diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5716 @@
+{
+  "best_metric": 0.4023992419242859,
+  "best_model_checkpoint": "/scratch/s3545881/dumped/translation/byt5/3124577/checkpoint-6600",
+  "epoch": 6.7736185383244205,
+  "global_step": 7600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5e-06,
+      "loss": 9.97,
+      "step": 10
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5e-06,
+      "loss": 9.212,
+      "step": 20
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 7.5e-06,
+      "loss": 7.8584,
+      "step": 30
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1e-05,
+      "loss": 6.3664,
+      "step": 40
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.25e-05,
+      "loss": 4.6717,
+      "step": 50
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5e-05,
+      "loss": 3.7212,
+      "step": 60
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.7500000000000002e-05,
+      "loss": 3.113,
+      "step": 70
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2e-05,
+      "loss": 2.6835,
+      "step": 80
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2.2499999999999998e-05,
+      "loss": 2.4711,
+      "step": 90
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.5e-05,
+      "loss": 2.3731,
+      "step": 100
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2.75e-05,
+      "loss": 2.1419,
+      "step": 110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3e-05,
+      "loss": 2.037,
+      "step": 120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.2500000000000004e-05,
+      "loss": 1.9651,
+      "step": 130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.5000000000000004e-05,
+      "loss": 1.8411,
+      "step": 140
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 3.75e-05,
+      "loss": 1.7802,
+      "step": 150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4e-05,
+      "loss": 1.7039,
+      "step": 160
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 4.25e-05,
+      "loss": 1.6342,
+      "step": 170
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.4999999999999996e-05,
+      "loss": 1.5809,
+      "step": 180
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 4.75e-05,
+      "loss": 1.5851,
+      "step": 190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5e-05,
+      "loss": 1.4769,
+      "step": 200
+    },
+    {
+      "epoch": 0.18,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        17716,
+        9437,
+        5749,
+        3729
+      ],
+      "eval_loss": 0.7528842687606812,
+      "eval_precisions": [
+        5.2827443075418365,
+        2.843172108773854,
+        1.7501240521049284,
+        1.1471304818624797
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 1613.712,
+      "eval_samples_per_second": 2.146,
+      "eval_score": 2.3433442773872275,
+      "eval_steps_per_second": 2.146,
+      "eval_sys_len": 335356,
+      "eval_totals": [
+        335356,
+        331918,
+        328491,
+        325072
+      ],
+      "step": 200
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 5.25e-05,
+      "loss": 1.5189,
+      "step": 210
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5.5e-05,
+      "loss": 1.442,
+      "step": 220
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 5.75e-05,
+      "loss": 1.4614,
+      "step": 230
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 6e-05,
+      "loss": 1.372,
+      "step": 240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 6.25e-05,
+      "loss": 1.382,
+      "step": 250
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 6.500000000000001e-05,
+      "loss": 1.3639,
+      "step": 260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 6.75e-05,
+      "loss": 1.418,
+      "step": 270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 7.000000000000001e-05,
+      "loss": 1.3478,
+      "step": 280
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 7.25e-05,
+      "loss": 1.3029,
+      "step": 290
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 7.5e-05,
+      "loss": 1.2836,
+      "step": 300
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 7.75e-05,
+      "loss": 1.2925,
+      "step": 310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8e-05,
+      "loss": 1.278,
+      "step": 320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 8.25e-05,
+      "loss": 1.2818,
+      "step": 330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 8.5e-05,
+      "loss": 1.2765,
+      "step": 340
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 8.75e-05,
+      "loss": 1.2529,
+      "step": 350
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 8.999999999999999e-05,
+      "loss": 1.2395,
+      "step": 360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.25e-05,
+      "loss": 1.2002,
+      "step": 370
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.5e-05,
+      "loss": 1.2723,
+      "step": 380
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.750000000000001e-05,
+      "loss": 1.2499,
+      "step": 390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 0.0001,
+      "loss": 1.2356,
+      "step": 400
+    },
+    {
+      "epoch": 0.36,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        23700,
+        14605,
+        9895,
+        7121
+      ],
+      "eval_loss": 0.602267324924469,
+      "eval_precisions": [
+        34.1729990050899,
+        22.165394363418372,
+        15.848482421718588,
+        12.06335761477215
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 852.9342,
+      "eval_samples_per_second": 4.06,
+      "eval_score": 19.50758960232416,
+      "eval_steps_per_second": 4.06,
+      "eval_sys_len": 69353,
+      "eval_totals": [
+        69353,
+        65891,
+        62435,
+        59030
+      ],
+      "step": 400
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.0001025,
+      "loss": 1.2241,
+      "step": 410
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 0.000105,
+      "loss": 1.209,
+      "step": 420
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 0.0001075,
+      "loss": 1.1935,
+      "step": 430
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 0.00011,
+      "loss": 1.1674,
+      "step": 440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 0.00011250000000000001,
+      "loss": 1.2343,
+      "step": 450
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 0.000115,
+      "loss": 1.1924,
+      "step": 460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 0.0001175,
+      "loss": 1.1366,
+      "step": 470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 0.00012,
+      "loss": 1.1779,
+      "step": 480
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 0.0001225,
+      "loss": 1.1417,
+      "step": 490
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.000125,
+      "loss": 1.1594,
+      "step": 500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 0.0001275,
+      "loss": 1.1466,
+      "step": 510
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 0.00013000000000000002,
+      "loss": 1.1283,
+      "step": 520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 0.00013250000000000002,
+      "loss": 1.0987,
+      "step": 530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 0.000135,
+      "loss": 1.1215,
+      "step": 540
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 0.0001375,
+      "loss": 1.1662,
+      "step": 550
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 0.00014000000000000001,
+      "loss": 1.1227,
+      "step": 560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 0.0001425,
+      "loss": 1.1494,
+      "step": 570
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 0.000145,
+      "loss": 1.1005,
+      "step": 580
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.0001475,
+      "loss": 1.1506,
+      "step": 590
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 0.00015,
+      "loss": 1.0997,
+      "step": 600
+    },
+    {
+      "epoch": 0.53,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        25462,
+        16355,
+        11456,
+        8410
+      ],
+      "eval_loss": 0.5470613241195679,
+      "eval_precisions": [
+        46.03091385700081,
+        31.541695595155442,
+        23.670888691447818,
+        18.699693156045715
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 585.2022,
+      "eval_samples_per_second": 5.918,
+      "eval_score": 28.313653908246078,
+      "eval_steps_per_second": 5.918,
+      "eval_sys_len": 55315,
+      "eval_totals": [
+        55315,
+        51852,
+        48397,
+        44974
+      ],
+      "step": 600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 0.0001525,
+      "loss": 1.1135,
+      "step": 610
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 0.000155,
+      "loss": 1.1367,
+      "step": 620
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 0.0001575,
+      "loss": 1.0833,
+      "step": 630
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 0.00016,
+      "loss": 1.104,
+      "step": 640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 0.00016250000000000002,
+      "loss": 1.0879,
+      "step": 650
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 0.000165,
+      "loss": 1.0817,
+      "step": 660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 0.0001675,
+      "loss": 1.0712,
+      "step": 670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.00017,
+      "loss": 1.1092,
+      "step": 680
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 0.0001725,
+      "loss": 1.0784,
+      "step": 690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 0.000175,
+      "loss": 1.0826,
+      "step": 700
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 0.0001775,
+      "loss": 1.0979,
+      "step": 710
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 0.00017999999999999998,
+      "loss": 1.0701,
+      "step": 720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 0.0001825,
+      "loss": 1.0797,
+      "step": 730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 0.000185,
+      "loss": 1.0733,
+      "step": 740
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 0.0001875,
+      "loss": 1.0547,
+      "step": 750
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 0.00019,
+      "loss": 1.0681,
+      "step": 760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 0.00019250000000000002,
+      "loss": 1.0683,
+      "step": 770
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.00019500000000000002,
+      "loss": 1.0525,
+      "step": 780
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 0.0001975,
+      "loss": 1.0493,
+      "step": 790
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 0.0002,
+      "loss": 1.0861,
+      "step": 800
+    },
+    {
+      "epoch": 0.71,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        26335,
+        17460,
+        12480,
+        9234
+      ],
+      "eval_loss": 0.5133824944496155,
+      "eval_precisions": [
+        50.52956751985878,
+        35.885314972767446,
+        27.611841231912916,
+        22.09989708733217
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 535.368,
+      "eval_samples_per_second": 6.468,
+      "eval_score": 32.43299018324529,
+      "eval_steps_per_second": 6.468,
+      "eval_sys_len": 52118,
+      "eval_totals": [
+        52118,
+        48655,
+        45198,
+        41783
+      ],
+      "step": 800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 0.00020250000000000002,
+      "loss": 1.0747,
+      "step": 810
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 0.000205,
+      "loss": 1.0245,
+      "step": 820
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 0.0002075,
+      "loss": 1.086,
+      "step": 830
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 0.00021,
+      "loss": 1.0777,
+      "step": 840
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 0.0002125,
+      "loss": 1.0618,
+      "step": 850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 0.000215,
+      "loss": 1.0328,
+      "step": 860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.0002175,
+      "loss": 1.0533,
+      "step": 870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 0.00022,
+      "loss": 1.0574,
+      "step": 880
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 0.00022250000000000001,
+      "loss": 1.0408,
+      "step": 890
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 0.00022500000000000002,
+      "loss": 1.0357,
+      "step": 900
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 0.0002275,
+      "loss": 1.0146,
+      "step": 910
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 0.00023,
+      "loss": 0.9986,
+      "step": 920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 0.0002325,
+      "loss": 1.0326,
+      "step": 930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 0.000235,
+      "loss": 1.0253,
+      "step": 940
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 0.0002375,
+      "loss": 0.9881,
+      "step": 950
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00024,
+      "loss": 1.0384,
+      "step": 960
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 0.00024249999999999999,
+      "loss": 1.0008,
+      "step": 970
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 0.000245,
+      "loss": 0.9963,
+      "step": 980
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 0.0002475,
+      "loss": 0.9914,
+      "step": 990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 0.00025,
