05,
      "loss": 0.2525,
      "rl_loss": -0.1072,
      "step": 1760
    },
    {
      "acc": 0.0,
      "epoch": 1.18,
      "learning_rate": 2e-05,
      "loss": 0.224,
      "rl_loss": -0.3149,
      "step": 1770
    },
    {
      "acc": 0.0,
      "epoch": 1.19,
      "learning_rate": 2e-05,
      "loss": 0.3125,
      "rl_loss": -0.2268,
      "step": 1780
    },
    {
      "acc": 0.0,
      "epoch": 1.19,
      "learning_rate": 2e-05,
      "loss": 0.2726,
      "rl_loss": -0.3026,
      "step": 1790
    },
    {
      "acc": 0.0,
      "epoch": 1.2,
      "learning_rate": 2e-05,
      "loss": 0.2067,
      "rl_loss": -0.0679,
      "step": 1800
    },
    {
      "acc": 0.0,
      "epoch": 1.21,
      "learning_rate": 2e-05,
      "loss": 0.2192,
      "rl_loss": -0.2496,
      "step": 1810
    },
    {
      "acc": 0.0,
      "epoch": 1.21,
      "learning_rate": 2e-05,
      "loss": 0.2192,
      "rl_loss": -0.3184,
      "step": 1820
    },
    {
      "acc": 0.0,
      "epoch": 1.22,
      "learning_rate": 2e-05,
      "loss": 0.2152,
      "rl_loss": -0.0508,
      "step": 1830
    },
    {
      "acc": 0.0,
      "epoch": 1.23,
      "learning_rate": 2e-05,
      "loss": 0.3504,
      "rl_loss": -0.1447,
      "step": 1840
    },
    {
      "acc": 0.0,
      "epoch": 1.23,
      "learning_rate": 2e-05,
      "loss": 0.2511,
      "rl_loss": -0.1003,
      "step": 1850
    },
    {
      "acc": 0.0,
      "epoch": 1.24,
      "learning_rate": 2e-05,
      "loss": 0.2549,
      "rl_loss": -0.2655,
      "step": 1860
    },
    {
      "acc": 0.0,
      "epoch": 1.25,
      "learning_rate": 2e-05,
      "loss": 0.2788,
      "rl_loss": -0.1448,
      "step": 1870
    },
    {
      "acc": 0.0,
      "epoch": 1.25,
      "learning_rate": 2e-05,
      "loss": 0.2365,
      "rl_loss": -0.3471,
      "step": 1880
    },
    {
      "acc": 0.0,
      "epoch": 1.26,
      "learning_rate": 2e-05,
      "loss": 0.2466,
      "rl_loss": -0.2494,
      "step": 1890
    },
    {
      "acc": 0.0,
      "epoch": 1.27,
      "learning_rate": 2e-05,
      "loss": 0.2206,
      "rl_loss": -0.4574,
      "step": 1900
    },
    {
      "acc": 0.0,
      "epoch": 1.27,
      "learning_rate": 2e-05,
      "loss": 0.3148,
      "rl_loss": -0.1643,
      "step": 1910
    },
    {
      "acc": 0.0,
      "epoch": 1.28,
      "learning_rate": 2e-05,
      "loss": 0.262,
      "rl_loss": -0.2892,
      "step": 1920
    },
    {
      "acc": 0.0,
      "epoch": 1.29,
      "learning_rate": 2e-05,
      "loss": 0.339,
      "rl_loss": -0.0317,
      "step": 1930
    },
    {
      "acc": 0.0,
      "epoch": 1.29,
      "learning_rate": 2e-05,
      "loss": 0.2181,
      "rl_loss": -0.2939,
      "step": 1940
    },
    {
      "acc": 0.0,
      "epoch": 1.3,
      "learning_rate": 2e-05,
      "loss": 0.3218,
      "rl_loss": -0.1217,
      "step": 1950
    },
    {
      "acc": 0.0,
      "epoch": 1.31,
      "learning_rate": 2e-05,
      "loss": 0.2099,
      "rl_loss": -0.4231,
      "step": 1960
    },
    {
      "acc": 0.0,
      "epoch": 1.31,
