diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,4830 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 436,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0022935779816513763,
+      "grad_norm": 1.7423042058944702,
+      "learning_rate": 0.0,
+      "loss": 0.1978,
+      "memory/device_reserved (GiB)": 62.78,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 1,
+      "tokens_per_second_per_gpu": 222.25
+    },
+    {
+      "epoch": 0.0045871559633027525,
+      "grad_norm": 1.9587920904159546,
+      "learning_rate": 4.7619047619047615e-06,
+      "loss": 0.2353,
+      "memory/device_reserved (GiB)": 62.86,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 2,
+      "tokens_per_second_per_gpu": 350.65
+    },
+    {
+      "epoch": 0.006880733944954129,
+      "grad_norm": 1.969442367553711,
+      "learning_rate": 9.523809523809523e-06,
+      "loss": 0.2242,
+      "memory/device_reserved (GiB)": 62.96,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 3,
+      "tokens_per_second_per_gpu": 323.39
+    },
+    {
+      "epoch": 0.009174311926605505,
+      "grad_norm": 2.038032054901123,
+      "learning_rate": 1.4285714285714285e-05,
+      "loss": 0.2386,
+      "memory/device_reserved (GiB)": 62.96,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 4,
+      "tokens_per_second_per_gpu": 360.19
+    },
+    {
+      "epoch": 0.011467889908256881,
+      "grad_norm": 1.8824621438980103,
+      "learning_rate": 1.9047619047619046e-05,
+      "loss": 0.2155,
+      "memory/device_reserved (GiB)": 62.96,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 5,
+      "tokens_per_second_per_gpu": 323.59
+    },
+    {
+      "epoch": 0.013761467889908258,
+      "grad_norm": 1.6938378810882568,
+      "learning_rate": 2.380952380952381e-05,
+      "loss": 0.2052,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 6,
+      "tokens_per_second_per_gpu": 377.66
+    },
+    {
+      "epoch": 0.016055045871559634,
+      "grad_norm": 1.7361139059066772,
+      "learning_rate": 2.857142857142857e-05,
+      "loss": 0.2156,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 7,
+      "tokens_per_second_per_gpu": 367.83
+    },
+    {
+      "epoch": 0.01834862385321101,
+      "grad_norm": 1.4934383630752563,
+      "learning_rate": 3.3333333333333335e-05,
+      "loss": 0.1775,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 8,
+      "tokens_per_second_per_gpu": 386.45
+    },
+    {
+      "epoch": 0.020642201834862386,
+      "grad_norm": 1.7436366081237793,
+      "learning_rate": 3.809523809523809e-05,
+      "loss": 0.1866,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 9,
+      "tokens_per_second_per_gpu": 361.96
+    },
+    {
+      "epoch": 0.022935779816513763,
+      "grad_norm": 1.4711122512817383,
+      "learning_rate": 4.2857142857142856e-05,
+      "loss": 0.1476,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 10,
+      "tokens_per_second_per_gpu": 436.99
+    },
+    {
+      "epoch": 0.02522935779816514,
+      "grad_norm": 1.5720685720443726,
+      "learning_rate": 4.761904761904762e-05,
+      "loss": 0.1586,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 11,
+      "tokens_per_second_per_gpu": 328.72
+    },
+    {
+      "epoch": 0.027522935779816515,
+      "grad_norm": 1.3835794925689697,
+      "learning_rate": 5.2380952380952384e-05,
+      "loss": 0.1379,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 12,
+      "tokens_per_second_per_gpu": 349.58
+    },
+    {
+      "epoch": 0.02981651376146789,
+      "grad_norm": 1.467041015625,
+      "learning_rate": 5.714285714285714e-05,
+      "loss": 0.1359,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 13,
+      "tokens_per_second_per_gpu": 387.82
+    },
+    {
+      "epoch": 0.03211009174311927,
+      "grad_norm": 1.4792145490646362,
+      "learning_rate": 6.19047619047619e-05,
+      "loss": 0.1276,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 14,
+      "tokens_per_second_per_gpu": 337.15
+    },
+    {
+      "epoch": 0.034403669724770644,
+      "grad_norm": 1.4552762508392334,
+      "learning_rate": 6.666666666666667e-05,
+      "loss": 0.1088,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 15,
+      "tokens_per_second_per_gpu": 392.76
+    },
+    {
+      "epoch": 0.03669724770642202,
+      "grad_norm": 1.3663415908813477,
+      "learning_rate": 7.142857142857143e-05,
+      "loss": 0.0987,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 16,
+      "tokens_per_second_per_gpu": 415.02
+    },
+    {
+      "epoch": 0.0389908256880734,
+      "grad_norm": 1.4881049394607544,
+      "learning_rate": 7.619047619047618e-05,
+      "loss": 0.1075,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 17,
+      "tokens_per_second_per_gpu": 335.22
+    },
+    {
+      "epoch": 0.04128440366972477,
+      "grad_norm": 1.348233938217163,
+      "learning_rate": 8.095238095238096e-05,
+      "loss": 0.0828,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 18,
+      "tokens_per_second_per_gpu": 345.31
+    },
+    {
+      "epoch": 0.04357798165137615,
+      "grad_norm": 1.3850582838058472,
+      "learning_rate": 8.571428571428571e-05,
+      "loss": 0.0793,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 19,
+      "tokens_per_second_per_gpu": 390.97
+    },
+    {
+      "epoch": 0.045871559633027525,
+      "grad_norm": 1.2875252962112427,
+      "learning_rate": 9.047619047619048e-05,
+      "loss": 0.0637,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 20,
+      "tokens_per_second_per_gpu": 336.71
+    },
+    {
+      "epoch": 0.0481651376146789,
+      "grad_norm": 1.2324072122573853,
+      "learning_rate": 9.523809523809524e-05,
+      "loss": 0.0606,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 21,
+      "tokens_per_second_per_gpu": 366.7
+    },
+    {
+      "epoch": 0.05045871559633028,
+      "grad_norm": 1.2610903978347778,
+      "learning_rate": 0.0001,
+      "loss": 0.0558,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 22,
+      "tokens_per_second_per_gpu": 332.44
+    },
+    {
+      "epoch": 0.052752293577981654,
+      "grad_norm": 0.9550244808197021,
+      "learning_rate": 9.999856734543933e-05,
+      "loss": 0.0498,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 23,
+      "tokens_per_second_per_gpu": 354.84
+    },
+    {
+      "epoch": 0.05504587155963303,
+      "grad_norm": 0.46502307057380676,
+      "learning_rate": 9.999426946385727e-05,
+      "loss": 0.0427,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 24,
+      "tokens_per_second_per_gpu": 437.43
+    },
+    {
+      "epoch": 0.05733944954128441,
+      "grad_norm": 0.3574303984642029,
+      "learning_rate": 9.998710660154898e-05,
+      "loss": 0.0412,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 25,
+      "tokens_per_second_per_gpu": 347.75
+    },
+    {
+      "epoch": 0.05963302752293578,
+      "grad_norm": 0.33508938550949097,
+      "learning_rate": 9.997707916899079e-05,
+      "loss": 0.0512,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 26,
+      "tokens_per_second_per_gpu": 371.44
+    },
+    {
+      "epoch": 0.06192660550458716,
+      "grad_norm": 0.3756350874900818,
+      "learning_rate": 9.996418774081658e-05,
+      "loss": 0.0434,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 27,
+      "tokens_per_second_per_gpu": 383.63
+    },
+    {
+      "epoch": 0.06422018348623854,
+      "grad_norm": 0.34942054748535156,
+      "learning_rate": 9.994843305578486e-05,
+      "loss": 0.0414,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 28,
+      "tokens_per_second_per_gpu": 320.82
+    },
+    {
+      "epoch": 0.06651376146788991,
+      "grad_norm": 0.3070521950721741,
+      "learning_rate": 9.99298160167365e-05,
+      "loss": 0.032,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 29,
+      "tokens_per_second_per_gpu": 309.7
+    },
+    {
+      "epoch": 0.06880733944954129,
+      "grad_norm": 0.33859875798225403,
+      "learning_rate": 9.990833769054293e-05,
+      "loss": 0.0375,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 30,
+      "tokens_per_second_per_gpu": 397.11
+    },
+    {
+      "epoch": 0.07110091743119266,
+      "grad_norm": 0.37161344289779663,
+      "learning_rate": 9.988399930804504e-05,
+      "loss": 0.0444,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 31,