+      "loss": 1.0244,
+      "step": 1000
+    },
+    {
+      "epoch": 0.89,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        26618,
+        17909,
+        12938,
+        9646
+      ],
+      "eval_loss": 0.4954930245876312,
+      "eval_precisions": [
+        51.9588514318062,
+        37.49319599715279,
+        29.198167497912483,
+        23.586658841940533
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 549.0088,
+      "eval_samples_per_second": 6.308,
+      "eval_score": 34.03364828263135,
+      "eval_steps_per_second": 6.308,
+      "eval_sys_len": 51229,
+      "eval_totals": [
+        51229,
+        47766,
+        44311,
+        40896
+      ],
+      "step": 1000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 0.0002525,
+      "loss": 1.0132,
+      "step": 1010
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 0.000255,
+      "loss": 0.9777,
+      "step": 1020
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 0.0002575,
+      "loss": 0.9896,
+      "step": 1030
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 0.00026000000000000003,
+      "loss": 0.9835,
+      "step": 1040
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00026250000000000004,
+      "loss": 0.9828,
+      "step": 1050
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 0.00026500000000000004,
+      "loss": 0.9894,
+      "step": 1060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 0.0002675,
+      "loss": 0.9767,
+      "step": 1070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 0.00027,
+      "loss": 0.9637,
+      "step": 1080
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 0.0002725,
+      "loss": 0.9551,
+      "step": 1090
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 0.000275,
+      "loss": 0.9858,
+      "step": 1100
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 0.0002775,
+      "loss": 1.015,
+      "step": 1110
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 0.00028000000000000003,
+      "loss": 0.9669,
+      "step": 1120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 0.0002825,
+      "loss": 0.954,
+      "step": 1130
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.000285,
+      "loss": 0.9628,
+      "step": 1140
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 0.0002875,
+      "loss": 0.9235,
+      "step": 1150
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 0.00029,
+      "loss": 0.9764,
+      "step": 1160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 0.0002925,
+      "loss": 0.9347,
+      "step": 1170
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 0.000295,
+      "loss": 0.9473,
+      "step": 1180
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 0.00029749999999999997,
+      "loss": 0.9664,
+      "step": 1190
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 0.0003,
+      "loss": 0.9572,
+      "step": 1200
+    },
+    {
+      "epoch": 1.07,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27385,
+        18748,
+        13686,
+        10276
+      ],
+      "eval_loss": 0.47938892245292664,
+      "eval_precisions": [
+        60.447201130143036,
+        44.80772448077245,
+        35.65362371698015,
+        29.386027624467385
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 508.8009,
+      "eval_samples_per_second": 6.806,
+      "eval_score": 41.04340766403553,
+      "eval_steps_per_second": 6.806,
+      "eval_sys_len": 45304,
+      "eval_totals": [
+        45304,
+        41841,
+        38386,
+        34969
+      ],
+      "step": 1200
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 0.0003025,
+      "loss": 0.9521,
+      "step": 1210
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 0.000305,
+      "loss": 0.9554,
+      "step": 1220
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 0.0003075,
+      "loss": 0.934,
+      "step": 1230
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.00031,
+      "loss": 0.9731,
+      "step": 1240
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 0.0003125,
+      "loss": 0.9477,
+      "step": 1250
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 0.000315,
+      "loss": 0.9497,
+      "step": 1260
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 0.0003175,
+      "loss": 0.8888,
+      "step": 1270
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 0.00032,
+      "loss": 0.9484,
+      "step": 1280
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 0.00032250000000000003,
+      "loss": 0.9342,
+      "step": 1290
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 0.00032500000000000004,
+      "loss": 0.8897,
+      "step": 1300
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 0.00032750000000000005,
+      "loss": 0.9814,
+      "step": 1310
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 0.00033,
+      "loss": 0.9506,
+      "step": 1320
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.0003325,
+      "loss": 0.9298,
+      "step": 1330
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 0.000335,
+      "loss": 0.938,
+      "step": 1340
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 0.0003375,
+      "loss": 0.951,
+      "step": 1350
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 0.00034,
+      "loss": 0.9584,
+      "step": 1360
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 0.00034250000000000003,
+      "loss": 0.9679,
+      "step": 1370
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 0.000345,
+      "loss": 0.9592,
+      "step": 1380
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 0.0003475,
+      "loss": 0.9307,
+      "step": 1390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 0.00035,
+      "loss": 0.9061,
+      "step": 1400
+    },
+    {
+      "epoch": 1.25,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27662,
+        18944,
+        13768,
+        10241
+      ],
+      "eval_loss": 0.48935189843177795,
+      "eval_precisions": [
+        65.6882999691292,
+        49.01676671496585,
+        39.120304597374556,
+        32.218586799219786
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 411.2631,
+      "eval_samples_per_second": 8.42,
+      "eval_score": 44.88335879592958,
+      "eval_steps_per_second": 8.42,
+      "eval_sys_len": 42111,
+      "eval_totals": [
+        42111,
+        38648,
+        35194,
+        31786
+      ],
+      "step": 1400
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 0.0003525,
+      "loss": 0.9391,
+      "step": 1410
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.000355,
+      "loss": 0.9575,
+      "step": 1420
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 0.0003575,
+      "loss": 0.9054,
+      "step": 1430
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 0.00035999999999999997,
+      "loss": 0.9524,
+      "step": 1440
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 0.0003625,
+      "loss": 0.9298,
+      "step": 1450
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 0.000365,
+      "loss": 0.9706,
+      "step": 1460
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 0.0003675,
+      "loss": 0.955,
+      "step": 1470
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 0.00037,
+      "loss": 0.9234,
+      "step": 1480
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 0.0003725,
+      "loss": 0.9129,
+      "step": 1490
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 0.000375,
+      "loss": 0.9137,
+      "step": 1500
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.0003775,
+      "loss": 0.9047,
+      "step": 1510
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 0.00038,
+      "loss": 0.9261,
+      "step": 1520
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 0.00038250000000000003,
+      "loss": 0.876,
+      "step": 1530
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 0.00038500000000000003,
+      "loss": 0.9451,
+      "step": 1540
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 0.00038750000000000004,
+      "loss": 0.9117,
+      "step": 1550
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 0.00039000000000000005,
+      "loss": 0.9202,
+      "step": 1560
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 0.0003925,
+      "loss": 0.9203,
+      "step": 1570
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 0.000395,
+      "loss": 0.9688,
+      "step": 1580
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 0.0003975,
+      "loss": 0.9265,
+      "step": 1590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0004,
+      "loss": 0.9381,
+      "step": 1600
+    },
+    {
+      "epoch": 1.43,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27320,
+        18959,
+        14016,
+        10644
+      ],
+      "eval_loss": 0.4612188935279846,
+      "eval_precisions": [
+        58.260294713496684,
+        43.654156113285744,
+        35.06191369606004,
+        29.116174740815712
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 502.6767,
+      "eval_samples_per_second": 6.889,
+      "eval_score": 40.14135718782399,
+      "eval_steps_per_second": 6.889,
+      "eval_sys_len": 46893,
+      "eval_totals": [
+        46893,
+        43430,
+        39975,
+        36557
+      ],
+      "step": 1600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 0.0004025,
+      "loss": 0.9329,
+      "step": 1610
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 0.00040500000000000003,
+      "loss": 0.9324,
+      "step": 1620
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 0.0004075,
+      "loss": 0.927,
+      "step": 1630
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 0.00041,
+      "loss": 0.9037,
+      "step": 1640
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 0.0004125,
+      "loss": 0.8935,
+      "step": 1650
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 0.000415,
+      "loss": 0.9072,
+      "step": 1660
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 0.0004175,
+      "loss": 0.8702,
+      "step": 1670
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 0.00042,
+      "loss": 0.9744,
+      "step": 1680
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 0.00042249999999999997,
+      "loss": 0.8935,
+      "step": 1690
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.000425,
+      "loss": 0.9277,
+      "step": 1700
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 0.0004275,
+      "loss": 0.8846,
+      "step": 1710
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 0.00043,
+      "loss": 0.8766,
+      "step": 1720
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 0.0004325,
+      "loss": 0.8836,
+      "step": 1730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 0.000435,
+      "loss": 0.921,
+      "step": 1740
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 0.0004375,
+      "loss": 0.9319,
+      "step": 1750
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 0.00044,
+      "loss": 0.9127,
+      "step": 1760
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 0.0004425,
+      "loss": 0.9045,
+      "step": 1770
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 0.00044500000000000003,
+      "loss": 0.918,
+      "step": 1780
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00044750000000000004,
+      "loss": 0.9119,
+      "step": 1790
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 0.00045000000000000004,
+      "loss": 0.8815,
+      "step": 1800
+    },
+    {
+      "epoch": 1.6,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28644,
+        20158,
+        15010,
+        11444
+      ],
+      "eval_loss": 0.45668816566467285,
+      "eval_precisions": [
+        68.8309503784693,
+        52.8360243237576,
+        43.25897746267796,
+        36.58099987213911
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 377.0997,
+      "eval_samples_per_second": 9.183,
+      "eval_score": 48.97917095230454,
+      "eval_steps_per_second": 9.183,
+      "eval_sys_len": 41615,
+      "eval_totals": [
+        41615,
+        38152,
+        34698,
+        31284
+      ],
+      "step": 1800
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 0.00045250000000000005,
+      "loss": 0.8805,
+      "step": 1810
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 0.000455,
+      "loss": 0.8728,
+      "step": 1820
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 0.0004575,
+      "loss": 0.885,
+      "step": 1830
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 0.00046,
+      "loss": 0.9132,
+      "step": 1840
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 0.0004625,
+      "loss": 0.8883,
+      "step": 1850
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 0.000465,
+      "loss": 0.8669,
+      "step": 1860
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 0.00046750000000000003,
+      "loss": 0.9288,
+      "step": 1870
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.00047,
+      "loss": 0.8605,
+      "step": 1880
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 0.0004725,
+      "loss": 0.8763,
+      "step": 1890
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 0.000475,
+      "loss": 0.8821,
+      "step": 1900
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 0.0004775,
+      "loss": 0.8632,
+      "step": 1910
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 0.00048,
+      "loss": 0.8763,
+      "step": 1920
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 0.0004825,
+      "loss": 0.8936,
+      "step": 1930
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 0.00048499999999999997,