      "learning_rate": 2e-05,
      "loss": 0.3301,
      "rl_loss": -0.1967,
      "step": 1970
    },
    {
      "acc": 0.0,
      "epoch": 1.32,
      "learning_rate": 2e-05,
      "loss": 0.3348,
      "rl_loss": -0.2999,
      "step": 1980
    },
    {
      "acc": 0.0,
      "epoch": 1.33,
      "learning_rate": 2e-05,
      "loss": 0.2197,
      "rl_loss": -0.2697,
      "step": 1990
    },
    {
      "acc": 0.0,
      "epoch": 1.33,
      "learning_rate": 2e-05,
      "loss": 0.2636,
      "rl_loss": -0.1534,
      "step": 2000
    },
    {
      "acc": 0.0,
      "epoch": 1.34,
      "learning_rate": 2e-05,
      "loss": 0.3831,
      "rl_loss": -0.3402,
      "step": 2010
    },
    {
      "acc": 0.0,
      "epoch": 1.35,
      "learning_rate": 2e-05,
      "loss": 0.2709,
      "rl_loss": -0.2815,
      "step": 2020
    },
    {
      "acc": 0.0,
      "epoch": 1.35,
      "learning_rate": 2e-05,
      "loss": 0.4735,
      "rl_loss": -0.1193,
      "step": 2030
    },
    {
      "acc": 0.0,
      "epoch": 1.36,
      "learning_rate": 2e-05,
      "loss": 0.3152,
      "rl_loss": -0.3396,
      "step": 2040
    },
    {
      "acc": 0.0,
      "epoch": 1.37,
      "learning_rate": 2e-05,
      "loss": 0.213,
      "rl_loss": -0.1602,
      "step": 2050
    },
    {
      "acc": 0.0,
      "epoch": 1.37,
      "learning_rate": 2e-05,
      "loss": 0.2118,
      "rl_loss": -0.2494,
      "step": 2060
    },
    {
      "acc": 0.0,
      "epoch": 1.38,
      "learning_rate": 2e-05,
      "loss": 0.2877,
      "rl_loss": -0.2309,
      "step": 2070
    },
    {
      "acc": 0.0,
      "epoch": 1.39,
      "learning_rate": 2e-05,
      "loss": 0.2999,
      "rl_loss": -0.2919,
      "step": 2080
    },
    {
      "acc": 0.0,
      "epoch": 1.39,
      "learning_rate": 2e-05,
      "loss": 0.3178,
      "rl_loss": -0.3203,
      "step": 2090
    },
    {
      "acc": 0.0,
      "epoch": 1.4,
      "learning_rate": 2e-05,
      "loss": 0.2346,
      "rl_loss": -0.3028,
      "step": 2100
    },
    {
      "acc": 0.0,
      "epoch": 1.41,
      "learning_rate": 2e-05,
      "loss": 0.2955,
      "rl_loss": -0.2407,
      "step": 2110
    },
    {
      "acc": 0.0,
      "epoch": 1.41,
      "learning_rate": 2e-05,
      "loss": 0.3054,
      "rl_loss": -0.3303,
      "step": 2120
    },
    {
      "acc": 0.0,
      "epoch": 1.42,
      "learning_rate": 2e-05,
      "loss": 0.221,
      "rl_loss": -0.0344,
      "step": 2130
    },
    {
      "acc": 0.0,
      "epoch": 1.43,
      "learning_rate": 2e-05,
      "loss": 0.2423,
      "rl_loss": -0.4771,
      "step": 2140
    },
    {
      "acc": 0.0,
      "epoch": 1.43,
      "learning_rate": 2e-05,
      "loss": 0.3343,
      "rl_loss": -0.7419,
      "step": 2150
    },
    {
      "acc": 0.0,
      "epoch": 1.44,
      "learning_rate": 2e-05,
      "loss": 0.2997,
      "rl_loss": -0.183,
      "step": 2160
    },
    {
      "acc": 0.0,
      "epoch": 1.45,
      "learning_rate": 2e-05,
      "loss": 0.2399,
      "rl_loss": -0.2342,
      "step": 2170
    },
    {
      "acc": 0.0,
      "epoch": 1.45,