+      "tokens_per_second_per_gpu": 293.8
+    },
+    {
+      "epoch": 0.07339449541284404,
+      "grad_norm": 0.2985033392906189,
+      "learning_rate": 9.985680226398261e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 32,
+      "tokens_per_second_per_gpu": 382.51
+    },
+    {
+      "epoch": 0.07568807339449542,
+      "grad_norm": 0.26349207758903503,
+      "learning_rate": 9.98267481169144e-05,
+      "loss": 0.0292,
+      "memory/device_reserved (GiB)": 63.04,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 33,
+      "tokens_per_second_per_gpu": 416.81
+    },
+    {
+      "epoch": 0.0779816513761468,
+      "grad_norm": 0.27664628624916077,
+      "learning_rate": 9.979383858912885e-05,
+      "loss": 0.0313,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 34,
+      "tokens_per_second_per_gpu": 450.68
+    },
+    {
+      "epoch": 0.08027522935779817,
+      "grad_norm": 0.33281639218330383,
+      "learning_rate": 9.975807556654537e-05,
+      "loss": 0.0326,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 35,
+      "tokens_per_second_per_gpu": 305.93
+    },
+    {
+      "epoch": 0.08256880733944955,
+      "grad_norm": 0.3006906509399414,
+      "learning_rate": 9.971946109860626e-05,
+      "loss": 0.0323,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 36,
+      "tokens_per_second_per_gpu": 304.57
+    },
+    {
+      "epoch": 0.08486238532110092,
+      "grad_norm": 0.3224690556526184,
+      "learning_rate": 9.967799739815925e-05,
+      "loss": 0.0354,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 37,
+      "tokens_per_second_per_gpu": 464.33
+    },
+    {
+      "epoch": 0.0871559633027523,
+      "grad_norm": 0.3223485052585602,
+      "learning_rate": 9.963368684133072e-05,
+      "loss": 0.0359,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 38,
+      "tokens_per_second_per_gpu": 319.52
+    },
+    {
+      "epoch": 0.08944954128440367,
+      "grad_norm": 0.24846729636192322,
+      "learning_rate": 9.958653196738954e-05,
+      "loss": 0.0309,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 39,
+      "tokens_per_second_per_gpu": 406.71
+    },
+    {
+      "epoch": 0.09174311926605505,
+      "grad_norm": 0.33822742104530334,
+      "learning_rate": 9.953653547860151e-05,
+      "loss": 0.0477,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 40,
+      "tokens_per_second_per_gpu": 322.05
+    },
+    {
+      "epoch": 0.09403669724770643,
+      "grad_norm": 0.32081446051597595,
+      "learning_rate": 9.948370024007454e-05,
+      "loss": 0.037,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 41,
+      "tokens_per_second_per_gpu": 416.25
+    },
+    {
+      "epoch": 0.0963302752293578,
+      "grad_norm": 0.3192990720272064,
+      "learning_rate": 9.942802927959443e-05,
+      "loss": 0.045,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 42,
+      "tokens_per_second_per_gpu": 320.22
+    },
+    {
+      "epoch": 0.09862385321100918,
+      "grad_norm": 0.31890812516212463,
+      "learning_rate": 9.936952578745142e-05,
+      "loss": 0.0385,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 43,
+      "tokens_per_second_per_gpu": 362.86
+    },
+    {
+      "epoch": 0.10091743119266056,
+      "grad_norm": 0.26405248045921326,
+      "learning_rate": 9.93081931162573e-05,
+      "loss": 0.0302,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 44,
+      "tokens_per_second_per_gpu": 382.0
+    },
+    {
+      "epoch": 0.10321100917431193,
+      "grad_norm": 0.25612977147102356,
+      "learning_rate": 9.92440347807533e-05,
+      "loss": 0.0316,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 45,
+      "tokens_per_second_per_gpu": 422.99
+    },
+    {
+      "epoch": 0.10550458715596331,
+      "grad_norm": 0.29121536016464233,
+      "learning_rate": 9.91770544576087e-05,
+      "loss": 0.0379,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 46,
+      "tokens_per_second_per_gpu": 338.46
+    },
+    {
+      "epoch": 0.10779816513761468,
+      "grad_norm": 0.27259379625320435,
+      "learning_rate": 9.910725598521013e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 47,
+      "tokens_per_second_per_gpu": 412.59
+    },
+    {
+      "epoch": 0.11009174311926606,
+      "grad_norm": 0.4391794502735138,
+      "learning_rate": 9.90346433634416e-05,
+      "loss": 0.0516,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 48,
+      "tokens_per_second_per_gpu": 282.61
+    },
+    {
+      "epoch": 0.11238532110091744,
+      "grad_norm": 0.34521281719207764,
+      "learning_rate": 9.89592207534552e-05,
+      "loss": 0.0335,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.04,
+      "memory/max_allocated (GiB)": 54.04,
+      "step": 49,
+      "tokens_per_second_per_gpu": 274.17
+    },
+    {
+      "epoch": 0.11467889908256881,
+      "grad_norm": 0.3050225079059601,
+      "learning_rate": 9.888099247743283e-05,
+      "loss": 0.0413,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 50,
+      "tokens_per_second_per_gpu": 376.17
+    },
+    {
+      "epoch": 0.11697247706422019,
+      "grad_norm": 0.2696986496448517,
+      "learning_rate": 9.879996301833833e-05,
+      "loss": 0.0332,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 51,
+      "tokens_per_second_per_gpu": 338.77
+    },
+    {
+      "epoch": 0.11926605504587157,
+      "grad_norm": 0.2705146372318268,
+      "learning_rate": 9.871613701966067e-05,
+      "loss": 0.0323,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 52,
+      "tokens_per_second_per_gpu": 448.16
+    },
+    {
+      "epoch": 0.12155963302752294,
+      "grad_norm": 0.34892207384109497,
+      "learning_rate": 9.862951928514782e-05,
+      "loss": 0.0416,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 53,
+      "tokens_per_second_per_gpu": 281.47
+    },
+    {
+      "epoch": 0.12385321100917432,
+      "grad_norm": 0.30306676030158997,
+      "learning_rate": 9.854011477853146e-05,
+      "loss": 0.0329,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 54,
+      "tokens_per_second_per_gpu": 393.01
+    },
+    {
+      "epoch": 0.12614678899082568,
+      "grad_norm": 0.2857978940010071,
+      "learning_rate": 9.844792862324258e-05,
+      "loss": 0.0315,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 55,
+      "tokens_per_second_per_gpu": 391.6
+    },
+    {
+      "epoch": 0.12844036697247707,
+      "grad_norm": 0.294310063123703,
+      "learning_rate": 9.835296610211779e-05,
+      "loss": 0.035,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 56,
+      "tokens_per_second_per_gpu": 401.46
+    },
+    {
+      "epoch": 0.13073394495412843,
+      "grad_norm": 0.248285710811615,
+      "learning_rate": 9.825523265709666e-05,
+      "loss": 0.0267,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 57,
+      "tokens_per_second_per_gpu": 400.95
+    },
+    {
+      "epoch": 0.13302752293577982,
+      "grad_norm": 0.2690254747867584,
+      "learning_rate": 9.815473388890983e-05,
+      "loss": 0.0326,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 58,
+      "tokens_per_second_per_gpu": 342.0
+    },
+    {
+      "epoch": 0.1353211009174312,
+      "grad_norm": 0.27940982580184937,
+      "learning_rate": 9.805147555675805e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 59,
+      "tokens_per_second_per_gpu": 407.45
+    },
+    {
+      "epoch": 0.13761467889908258,
+      "grad_norm": 0.23285861313343048,
+      "learning_rate": 9.794546357798208e-05,
+      "loss": 0.0269,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 60,
+      "tokens_per_second_per_gpu": 395.56
+    },
+    {
+      "epoch": 0.13990825688073394,
+      "grad_norm": 0.2705455720424652,
+      "learning_rate": 9.783670402772379e-05,
+      "loss": 0.0299,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 61,
+      "tokens_per_second_per_gpu": 404.42
+    },
+    {
+      "epoch": 0.14220183486238533,
+      "grad_norm": 0.30998194217681885,
+      "learning_rate": 9.772520313857775e-05,
+      "loss": 0.0432,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 62,
+      "tokens_per_second_per_gpu": 346.21
+    },
+    {
+      "epoch": 0.1444954128440367,
+      "grad_norm": 0.30185559391975403,
+      "learning_rate": 9.761096730023432e-05,
+      "loss": 0.0379,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 63,