+      "loss": 0.9197,
+      "step": 1940
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 0.0004875,
+      "loss": 0.8868,
+      "step": 1950
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 0.00049,
+      "loss": 0.8543,
+      "step": 1960
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.0004925,
+      "loss": 0.8487,
+      "step": 1970
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 0.000495,
+      "loss": 0.8654,
+      "step": 1980
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 0.0004975,
+      "loss": 0.878,
+      "step": 1990
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 0.0005,
+      "loss": 0.8889,
+      "step": 2000
+    },
+    {
+      "epoch": 1.78,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28627,
+        20189,
+        15089,
+        11523
+      ],
+      "eval_loss": 0.45155465602874756,
+      "eval_precisions": [
+        68.39401758409785,
+        52.58510666006824,
+        43.185460789925585,
+        36.551942902458364
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 367.2279,
+      "eval_samples_per_second": 9.43,
+      "eval_score": 48.812631909516725,
+      "eval_steps_per_second": 9.43,
+      "eval_sys_len": 41856,
+      "eval_totals": [
+        41856,
+        38393,
+        34940,
+        31525
+      ],
+      "step": 2000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 0.0005024999999999999,
+      "loss": 0.8704,
+      "step": 2010
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 0.000505,
+      "loss": 0.8952,
+      "step": 2020
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 0.0005074999999999999,
+      "loss": 0.8621,
+      "step": 2030
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 0.00051,
+      "loss": 0.8965,
+      "step": 2040
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 0.0005124999999999999,
+      "loss": 0.8919,
+      "step": 2050
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.000515,
+      "loss": 0.888,
+      "step": 2060
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 0.0005175,
+      "loss": 0.8876,
+      "step": 2070
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 0.0005200000000000001,
+      "loss": 0.8368,
+      "step": 2080
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 0.0005225,
+      "loss": 0.8681,
+      "step": 2090
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 0.0005250000000000001,
+      "loss": 0.9107,
+      "step": 2100
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 0.0005275,
+      "loss": 0.9227,
+      "step": 2110
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 0.0005300000000000001,
+      "loss": 0.9045,
+      "step": 2120
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 0.0005325,
+      "loss": 0.8744,
+      "step": 2130
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 0.000535,
+      "loss": 0.9088,
+      "step": 2140
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 0.0005375,
+      "loss": 0.8861,
+      "step": 2150
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.00054,
+      "loss": 0.8542,
+      "step": 2160
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 0.0005425,
+      "loss": 0.915,
+      "step": 2170
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 0.000545,
+      "loss": 0.8839,
+      "step": 2180
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 0.0005475,
+      "loss": 0.8698,
+      "step": 2190
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 0.00055,
+      "loss": 0.8598,
+      "step": 2200
+    },
+    {
+      "epoch": 1.96,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27573,
+        19348,
+        14395,
+        10963
+      ],
+      "eval_loss": 0.4470996558666229,
+      "eval_precisions": [
+        55.13717805151176,
+        41.56837469115909,
+        33.40449725013343,
+        27.63061723416589
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 526.7436,
+      "eval_samples_per_second": 6.574,
+      "eval_score": 38.137373171080895,
+      "eval_steps_per_second": 6.574,
+      "eval_sys_len": 50008,
+      "eval_totals": [
+        50008,
+        46545,
+        43093,
+        39677
+      ],
+      "step": 2200
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 0.0005525,
+      "loss": 0.8728,
+      "step": 2210
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 0.000555,
+      "loss": 0.9058,
+      "step": 2220
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 0.0005575,
+      "loss": 0.8776,
+      "step": 2230
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 0.0005600000000000001,
+      "loss": 0.8669,
+      "step": 2240
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.0005625000000000001,
+      "loss": 0.8562,
+      "step": 2250
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 0.000565,
+      "loss": 0.7894,
+      "step": 2260
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 0.0005675,
+      "loss": 0.832,
+      "step": 2270
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 0.00057,
+      "loss": 0.8437,
+      "step": 2280
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 0.0005725,
+      "loss": 0.8788,
+      "step": 2290
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 0.000575,
+      "loss": 0.8238,
+      "step": 2300
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 0.0005775,
+      "loss": 0.8248,
+      "step": 2310
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 0.00058,
+      "loss": 0.8653,
+      "step": 2320
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 0.0005825,
+      "loss": 0.8257,
+      "step": 2330
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.000585,
+      "loss": 0.8608,
+      "step": 2340
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 0.0005875,
+      "loss": 0.8483,
+      "step": 2350
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 0.00059,
+      "loss": 0.8503,
+      "step": 2360
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 0.0005925,
+      "loss": 0.8306,
+      "step": 2370
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 0.0005949999999999999,
+      "loss": 0.8413,
+      "step": 2380
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 0.0005975,
+      "loss": 0.8448,
+      "step": 2390
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 0.0006,
+      "loss": 0.8693,
+      "step": 2400
+    },
+    {
+      "epoch": 2.14,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27425,
+        19232,
+        14323,
+        10947
+      ],
+      "eval_loss": 0.44469693303108215,
+      "eval_precisions": [
+        52.27294386734013,
+        39.24737765805477,
+        31.448708940804497,
+        25.99002849002849
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 633.5059,
+      "eval_samples_per_second": 5.466,
+      "eval_score": 35.98523101021763,
+      "eval_steps_per_second": 5.466,
+      "eval_sys_len": 52465,
+      "eval_totals": [
+        52465,
+        49002,
+        45544,
+        42120
+      ],
+      "step": 2400
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 0.0006025000000000001,
+      "loss": 0.8079,
+      "step": 2410
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 0.000605,
+      "loss": 0.8713,
+      "step": 2420
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.0006075000000000001,
+      "loss": 0.8728,
+      "step": 2430
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 0.00061,
+      "loss": 0.8336,
+      "step": 2440
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 0.0006125000000000001,
+      "loss": 0.8426,
+      "step": 2450
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 0.000615,
+      "loss": 0.8462,
+      "step": 2460
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 0.0006175000000000001,
+      "loss": 0.8412,
+      "step": 2470
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 0.00062,
+      "loss": 0.8114,
+      "step": 2480
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 0.0006225000000000001,
+      "loss": 0.826,
+      "step": 2490
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 0.000625,
+      "loss": 0.8248,
+      "step": 2500
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 0.0006274999999999999,
+      "loss": 0.8524,
+      "step": 2510
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.00063,
+      "loss": 0.8695,
+      "step": 2520
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 0.0006324999999999999,
+      "loss": 0.826,
+      "step": 2530
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 0.000635,
+      "loss": 0.8171,
+      "step": 2540
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 0.0006374999999999999,
+      "loss": 0.8296,
+      "step": 2550
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 0.00064,
+      "loss": 0.8438,
+      "step": 2560
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 0.0006425,
+      "loss": 0.8327,
+      "step": 2570
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 0.0006450000000000001,
+      "loss": 0.8357,
+      "step": 2580
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 0.0006475,
+      "loss": 0.8652,
+      "step": 2590
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 0.0006500000000000001,
+      "loss": 0.8196,
+      "step": 2600
+    },
+    {
+      "epoch": 2.32,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28513,
+        20270,
+        15211,
+        11701
+      ],
+      "eval_loss": 0.43805596232414246,
+      "eval_precisions": [
+        66.24459829933554,
+        51.21402764092069,
+        42.11356903568759,
+        35.78834684202477
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 409.6535,
+      "eval_samples_per_second": 8.453,
+      "eval_score": 47.55276474687039,
+      "eval_steps_per_second": 8.453,
+      "eval_sys_len": 43042,
+      "eval_totals": [
+        43042,
+        39579,
+        36119,
+        32695
+      ],
+      "step": 2600
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 0.0006525,
+      "loss": 0.8485,
+      "step": 2610
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0006550000000000001,
+      "loss": 0.8067,
+      "step": 2620
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 0.0006575,
+      "loss": 0.8607,
+      "step": 2630
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 0.00066,
+      "loss": 0.8362,
+      "step": 2640
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 0.0006625,
+      "loss": 0.8426,
+      "step": 2650
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 0.000665,
+      "loss": 0.8327,
+      "step": 2660
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 0.0006675,
+      "loss": 0.8404,
+      "step": 2670
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 0.00067,
+      "loss": 0.8353,
+      "step": 2680
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 0.0006725,
+      "loss": 0.8576,
+      "step": 2690
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 0.000675,
+      "loss": 0.8357,
+      "step": 2700
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.0006775,
+      "loss": 0.8328,
+      "step": 2710
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 0.00068,
+      "loss": 0.8499,
+      "step": 2720
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 0.0006825000000000001,
+      "loss": 0.8171,
+      "step": 2730
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 0.0006850000000000001,
+      "loss": 0.8372,
+      "step": 2740
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 0.0006875,
+      "loss": 0.8542,
+      "step": 2750
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 0.00069,
+      "loss": 0.8209,
+      "step": 2760
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 0.0006925,
+      "loss": 0.8048,
+      "step": 2770
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 0.000695,
+      "loss": 0.8162,
+      "step": 2780
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 0.0006975,
+      "loss": 0.8293,
+      "step": 2790
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0007,
+      "loss": 0.8423,
+      "step": 2800
+    },
+    {
+      "epoch": 2.5,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28571,
+        20383,
+        15384,
+        11858
+      ],
+      "eval_loss": 0.43379348516464233,
+      "eval_precisions": [
+        65.67744011769574,
+        50.907864831789006,
+        42.046572646769434,
+        35.745937961595274
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 414.063,
+      "eval_samples_per_second": 8.363,
+      "eval_score": 47.346687432735244,
+      "eval_steps_per_second": 8.363,
+      "eval_sys_len": 43502,
+      "eval_totals": [
+        43502,
+        40039,
+        36588,
+        33173
+      ],
+      "step": 2800
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 0.0007025,
+      "loss": 0.8276,
+      "step": 2810
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 0.000705,
+      "loss": 0.8094,
+      "step": 2820
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 0.0007075,
+      "loss": 0.8435,
+      "step": 2830
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 0.00071,
+      "loss": 0.8609,
+      "step": 2840
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 0.0007125,
+      "loss": 0.8234,
+      "step": 2850
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 0.000715,
+      "loss": 0.8359,
+      "step": 2860
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 0.0007175,
+      "loss": 0.844,
+      "step": 2870