      "learning_rate": 2e-05,
      "loss": 0.2391,
      "rl_loss": -0.3045,
      "step": 2180
    },
    {
      "acc": 0.0,
      "epoch": 1.46,
      "learning_rate": 2e-05,
      "loss": 0.4321,
      "rl_loss": -0.3114,
      "step": 2190
    },
    {
      "acc": 0.0,
      "epoch": 1.47,
      "learning_rate": 2e-05,
      "loss": 0.2387,
      "rl_loss": -0.2203,
      "step": 2200
    },
    {
      "acc": 0.0,
      "epoch": 1.47,
      "learning_rate": 2e-05,
      "loss": 0.2195,
      "rl_loss": -0.1261,
      "step": 2210
    },
    {
      "acc": 0.0,
      "epoch": 1.48,
      "learning_rate": 2e-05,
      "loss": 0.2744,
      "rl_loss": -0.2352,
      "step": 2220
    },
    {
      "acc": 0.0,
      "epoch": 1.49,
      "learning_rate": 2e-05,
      "loss": 0.56,
      "rl_loss": -0.4791,
      "step": 2230
    },
    {
      "acc": 0.0,
      "epoch": 1.49,
      "learning_rate": 2e-05,
      "loss": 0.2127,
      "rl_loss": -0.0956,
      "step": 2240
    },
    {
      "acc": 0.0,
      "epoch": 1.5,
      "learning_rate": 2e-05,
      "loss": 0.3765,
      "rl_loss": -0.1513,
      "step": 2250
    },
    {
      "acc": 0.0,
      "epoch": 1.51,
      "learning_rate": 2e-05,
      "loss": 0.2626,
      "rl_loss": -0.0945,
      "step": 2260
    },
    {
      "acc": 0.0,
      "epoch": 1.51,
      "learning_rate": 2e-05,
      "loss": 0.2584,
      "rl_loss": -0.2644,
      "step": 2270
    },
    {
      "acc": 0.0,
      "epoch": 1.52,
      "learning_rate": 2e-05,
      "loss": 0.3124,
      "rl_loss": -0.1856,
      "step": 2280
    },
    {
      "acc": 0.0,
      "epoch": 1.53,
      "learning_rate": 2e-05,
      "loss": 0.3314,
      "rl_loss": -0.1497,
      "step": 2290
    },
    {
      "acc": 0.0,
      "epoch": 1.53,
      "learning_rate": 2e-05,
      "loss": 0.222,
      "rl_loss": -0.2127,
      "step": 2300
    },
    {
      "acc": 0.0,
      "epoch": 1.54,
      "learning_rate": 2e-05,
      "loss": 0.2053,
      "rl_loss": -0.2785,
      "step": 2310
    },
    {
      "acc": 0.0,
      "epoch": 1.55,
      "learning_rate": 2e-05,
      "loss": 0.2848,
      "rl_loss": -0.1348,
      "step": 2320
    },
    {
      "acc": 0.0,
      "epoch": 1.55,
      "learning_rate": 2e-05,
      "loss": 0.2008,
      "rl_loss": -0.3603,
      "step": 2330
    },
    {
      "acc": 0.0,
      "epoch": 1.56,
      "learning_rate": 2e-05,
      "loss": 0.334,
      "rl_loss": -0.1207,
      "step": 2340
    },
    {
      "acc": 0.0,
      "epoch": 1.57,
      "learning_rate": 2e-05,
      "loss": 0.2366,
      "rl_loss": -0.556,
      "step": 2350
    },
    {
      "acc": 0.0,
      "epoch": 1.57,
      "learning_rate": 2e-05,
      "loss": 0.3089,
      "rl_loss": -0.1413,
      "step": 2360
    },
    {
      "acc": 0.0,
      "epoch": 1.58,
      "learning_rate": 2e-05,
      "loss": 0.2942,
      "rl_loss": -0.1651,
      "step": 2370
    },
    {
      "acc": 0.0,
      "epoch": 1.59,
      "learning_rate": 2e-05,
      "loss": 0.2984,
      "rl_loss": -0.3665,
      "step": 2380
    },
    {
      "acc": 0.0,