+      "tokens_per_second_per_gpu": 385.68
+    },
+    {
+      "epoch": 0.14678899082568808,
+      "grad_norm": 0.2769138813018799,
+      "learning_rate": 9.749400305911322e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 64,
+      "tokens_per_second_per_gpu": 405.39
+    },
+    {
+      "epoch": 0.14908256880733944,
+      "grad_norm": 0.2720226049423218,
+      "learning_rate": 9.737431711798864e-05,
+      "loss": 0.03,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 65,
+      "tokens_per_second_per_gpu": 377.56
+    },
+    {
+      "epoch": 0.15137614678899083,
+      "grad_norm": 0.31610751152038574,
+      "learning_rate": 9.725191633560491e-05,
+      "loss": 0.0429,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 66,
+      "tokens_per_second_per_gpu": 352.42
+    },
+    {
+      "epoch": 0.1536697247706422,
+      "grad_norm": 0.33488714694976807,
+      "learning_rate": 9.712680772628364e-05,
+      "loss": 0.0361,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 67,
+      "tokens_per_second_per_gpu": 375.77
+    },
+    {
+      "epoch": 0.1559633027522936,
+      "grad_norm": 0.35891827940940857,
+      "learning_rate": 9.69989984595216e-05,
+      "loss": 0.0361,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 68,
+      "tokens_per_second_per_gpu": 356.9
+    },
+    {
+      "epoch": 0.15825688073394495,
+      "grad_norm": 0.2968045175075531,
+      "learning_rate": 9.686849585957994e-05,
+      "loss": 0.0329,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 69,
+      "tokens_per_second_per_gpu": 367.75
+    },
+    {
+      "epoch": 0.16055045871559634,
+      "grad_norm": 0.2232455462217331,
+      "learning_rate": 9.673530740506447e-05,
+      "loss": 0.0314,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 70,
+      "tokens_per_second_per_gpu": 442.71
+    },
+    {
+      "epoch": 0.1628440366972477,
+      "grad_norm": 0.30435213446617126,
+      "learning_rate": 9.659944072849707e-05,
+      "loss": 0.0456,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 71,
+      "tokens_per_second_per_gpu": 403.95
+    },
+    {
+      "epoch": 0.1651376146788991,
+      "grad_norm": 0.28937217593193054,
+      "learning_rate": 9.646090361587827e-05,
+      "loss": 0.0334,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 72,
+      "tokens_per_second_per_gpu": 327.53
+    },
+    {
+      "epoch": 0.16743119266055045,
+      "grad_norm": 0.30293306708335876,
+      "learning_rate": 9.631970400624113e-05,
+      "loss": 0.0337,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 73,
+      "tokens_per_second_per_gpu": 273.81
+    },
+    {
+      "epoch": 0.16972477064220184,
+      "grad_norm": 0.3061770796775818,
+      "learning_rate": 9.617584999119625e-05,
+      "loss": 0.0316,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 74,
+      "tokens_per_second_per_gpu": 347.9
+    },
+    {
+      "epoch": 0.1720183486238532,
+      "grad_norm": 0.33200153708457947,
+      "learning_rate": 9.602934981446803e-05,
+      "loss": 0.0321,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 75,
+      "tokens_per_second_per_gpu": 463.06
+    },
+    {
+      "epoch": 0.1743119266055046,
+      "grad_norm": 0.3321312963962555,
+      "learning_rate": 9.588021187142235e-05,
+      "loss": 0.0415,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 76,
+      "tokens_per_second_per_gpu": 371.73
+    },
+    {
+      "epoch": 0.17660550458715596,
+      "grad_norm": 0.3035199046134949,
+      "learning_rate": 9.572844470858537e-05,
+      "loss": 0.0351,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 77,
+      "tokens_per_second_per_gpu": 403.09
+    },
+    {
+      "epoch": 0.17889908256880735,
+      "grad_norm": 0.2940441071987152,
+      "learning_rate": 9.557405702315381e-05,
+      "loss": 0.0292,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 78,
+      "tokens_per_second_per_gpu": 414.16
+    },
+    {
+      "epoch": 0.1811926605504587,
+      "grad_norm": 0.26305070519447327,
+      "learning_rate": 9.541705766249655e-05,
+      "loss": 0.0287,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 79,
+      "tokens_per_second_per_gpu": 427.63
+    },
+    {
+      "epoch": 0.1834862385321101,
+      "grad_norm": 0.25760307908058167,
+      "learning_rate": 9.525745562364756e-05,
+      "loss": 0.0309,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 80,
+      "tokens_per_second_per_gpu": 333.83
+    },
+    {
+      "epoch": 0.18577981651376146,
+      "grad_norm": 0.23999342322349548,
+      "learning_rate": 9.509526005279044e-05,
+      "loss": 0.0258,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 81,
+      "tokens_per_second_per_gpu": 361.37
+    },
+    {
+      "epoch": 0.18807339449541285,
+      "grad_norm": 0.26708847284317017,
+      "learning_rate": 9.493048024473412e-05,
+      "loss": 0.0367,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 82,
+      "tokens_per_second_per_gpu": 349.78
+    },
+    {
+      "epoch": 0.19036697247706422,
+      "grad_norm": 0.3237634301185608,
+      "learning_rate": 9.476312564238034e-05,
+      "loss": 0.0239,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 83,
+      "tokens_per_second_per_gpu": 320.89
+    },
+    {
+      "epoch": 0.1926605504587156,
+      "grad_norm": 0.2014906406402588,
+      "learning_rate": 9.459320583618252e-05,
+      "loss": 0.0238,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 84,
+      "tokens_per_second_per_gpu": 534.67
+    },
+    {
+      "epoch": 0.19495412844036697,
+      "grad_norm": 0.27790626883506775,
+      "learning_rate": 9.442073056359604e-05,
+      "loss": 0.0347,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 85,
+      "tokens_per_second_per_gpu": 421.73
+    },
+    {
+      "epoch": 0.19724770642201836,
+      "grad_norm": 0.2846669554710388,
+      "learning_rate": 9.424570970852034e-05,
+      "loss": 0.0383,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 86,
+      "tokens_per_second_per_gpu": 373.72
+    },
+    {
+      "epoch": 0.19954128440366972,
+      "grad_norm": 0.24163806438446045,
+      "learning_rate": 9.406815330073244e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 87,
+      "tokens_per_second_per_gpu": 405.43
+    },
+    {
+      "epoch": 0.2018348623853211,
+      "grad_norm": 0.28796976804733276,
+      "learning_rate": 9.388807151531229e-05,
+      "loss": 0.0381,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 88,
+      "tokens_per_second_per_gpu": 305.92
+    },
+    {
+      "epoch": 0.20412844036697247,
+      "grad_norm": 0.29206299781799316,
+      "learning_rate": 9.37054746720595e-05,
+      "loss": 0.0293,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 89,
+      "tokens_per_second_per_gpu": 359.51
+    },
+    {
+      "epoch": 0.20642201834862386,
+      "grad_norm": 0.2992270588874817,
+      "learning_rate": 9.352037323490208e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 90,
+      "tokens_per_second_per_gpu": 345.5
+    },
+    {
+      "epoch": 0.20871559633027523,
+      "grad_norm": 0.3274737298488617,
+      "learning_rate": 9.333277781129678e-05,
+      "loss": 0.0398,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 91,
+      "tokens_per_second_per_gpu": 339.92
+    },
+    {
+      "epoch": 0.21100917431192662,
+      "grad_norm": 0.25112593173980713,
+      "learning_rate": 9.314269915162114e-05,
+      "loss": 0.0262,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 92,
+      "tokens_per_second_per_gpu": 394.8
+    },
+    {
+      "epoch": 0.21330275229357798,
+      "grad_norm": 0.259998083114624,
+      "learning_rate": 9.295014814855753e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 93,
+      "tokens_per_second_per_gpu": 371.57
+    },
+    {
+      "epoch": 0.21559633027522937,
+      "grad_norm": 0.2672531008720398,
+      "learning_rate": 9.275513583646884e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 94,
+      "tokens_per_second_per_gpu": 336.64
+    },
+    {
+      "epoch": 0.21788990825688073,
+      "grad_norm": 0.3032653331756592,
+      "learning_rate": 9.255767339076622e-05,
+      "loss": 0.0349,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 95,
+      "tokens_per_second_per_gpu": 373.25
+    },
+    {
+      "epoch": 0.22018348623853212,
+      "grad_norm": 0.2861638367176056,