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 0.0007199999999999999,
+      "loss": 0.8147,
+      "step": 2880
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.0007225,
+      "loss": 0.8016,
+      "step": 2890
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 0.000725,
+      "loss": 0.8507,
+      "step": 2900
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 0.0007275000000000001,
+      "loss": 0.8298,
+      "step": 2910
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 0.00073,
+      "loss": 0.8564,
+      "step": 2920
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 0.0007325000000000001,
+      "loss": 0.79,
+      "step": 2930
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 0.000735,
+      "loss": 0.8021,
+      "step": 2940
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 0.0007375000000000001,
+      "loss": 0.8208,
+      "step": 2950
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 0.00074,
+      "loss": 0.8673,
+      "step": 2960
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 0.0007425000000000001,
+      "loss": 0.8469,
+      "step": 2970
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.000745,
+      "loss": 0.8596,
+      "step": 2980
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 0.0007475000000000001,
+      "loss": 0.8321,
+      "step": 2990
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 0.00075,
+      "loss": 0.7965,
+      "step": 3000
+    },
+    {
+      "epoch": 2.67,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27367,
+        19330,
+        14468,
+        11113
+      ],
+      "eval_loss": 0.433325856924057,
+      "eval_precisions": [
+        47.69681231155341,
+        35.853396149423155,
+        28.673352094811527,
+        23.624574829931973
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 723.5054,
+      "eval_samples_per_second": 4.786,
+      "eval_score": 32.80692539007455,
+      "eval_steps_per_second": 4.786,
+      "eval_sys_len": 57377,
+      "eval_totals": [
+        57377,
+        53914,
+        50458,
+        47040
+      ],
+      "step": 3000
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 0.0007524999999999999,
+      "loss": 0.8643,
+      "step": 3010
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 0.000755,
+      "loss": 0.8288,
+      "step": 3020
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 0.0007574999999999999,
+      "loss": 0.8361,
+      "step": 3030
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 0.00076,
+      "loss": 0.8157,
+      "step": 3040
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 0.0007624999999999999,
+      "loss": 0.8668,
+      "step": 3050
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 0.0007650000000000001,
+      "loss": 0.8379,
+      "step": 3060
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 0.0007675,
+      "loss": 0.8419,
+      "step": 3070
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0007700000000000001,
+      "loss": 0.8147,
+      "step": 3080
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 0.0007725,
+      "loss": 0.8318,
+      "step": 3090
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 0.0007750000000000001,
+      "loss": 0.7908,
+      "step": 3100
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 0.0007775,
+      "loss": 0.8331,
+      "step": 3110
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 0.0007800000000000001,
+      "loss": 0.8459,
+      "step": 3120
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 0.0007825,
+      "loss": 0.8177,
+      "step": 3130
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 0.000785,
+      "loss": 0.8354,
+      "step": 3140
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 0.0007875,
+      "loss": 0.8164,
+      "step": 3150
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 0.00079,
+      "loss": 0.8422,
+      "step": 3160
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.0007925,
+      "loss": 0.783,
+      "step": 3170
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 0.000795,
+      "loss": 0.8223,
+      "step": 3180
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 0.0007975,
+      "loss": 0.8012,
+      "step": 3190
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 0.0008,
+      "loss": 0.8108,
+      "step": 3200
+    },
+    {
+      "epoch": 2.85,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28853,
+        20613,
+        15542,
+        11972
+      ],
+      "eval_loss": 0.4307364523410797,
+      "eval_precisions": [
+        68.28465944052634,
+        53.138614627104225,
+        43.98347294543808,
+        37.508615828059405
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 423.5558,
+      "eval_samples_per_second": 8.176,
+      "eval_score": 49.463914513865696,
+      "eval_steps_per_second": 8.176,
+      "eval_sys_len": 42254,
+      "eval_totals": [
+        42254,
+        38791,
+        35336,
+        31918
+      ],
+      "step": 3200
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 0.0008025,
+      "loss": 0.8126,
+      "step": 3210
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 0.000805,
+      "loss": 0.8282,
+      "step": 3220
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 0.0008075000000000001,
+      "loss": 0.8188,
+      "step": 3230
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 0.0008100000000000001,
+      "loss": 0.7631,
+      "step": 3240
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 0.0008125000000000001,
+      "loss": 0.8016,
+      "step": 3250
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.000815,
+      "loss": 0.8385,
+      "step": 3260
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 0.0008175,
+      "loss": 0.8138,
+      "step": 3270
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 0.00082,
+      "loss": 0.7849,
+      "step": 3280
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 0.0008225,
+      "loss": 0.8189,
+      "step": 3290
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 0.000825,
+      "loss": 0.8385,
+      "step": 3300
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 0.0008275,
+      "loss": 0.7924,
+      "step": 3310
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 0.00083,
+      "loss": 0.8227,
+      "step": 3320
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 0.0008325,
+      "loss": 0.8135,
+      "step": 3330
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 0.000835,
+      "loss": 0.7983,
+      "step": 3340
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.0008375,
+      "loss": 0.8417,
+      "step": 3350
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 0.00084,
+      "loss": 0.8202,
+      "step": 3360
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 0.0008425,
+      "loss": 0.8205,
+      "step": 3370
+    },
+    {
+      "epoch": 3.01,
+      "learning_rate": 0.0008449999999999999,
+      "loss": 0.7563,
+      "step": 3380
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 0.0008475000000000001,
+      "loss": 0.8087,
+      "step": 3390
+    },
+    {
+      "epoch": 3.03,
+      "learning_rate": 0.00085,
+      "loss": 0.7227,
+      "step": 3400
+    },
+    {
+      "epoch": 3.03,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28393,
+        20333,
+        15377,
+        11915
+      ],
+      "eval_loss": 0.4285724461078644,
+      "eval_precisions": [
+        66.42103539429668,
+        51.75898584665513,
+        42.91774819280471,
+        36.76447900274615
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 379.2685,
+      "eval_samples_per_second": 9.131,
+      "eval_score": 48.260207590301434,
+      "eval_steps_per_second": 9.131,
+      "eval_sys_len": 42747,
+      "eval_totals": [
+        42747,
+        39284,
+        35829,
+        32409
+      ],
+      "step": 3400
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 0.0008525000000000001,
+      "loss": 0.7961,
+      "step": 3410
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 0.000855,
+      "loss": 0.8006,
+      "step": 3420
+    },
+    {
+      "epoch": 3.06,
+      "learning_rate": 0.0008575000000000001,
+      "loss": 0.7477,
+      "step": 3430
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.00086,
+      "loss": 0.7639,
+      "step": 3440
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 0.0008625000000000001,
+      "loss": 0.7835,
+      "step": 3450
+    },
+    {
+      "epoch": 3.08,
+      "learning_rate": 0.000865,
+      "loss": 0.7763,
+      "step": 3460
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 0.0008675000000000001,
+      "loss": 0.8046,
+      "step": 3470
+    },
+    {
+      "epoch": 3.1,
+      "learning_rate": 0.00087,
+      "loss": 0.7509,
+      "step": 3480
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 0.0008725000000000001,
+      "loss": 0.781,
+      "step": 3490
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 0.000875,
+      "loss": 0.7989,
+      "step": 3500
+    },
+    {
+      "epoch": 3.13,
+      "learning_rate": 0.0008774999999999999,
+      "loss": 0.8082,
+      "step": 3510
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 0.00088,
+      "loss": 0.7752,
+      "step": 3520
+    },
+    {
+      "epoch": 3.15,
+      "learning_rate": 0.0008824999999999999,
+      "loss": 0.7655,
+      "step": 3530
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.000885,
+      "loss": 0.8013,
+      "step": 3540
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 0.0008874999999999999,
+      "loss": 0.7823,
+      "step": 3550
+    },
+    {
+      "epoch": 3.17,
+      "learning_rate": 0.0008900000000000001,
+      "loss": 0.8154,
+      "step": 3560
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 0.0008925,
+      "loss": 0.7932,
+      "step": 3570
+    },
+    {
+      "epoch": 3.19,
+      "learning_rate": 0.0008950000000000001,
+      "loss": 0.7951,
+      "step": 3580
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 0.0008975,
+      "loss": 0.7676,
+      "step": 3590
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 0.0009000000000000001,
+      "loss": 0.7851,
+      "step": 3600
+    },
+    {
+      "epoch": 3.21,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28905,
+        20753,
+        15720,
+        12165
+      ],
+      "eval_loss": 0.4306156039237976,
+      "eval_precisions": [
+        68.80668428193958,
+        53.83956830799564,
+        44.78887685908029,
+        38.38508140855736
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 368.5897,
+      "eval_samples_per_second": 9.395,
+      "eval_score": 50.23613389703178,
+      "eval_steps_per_second": 9.395,
+      "eval_sys_len": 42009,
+      "eval_totals": [
+        42009,
+        38546,
+        35098,
+        31692
+      ],
+      "step": 3600
+    },
+    {
+      "epoch": 3.22,
+      "learning_rate": 0.0009025,
+      "loss": 0.8031,
+      "step": 3610
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 0.0009050000000000001,
+      "loss": 0.8191,
+      "step": 3620
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.0009075,
+      "loss": 0.7885,
+      "step": 3630
+    },
+    {
+      "epoch": 3.24,
+      "learning_rate": 0.00091,
+      "loss": 0.8008,
+      "step": 3640
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 0.0009125,
+      "loss": 0.8004,
+      "step": 3650
+    },
+    {
+      "epoch": 3.26,
+      "learning_rate": 0.000915,
+      "loss": 0.8011,
+      "step": 3660
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 0.0009175,
+      "loss": 0.8327,
+      "step": 3670
+    },
+    {
+      "epoch": 3.28,
+      "learning_rate": 0.00092,
+      "loss": 0.7965,
+      "step": 3680
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 0.0009225,
+      "loss": 0.8453,
+      "step": 3690
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 0.000925,
+      "loss": 0.7944,
+      "step": 3700
+    },
+    {
+      "epoch": 3.31,
+      "learning_rate": 0.0009275,
+      "loss": 0.7959,
+      "step": 3710
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.00093,
+      "loss": 0.7691,
+      "step": 3720
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 0.0009325000000000001,
+      "loss": 0.805,
+      "step": 3730
+    },
+    {
+      "epoch": 3.33,
+      "learning_rate": 0.0009350000000000001,
+      "loss": 0.7473,
+      "step": 3740
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 0.0009375,
+      "loss": 0.7917,
+      "step": 3750
+    },
+    {
+      "epoch": 3.35,
+      "learning_rate": 0.00094,
+      "loss": 0.8023,
+      "step": 3760
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 0.0009425,
+      "loss": 0.7892,
+      "step": 3770
+    },
+    {
+      "epoch": 3.37,
+      "learning_rate": 0.000945,
+      "loss": 0.785,
+      "step": 3780
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 0.0009475,
+      "loss": 0.792,
+      "step": 3790
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 0.00095,
+      "loss": 0.7561,
+      "step": 3800
+    },
+    {
+      "epoch": 3.39,
+      "eval_bp": 0.9827426623868254,
+      "eval_counts": [
+        28866,
+        20722,
+        15692,
+        12126
+      ],
+      "eval_loss": 0.42703622579574585,
+      "eval_precisions": [
+        70.9744043667478,
+        55.692324231348096,
+        46.4851734455076,
+        39.96177168468231
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 313.0426,
+      "eval_samples_per_second": 11.062,
+      "eval_score": 51.15686029358077,
+      "eval_steps_per_second": 11.062,