      "epoch": 1.59,
      "learning_rate": 2e-05,
      "loss": 0.2923,
      "rl_loss": -0.1287,
      "step": 2390
    },
    {
      "acc": 0.0,
      "epoch": 1.6,
      "learning_rate": 2e-05,
      "loss": 0.2936,
      "rl_loss": -0.2642,
      "step": 2400
    },
    {
      "acc": 0.0,
      "epoch": 1.61,
      "learning_rate": 2e-05,
      "loss": 0.2817,
      "rl_loss": -0.0497,
      "step": 2410
    },
    {
      "acc": 0.0,
      "epoch": 1.61,
      "learning_rate": 2e-05,
      "loss": 0.2176,
      "rl_loss": -0.1427,
      "step": 2420
    },
    {
      "acc": 0.0,
      "epoch": 1.62,
      "learning_rate": 2e-05,
      "loss": 0.2181,
      "rl_loss": -0.2153,
      "step": 2430
    },
    {
      "acc": 0.0,
      "epoch": 1.63,
      "learning_rate": 2e-05,
      "loss": 0.353,
      "rl_loss": -0.0923,
      "step": 2440
    },
    {
      "acc": 0.0,
      "epoch": 1.63,
      "learning_rate": 2e-05,
      "loss": 0.2725,
      "rl_loss": -0.1337,
      "step": 2450
    },
    {
      "acc": 0.0,
      "epoch": 1.64,
      "learning_rate": 2e-05,
      "loss": 0.3012,
      "rl_loss": -0.2073,
      "step": 2460
    },
    {
      "acc": 0.0,
      "epoch": 1.65,
      "learning_rate": 2e-05,
      "loss": 0.4572,
      "rl_loss": -0.2452,
      "step": 2470
    },
    {
      "acc": 0.0,
      "epoch": 1.65,
      "learning_rate": 2e-05,
      "loss": 0.3193,
      "rl_loss": -0.322,
      "step": 2480
    },
    {
      "acc": 0.0,
      "epoch": 1.66,
      "learning_rate": 2e-05,
      "loss": 0.2718,
      "rl_loss": -0.1434,
      "step": 2490
    },
    {
      "acc": 0.0,
      "epoch": 1.67,
      "learning_rate": 2e-05,
      "loss": 0.2396,
      "rl_loss": -0.3019,
      "step": 2500
    },
    {
      "acc": 0.0,
      "epoch": 1.67,
      "learning_rate": 2e-05,
      "loss": 0.2117,
      "rl_loss": -0.1138,
      "step": 2510
    },
    {
      "acc": 0.0,
      "epoch": 1.68,
      "learning_rate": 2e-05,
      "loss": 0.305,
      "rl_loss": -0.1993,
      "step": 2520
    },
    {
      "acc": 0.0,
      "epoch": 1.69,
      "learning_rate": 2e-05,
      "loss": 0.2979,
      "rl_loss": -0.1591,
      "step": 2530
    },
    {
      "acc": 0.0,
      "epoch": 1.69,
      "learning_rate": 2e-05,
      "loss": 0.2312,
      "rl_loss": -0.0862,
      "step": 2540
    },
    {
      "acc": 0.0,
      "epoch": 1.7,
      "learning_rate": 2e-05,
      "loss": 0.2726,
      "rl_loss": -0.3785,
      "step": 2550
    },
    {
      "acc": 0.0,
      "epoch": 1.71,
      "learning_rate": 2e-05,
      "loss": 0.3373,
      "rl_loss": -0.0622,
      "step": 2560
    },
    {
      "acc": 0.0,
      "epoch": 1.71,
      "learning_rate": 2e-05,
      "loss": 0.2486,
      "rl_loss": -0.0971,
      "step": 2570
    },
    {
      "acc": 0.0,
      "epoch": 1.72,
      "learning_rate": 2e-05,
      "loss": 0.2828,
      "rl_loss": -0.1688,
      "step": 2580
    },
    {
      "acc": 0.0,
      "epoch": 1.73,
      "learning_rate": 2e-05,
      "loss": 0.2374,
      "rl_loss": -0.1622,
      "step": 2590
    },
    {