+      "learning_rate": 9.23577721272686e-05,
+      "loss": 0.0605,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 96,
+      "tokens_per_second_per_gpu": 424.59
+    },
+    {
+      "epoch": 0.22247706422018348,
+      "grad_norm": 0.29100459814071655,
+      "learning_rate": 9.215544350155422e-05,
+      "loss": 0.04,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 97,
+      "tokens_per_second_per_gpu": 375.12
+    },
+    {
+      "epoch": 0.22477064220183487,
+      "grad_norm": 0.2690049111843109,
+      "learning_rate": 9.195069910830427e-05,
+      "loss": 0.0344,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 98,
+      "tokens_per_second_per_gpu": 356.89
+    },
+    {
+      "epoch": 0.22706422018348624,
+      "grad_norm": 0.2193092703819275,
+      "learning_rate": 9.174355068063828e-05,
+      "loss": 0.0281,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 99,
+      "tokens_per_second_per_gpu": 364.88
+    },
+    {
+      "epoch": 0.22935779816513763,
+      "grad_norm": 0.24899783730506897,
+      "learning_rate": 9.15340100894418e-05,
+      "loss": 0.0318,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 100,
+      "tokens_per_second_per_gpu": 352.49
+    },
+    {
+      "epoch": 0.231651376146789,
+      "grad_norm": 0.2679343819618225,
+      "learning_rate": 9.132208934268622e-05,
+      "loss": 0.0259,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 101,
+      "tokens_per_second_per_gpu": 428.81
+    },
+    {
+      "epoch": 0.23394495412844038,
+      "grad_norm": 0.31847280263900757,
+      "learning_rate": 9.110780058474052e-05,
+      "loss": 0.0367,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 102,
+      "tokens_per_second_per_gpu": 387.83
+    },
+    {
+      "epoch": 0.23623853211009174,
+      "grad_norm": 0.2766435742378235,
+      "learning_rate": 9.08911560956753e-05,
+      "loss": 0.0369,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 103,
+      "tokens_per_second_per_gpu": 449.4
+    },
+    {
+      "epoch": 0.23853211009174313,
+      "grad_norm": 0.25810885429382324,
+      "learning_rate": 9.067216829055922e-05,
+      "loss": 0.0298,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 104,
+      "tokens_per_second_per_gpu": 387.04
+    },
+    {
+      "epoch": 0.2408256880733945,
+      "grad_norm": 0.2696630656719208,
+      "learning_rate": 9.045084971874738e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 105,
+      "tokens_per_second_per_gpu": 367.73
+    },
+    {
+      "epoch": 0.24311926605504589,
+      "grad_norm": 0.3242337107658386,
+      "learning_rate": 9.022721306316222e-05,
+      "loss": 0.0413,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 106,
+      "tokens_per_second_per_gpu": 261.31
+    },
+    {
+      "epoch": 0.24541284403669725,
+      "grad_norm": 0.3147953748703003,
+      "learning_rate": 9.000127113956674e-05,
+      "loss": 0.0385,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 107,
+      "tokens_per_second_per_gpu": 295.27
+    },
+    {
+      "epoch": 0.24770642201834864,
+      "grad_norm": 0.31290891766548157,
+      "learning_rate": 8.977303689583e-05,
+      "loss": 0.0388,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 108,
+      "tokens_per_second_per_gpu": 333.94
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 0.2813985049724579,
+      "learning_rate": 8.954252341118523e-05,
+      "loss": 0.0362,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 109,
+      "tokens_per_second_per_gpu": 340.1
+    },
+    {
+      "epoch": 0.25229357798165136,
+      "grad_norm": 0.2733946442604065,
+      "learning_rate": 8.930974389548023e-05,
+      "loss": 0.0369,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 110,
+      "tokens_per_second_per_gpu": 412.96
+    },
+    {
+      "epoch": 0.2545871559633027,
+      "grad_norm": 0.271311491727829,
+      "learning_rate": 8.90747116884204e-05,
+      "loss": 0.0329,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 111,
+      "tokens_per_second_per_gpu": 385.43
+    },
+    {
+      "epoch": 0.25688073394495414,
+      "grad_norm": 0.3098628520965576,
+      "learning_rate": 8.883744025880428e-05,
+      "loss": 0.0424,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 112,
+      "tokens_per_second_per_gpu": 355.56
+    },
+    {
+      "epoch": 0.2591743119266055,
+      "grad_norm": 0.25453096628189087,
+      "learning_rate": 8.859794320375168e-05,
+      "loss": 0.0292,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 113,
+      "tokens_per_second_per_gpu": 373.16
+    },
+    {
+      "epoch": 0.26146788990825687,
+      "grad_norm": 0.2808026373386383,
+      "learning_rate": 8.835623424792452e-05,
+      "loss": 0.032,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 114,
+      "tokens_per_second_per_gpu": 309.26
+    },
+    {
+      "epoch": 0.26376146788990823,
+      "grad_norm": 0.21891231834888458,
+      "learning_rate": 8.811232724274035e-05,
+      "loss": 0.0257,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 115,
+      "tokens_per_second_per_gpu": 416.92
+    },
+    {
+      "epoch": 0.26605504587155965,
+      "grad_norm": 0.26432308554649353,
+      "learning_rate": 8.786623616557847e-05,
+      "loss": 0.0347,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 116,
+      "tokens_per_second_per_gpu": 379.31
+    },
+    {
+      "epoch": 0.268348623853211,
+      "grad_norm": 0.25801753997802734,
+      "learning_rate": 8.761797511897906e-05,
+      "loss": 0.0279,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 117,
+      "tokens_per_second_per_gpu": 363.95
+    },
+    {
+      "epoch": 0.2706422018348624,
+      "grad_norm": 0.27979329228401184,
+      "learning_rate": 8.736755832983497e-05,
+      "loss": 0.0358,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 118,
+      "tokens_per_second_per_gpu": 307.44
+    },
+    {
+      "epoch": 0.27293577981651373,
+      "grad_norm": 0.2807314693927765,
+      "learning_rate": 8.711500014857634e-05,
+      "loss": 0.0355,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 119,
+      "tokens_per_second_per_gpu": 315.85
+    },
+    {
+      "epoch": 0.27522935779816515,
+      "grad_norm": 0.2589738368988037,
+      "learning_rate": 8.686031504834843e-05,
+      "loss": 0.0332,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 120,
+      "tokens_per_second_per_gpu": 370.06
+    },
+    {
+      "epoch": 0.2775229357798165,
+      "grad_norm": 0.2834858000278473,
+      "learning_rate": 8.660351762418203e-05,
+      "loss": 0.04,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 121,
+      "tokens_per_second_per_gpu": 423.2
+    },
+    {
+      "epoch": 0.2798165137614679,
+      "grad_norm": 0.2670377790927887,
+      "learning_rate": 8.634462259215719e-05,
+      "loss": 0.0283,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 122,
+      "tokens_per_second_per_gpu": 303.03
+    },
+    {
+      "epoch": 0.28211009174311924,
+      "grad_norm": 0.28685668110847473,
+      "learning_rate": 8.608364478855983e-05,
+      "loss": 0.0392,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 123,
+      "tokens_per_second_per_gpu": 413.82
+    },
+    {
+      "epoch": 0.28440366972477066,
+      "grad_norm": 0.23065194487571716,
+      "learning_rate": 8.58205991690316e-05,
+      "loss": 0.0308,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 124,
+      "tokens_per_second_per_gpu": 388.34
+    },
+    {
+      "epoch": 0.286697247706422,
+      "grad_norm": 0.26674139499664307,
+      "learning_rate": 8.555550080771273e-05,
+      "loss": 0.0334,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 125,
+      "tokens_per_second_per_gpu": 358.66
+    },
+    {
+      "epoch": 0.2889908256880734,
+      "grad_norm": 0.2913675308227539,
+      "learning_rate": 8.528836489637828e-05,
+      "loss": 0.0419,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 126,
+      "tokens_per_second_per_gpu": 256.07
+    },
+    {
+      "epoch": 0.29128440366972475,
+      "grad_norm": 0.2554248869419098,
+      "learning_rate": 8.501920674356754e-05,
+      "loss": 0.0287,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 127,
+      "tokens_per_second_per_gpu": 369.85
+    },
+    {
+      "epoch": 0.29357798165137616,
+      "grad_norm": 0.2390403300523758,
+      "learning_rate": 8.47480417737067e-05,