+      "eval_sys_len": 40671,
+      "eval_totals": [
+        40671,
+        37208,
+        33757,
+        30344
+      ],
+      "step": 3800
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.0009525,
+      "loss": 0.8078,
+      "step": 3810
+    },
+    {
+      "epoch": 3.4,
+      "learning_rate": 0.000955,
+      "loss": 0.8172,
+      "step": 3820
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 0.0009575,
+      "loss": 0.7645,
+      "step": 3830
+    },
+    {
+      "epoch": 3.42,
+      "learning_rate": 0.00096,
+      "loss": 0.7616,
+      "step": 3840
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 0.0009625,
+      "loss": 0.7658,
+      "step": 3850
+    },
+    {
+      "epoch": 3.44,
+      "learning_rate": 0.000965,
+      "loss": 0.7459,
+      "step": 3860
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 0.0009675,
+      "loss": 0.7666,
+      "step": 3870
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 0.0009699999999999999,
+      "loss": 0.7894,
+      "step": 3880
+    },
+    {
+      "epoch": 3.47,
+      "learning_rate": 0.0009725000000000001,
+      "loss": 0.7572,
+      "step": 3890
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.000975,
+      "loss": 0.7827,
+      "step": 3900
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 0.0009775,
+      "loss": 0.8275,
+      "step": 3910
+    },
+    {
+      "epoch": 3.49,
+      "learning_rate": 0.00098,
+      "loss": 0.7848,
+      "step": 3920
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 0.0009825,
+      "loss": 0.755,
+      "step": 3930
+    },
+    {
+      "epoch": 3.51,
+      "learning_rate": 0.000985,
+      "loss": 0.7603,
+      "step": 3940
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 0.0009875,
+      "loss": 0.7703,
+      "step": 3950
+    },
+    {
+      "epoch": 3.53,
+      "learning_rate": 0.00099,
+      "loss": 0.7523,
+      "step": 3960
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 0.0009925000000000001,
+      "loss": 0.8317,
+      "step": 3970
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 0.000995,
+      "loss": 0.7548,
+      "step": 3980
+    },
+    {
+      "epoch": 3.56,
+      "learning_rate": 0.0009975000000000001,
+      "loss": 0.8015,
+      "step": 3990
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.001,
+      "loss": 0.8206,
+      "step": 4000
+    },
+    {
+      "epoch": 3.57,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27552,
+        19510,
+        14673,
+        11285
+      ],
+      "eval_loss": 0.4267815351486206,
+      "eval_precisions": [
+        57.07538375489404,
+        43.53938852934613,
+        35.480594849474066,
+        29.74825359166996
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 541.444,
+      "eval_samples_per_second": 6.396,
+      "eval_score": 40.24351285668535,
+      "eval_steps_per_second": 6.396,
+      "eval_sys_len": 48273,
+      "eval_totals": [
+        48273,
+        44810,
+        41355,
+        37935
+      ],
+      "step": 4000
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 0.001,
+      "loss": 0.78,
+      "step": 4010
+    },
+    {
+      "epoch": 3.58,
+      "learning_rate": 0.001,
+      "loss": 0.8166,
+      "step": 4020
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 0.001,
+      "loss": 0.756,
+      "step": 4030
+    },
+    {
+      "epoch": 3.6,
+      "learning_rate": 0.001,
+      "loss": 0.777,
+      "step": 4040
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 0.001,
+      "loss": 0.8195,
+      "step": 4050
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 0.001,
+      "loss": 0.7597,
+      "step": 4060
+    },
+    {
+      "epoch": 3.63,
+      "learning_rate": 0.001,
+      "loss": 0.7881,
+      "step": 4070
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 0.001,
+      "loss": 0.7737,
+      "step": 4080
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.001,
+      "loss": 0.8094,
+      "step": 4090
+    },
+    {
+      "epoch": 3.65,
+      "learning_rate": 0.001,
+      "loss": 0.8164,
+      "step": 4100
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 0.001,
+      "loss": 0.7642,
+      "step": 4110
+    },
+    {
+      "epoch": 3.67,
+      "learning_rate": 0.001,
+      "loss": 0.7943,
+      "step": 4120
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 0.001,
+      "loss": 0.8135,
+      "step": 4130
+    },
+    {
+      "epoch": 3.69,
+      "learning_rate": 0.001,
+      "loss": 0.7656,
+      "step": 4140
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 0.001,
+      "loss": 0.8099,
+      "step": 4150
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 0.001,
+      "loss": 0.7723,
+      "step": 4160
+    },
+    {
+      "epoch": 3.72,
+      "learning_rate": 0.001,
+      "loss": 0.8023,
+      "step": 4170
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.001,
+      "loss": 0.7847,
+      "step": 4180
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 0.001,
+      "loss": 0.7681,
+      "step": 4190
+    },
+    {
+      "epoch": 3.74,
+      "learning_rate": 0.001,
+      "loss": 0.7777,
+      "step": 4200
+    },
+    {
+      "epoch": 3.74,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        29035,
+        20912,
+        15887,
+        12288
+      ],
+      "eval_loss": 0.41903844475746155,
+      "eval_precisions": [
+        68.37717542330971,
+        53.62051282051282,
+        44.69545646363764,
+        38.24939301500343
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 387.8999,
+      "eval_samples_per_second": 8.928,
+      "eval_score": 50.03599162186965,
+      "eval_steps_per_second": 8.928,
+      "eval_sys_len": 42463,
+      "eval_totals": [
+        42463,
+        39000,
+        35545,
+        32126
+      ],
+      "step": 4200
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 0.001,
+      "loss": 0.7665,
+      "step": 4210
+    },
+    {
+      "epoch": 3.76,
+      "learning_rate": 0.001,
+      "loss": 0.8047,
+      "step": 4220
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 0.001,
+      "loss": 0.781,
+      "step": 4230
+    },
+    {
+      "epoch": 3.78,
+      "learning_rate": 0.001,
+      "loss": 0.7649,
+      "step": 4240
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 0.001,
+      "loss": 0.7847,
+      "step": 4250
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 0.001,
+      "loss": 0.7721,
+      "step": 4260
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.001,
+      "loss": 0.758,
+      "step": 4270
+    },
+    {
+      "epoch": 3.81,
+      "learning_rate": 0.001,
+      "loss": 0.7738,
+      "step": 4280
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 0.001,
+      "loss": 0.7786,
+      "step": 4290
+    },
+    {
+      "epoch": 3.83,
+      "learning_rate": 0.001,
+      "loss": 0.8155,
+      "step": 4300
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 0.001,
+      "loss": 0.7831,
+      "step": 4310
+    },
+    {
+      "epoch": 3.85,
+      "learning_rate": 0.001,
+      "loss": 0.7836,
+      "step": 4320
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 0.001,
+      "loss": 0.756,
+      "step": 4330
+    },
+    {
+      "epoch": 3.87,
+      "learning_rate": 0.001,
+      "loss": 0.7785,
+      "step": 4340
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 0.001,
+      "loss": 0.8093,
+      "step": 4350
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.001,
+      "loss": 0.7747,
+      "step": 4360
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 0.001,
+      "loss": 0.78,
+      "step": 4370
+    },
+    {
+      "epoch": 3.9,
+      "learning_rate": 0.001,
+      "loss": 0.7726,
+      "step": 4380
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 0.001,
+      "loss": 0.8045,
+      "step": 4390
+    },
+    {
+      "epoch": 3.92,
+      "learning_rate": 0.001,
+      "loss": 0.7241,
+      "step": 4400
+    },
+    {
+      "epoch": 3.92,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28666,
+        20580,
+        15602,
+        12089
+      ],
+      "eval_loss": 0.41805002093315125,
+      "eval_precisions": [
+        67.87743890888426,
+        53.08364930743635,
+        44.174523627509274,
+        37.8894251864853
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 376.5087,
+      "eval_samples_per_second": 9.198,
+      "eval_score": 49.55574706503557,
+      "eval_steps_per_second": 9.198,
+      "eval_sys_len": 42232,
+      "eval_totals": [
+        42232,
+        38769,
+        35319,
+        31906
+      ],
+      "step": 4400
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 0.001,
+      "loss": 0.7529,
+      "step": 4410
+    },
+    {
+      "epoch": 3.94,
+      "learning_rate": 0.001,
+      "loss": 0.7707,
+      "step": 4420
+    },
+    {
+      "epoch": 3.95,
+      "learning_rate": 0.001,
+      "loss": 0.7616,
+      "step": 4430
+    },
+    {
+      "epoch": 3.96,
+      "learning_rate": 0.001,
+      "loss": 0.7536,
+      "step": 4440
+    },
+    {
+      "epoch": 3.97,
+      "learning_rate": 0.001,
+      "loss": 0.7813,
+      "step": 4450
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.001,
+      "loss": 0.8368,
+      "step": 4460
+    },
+    {
+      "epoch": 3.98,
+      "learning_rate": 0.001,
+      "loss": 0.8251,
+      "step": 4470
+    },
+    {
+      "epoch": 3.99,
+      "learning_rate": 0.001,
+      "loss": 0.7924,
+      "step": 4480
+    },
+    {
+      "epoch": 4.0,
+      "learning_rate": 0.001,
+      "loss": 0.7777,
+      "step": 4490
+    },
+    {
+      "epoch": 4.01,
+      "learning_rate": 0.001,
+      "loss": 0.7578,
+      "step": 4500
+    },
+    {
+      "epoch": 4.02,
+      "learning_rate": 0.001,
+      "loss": 0.675,
+      "step": 4510
+    },
+    {
+      "epoch": 4.03,
+      "learning_rate": 0.001,
+      "loss": 0.7593,
+      "step": 4520
+    },
+    {
+      "epoch": 4.04,
+      "learning_rate": 0.001,
+      "loss": 0.7106,
+      "step": 4530
+    },
+    {
+      "epoch": 4.05,
+      "learning_rate": 0.001,
+      "loss": 0.7564,
+      "step": 4540
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.001,
+      "loss": 0.7051,
+      "step": 4550
+    },
+    {
+      "epoch": 4.06,
+      "learning_rate": 0.001,
+      "loss": 0.724,
+      "step": 4560
+    },
+    {
+      "epoch": 4.07,
+      "learning_rate": 0.001,
+      "loss": 0.7047,
+      "step": 4570
+    },
+    {
+      "epoch": 4.08,
+      "learning_rate": 0.001,
+      "loss": 0.7605,
+      "step": 4580
+    },
+    {
+      "epoch": 4.09,
+      "learning_rate": 0.001,
+      "loss": 0.7314,
+      "step": 4590
+    },
+    {
+      "epoch": 4.1,
+      "learning_rate": 0.001,
+      "loss": 0.7592,
+      "step": 4600
+    },
+    {
+      "epoch": 4.1,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28474,
+        20378,
+        15394,
+        11847
+      ],
+      "eval_loss": 0.4237304925918579,
+      "eval_precisions": [
+        56.65791149315505,
+        43.54924881926784,
+        35.520790068761826,
+        29.678340598226363
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 584.8688,
+      "eval_samples_per_second": 5.921,
+      "eval_score": 40.15973749581189,
+      "eval_steps_per_second": 5.921,
+      "eval_sys_len": 50256,
+      "eval_totals": [
+        50256,
+        46793,
+        43338,
+        39918
+      ],
+      "step": 4600
+    },
+    {
+      "epoch": 4.11,
+      "learning_rate": 0.001,
+      "loss": 0.7484,
+      "step": 4610
+    },
+    {
+      "epoch": 4.12,
+      "learning_rate": 0.001,
+      "loss": 0.6936,
+      "step": 4620
+    },
+    {
+      "epoch": 4.13,
+      "learning_rate": 0.001,
+      "loss": 0.7071,
+      "step": 4630
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.001,
+      "loss": 0.7151,
+      "step": 4640
+    },
+    {
+      "epoch": 4.14,
+      "learning_rate": 0.001,
+      "loss": 0.7612,
+      "step": 4650
+    },
+    {
+      "epoch": 4.15,
+      "learning_rate": 0.001,
+      "loss": 0.7635,
+      "step": 4660
+    },
+    {
+      "epoch": 4.16,
+      "learning_rate": 0.001,
+      "loss": 0.7501,
+      "step": 4670
+    },
+    {
+      "epoch": 4.17,
+      "learning_rate": 0.001,
+      "loss": 0.7595,
+      "step": 4680
+    },
+    {
+      "epoch": 4.18,
+      "learning_rate": 0.001,
+      "loss": 0.7207,
+      "step": 4690
+    },
+    {
+      "epoch": 4.19,
+      "learning_rate": 0.001,
+      "loss": 0.7126,
+      "step": 4700
+    },
+    {
+      "epoch": 4.2,
+      "learning_rate": 0.001,
+      "loss": 0.7357,
+      "step": 4710
+    },
+    {
+      "epoch": 4.21,
+      "learning_rate": 0.001,
+      "loss": 0.7577,
+      "step": 4720
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.001,
+      "loss": 0.7425,
+      "step": 4730
+    },
+    {
+      "epoch": 4.22,
+      "learning_rate": 0.001,
+      "loss": 0.7373,
+      "step": 4740
+    },
+    {
+      "epoch": 4.23,
+      "learning_rate": 0.001,
+      "loss": 0.7281,
+      "step": 4750
+    },
+    {
+      "epoch": 4.24,
+      "learning_rate": 0.001,
+      "loss": 0.7504,
+      "step": 4760
+    },
+    {
+      "epoch": 4.25,
+      "learning_rate": 0.001,
+      "loss": 0.7454,
+      "step": 4770
+    },
+    {
+      "epoch": 4.26,
+      "learning_rate": 0.001,
+      "loss": 0.7439,
+      "step": 4780
+    },
+    {
+      "epoch": 4.27,
+      "learning_rate": 0.001,
+      "loss": 0.7349,
+      "step": 4790
+    },
+    {
+      "epoch": 4.28,
+      "learning_rate": 0.001,