      "acc": 0.0,
      "epoch": 1.73,
      "learning_rate": 2e-05,
      "loss": 0.428,
      "rl_loss": -0.2429,
      "step": 2600
    },
    {
      "acc": 0.0,
      "epoch": 1.74,
      "learning_rate": 2e-05,
      "loss": 0.265,
      "rl_loss": -0.314,
      "step": 2610
    },
    {
      "acc": 0.0,
      "epoch": 1.75,
      "learning_rate": 2e-05,
      "loss": 0.3008,
      "rl_loss": -0.2175,
      "step": 2620
    },
    {
      "acc": 0.0,
      "epoch": 1.75,
      "learning_rate": 2e-05,
      "loss": 0.212,
      "rl_loss": -0.2465,
      "step": 2630
    },
    {
      "acc": 0.0,
      "epoch": 1.76,
      "learning_rate": 2e-05,
      "loss": 0.2471,
      "rl_loss": -0.0837,
      "step": 2640
    },
    {
      "acc": 0.0,
      "epoch": 1.77,
      "learning_rate": 2e-05,
      "loss": 0.2352,
      "rl_loss": -0.0888,
      "step": 2650
    },
    {
      "acc": 0.0,
      "epoch": 1.77,
      "learning_rate": 2e-05,
      "loss": 0.4229,
      "rl_loss": -0.3808,
      "step": 2660
    },
    {
      "acc": 0.0,
      "epoch": 1.78,
      "learning_rate": 2e-05,
      "loss": 0.2738,
      "rl_loss": -0.1199,
      "step": 2670
    },
    {
      "acc": 0.0,
      "epoch": 1.79,
      "learning_rate": 2e-05,
      "loss": 0.2162,
      "rl_loss": -0.0656,
      "step": 2680
    },
    {
      "acc": 0.0,
      "epoch": 1.79,
      "learning_rate": 2e-05,
      "loss": 0.246,
      "rl_loss": -0.2123,
      "step": 2690
    },
    {
      "acc": 0.0,
      "epoch": 1.8,
      "learning_rate": 2e-05,
      "loss": 0.2648,
      "rl_loss": -0.2322,
      "step": 2700
    },
    {
      "acc": 0.0,
      "epoch": 1.81,
      "learning_rate": 2e-05,
      "loss": 0.2433,
      "rl_loss": -0.0837,
      "step": 2710
    },
    {
      "acc": 0.0,
      "epoch": 1.81,
      "learning_rate": 2e-05,
      "loss": 0.269,
      "rl_loss": -0.0561,
      "step": 2720
    },
    {
      "acc": 0.0,
      "epoch": 1.82,
      "learning_rate": 2e-05,
      "loss": 0.2715,
      "rl_loss": -0.3806,
      "step": 2730
    },
    {
      "acc": 0.0,
      "epoch": 1.83,
      "learning_rate": 2e-05,
      "loss": 0.3139,
      "rl_loss": -0.0817,
      "step": 2740
    },
    {
      "acc": 0.0,
      "epoch": 1.83,
      "learning_rate": 2e-05,
      "loss": 0.2513,
      "rl_loss": -0.2584,
      "step": 2750
    },
    {
      "acc": 0.0,
      "epoch": 1.84,
      "learning_rate": 2e-05,
      "loss": 0.3049,
      "rl_loss": -0.2728,
      "step": 2760
    },
    {
      "acc": 0.0,
      "epoch": 1.85,
      "learning_rate": 2e-05,
      "loss": 0.228,
      "rl_loss": -0.165,
      "step": 2770
    },
    {
      "acc": 0.0,
      "epoch": 1.85,
      "learning_rate": 2e-05,
      "loss": 0.2464,
      "rl_loss": -0.2593,
      "step": 2780
    },
    {
      "acc": 0.0,
      "epoch": 1.86,
      "learning_rate": 2e-05,
      "loss": 0.3339,
      "rl_loss": -0.1801,
      "step": 2790
    },
    {
      "acc": 0.0,
      "epoch": 1.87,
      "learning_rate": 2e-05,
      "loss": 0.2498,
      "rl_loss": -0.0794,
      "step": 2800
    },