+      "loss": 0.029,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 128,
+      "tokens_per_second_per_gpu": 385.26
+    },
+    {
+      "epoch": 0.2958715596330275,
+      "grad_norm": 0.2914634943008423,
+      "learning_rate": 8.447488552622498e-05,
+      "loss": 0.0352,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 129,
+      "tokens_per_second_per_gpu": 344.72
+    },
+    {
+      "epoch": 0.2981651376146789,
+      "grad_norm": 0.25486281514167786,
+      "learning_rate": 8.419975365466415e-05,
+      "loss": 0.0299,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 130,
+      "tokens_per_second_per_gpu": 333.26
+    },
+    {
+      "epoch": 0.30045871559633025,
+      "grad_norm": 0.27855366468429565,
+      "learning_rate": 8.392266192578143e-05,
+      "loss": 0.0376,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 131,
+      "tokens_per_second_per_gpu": 348.74
+    },
+    {
+      "epoch": 0.30275229357798167,
+      "grad_norm": 0.2386406511068344,
+      "learning_rate": 8.364362621864595e-05,
+      "loss": 0.0343,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 132,
+      "tokens_per_second_per_gpu": 351.58
+    },
+    {
+      "epoch": 0.30504587155963303,
+      "grad_norm": 0.24866312742233276,
+      "learning_rate": 8.336266252372889e-05,
+      "loss": 0.0342,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 133,
+      "tokens_per_second_per_gpu": 403.38
+    },
+    {
+      "epoch": 0.3073394495412844,
+      "grad_norm": 0.22284211218357086,
+      "learning_rate": 8.307978694198699e-05,
+      "loss": 0.0323,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 134,
+      "tokens_per_second_per_gpu": 326.88
+    },
+    {
+      "epoch": 0.30963302752293576,
+      "grad_norm": 0.23460738360881805,
+      "learning_rate": 8.279501568393994e-05,
+      "loss": 0.0314,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 135,
+      "tokens_per_second_per_gpu": 319.95
+    },
+    {
+      "epoch": 0.3119266055045872,
+      "grad_norm": 0.25024035573005676,
+      "learning_rate": 8.250836506874142e-05,
+      "loss": 0.0338,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 136,
+      "tokens_per_second_per_gpu": 379.96
+    },
+    {
+      "epoch": 0.31422018348623854,
+      "grad_norm": 0.2852768898010254,
+      "learning_rate": 8.221985152324385e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 137,
+      "tokens_per_second_per_gpu": 383.41
+    },
+    {
+      "epoch": 0.3165137614678899,
+      "grad_norm": 0.22059263288974762,
+      "learning_rate": 8.192949158105713e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 138,
+      "tokens_per_second_per_gpu": 310.55
+    },
+    {
+      "epoch": 0.31880733944954126,
+      "grad_norm": 0.2608126699924469,
+      "learning_rate": 8.163730188160105e-05,
+      "loss": 0.0362,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 139,
+      "tokens_per_second_per_gpu": 372.07
+    },
+    {
+      "epoch": 0.3211009174311927,
+      "grad_norm": 0.26272955536842346,
+      "learning_rate": 8.134329916915184e-05,
+      "loss": 0.0377,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 140,
+      "tokens_per_second_per_gpu": 319.68
+    },
+    {
+      "epoch": 0.32339449541284404,
+      "grad_norm": 0.23061028122901917,
+      "learning_rate": 8.104750029188257e-05,
+      "loss": 0.0326,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 141,
+      "tokens_per_second_per_gpu": 467.38
+    },
+    {
+      "epoch": 0.3256880733944954,
+      "grad_norm": 0.221657395362854,
+      "learning_rate": 8.074992220089769e-05,
+      "loss": 0.0309,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 142,
+      "tokens_per_second_per_gpu": 384.81
+    },
+    {
+      "epoch": 0.32798165137614677,
+      "grad_norm": 0.24493619799613953,
+      "learning_rate": 8.045058194926153e-05,
+      "loss": 0.0327,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 143,
+      "tokens_per_second_per_gpu": 358.09
+    },
+    {
+      "epoch": 0.3302752293577982,
+      "grad_norm": 0.25569525361061096,
+      "learning_rate": 8.014949669102117e-05,
+      "loss": 0.0311,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 144,
+      "tokens_per_second_per_gpu": 364.26
+    },
+    {
+      "epoch": 0.33256880733944955,
+      "grad_norm": 0.2208516001701355,
+      "learning_rate": 7.984668368022335e-05,
+      "loss": 0.0278,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 145,
+      "tokens_per_second_per_gpu": 262.85
+    },
+    {
+      "epoch": 0.3348623853211009,
+      "grad_norm": 0.3040093779563904,
+      "learning_rate": 7.954216026992571e-05,
+      "loss": 0.0342,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 146,
+      "tokens_per_second_per_gpu": 372.26
+    },
+    {
+      "epoch": 0.33715596330275227,
+      "grad_norm": 0.23538342118263245,
+      "learning_rate": 7.923594391120236e-05,
+      "loss": 0.0354,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 147,
+      "tokens_per_second_per_gpu": 438.13
+    },
+    {
+      "epoch": 0.3394495412844037,
+      "grad_norm": 0.23173239827156067,
+      "learning_rate": 7.892805215214381e-05,
+      "loss": 0.0332,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 148,
+      "tokens_per_second_per_gpu": 339.88
+    },
+    {
+      "epoch": 0.34174311926605505,
+      "grad_norm": 0.22046047449111938,
+      "learning_rate": 7.861850263685134e-05,
+      "loss": 0.0309,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 149,
+      "tokens_per_second_per_gpu": 412.55
+    },
+    {
+      "epoch": 0.3440366972477064,
+      "grad_norm": 0.22880156338214874,
+      "learning_rate": 7.830731310442599e-05,
+      "loss": 0.0285,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 150,
+      "tokens_per_second_per_gpu": 337.23
+    },
+    {
+      "epoch": 0.3463302752293578,
+      "grad_norm": 0.2623577117919922,
+      "learning_rate": 7.799450138795185e-05,
+      "loss": 0.0316,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 151,
+      "tokens_per_second_per_gpu": 324.27
+    },
+    {
+      "epoch": 0.3486238532110092,
+      "grad_norm": 0.23668667674064636,
+      "learning_rate": 7.768008541347423e-05,
+      "loss": 0.0285,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 152,
+      "tokens_per_second_per_gpu": 353.61
+    },
+    {
+      "epoch": 0.35091743119266056,
+      "grad_norm": 0.23275883495807648,
+      "learning_rate": 7.73640831989723e-05,
+      "loss": 0.0301,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 153,
+      "tokens_per_second_per_gpu": 406.13
+    },
+    {
+      "epoch": 0.3532110091743119,
+      "grad_norm": 0.21618887782096863,
+      "learning_rate": 7.704651285332663e-05,
+      "loss": 0.0282,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 154,
+      "tokens_per_second_per_gpu": 365.9
+    },
+    {
+      "epoch": 0.3555045871559633,
+      "grad_norm": 0.23415431380271912,
+      "learning_rate": 7.672739257528134e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 155,
+      "tokens_per_second_per_gpu": 437.09
+    },
+    {
+      "epoch": 0.3577981651376147,
+      "grad_norm": 0.28743037581443787,
+      "learning_rate": 7.640674065240136e-05,
+      "loss": 0.0359,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 156,
+      "tokens_per_second_per_gpu": 289.72
+    },
+    {
+      "epoch": 0.36009174311926606,
+      "grad_norm": 0.22749994695186615,
+      "learning_rate": 7.608457546002424e-05,
+      "loss": 0.0275,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 157,
+      "tokens_per_second_per_gpu": 274.83
+    },
+    {
+      "epoch": 0.3623853211009174,
+      "grad_norm": 0.2272910177707672,
+      "learning_rate": 7.576091546020725e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 158,
+      "tokens_per_second_per_gpu": 339.03
+    },
+    {
+      "epoch": 0.3646788990825688,
+      "grad_norm": 0.24701513350009918,
+      "learning_rate": 7.543577920066944e-05,
+      "loss": 0.0372,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 159,
+      "tokens_per_second_per_gpu": 400.2
+    },
+    {
+      "epoch": 0.3669724770642202,
+      "grad_norm": 0.21798241138458252,
+      "learning_rate": 7.510918531372857e-05,
+      "loss": 0.0354,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 160,
+      "tokens_per_second_per_gpu": 317.01
+    },
+    {