+      "loss": 0.7379,
+      "step": 4800
+    },
+    {
+      "epoch": 4.28,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27615,
+        19598,
+        14752,
+        11330
+      ],
+      "eval_loss": 0.4184766411781311,
+      "eval_precisions": [
+        50.95676563393796,
+        38.63197319140548,
+        31.2072941126695,
+        25.83985221337834
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 631.7986,
+      "eval_samples_per_second": 5.481,
+      "eval_score": 35.49553457595366,
+      "eval_steps_per_second": 5.481,
+      "eval_sys_len": 54193,
+      "eval_totals": [
+        54193,
+        50730,
+        47271,
+        43847
+      ],
+      "step": 4800
+    },
+    {
+      "epoch": 4.29,
+      "learning_rate": 0.001,
+      "loss": 0.7364,
+      "step": 4810
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.001,
+      "loss": 0.7491,
+      "step": 4820
+    },
+    {
+      "epoch": 4.3,
+      "learning_rate": 0.001,
+      "loss": 0.7557,
+      "step": 4830
+    },
+    {
+      "epoch": 4.31,
+      "learning_rate": 0.001,
+      "loss": 0.7362,
+      "step": 4840
+    },
+    {
+      "epoch": 4.32,
+      "learning_rate": 0.001,
+      "loss": 0.7544,
+      "step": 4850
+    },
+    {
+      "epoch": 4.33,
+      "learning_rate": 0.001,
+      "loss": 0.7076,
+      "step": 4860
+    },
+    {
+      "epoch": 4.34,
+      "learning_rate": 0.001,
+      "loss": 0.7698,
+      "step": 4870
+    },
+    {
+      "epoch": 4.35,
+      "learning_rate": 0.001,
+      "loss": 0.7554,
+      "step": 4880
+    },
+    {
+      "epoch": 4.36,
+      "learning_rate": 0.001,
+      "loss": 0.7318,
+      "step": 4890
+    },
+    {
+      "epoch": 4.37,
+      "learning_rate": 0.001,
+      "loss": 0.7667,
+      "step": 4900
+    },
+    {
+      "epoch": 4.38,
+      "learning_rate": 0.001,
+      "loss": 0.7311,
+      "step": 4910
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.001,
+      "loss": 0.7199,
+      "step": 4920
+    },
+    {
+      "epoch": 4.39,
+      "learning_rate": 0.001,
+      "loss": 0.7221,
+      "step": 4930
+    },
+    {
+      "epoch": 4.4,
+      "learning_rate": 0.001,
+      "loss": 0.744,
+      "step": 4940
+    },
+    {
+      "epoch": 4.41,
+      "learning_rate": 0.001,
+      "loss": 0.721,
+      "step": 4950
+    },
+    {
+      "epoch": 4.42,
+      "learning_rate": 0.001,
+      "loss": 0.7239,
+      "step": 4960
+    },
+    {
+      "epoch": 4.43,
+      "learning_rate": 0.001,
+      "loss": 0.7995,
+      "step": 4970
+    },
+    {
+      "epoch": 4.44,
+      "learning_rate": 0.001,
+      "loss": 0.7436,
+      "step": 4980
+    },
+    {
+      "epoch": 4.45,
+      "learning_rate": 0.001,
+      "loss": 0.7742,
+      "step": 4990
+    },
+    {
+      "epoch": 4.46,
+      "learning_rate": 0.001,
+      "loss": 0.7488,
+      "step": 5000
+    },
+    {
+      "epoch": 4.46,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28036,
+        19955,
+        14999,
+        11506
+      ],
+      "eval_loss": 0.41945788264274597,
+      "eval_precisions": [
+        58.80773587280279,
+        45.1358259256746,
+        36.801040312093626,
+        30.811664836783333
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 519.6429,
+      "eval_samples_per_second": 6.664,
+      "eval_score": 41.65170448536966,
+      "eval_steps_per_second": 6.664,
+      "eval_sys_len": 47674,
+      "eval_totals": [
+        47674,
+        44211,
+        40757,
+        37343
+      ],
+      "step": 5000
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.001,
+      "loss": 0.7725,
+      "step": 5010
+    },
+    {
+      "epoch": 4.47,
+      "learning_rate": 0.001,
+      "loss": 0.7873,
+      "step": 5020
+    },
+    {
+      "epoch": 4.48,
+      "learning_rate": 0.001,
+      "loss": 0.6723,
+      "step": 5030
+    },
+    {
+      "epoch": 4.49,
+      "learning_rate": 0.001,
+      "loss": 0.7106,
+      "step": 5040
+    },
+    {
+      "epoch": 4.5,
+      "learning_rate": 0.001,
+      "loss": 0.7208,
+      "step": 5050
+    },
+    {
+      "epoch": 4.51,
+      "learning_rate": 0.001,
+      "loss": 0.7665,
+      "step": 5060
+    },
+    {
+      "epoch": 4.52,
+      "learning_rate": 0.001,
+      "loss": 0.7777,
+      "step": 5070
+    },
+    {
+      "epoch": 4.53,
+      "learning_rate": 0.001,
+      "loss": 0.7249,
+      "step": 5080
+    },
+    {
+      "epoch": 4.54,
+      "learning_rate": 0.001,
+      "loss": 0.7591,
+      "step": 5090
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.001,
+      "loss": 0.7441,
+      "step": 5100
+    },
+    {
+      "epoch": 4.55,
+      "learning_rate": 0.001,
+      "loss": 0.7468,
+      "step": 5110
+    },
+    {
+      "epoch": 4.56,
+      "learning_rate": 0.001,
+      "loss": 0.7277,
+      "step": 5120
+    },
+    {
+      "epoch": 4.57,
+      "learning_rate": 0.001,
+      "loss": 0.7283,
+      "step": 5130
+    },
+    {
+      "epoch": 4.58,
+      "learning_rate": 0.001,
+      "loss": 0.7538,
+      "step": 5140
+    },
+    {
+      "epoch": 4.59,
+      "learning_rate": 0.001,
+      "loss": 0.7294,
+      "step": 5150
+    },
+    {
+      "epoch": 4.6,
+      "learning_rate": 0.001,
+      "loss": 0.7511,
+      "step": 5160
+    },
+    {
+      "epoch": 4.61,
+      "learning_rate": 0.001,
+      "loss": 0.7919,
+      "step": 5170
+    },
+    {
+      "epoch": 4.62,
+      "learning_rate": 0.001,
+      "loss": 0.7084,
+      "step": 5180
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.001,
+      "loss": 0.7445,
+      "step": 5190
+    },
+    {
+      "epoch": 4.63,
+      "learning_rate": 0.001,
+      "loss": 0.7782,
+      "step": 5200
+    },
+    {
+      "epoch": 4.63,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28268,
+        20306,
+        15368,
+        11874
+      ],
+      "eval_loss": 0.4119018316268921,
+      "eval_precisions": [
+        65.29311220954405,
+        50.98039215686274,
+        42.248797250859106,
+        36.02876475407349
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 400.7856,
+      "eval_samples_per_second": 8.641,
+      "eval_score": 47.44424785372734,
+      "eval_steps_per_second": 8.641,
+      "eval_sys_len": 43294,
+      "eval_totals": [
+        43294,
+        39831,
+        36375,
+        32957
+      ],
+      "step": 5200
+    },
+    {
+      "epoch": 4.64,
+      "learning_rate": 0.001,
+      "loss": 0.7814,
+      "step": 5210
+    },
+    {
+      "epoch": 4.65,
+      "learning_rate": 0.001,
+      "loss": 0.763,
+      "step": 5220
+    },
+    {
+      "epoch": 4.66,
+      "learning_rate": 0.001,
+      "loss": 0.7003,
+      "step": 5230
+    },
+    {
+      "epoch": 4.67,
+      "learning_rate": 0.001,
+      "loss": 0.7538,
+      "step": 5240
+    },
+    {
+      "epoch": 4.68,
+      "learning_rate": 0.001,
+      "loss": 0.6957,
+      "step": 5250
+    },
+    {
+      "epoch": 4.69,
+      "learning_rate": 0.001,
+      "loss": 0.7334,
+      "step": 5260
+    },
+    {
+      "epoch": 4.7,
+      "learning_rate": 0.001,
+      "loss": 0.7609,
+      "step": 5270
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.001,
+      "loss": 0.7183,
+      "step": 5280
+    },
+    {
+      "epoch": 4.71,
+      "learning_rate": 0.001,
+      "loss": 0.7249,
+      "step": 5290
+    },
+    {
+      "epoch": 4.72,
+      "learning_rate": 0.001,
+      "loss": 0.7514,
+      "step": 5300
+    },
+    {
+      "epoch": 4.73,
+      "learning_rate": 0.001,
+      "loss": 0.7246,
+      "step": 5310
+    },
+    {
+      "epoch": 4.74,
+      "learning_rate": 0.001,
+      "loss": 0.7183,
+      "step": 5320
+    },
+    {
+      "epoch": 4.75,
+      "learning_rate": 0.001,
+      "loss": 0.7347,
+      "step": 5330
+    },
+    {
+      "epoch": 4.76,
+      "learning_rate": 0.001,
+      "loss": 0.7779,
+      "step": 5340
+    },
+    {
+      "epoch": 4.77,
+      "learning_rate": 0.001,
+      "loss": 0.7029,
+      "step": 5350
+    },
+    {
+      "epoch": 4.78,
+      "learning_rate": 0.001,
+      "loss": 0.7365,
+      "step": 5360
+    },
+    {
+      "epoch": 4.79,
+      "learning_rate": 0.001,
+      "loss": 0.7072,
+      "step": 5370
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.001,
+      "loss": 0.7292,
+      "step": 5380
+    },
+    {
+      "epoch": 4.8,
+      "learning_rate": 0.001,
+      "loss": 0.7763,
+      "step": 5390
+    },
+    {
+      "epoch": 4.81,
+      "learning_rate": 0.001,
+      "loss": 0.7358,
+      "step": 5400
+    },
+    {
+      "epoch": 4.81,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28311,
+        20157,
+        15171,
+        11654
+      ],
+      "eval_loss": 0.4105658531188965,
+      "eval_precisions": [
+        61.92528107091299,
+        47.703230386936454,
+        39.0985000773156,
+        32.9339286723563
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 468.4336,
+      "eval_samples_per_second": 7.393,
+      "eval_score": 44.16262233243579,
+      "eval_steps_per_second": 7.393,
+      "eval_sys_len": 45718,
+      "eval_totals": [
+        45718,
+        42255,
+        38802,
+        35386
+      ],
+      "step": 5400
+    },
+    {
+      "epoch": 4.82,
+      "learning_rate": 0.001,
+      "loss": 0.7607,
+      "step": 5410
+    },
+    {
+      "epoch": 4.83,
+      "learning_rate": 0.001,
+      "loss": 0.7039,
+      "step": 5420
+    },
+    {
+      "epoch": 4.84,
+      "learning_rate": 0.001,
+      "loss": 0.7188,
+      "step": 5430
+    },
+    {
+      "epoch": 4.85,
+      "learning_rate": 0.001,
+      "loss": 0.7493,
+      "step": 5440
+    },
+    {
+      "epoch": 4.86,
+      "learning_rate": 0.001,
+      "loss": 0.741,
+      "step": 5450
+    },
+    {
+      "epoch": 4.87,
+      "learning_rate": 0.001,
+      "loss": 0.7424,
+      "step": 5460
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.001,
+      "loss": 0.7076,
+      "step": 5470
+    },
+    {
+      "epoch": 4.88,
+      "learning_rate": 0.001,
+      "loss": 0.7188,
+      "step": 5480
+    },
+    {
+      "epoch": 4.89,
+      "learning_rate": 0.001,
+      "loss": 0.7182,
+      "step": 5490
+    },
+    {
+      "epoch": 4.9,
+      "learning_rate": 0.001,
+      "loss": 0.7285,
+      "step": 5500
+    },
+    {
+      "epoch": 4.91,
+      "learning_rate": 0.001,
+      "loss": 0.7231,
+      "step": 5510
+    },
+    {
+      "epoch": 4.92,
+      "learning_rate": 0.001,
+      "loss": 0.718,
+      "step": 5520
+    },
+    {
+      "epoch": 4.93,
+      "learning_rate": 0.001,
+      "loss": 0.7232,
+      "step": 5530
+    },
+    {
+      "epoch": 4.94,
+      "learning_rate": 0.001,
+      "loss": 0.7678,
+      "step": 5540
+    },
+    {
+      "epoch": 4.95,
+      "learning_rate": 0.001,
+      "loss": 0.7502,
+      "step": 5550
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.001,
+      "loss": 0.742,
+      "step": 5560
+    },
+    {
+      "epoch": 4.96,
+      "learning_rate": 0.001,
+      "loss": 0.7435,
+      "step": 5570
+    },
+    {
+      "epoch": 4.97,
+      "learning_rate": 0.001,
+      "loss": 0.7489,
+      "step": 5580
+    },
+    {
+      "epoch": 4.98,
+      "learning_rate": 0.001,
+      "loss": 0.7244,
+      "step": 5590
+    },
+    {
+      "epoch": 4.99,
+      "learning_rate": 0.001,
+      "loss": 0.7587,
+      "step": 5600
+    },
+    {
+      "epoch": 4.99,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28615,
+        20649,
+        15678,
+        12143
+      ],
+      "eval_loss": 0.4097524583339691,
+      "eval_precisions": [
+        62.32711114983337,
+        48.64540143234075,
+        40.209279064399475,
+        34.13448023837634
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 437.6071,
+      "eval_samples_per_second": 7.913,
+      "eval_score": 45.16578809030254,
+      "eval_steps_per_second": 7.913,
+      "eval_sys_len": 45911,
+      "eval_totals": [
+        45911,
+        42448,
+        38991,
+        35574
+      ],
+      "step": 5600
+    },
+    {
+      "epoch": 5.0,
+      "learning_rate": 0.001,
+      "loss": 0.7357,
+      "step": 5610
+    },
+    {
+      "epoch": 5.01,
+      "learning_rate": 0.001,
+      "loss": 0.6867,
+      "step": 5620
+    },
+    {
+      "epoch": 5.02,
+      "learning_rate": 0.001,
+      "loss": 0.6787,
+      "step": 5630
+    },
+    {
+      "epoch": 5.03,
+      "learning_rate": 0.001,
+      "loss": 0.7284,
+      "step": 5640
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.001,
+      "loss": 0.6993,
+      "step": 5650
+    },
+    {
+      "epoch": 5.04,
+      "learning_rate": 0.001,
+      "loss": 0.7037,
+      "step": 5660
+    },
+    {
+      "epoch": 5.05,
+      "learning_rate": 0.001,
+      "loss": 0.6966,
+      "step": 5670
+    },
+    {
+      "epoch": 5.06,
+      "learning_rate": 0.001,
+      "loss": 0.71,
+      "step": 5680
+    },
+    {
+      "epoch": 5.07,
+      "learning_rate": 0.001,
+      "loss": 0.6992,
+      "step": 5690
+    },
+    {
+      "epoch": 5.08,
+      "learning_rate": 0.001,
+      "loss": 0.6615,
+      "step": 5700
+    },
+    {
+      "epoch": 5.09,
+      "learning_rate": 0.001,
+      "loss": 0.6818,
+      "step": 5710
+    },
+    {
+      "epoch": 5.1,
+      "learning_rate": 0.001,
+      "loss": 0.6957,
+      "step": 5720
+    },
+    {
+      "epoch": 5.11,
+      "learning_rate": 0.001,
+      "loss": 0.6872,
+      "step": 5730
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.001,