    {
      "acc": 0.0,
      "epoch": 1.87,
      "learning_rate": 2e-05,
      "loss": 0.2176,
      "rl_loss": -0.2592,
      "step": 2810
    },
    {
      "acc": 0.0,
      "epoch": 1.88,
      "learning_rate": 2e-05,
      "loss": 0.2339,
      "rl_loss": -0.1273,
      "step": 2820
    },
    {
      "acc": 0.0,
      "epoch": 1.89,
      "learning_rate": 2e-05,
      "loss": 0.3225,
      "rl_loss": -0.2312,
      "step": 2830
    },
    {
      "acc": 0.0,
      "epoch": 1.89,
      "learning_rate": 2e-05,
      "loss": 0.3005,
      "rl_loss": -0.2849,
      "step": 2840
    },
    {
      "acc": 0.0,
      "epoch": 1.9,
      "learning_rate": 2e-05,
      "loss": 0.3244,
      "rl_loss": -0.4541,
      "step": 2850
    },
    {
      "acc": 0.0,
      "epoch": 1.91,
      "learning_rate": 2e-05,
      "loss": 0.2361,
      "rl_loss": -0.4858,
      "step": 2860
    },
    {
      "acc": 0.0,
      "epoch": 1.91,
      "learning_rate": 2e-05,
      "loss": 0.2548,
      "rl_loss": -0.0967,
      "step": 2870
    },
    {
      "acc": 0.0,
      "epoch": 1.92,
      "learning_rate": 2e-05,
      "loss": 0.2556,
      "rl_loss": -0.4954,
      "step": 2880
    },
    {
      "acc": 0.0,
      "epoch": 1.93,
      "learning_rate": 2e-05,
      "loss": 0.2572,
      "rl_loss": -0.19,
      "step": 2890
    },
    {
      "acc": 0.0,
      "epoch": 1.93,
      "learning_rate": 2e-05,
      "loss": 0.3286,
      "rl_loss": -0.2709,
      "step": 2900
    },
    {
      "acc": 0.0,
      "epoch": 1.94,
      "learning_rate": 2e-05,
      "loss": 0.2768,
      "rl_loss": -0.1072,
      "step": 2910
    },
    {
      "acc": 0.0,
      "epoch": 1.95,
      "learning_rate": 2e-05,
      "loss": 0.2248,
      "rl_loss": -0.4267,
      "step": 2920
    },
    {
      "acc": 0.0,
      "epoch": 1.95,
      "learning_rate": 2e-05,
      "loss": 0.2976,
      "rl_loss": -0.3855,
      "step": 2930
    },
    {
      "acc": 0.0,
      "epoch": 1.96,
      "learning_rate": 2e-05,
      "loss": 0.3344,
      "rl_loss": -0.1849,
      "step": 2940
    },
    {
      "acc": 0.0,
      "epoch": 1.97,
      "learning_rate": 2e-05,
      "loss": 0.2909,
      "rl_loss": -0.4777,
      "step": 2950
    },
    {
      "acc": 0.0,
      "epoch": 1.97,
      "learning_rate": 2e-05,
      "loss": 0.368,
      "rl_loss": -0.5148,
      "step": 2960
    },
    {
      "acc": 0.0,
      "epoch": 1.98,
      "learning_rate": 2e-05,
      "loss": 0.2897,
      "rl_loss": -0.4314,
      "step": 2970
    },
    {
      "acc": 0.0,
      "epoch": 1.99,
      "learning_rate": 2e-05,
      "loss": 0.3902,
      "rl_loss": -0.3644,
      "step": 2980
    },
    {
      "acc": 0.0,
      "epoch": 1.99,
      "learning_rate": 2e-05,
      "loss": 0.2532,
      "rl_loss": -0.5152,
      "step": 2990
    },
    {
      "acc": 0.0,
      "epoch": 2.0,
      "learning_rate": 2e-05,
      "loss": 0.2424,
      "rl_loss": -0.534,
      "step": 3000
    }
  ],
  "max_steps": 14990,
  "num_train_epochs": 10,
  "total_flos": 4.873690607616e+17,
  "trial_name": null,
  "trial_params": null
}