+      "epoch": 0.36926605504587157,
+      "grad_norm": 0.218458890914917,
+      "learning_rate": 7.478115251523352e-05,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 161,
+      "tokens_per_second_per_gpu": 359.48
+    },
+    {
+      "epoch": 0.37155963302752293,
+      "grad_norm": 0.21478238701820374,
+      "learning_rate": 7.445169960349167e-05,
+      "loss": 0.024,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 162,
+      "tokens_per_second_per_gpu": 384.53
+    },
+    {
+      "epoch": 0.3738532110091743,
+      "grad_norm": 0.20170234143733978,
+      "learning_rate": 7.412084545819168e-05,
+      "loss": 0.0298,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 163,
+      "tokens_per_second_per_gpu": 385.9
+    },
+    {
+      "epoch": 0.3761467889908257,
+      "grad_norm": 0.21560414135456085,
+      "learning_rate": 7.378860903932159e-05,
+      "loss": 0.0323,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 164,
+      "tokens_per_second_per_gpu": 335.57
+    },
+    {
+      "epoch": 0.37844036697247707,
+      "grad_norm": 0.22123412787914276,
+      "learning_rate": 7.34550093860822e-05,
+      "loss": 0.042,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 165,
+      "tokens_per_second_per_gpu": 408.47
+    },
+    {
+      "epoch": 0.38073394495412843,
+      "grad_norm": 0.22786198556423187,
+      "learning_rate": 7.31200656157961e-05,
+      "loss": 0.0313,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 166,
+      "tokens_per_second_per_gpu": 299.4
+    },
+    {
+      "epoch": 0.3830275229357798,
+      "grad_norm": 0.19029876589775085,
+      "learning_rate": 7.278379692281208e-05,
+      "loss": 0.0316,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 167,
+      "tokens_per_second_per_gpu": 361.02
+    },
+    {
+      "epoch": 0.3853211009174312,
+      "grad_norm": 0.1974560171365738,
+      "learning_rate": 7.244622257740523e-05,
+      "loss": 0.0277,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 168,
+      "tokens_per_second_per_gpu": 382.0
+    },
+    {
+      "epoch": 0.3876146788990826,
+      "grad_norm": 0.18544377386569977,
+      "learning_rate": 7.210736192467256e-05,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 169,
+      "tokens_per_second_per_gpu": 392.7
+    },
+    {
+      "epoch": 0.38990825688073394,
+      "grad_norm": 0.2139897495508194,
+      "learning_rate": 7.176723438342446e-05,
+      "loss": 0.032,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 170,
+      "tokens_per_second_per_gpu": 373.71
+    },
+    {
+      "epoch": 0.3922018348623853,
+      "grad_norm": 0.19245761632919312,
+      "learning_rate": 7.142585944507185e-05,
+      "loss": 0.0252,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 171,
+      "tokens_per_second_per_gpu": 377.45
+    },
+    {
+      "epoch": 0.3944954128440367,
+      "grad_norm": 0.2111990749835968,
+      "learning_rate": 7.10832566725092e-05,
+      "loss": 0.0246,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 172,
+      "tokens_per_second_per_gpu": 336.9
+    },
+    {
+      "epoch": 0.3967889908256881,
+      "grad_norm": 0.21643568575382233,
+      "learning_rate": 7.073944569899354e-05,
+      "loss": 0.0385,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 173,
+      "tokens_per_second_per_gpu": 412.94
+    },
+    {
+      "epoch": 0.39908256880733944,
+      "grad_norm": 0.24215413630008698,
+      "learning_rate": 7.039444622701922e-05,
+      "loss": 0.0387,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 174,
+      "tokens_per_second_per_gpu": 334.89
+    },
+    {
+      "epoch": 0.4013761467889908,
+      "grad_norm": 0.20528389513492584,
+      "learning_rate": 7.00482780271889e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 175,
+      "tokens_per_second_per_gpu": 339.13
+    },
+    {
+      "epoch": 0.4036697247706422,
+      "grad_norm": 0.196381613612175,
+      "learning_rate": 6.97009609370806e-05,
+      "loss": 0.0326,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 176,
+      "tokens_per_second_per_gpu": 375.95
+    },
+    {
+      "epoch": 0.4059633027522936,
+      "grad_norm": 0.18786925077438354,
+      "learning_rate": 6.935251486011087e-05,
+      "loss": 0.0243,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 177,
+      "tokens_per_second_per_gpu": 331.45
+    },
+    {
+      "epoch": 0.40825688073394495,
+      "grad_norm": 0.16835825145244598,
+      "learning_rate": 6.900295976439413e-05,
+      "loss": 0.0262,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 178,
+      "tokens_per_second_per_gpu": 340.03
+    },
+    {
+      "epoch": 0.4105504587155963,
+      "grad_norm": 0.2048378884792328,
+      "learning_rate": 6.865231568159846e-05,
+      "loss": 0.0343,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 179,
+      "tokens_per_second_per_gpu": 395.65
+    },
+    {
+      "epoch": 0.41284403669724773,
+      "grad_norm": 0.2247331142425537,
+      "learning_rate": 6.830060270579768e-05,
+      "loss": 0.0315,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 180,
+      "tokens_per_second_per_gpu": 353.54
+    },
+    {
+      "epoch": 0.4151376146788991,
+      "grad_norm": 0.1774366796016693,
+      "learning_rate": 6.794784099231972e-05,
+      "loss": 0.0264,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 181,
+      "tokens_per_second_per_gpu": 321.16
+    },
+    {
+      "epoch": 0.41743119266055045,
+      "grad_norm": 0.18517284095287323,
+      "learning_rate": 6.759405075659166e-05,
+      "loss": 0.029,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 182,
+      "tokens_per_second_per_gpu": 340.2
+    },
+    {
+      "epoch": 0.4197247706422018,
+      "grad_norm": 0.18091663718223572,
+      "learning_rate": 6.723925227298132e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 183,
+      "tokens_per_second_per_gpu": 325.55
+    },
+    {
+      "epoch": 0.42201834862385323,
+      "grad_norm": 0.2186087667942047,
+      "learning_rate": 6.688346587363533e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 184,
+      "tokens_per_second_per_gpu": 373.33
+    },
+    {
+      "epoch": 0.4243119266055046,
+      "grad_norm": 0.17287197709083557,
+      "learning_rate": 6.652671194731396e-05,
+      "loss": 0.0252,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 185,
+      "tokens_per_second_per_gpu": 358.01
+    },
+    {
+      "epoch": 0.42660550458715596,
+      "grad_norm": 0.19738338887691498,
+      "learning_rate": 6.616901093822283e-05,
+      "loss": 0.0345,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 186,
+      "tokens_per_second_per_gpu": 361.15
+    },
+    {
+      "epoch": 0.4288990825688073,
+      "grad_norm": 0.19579005241394043,
+      "learning_rate": 6.58103833448412e-05,
+      "loss": 0.0224,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 187,
+      "tokens_per_second_per_gpu": 363.04
+    },
+    {
+      "epoch": 0.43119266055045874,
+      "grad_norm": 0.19978734850883484,
+      "learning_rate": 6.545084971874738e-05,
+      "loss": 0.0305,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 188,
+      "tokens_per_second_per_gpu": 453.6
+    },
+    {
+      "epoch": 0.4334862385321101,
+      "grad_norm": 0.20251895487308502,
+      "learning_rate": 6.509043066344092e-05,
+      "loss": 0.0356,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 189,
+      "tokens_per_second_per_gpu": 413.42
+    },
+    {
+      "epoch": 0.43577981651376146,
+      "grad_norm": 0.23251071572303772,
+      "learning_rate": 6.472914683316195e-05,
+      "loss": 0.0393,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 190,
+      "tokens_per_second_per_gpu": 359.47
+    },
+    {
+      "epoch": 0.4380733944954128,
+      "grad_norm": 0.1756441444158554,
+      "learning_rate": 6.436701893170756e-05,
+      "loss": 0.0264,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 191,
+      "tokens_per_second_per_gpu": 367.32
+    },
+    {
+      "epoch": 0.44036697247706424,
+      "grad_norm": 0.19699731469154358,
+      "learning_rate": 6.400406771124536e-05,
+      "loss": 0.0302,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 192,
+      "tokens_per_second_per_gpu": 322.21
+    },
+    {
+      "epoch": 0.4426605504587156,
+      "grad_norm": 0.22188176214694977,
+      "learning_rate": 6.364031397112416e-05,