+      "loss": 0.6464,
+      "step": 5740
+    },
+    {
+      "epoch": 5.12,
+      "learning_rate": 0.001,
+      "loss": 0.6667,
+      "step": 5750
+    },
+    {
+      "epoch": 5.13,
+      "learning_rate": 0.001,
+      "loss": 0.7373,
+      "step": 5760
+    },
+    {
+      "epoch": 5.14,
+      "learning_rate": 0.001,
+      "loss": 0.692,
+      "step": 5770
+    },
+    {
+      "epoch": 5.15,
+      "learning_rate": 0.001,
+      "loss": 0.7055,
+      "step": 5780
+    },
+    {
+      "epoch": 5.16,
+      "learning_rate": 0.001,
+      "loss": 0.7528,
+      "step": 5790
+    },
+    {
+      "epoch": 5.17,
+      "learning_rate": 0.001,
+      "loss": 0.6792,
+      "step": 5800
+    },
+    {
+      "epoch": 5.17,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        27651,
+        19813,
+        15002,
+        11593
+      ],
+      "eval_loss": 0.4170142412185669,
+      "eval_precisions": [
+        62.127305817062485,
+        48.2725855179807,
+        39.915921668795235,
+        33.93735362997658
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 414.7853,
+      "eval_samples_per_second": 8.349,
+      "eval_score": 44.89539574981961,
+      "eval_steps_per_second": 8.349,
+      "eval_sys_len": 44507,
+      "eval_totals": [
+        44507,
+        41044,
+        37584,
+        34160
+      ],
+      "step": 5800
+    },
+    {
+      "epoch": 5.18,
+      "learning_rate": 0.001,
+      "loss": 0.715,
+      "step": 5810
+    },
+    {
+      "epoch": 5.19,
+      "learning_rate": 0.001,
+      "loss": 0.6674,
+      "step": 5820
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.001,
+      "loss": 0.6976,
+      "step": 5830
+    },
+    {
+      "epoch": 5.2,
+      "learning_rate": 0.001,
+      "loss": 0.6825,
+      "step": 5840
+    },
+    {
+      "epoch": 5.21,
+      "learning_rate": 0.001,
+      "loss": 0.6884,
+      "step": 5850
+    },
+    {
+      "epoch": 5.22,
+      "learning_rate": 0.001,
+      "loss": 0.6909,
+      "step": 5860
+    },
+    {
+      "epoch": 5.23,
+      "learning_rate": 0.001,
+      "loss": 0.6839,
+      "step": 5870
+    },
+    {
+      "epoch": 5.24,
+      "learning_rate": 0.001,
+      "loss": 0.7047,
+      "step": 5880
+    },
+    {
+      "epoch": 5.25,
+      "learning_rate": 0.001,
+      "loss": 0.7109,
+      "step": 5890
+    },
+    {
+      "epoch": 5.26,
+      "learning_rate": 0.001,
+      "loss": 0.7006,
+      "step": 5900
+    },
+    {
+      "epoch": 5.27,
+      "learning_rate": 0.001,
+      "loss": 0.7107,
+      "step": 5910
+    },
+    {
+      "epoch": 5.28,
+      "learning_rate": 0.001,
+      "loss": 0.6975,
+      "step": 5920
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.001,
+      "loss": 0.7228,
+      "step": 5930
+    },
+    {
+      "epoch": 5.29,
+      "learning_rate": 0.001,
+      "loss": 0.6847,
+      "step": 5940
+    },
+    {
+      "epoch": 5.3,
+      "learning_rate": 0.001,
+      "loss": 0.693,
+      "step": 5950
+    },
+    {
+      "epoch": 5.31,
+      "learning_rate": 0.001,
+      "loss": 0.6744,
+      "step": 5960
+    },
+    {
+      "epoch": 5.32,
+      "learning_rate": 0.001,
+      "loss": 0.7602,
+      "step": 5970
+    },
+    {
+      "epoch": 5.33,
+      "learning_rate": 0.001,
+      "loss": 0.6582,
+      "step": 5980
+    },
+    {
+      "epoch": 5.34,
+      "learning_rate": 0.001,
+      "loss": 0.6723,
+      "step": 5990
+    },
+    {
+      "epoch": 5.35,
+      "learning_rate": 0.001,
+      "loss": 0.6927,
+      "step": 6000
+    },
+    {
+      "epoch": 5.35,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28469,
+        20488,
+        15559,
+        12021
+      ],
+      "eval_loss": 0.4135335087776184,
+      "eval_precisions": [
+        65.44296813939589,
+        51.170109143585,
+        42.533008939066725,
+        36.24713544807623
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 396.1878,
+      "eval_samples_per_second": 8.741,
+      "eval_score": 47.667215417697754,
+      "eval_steps_per_second": 8.741,
+      "eval_sys_len": 43502,
+      "eval_totals": [
+        43502,
+        40039,
+        36581,
+        33164
+      ],
+      "step": 6000
+    },
+    {
+      "epoch": 5.36,
+      "learning_rate": 0.001,
+      "loss": 0.7059,
+      "step": 6010
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.001,
+      "loss": 0.6644,
+      "step": 6020
+    },
+    {
+      "epoch": 5.37,
+      "learning_rate": 0.001,
+      "loss": 0.6709,
+      "step": 6030
+    },
+    {
+      "epoch": 5.38,
+      "learning_rate": 0.001,
+      "loss": 0.6865,
+      "step": 6040
+    },
+    {
+      "epoch": 5.39,
+      "learning_rate": 0.001,
+      "loss": 0.6894,
+      "step": 6050
+    },
+    {
+      "epoch": 5.4,
+      "learning_rate": 0.001,
+      "loss": 0.6821,
+      "step": 6060
+    },
+    {
+      "epoch": 5.41,
+      "learning_rate": 0.001,
+      "loss": 0.7082,
+      "step": 6070
+    },
+    {
+      "epoch": 5.42,
+      "learning_rate": 0.001,
+      "loss": 0.694,
+      "step": 6080
+    },
+    {
+      "epoch": 5.43,
+      "learning_rate": 0.001,
+      "loss": 0.7056,
+      "step": 6090
+    },
+    {
+      "epoch": 5.44,
+      "learning_rate": 0.001,
+      "loss": 0.7194,
+      "step": 6100
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.001,
+      "loss": 0.6988,
+      "step": 6110
+    },
+    {
+      "epoch": 5.45,
+      "learning_rate": 0.001,
+      "loss": 0.7332,
+      "step": 6120
+    },
+    {
+      "epoch": 5.46,
+      "learning_rate": 0.001,
+      "loss": 0.6843,
+      "step": 6130
+    },
+    {
+      "epoch": 5.47,
+      "learning_rate": 0.001,
+      "loss": 0.7271,
+      "step": 6140
+    },
+    {
+      "epoch": 5.48,
+      "learning_rate": 0.001,
+      "loss": 0.7122,
+      "step": 6150
+    },
+    {
+      "epoch": 5.49,
+      "learning_rate": 0.001,
+      "loss": 0.7164,
+      "step": 6160
+    },
+    {
+      "epoch": 5.5,
+      "learning_rate": 0.001,
+      "loss": 0.7547,
+      "step": 6170
+    },
+    {
+      "epoch": 5.51,
+      "learning_rate": 0.001,
+      "loss": 0.6799,
+      "step": 6180
+    },
+    {
+      "epoch": 5.52,
+      "learning_rate": 0.001,
+      "loss": 0.7032,
+      "step": 6190
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.001,
+      "loss": 0.6984,
+      "step": 6200
+    },
+    {
+      "epoch": 5.53,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28063,
+        20180,
+        15363,
+        11911
+      ],
+      "eval_loss": 0.4087334871292114,
+      "eval_precisions": [
+        59.575416622439235,
+        46.239860684661565,
+        38.22402468152866,
+        32.38620914677253
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 443.9736,
+      "eval_samples_per_second": 7.8,
+      "eval_score": 42.97293668517703,
+      "eval_steps_per_second": 7.8,
+      "eval_sys_len": 47105,
+      "eval_totals": [
+        47105,
+        43642,
+        40192,
+        36778
+      ],
+      "step": 6200
+    },
+    {
+      "epoch": 5.53,
+      "learning_rate": 0.001,
+      "loss": 0.7198,
+      "step": 6210
+    },
+    {
+      "epoch": 5.54,
+      "learning_rate": 0.001,
+      "loss": 0.6764,
+      "step": 6220
+    },
+    {
+      "epoch": 5.55,
+      "learning_rate": 0.001,
+      "loss": 0.7239,
+      "step": 6230
+    },
+    {
+      "epoch": 5.56,
+      "learning_rate": 0.001,
+      "loss": 0.6859,
+      "step": 6240
+    },
+    {
+      "epoch": 5.57,
+      "learning_rate": 0.001,
+      "loss": 0.7203,
+      "step": 6250
+    },
+    {
+      "epoch": 5.58,
+      "learning_rate": 0.001,
+      "loss": 0.7088,
+      "step": 6260
+    },
+    {
+      "epoch": 5.59,
+      "learning_rate": 0.001,
+      "loss": 0.689,
+      "step": 6270
+    },
+    {
+      "epoch": 5.6,
+      "learning_rate": 0.001,
+      "loss": 0.6946,
+      "step": 6280
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.001,
+      "loss": 0.6679,
+      "step": 6290
+    },
+    {
+      "epoch": 5.61,
+      "learning_rate": 0.001,
+      "loss": 0.7237,
+      "step": 6300
+    },
+    {
+      "epoch": 5.62,
+      "learning_rate": 0.001,
+      "loss": 0.7139,
+      "step": 6310
+    },
+    {
+      "epoch": 5.63,
+      "learning_rate": 0.001,
+      "loss": 0.7018,
+      "step": 6320
+    },
+    {
+      "epoch": 5.64,
+      "learning_rate": 0.001,
+      "loss": 0.7296,
+      "step": 6330
+    },
+    {
+      "epoch": 5.65,
+      "learning_rate": 0.001,
+      "loss": 0.6959,
+      "step": 6340
+    },
+    {
+      "epoch": 5.66,
+      "learning_rate": 0.001,
+      "loss": 0.7384,
+      "step": 6350
+    },
+    {
+      "epoch": 5.67,
+      "learning_rate": 0.001,
+      "loss": 0.7346,
+      "step": 6360
+    },
+    {
+      "epoch": 5.68,
+      "learning_rate": 0.001,
+      "loss": 0.7235,
+      "step": 6370
+    },
+    {
+      "epoch": 5.69,
+      "learning_rate": 0.001,
+      "loss": 0.7139,
+      "step": 6380
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.001,
+      "loss": 0.7185,
+      "step": 6390
+    },
+    {
+      "epoch": 5.7,
+      "learning_rate": 0.001,
+      "loss": 0.7465,
+      "step": 6400
+    },
+    {
+      "epoch": 5.7,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28601,
+        20618,
+        15662,
+        12099
+      ],
+      "eval_loss": 0.4099283218383789,
+      "eval_precisions": [
+        58.90676168310918,
+        45.726325127522735,
+        37.62009992313605,
+        31.665314454709623
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 502.1342,
+      "eval_samples_per_second": 6.897,
+      "eval_score": 42.32371679075985,
+      "eval_steps_per_second": 6.897,
+      "eval_sys_len": 48553,
+      "eval_totals": [
+        48553,
+        45090,
+        41632,
+        38209
+      ],
+      "step": 6400
+    },
+    {
+      "epoch": 5.71,
+      "learning_rate": 0.001,
+      "loss": 0.7045,
+      "step": 6410
+    },
+    {
+      "epoch": 5.72,
+      "learning_rate": 0.001,
+      "loss": 0.664,
+      "step": 6420
+    },
+    {
+      "epoch": 5.73,
+      "learning_rate": 0.001,
+      "loss": 0.7009,
+      "step": 6430
+    },
+    {
+      "epoch": 5.74,
+      "learning_rate": 0.001,
+      "loss": 0.7025,
+      "step": 6440
+    },
+    {
+      "epoch": 5.75,
+      "learning_rate": 0.001,
+      "loss": 0.7353,
+      "step": 6450
+    },
+    {
+      "epoch": 5.76,
+      "learning_rate": 0.001,
+      "loss": 0.6896,
+      "step": 6460
+    },
+    {
+      "epoch": 5.77,
+      "learning_rate": 0.001,
+      "loss": 0.6899,
+      "step": 6470
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.001,
+      "loss": 0.6815,
+      "step": 6480
+    },
+    {
+      "epoch": 5.78,
+      "learning_rate": 0.001,
+      "loss": 0.704,
+      "step": 6490
+    },
+    {
+      "epoch": 5.79,
+      "learning_rate": 0.001,
+      "loss": 0.7097,
+      "step": 6500
+    },
+    {
+      "epoch": 5.8,
+      "learning_rate": 0.001,
+      "loss": 0.6945,
+      "step": 6510
+    },
+    {
+      "epoch": 5.81,
+      "learning_rate": 0.001,
+      "loss": 0.7193,
+      "step": 6520
+    },
+    {
+      "epoch": 5.82,
+      "learning_rate": 0.001,
+      "loss": 0.7148,
+      "step": 6530
+    },
+    {
+      "epoch": 5.83,
+      "learning_rate": 0.001,
+      "loss": 0.7248,
+      "step": 6540
+    },
+    {
+      "epoch": 5.84,
+      "learning_rate": 0.001,
+      "loss": 0.6939,
+      "step": 6550
+    },
+    {
+      "epoch": 5.85,
+      "learning_rate": 0.001,
+      "loss": 0.6806,
+      "step": 6560
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.001,
+      "loss": 0.692,
+      "step": 6570
+    },
+    {
+      "epoch": 5.86,
+      "learning_rate": 0.001,
+      "loss": 0.7103,
+      "step": 6580
+    },
+    {
+      "epoch": 5.87,
+      "learning_rate": 0.001,
+      "loss": 0.7486,
+      "step": 6590
+    },
+    {
+      "epoch": 5.88,
+      "learning_rate": 0.001,
+      "loss": 0.726,
+      "step": 6600
+    },
+    {
+      "epoch": 5.88,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28944,
+        21003,
+        16051,
+        12511
+      ],
+      "eval_loss": 0.4023992419242859,
+      "eval_precisions": [
+        65.1891891891892,
+        51.30566480201285,
+        42.819794584500464,
+        36.71930030523597
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 409.3565,
+      "eval_samples_per_second": 8.46,
+      "eval_score": 47.88725784349983,
+      "eval_steps_per_second": 8.46,
+      "eval_sys_len": 44400,
+      "eval_totals": [
+        44400,
+        40937,
+        37485,
+        34072
+      ],
+      "step": 6600
+    },
+    {
+      "epoch": 5.89,
+      "learning_rate": 0.001,
+      "loss": 0.7087,
+      "step": 6610
+    },
+    {
+      "epoch": 5.9,
+      "learning_rate": 0.001,
+      "loss": 0.7224,
+      "step": 6620
+    },
+    {
+      "epoch": 5.91,
+      "learning_rate": 0.001,
+      "loss": 0.676,
+      "step": 6630
+    },
+    {
+      "epoch": 5.92,
+      "learning_rate": 0.001,
+      "loss": 0.6951,
+      "step": 6640
+    },
+    {
+      "epoch": 5.93,
+      "learning_rate": 0.001,
+      "loss": 0.7293,
+      "step": 6650
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.001,
+      "loss": 0.734,
+      "step": 6660
+    },
+    {
+      "epoch": 5.94,
+      "learning_rate": 0.001,
+      "loss": 0.6968,
+      "step": 6670
+    },
+    {
+      "epoch": 5.95,
+      "learning_rate": 0.001,