+      "loss": 0.0332,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 193,
+      "tokens_per_second_per_gpu": 357.33
+    },
+    {
+      "epoch": 0.44495412844036697,
+      "grad_norm": 0.2094058096408844,
+      "learning_rate": 6.327577855668216e-05,
+      "loss": 0.0303,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 194,
+      "tokens_per_second_per_gpu": 413.1
+    },
+    {
+      "epoch": 0.44724770642201833,
+      "grad_norm": 0.22718685865402222,
+      "learning_rate": 6.291048235805234e-05,
+      "loss": 0.0343,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 195,
+      "tokens_per_second_per_gpu": 343.5
+    },
+    {
+      "epoch": 0.44954128440366975,
+      "grad_norm": 0.21585902571678162,
+      "learning_rate": 6.254444630896529e-05,
+      "loss": 0.0293,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 196,
+      "tokens_per_second_per_gpu": 294.22
+    },
+    {
+      "epoch": 0.4518348623853211,
+      "grad_norm": 0.18017566204071045,
+      "learning_rate": 6.21776913855496e-05,
+      "loss": 0.0244,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 197,
+      "tokens_per_second_per_gpu": 424.71
+    },
+    {
+      "epoch": 0.4541284403669725,
+      "grad_norm": 0.18264047801494598,
+      "learning_rate": 6.181023860512984e-05,
+      "loss": 0.0299,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 198,
+      "tokens_per_second_per_gpu": 380.31
+    },
+    {
+      "epoch": 0.45642201834862384,
+      "grad_norm": 0.19493581354618073,
+      "learning_rate": 6.144210902502207e-05,
+      "loss": 0.0278,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 199,
+      "tokens_per_second_per_gpu": 449.07
+    },
+    {
+      "epoch": 0.45871559633027525,
+      "grad_norm": 0.19989101588726044,
+      "learning_rate": 6.107332374132715e-05,
+      "loss": 0.0333,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 200,
+      "tokens_per_second_per_gpu": 390.66
+    },
+    {
+      "epoch": 0.4610091743119266,
+      "grad_norm": 0.17109017074108124,
+      "learning_rate": 6.0703903887721837e-05,
+      "loss": 0.0276,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 201,
+      "tokens_per_second_per_gpu": 396.01
+    },
+    {
+      "epoch": 0.463302752293578,
+      "grad_norm": 0.25124529004096985,
+      "learning_rate": 6.0333870634247645e-05,
+      "loss": 0.0333,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 202,
+      "tokens_per_second_per_gpu": 319.95
+    },
+    {
+      "epoch": 0.46559633027522934,
+      "grad_norm": 0.1988801658153534,
+      "learning_rate": 5.9963245186097725e-05,
+      "loss": 0.0341,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 203,
+      "tokens_per_second_per_gpu": 352.34
+    },
+    {
+      "epoch": 0.46788990825688076,
+      "grad_norm": 0.17253847420215607,
+      "learning_rate": 5.95920487824016e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 204,
+      "tokens_per_second_per_gpu": 354.16
+    },
+    {
+      "epoch": 0.4701834862385321,
+      "grad_norm": 0.17138919234275818,
+      "learning_rate": 5.922030269500809e-05,
+      "loss": 0.0237,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 205,
+      "tokens_per_second_per_gpu": 299.54
+    },
+    {
+      "epoch": 0.4724770642201835,
+      "grad_norm": 0.16596569120883942,
+      "learning_rate": 5.8848028227266325e-05,
+      "loss": 0.0254,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 206,
+      "tokens_per_second_per_gpu": 403.01
+    },
+    {
+      "epoch": 0.47477064220183485,
+      "grad_norm": 0.17839401960372925,
+      "learning_rate": 5.847524671280484e-05,
+      "loss": 0.0301,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 207,
+      "tokens_per_second_per_gpu": 345.59
+    },
+    {
+      "epoch": 0.47706422018348627,
+      "grad_norm": 0.13576573133468628,
+      "learning_rate": 5.810197951430911e-05,
+      "loss": 0.0202,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 208,
+      "tokens_per_second_per_gpu": 418.18
+    },
+    {
+      "epoch": 0.4793577981651376,
+      "grad_norm": 0.1716591864824295,
+      "learning_rate": 5.772824802229733e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 209,
+      "tokens_per_second_per_gpu": 305.6
+    },
+    {
+      "epoch": 0.481651376146789,
+      "grad_norm": 0.18199452757835388,
+      "learning_rate": 5.735407365389453e-05,
+      "loss": 0.0307,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 210,
+      "tokens_per_second_per_gpu": 364.43
+    },
+    {
+      "epoch": 0.48394495412844035,
+      "grad_norm": 0.15321891009807587,
+      "learning_rate": 5.697947785160532e-05,
+      "loss": 0.0206,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 211,
+      "tokens_per_second_per_gpu": 338.89
+    },
+    {
+      "epoch": 0.48623853211009177,
+      "grad_norm": 0.1617603749036789,
+      "learning_rate": 5.660448208208513e-05,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 212,
+      "tokens_per_second_per_gpu": 364.86
+    },
+    {
+      "epoch": 0.48853211009174313,
+      "grad_norm": 0.1874418407678604,
+      "learning_rate": 5.622910783490988e-05,
+      "loss": 0.0314,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 213,
+      "tokens_per_second_per_gpu": 328.29
+    },
+    {
+      "epoch": 0.4908256880733945,
+      "grad_norm": 0.19621190428733826,
+      "learning_rate": 5.585337662134471e-05,
+      "loss": 0.0339,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 214,
+      "tokens_per_second_per_gpu": 326.48
+    },
+    {
+      "epoch": 0.49311926605504586,
+      "grad_norm": 0.15886133909225464,
+      "learning_rate": 5.5477309973111046e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 215,
+      "tokens_per_second_per_gpu": 338.47
+    },
+    {
+      "epoch": 0.4954128440366973,
+      "grad_norm": 0.16169118881225586,
+      "learning_rate": 5.510092944115286e-05,
+      "loss": 0.0268,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 216,
+      "tokens_per_second_per_gpu": 409.66
+    },
+    {
+      "epoch": 0.49770642201834864,
+      "grad_norm": 0.17983202636241913,
+      "learning_rate": 5.472425659440157e-05,
+      "loss": 0.0276,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 217,
+      "tokens_per_second_per_gpu": 325.12
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 0.21549560129642487,
+      "learning_rate": 5.4347313018540056e-05,
+      "loss": 0.0288,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 218,
+      "tokens_per_second_per_gpu": 386.96
+    },
+    {
+      "epoch": 0.5022935779816514,
+      "grad_norm": 0.25012025237083435,
+      "learning_rate": 5.397012031476562e-05,
+      "loss": 0.0434,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 219,
+      "tokens_per_second_per_gpu": 333.39
+    },
+    {
+      "epoch": 0.5045871559633027,
+      "grad_norm": 0.14563509821891785,
+      "learning_rate": 5.359270009855216e-05,
+      "loss": 0.0217,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 220,
+      "tokens_per_second_per_gpu": 444.11
+    },
+    {
+      "epoch": 0.5068807339449541,
+      "grad_norm": 0.15768077969551086,
+      "learning_rate": 5.321507399841148e-05,
+      "loss": 0.0239,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 221,
+      "tokens_per_second_per_gpu": 381.48
+    },
+    {
+      "epoch": 0.5091743119266054,
+      "grad_norm": 0.18575942516326904,
+      "learning_rate": 5.2837263654653715e-05,
+      "loss": 0.0335,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 222,
+      "tokens_per_second_per_gpu": 293.14
+    },
+    {
+      "epoch": 0.5114678899082569,
+      "grad_norm": 0.21634536981582642,
+      "learning_rate": 5.2459290718147344e-05,
+      "loss": 0.0372,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 223,
+      "tokens_per_second_per_gpu": 426.22
+    },
+    {
+      "epoch": 0.5137614678899083,
+      "grad_norm": 0.16426831483840942,
+      "learning_rate": 5.2081176849078464e-05,
+      "loss": 0.0294,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 224,
+      "tokens_per_second_per_gpu": 381.32
+    },
+    {
+      "epoch": 0.5160550458715596,
+      "grad_norm": 0.17062143981456757,
+      "learning_rate": 5.170294371570939e-05,
+      "loss": 0.0276,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 225,
+      "tokens_per_second_per_gpu": 350.54