+      "loss": 0.6946,
+      "step": 6680
+    },
+    {
+      "epoch": 5.96,
+      "learning_rate": 0.001,
+      "loss": 0.6691,
+      "step": 6690
+    },
+    {
+      "epoch": 5.97,
+      "learning_rate": 0.001,
+      "loss": 0.7026,
+      "step": 6700
+    },
+    {
+      "epoch": 5.98,
+      "learning_rate": 0.001,
+      "loss": 0.7264,
+      "step": 6710
+    },
+    {
+      "epoch": 5.99,
+      "learning_rate": 0.001,
+      "loss": 0.7092,
+      "step": 6720
+    },
+    {
+      "epoch": 6.0,
+      "learning_rate": 0.001,
+      "loss": 0.635,
+      "step": 6730
+    },
+    {
+      "epoch": 6.01,
+      "learning_rate": 0.001,
+      "loss": 0.6688,
+      "step": 6740
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.001,
+      "loss": 0.6522,
+      "step": 6750
+    },
+    {
+      "epoch": 6.02,
+      "learning_rate": 0.001,
+      "loss": 0.6694,
+      "step": 6760
+    },
+    {
+      "epoch": 6.03,
+      "learning_rate": 0.001,
+      "loss": 0.6707,
+      "step": 6770
+    },
+    {
+      "epoch": 6.04,
+      "learning_rate": 0.001,
+      "loss": 0.6127,
+      "step": 6780
+    },
+    {
+      "epoch": 6.05,
+      "learning_rate": 0.001,
+      "loss": 0.6742,
+      "step": 6790
+    },
+    {
+      "epoch": 6.06,
+      "learning_rate": 0.001,
+      "loss": 0.6481,
+      "step": 6800
+    },
+    {
+      "epoch": 6.06,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28255,
+        20328,
+        15387,
+        11879
+      ],
+      "eval_loss": 0.407896488904953,
+      "eval_precisions": [
+        61.8677468797898,
+        48.16262705238468,
+        39.704288589564946,
+        33.61251803853881
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 451.1476,
+      "eval_samples_per_second": 7.676,
+      "eval_score": 44.655836980011564,
+      "eval_steps_per_second": 7.676,
+      "eval_sys_len": 45670,
+      "eval_totals": [
+        45670,
+        42207,
+        38754,
+        35341
+      ],
+      "step": 6800
+    },
+    {
+      "epoch": 6.07,
+      "learning_rate": 0.001,
+      "loss": 0.6797,
+      "step": 6810
+    },
+    {
+      "epoch": 6.08,
+      "learning_rate": 0.001,
+      "loss": 0.6612,
+      "step": 6820
+    },
+    {
+      "epoch": 6.09,
+      "learning_rate": 0.001,
+      "loss": 0.6498,
+      "step": 6830
+    },
+    {
+      "epoch": 6.1,
+      "learning_rate": 0.001,
+      "loss": 0.6577,
+      "step": 6840
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.001,
+      "loss": 0.6477,
+      "step": 6850
+    },
+    {
+      "epoch": 6.11,
+      "learning_rate": 0.001,
+      "loss": 0.6514,
+      "step": 6860
+    },
+    {
+      "epoch": 6.12,
+      "learning_rate": 0.001,
+      "loss": 0.6663,
+      "step": 6870
+    },
+    {
+      "epoch": 6.13,
+      "learning_rate": 0.001,
+      "loss": 0.6274,
+      "step": 6880
+    },
+    {
+      "epoch": 6.14,
+      "learning_rate": 0.001,
+      "loss": 0.6409,
+      "step": 6890
+    },
+    {
+      "epoch": 6.15,
+      "learning_rate": 0.001,
+      "loss": 0.644,
+      "step": 6900
+    },
+    {
+      "epoch": 6.16,
+      "learning_rate": 0.001,
+      "loss": 0.6545,
+      "step": 6910
+    },
+    {
+      "epoch": 6.17,
+      "learning_rate": 0.001,
+      "loss": 0.679,
+      "step": 6920
+    },
+    {
+      "epoch": 6.18,
+      "learning_rate": 0.001,
+      "loss": 0.6643,
+      "step": 6930
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.001,
+      "loss": 0.6624,
+      "step": 6940
+    },
+    {
+      "epoch": 6.19,
+      "learning_rate": 0.001,
+      "loss": 0.6418,
+      "step": 6950
+    },
+    {
+      "epoch": 6.2,
+      "learning_rate": 0.001,
+      "loss": 0.6474,
+      "step": 6960
+    },
+    {
+      "epoch": 6.21,
+      "learning_rate": 0.001,
+      "loss": 0.6754,
+      "step": 6970
+    },
+    {
+      "epoch": 6.22,
+      "learning_rate": 0.001,
+      "loss": 0.687,
+      "step": 6980
+    },
+    {
+      "epoch": 6.23,
+      "learning_rate": 0.001,
+      "loss": 0.6862,
+      "step": 6990
+    },
+    {
+      "epoch": 6.24,
+      "learning_rate": 0.001,
+      "loss": 0.6861,
+      "step": 7000
+    },
+    {
+      "epoch": 6.24,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28664,
+        20760,
+        15826,
+        12278
+      ],
+      "eval_loss": 0.41178786754608154,
+      "eval_precisions": [
+        58.025466102552684,
+        45.19331243469175,
+        37.256055933520095,
+        31.434496530889167
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 530.1653,
+      "eval_samples_per_second": 6.532,
+      "eval_score": 41.862392893483864,
+      "eval_steps_per_second": 6.532,
+      "eval_sys_len": 49399,
+      "eval_totals": [
+        49399,
+        45936,
+        42479,
+        39059
+      ],
+      "step": 7000
+    },
+    {
+      "epoch": 6.25,
+      "learning_rate": 0.001,
+      "loss": 0.6054,
+      "step": 7010
+    },
+    {
+      "epoch": 6.26,
+      "learning_rate": 0.001,
+      "loss": 0.6685,
+      "step": 7020
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.001,
+      "loss": 0.6526,
+      "step": 7030
+    },
+    {
+      "epoch": 6.27,
+      "learning_rate": 0.001,
+      "loss": 0.6443,
+      "step": 7040
+    },
+    {
+      "epoch": 6.28,
+      "learning_rate": 0.001,
+      "loss": 0.6378,
+      "step": 7050
+    },
+    {
+      "epoch": 6.29,
+      "learning_rate": 0.001,
+      "loss": 0.6647,
+      "step": 7060
+    },
+    {
+      "epoch": 6.3,
+      "learning_rate": 0.001,
+      "loss": 0.644,
+      "step": 7070
+    },
+    {
+      "epoch": 6.31,
+      "learning_rate": 0.001,
+      "loss": 0.7003,
+      "step": 7080
+    },
+    {
+      "epoch": 6.32,
+      "learning_rate": 0.001,
+      "loss": 0.6819,
+      "step": 7090
+    },
+    {
+      "epoch": 6.33,
+      "learning_rate": 0.001,
+      "loss": 0.6532,
+      "step": 7100
+    },
+    {
+      "epoch": 6.34,
+      "learning_rate": 0.001,
+      "loss": 0.6615,
+      "step": 7110
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.001,
+      "loss": 0.6749,
+      "step": 7120
+    },
+    {
+      "epoch": 6.35,
+      "learning_rate": 0.001,
+      "loss": 0.6733,
+      "step": 7130
+    },
+    {
+      "epoch": 6.36,
+      "learning_rate": 0.001,
+      "loss": 0.6569,
+      "step": 7140
+    },
+    {
+      "epoch": 6.37,
+      "learning_rate": 0.001,
+      "loss": 0.6881,
+      "step": 7150
+    },
+    {
+      "epoch": 6.38,
+      "learning_rate": 0.001,
+      "loss": 0.6938,
+      "step": 7160
+    },
+    {
+      "epoch": 6.39,
+      "learning_rate": 0.001,
+      "loss": 0.6505,
+      "step": 7170
+    },
+    {
+      "epoch": 6.4,
+      "learning_rate": 0.001,
+      "loss": 0.6582,
+      "step": 7180
+    },
+    {
+      "epoch": 6.41,
+      "learning_rate": 0.001,
+      "loss": 0.6855,
+      "step": 7190
+    },
+    {
+      "epoch": 6.42,
+      "learning_rate": 0.001,
+      "loss": 0.6636,
+      "step": 7200
+    },
+    {
+      "epoch": 6.42,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28586,
+        20643,
+        15705,
+        12185
+      ],
+      "eval_loss": 0.4073503613471985,
+      "eval_precisions": [
+        63.47789398885262,
+        49.658407505412555,
+        41.19884575026233,
+        35.1071799008874
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 433.3648,
+      "eval_samples_per_second": 7.991,
+      "eval_score": 46.20873428545001,
+      "eval_steps_per_second": 7.991,
+      "eval_sys_len": 45033,
+      "eval_totals": [
+        45033,
+        41570,
+        38120,
+        34708
+      ],
+      "step": 7200
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.001,
+      "loss": 0.6498,
+      "step": 7210
+    },
+    {
+      "epoch": 6.43,
+      "learning_rate": 0.001,
+      "loss": 0.6536,
+      "step": 7220
+    },
+    {
+      "epoch": 6.44,
+      "learning_rate": 0.001,
+      "loss": 0.6747,
+      "step": 7230
+    },
+    {
+      "epoch": 6.45,
+      "learning_rate": 0.001,
+      "loss": 0.6563,
+      "step": 7240
+    },
+    {
+      "epoch": 6.46,
+      "learning_rate": 0.001,
+      "loss": 0.6853,
+      "step": 7250
+    },
+    {
+      "epoch": 6.47,
+      "learning_rate": 0.001,
+      "loss": 0.6742,
+      "step": 7260
+    },
+    {
+      "epoch": 6.48,
+      "learning_rate": 0.001,
+      "loss": 0.666,
+      "step": 7270
+    },
+    {
+      "epoch": 6.49,
+      "learning_rate": 0.001,
+      "loss": 0.6618,
+      "step": 7280
+    },
+    {
+      "epoch": 6.5,
+      "learning_rate": 0.001,
+      "loss": 0.6849,
+      "step": 7290
+    },
+    {
+      "epoch": 6.51,
+      "learning_rate": 0.001,
+      "loss": 0.6696,
+      "step": 7300
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.001,
+      "loss": 0.6524,
+      "step": 7310
+    },
+    {
+      "epoch": 6.52,
+      "learning_rate": 0.001,
+      "loss": 0.683,
+      "step": 7320
+    },
+    {
+      "epoch": 6.53,
+      "learning_rate": 0.001,
+      "loss": 0.679,
+      "step": 7330
+    },
+    {
+      "epoch": 6.54,
+      "learning_rate": 0.001,
+      "loss": 0.6418,
+      "step": 7340
+    },
+    {
+      "epoch": 6.55,
+      "learning_rate": 0.001,
+      "loss": 0.6403,
+      "step": 7350
+    },
+    {
+      "epoch": 6.56,
+      "learning_rate": 0.001,
+      "loss": 0.681,
+      "step": 7360
+    },
+    {
+      "epoch": 6.57,
+      "learning_rate": 0.001,
+      "loss": 0.6369,
+      "step": 7370
+    },
+    {
+      "epoch": 6.58,
+      "learning_rate": 0.001,
+      "loss": 0.6728,
+      "step": 7380
+    },
+    {
+      "epoch": 6.59,
+      "learning_rate": 0.001,
+      "loss": 0.6444,
+      "step": 7390
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.001,
+      "loss": 0.6733,
+      "step": 7400
+    },
+    {
+      "epoch": 6.6,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        29063,
+        21171,
+        16212,
+        12650
+      ],
+      "eval_loss": 0.40429091453552246,
+      "eval_precisions": [
+        56.77476069544833,
+        44.358539191652525,
+        36.620736390332056,
+        30.966952264381884
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 459.9316,
+      "eval_samples_per_second": 7.529,
+      "eval_score": 41.109246958331305,
+      "eval_steps_per_second": 7.529,
+      "eval_sys_len": 51190,
+      "eval_totals": [
+        51190,
+        47727,
+        44270,
+        40850
+      ],
+      "step": 7400
+    },
+    {
+      "epoch": 6.6,
+      "learning_rate": 0.001,
+      "loss": 0.6469,
+      "step": 7410
+    },
+    {
+      "epoch": 6.61,
+      "learning_rate": 0.001,
+      "loss": 0.6875,
+      "step": 7420
+    },
+    {
+      "epoch": 6.62,
+      "learning_rate": 0.001,
+      "loss": 0.6756,
+      "step": 7430
+    },
+    {
+      "epoch": 6.63,
+      "learning_rate": 0.001,
+      "loss": 0.688,
+      "step": 7440
+    },
+    {
+      "epoch": 6.64,
+      "learning_rate": 0.001,
+      "loss": 0.6725,
+      "step": 7450
+    },
+    {
+      "epoch": 6.65,
+      "learning_rate": 0.001,
+      "loss": 0.6597,
+      "step": 7460
+    },
+    {
+      "epoch": 6.66,
+      "learning_rate": 0.001,
+      "loss": 0.7135,
+      "step": 7470
+    },
+    {
+      "epoch": 6.67,
+      "learning_rate": 0.001,
+      "loss": 0.6923,
+      "step": 7480
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.001,
+      "loss": 0.6674,
+      "step": 7490
+    },
+    {
+      "epoch": 6.68,
+      "learning_rate": 0.001,
+      "loss": 0.6743,
+      "step": 7500
+    },
+    {
+      "epoch": 6.69,
+      "learning_rate": 0.001,
+      "loss": 0.6706,
+      "step": 7510
+    },
+    {
+      "epoch": 6.7,
+      "learning_rate": 0.001,
+      "loss": 0.7095,
+      "step": 7520
+    },
+    {
+      "epoch": 6.71,
+      "learning_rate": 0.001,
+      "loss": 0.6835,
+      "step": 7530
+    },
+    {
+      "epoch": 6.72,
+      "learning_rate": 0.001,
+      "loss": 0.6735,
+      "step": 7540
+    },
+    {
+      "epoch": 6.73,
+      "learning_rate": 0.001,
+      "loss": 0.6785,
+      "step": 7550
+    },
+    {
+      "epoch": 6.74,
+      "learning_rate": 0.001,
+      "loss": 0.6595,
+      "step": 7560
+    },
+    {
+      "epoch": 6.75,
+      "learning_rate": 0.001,
+      "loss": 0.6841,
+      "step": 7570
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.001,
+      "loss": 0.7051,
+      "step": 7580
+    },
+    {
+      "epoch": 6.76,
+      "learning_rate": 0.001,
+      "loss": 0.6606,
+      "step": 7590
+    },
+    {
+      "epoch": 6.77,
+      "learning_rate": 0.001,
+      "loss": 0.7023,
+      "step": 7600
+    },
+    {
+      "epoch": 6.77,
+      "eval_bp": 1.0,
+      "eval_counts": [
+        28473,
+        20549,
+        15671,
+        12152
+      ],
+      "eval_loss": 0.4036000967025757,
+      "eval_precisions": [
+        63.1624481465871,
+        49.37764321414841,
+        41.066561844863735,
+        34.97985031663788
+      ],
+      "eval_ref_len": 41379,
+      "eval_runtime": 429.7544,
+      "eval_samples_per_second": 8.058,
+      "eval_score": 46.00699850503643,
+      "eval_steps_per_second": 8.058,
+      "eval_sys_len": 45079,
+      "eval_totals": [
+        45079,
+        41616,
+        38160,
+        34740
+      ],
+      "step": 7600
+    }
+  ],
+  "max_steps": 280500,
+  "num_train_epochs": 250,
+  "total_flos": 9.355029219235507e+16,
+  "trial_name": null,
+  "trial_params": null
+}