+    },
+    {
+      "epoch": 0.518348623853211,
+      "grad_norm": 0.19721156358718872,
+      "learning_rate": 5.132461299313709e-05,
+      "loss": 0.0332,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 226,
+      "tokens_per_second_per_gpu": 370.67
+    },
+    {
+      "epoch": 0.5206422018348624,
+      "grad_norm": 0.18919353187084198,
+      "learning_rate": 5.094620636205095e-05,
+      "loss": 0.031,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 227,
+      "tokens_per_second_per_gpu": 313.3
+    },
+    {
+      "epoch": 0.5229357798165137,
+      "grad_norm": 0.1820768415927887,
+      "learning_rate": 5.056774550749043e-05,
+      "loss": 0.023,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 228,
+      "tokens_per_second_per_gpu": 273.59
+    },
+    {
+      "epoch": 0.5252293577981652,
+      "grad_norm": 0.16542920470237732,
+      "learning_rate": 5.018925211760227e-05,
+      "loss": 0.0264,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 229,
+      "tokens_per_second_per_gpu": 366.63
+    },
+    {
+      "epoch": 0.5275229357798165,
+      "grad_norm": 0.16371333599090576,
+      "learning_rate": 4.981074788239773e-05,
+      "loss": 0.0228,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 230,
+      "tokens_per_second_per_gpu": 357.17
+    },
+    {
+      "epoch": 0.5298165137614679,
+      "grad_norm": 0.20354929566383362,
+      "learning_rate": 4.943225449250958e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 231,
+      "tokens_per_second_per_gpu": 387.38
+    },
+    {
+      "epoch": 0.5321100917431193,
+      "grad_norm": 0.2108154594898224,
+      "learning_rate": 4.9053793637949067e-05,
+      "loss": 0.0287,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 232,
+      "tokens_per_second_per_gpu": 333.1
+    },
+    {
+      "epoch": 0.5344036697247706,
+      "grad_norm": 0.17412519454956055,
+      "learning_rate": 4.8675387006862914e-05,
+      "loss": 0.033,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 233,
+      "tokens_per_second_per_gpu": 478.3
+    },
+    {
+      "epoch": 0.536697247706422,
+      "grad_norm": 0.1947973668575287,
+      "learning_rate": 4.829705628429061e-05,
+      "loss": 0.0339,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 234,
+      "tokens_per_second_per_gpu": 311.52
+    },
+    {
+      "epoch": 0.5389908256880734,
+      "grad_norm": 0.15672098100185394,
+      "learning_rate": 4.7918823150921555e-05,
+      "loss": 0.022,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 235,
+      "tokens_per_second_per_gpu": 319.97
+    },
+    {
+      "epoch": 0.5412844036697247,
+      "grad_norm": 0.1515381783246994,
+      "learning_rate": 4.754070928185266e-05,
+      "loss": 0.0241,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 236,
+      "tokens_per_second_per_gpu": 389.31
+    },
+    {
+      "epoch": 0.5435779816513762,
+      "grad_norm": 0.1594814956188202,
+      "learning_rate": 4.7162736345346303e-05,
+      "loss": 0.0232,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 237,
+      "tokens_per_second_per_gpu": 399.01
+    },
+    {
+      "epoch": 0.5458715596330275,
+      "grad_norm": 0.208835169672966,
+      "learning_rate": 4.6784926001588544e-05,
+      "loss": 0.042,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 238,
+      "tokens_per_second_per_gpu": 381.98
+    },
+    {
+      "epoch": 0.5481651376146789,
+      "grad_norm": 0.15975874662399292,
+      "learning_rate": 4.640729990144784e-05,
+      "loss": 0.0266,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 239,
+      "tokens_per_second_per_gpu": 420.2
+    },
+    {
+      "epoch": 0.5504587155963303,
+      "grad_norm": 0.1748957335948944,
+      "learning_rate": 4.6029879685234395e-05,
+      "loss": 0.0293,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 240,
+      "tokens_per_second_per_gpu": 393.42
+    },
+    {
+      "epoch": 0.5527522935779816,
+      "grad_norm": 0.15383028984069824,
+      "learning_rate": 4.565268698145997e-05,
+      "loss": 0.026,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 241,
+      "tokens_per_second_per_gpu": 409.27
+    },
+    {
+      "epoch": 0.555045871559633,
+      "grad_norm": 0.19024741649627686,
+      "learning_rate": 4.527574340559844e-05,
+      "loss": 0.0346,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 242,
+      "tokens_per_second_per_gpu": 340.52
+    },
+    {
+      "epoch": 0.5573394495412844,
+      "grad_norm": 0.15886802971363068,
+      "learning_rate": 4.4899070558847154e-05,
+      "loss": 0.029,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 243,
+      "tokens_per_second_per_gpu": 409.12
+    },
+    {
+      "epoch": 0.5596330275229358,
+      "grad_norm": 0.1559331864118576,
+      "learning_rate": 4.452269002688897e-05,
+      "loss": 0.0223,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 244,
+      "tokens_per_second_per_gpu": 271.44
+    },
+    {
+      "epoch": 0.5619266055045872,
+      "grad_norm": 0.1751302182674408,
+      "learning_rate": 4.4146623378655296e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.04,
+      "memory/max_allocated (GiB)": 54.04,
+      "step": 245,
+      "tokens_per_second_per_gpu": 280.17
+    },
+    {
+      "epoch": 0.5642201834862385,
+      "grad_norm": 0.1595376431941986,
+      "learning_rate": 4.3770892165090126e-05,
+      "loss": 0.0234,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 246,
+      "tokens_per_second_per_gpu": 344.19
+    },
+    {
+      "epoch": 0.5665137614678899,
+      "grad_norm": 0.16330555081367493,
+      "learning_rate": 4.3395517917914895e-05,
+      "loss": 0.0276,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 247,
+      "tokens_per_second_per_gpu": 464.2
+    },
+    {
+      "epoch": 0.5688073394495413,
+      "grad_norm": 0.16099631786346436,
+      "learning_rate": 4.3020522148394676e-05,
+      "loss": 0.0273,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 248,
+      "tokens_per_second_per_gpu": 354.57
+    },
+    {
+      "epoch": 0.5711009174311926,
+      "grad_norm": 0.18867036700248718,
+      "learning_rate": 4.2645926346105484e-05,
+      "loss": 0.0312,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 249,
+      "tokens_per_second_per_gpu": 297.48
+    },
+    {
+      "epoch": 0.573394495412844,
+      "grad_norm": 0.1458401381969452,
+      "learning_rate": 4.22717519777027e-05,
+      "loss": 0.0248,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 250,
+      "tokens_per_second_per_gpu": 359.95
+    },
+    {
+      "epoch": 0.5756880733944955,
+      "grad_norm": 0.14756208658218384,
+      "learning_rate": 4.189802048569089e-05,
+      "loss": 0.0242,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 251,
+      "tokens_per_second_per_gpu": 413.15
+    },
+    {
+      "epoch": 0.5779816513761468,
+      "grad_norm": 0.14128528535366058,
+      "learning_rate": 4.1524753287195165e-05,
+      "loss": 0.0186,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 252,
+      "tokens_per_second_per_gpu": 410.14
+    },
+    {
+      "epoch": 0.5802752293577982,
+      "grad_norm": 0.18493594229221344,
+      "learning_rate": 4.1151971772733686e-05,
+      "loss": 0.0322,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 253,
+      "tokens_per_second_per_gpu": 362.56
+    },
+    {
+      "epoch": 0.5825688073394495,
+      "grad_norm": 0.1811840981245041,
+      "learning_rate": 4.07796973049919e-05,
+      "loss": 0.0296,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 254,
+      "tokens_per_second_per_gpu": 374.13
+    },
+    {
+      "epoch": 0.5848623853211009,
+      "grad_norm": 0.18642045557498932,
+      "learning_rate": 4.04079512175984e-05,
+      "loss": 0.0319,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 255,
+      "tokens_per_second_per_gpu": 321.44
+    },
+    {
+      "epoch": 0.5871559633027523,
+      "grad_norm": 0.19729453325271606,
+      "learning_rate": 4.003675481390228e-05,
+      "loss": 0.0438,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 256,
+      "tokens_per_second_per_gpu": 347.19
+    },
+    {
+      "epoch": 0.5894495412844036,
+      "grad_norm": 0.150944784283638,
+      "learning_rate": 3.966612936575235e-05,
+      "loss": 0.0248,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 257,
+      "tokens_per_second_per_gpu": 331.37
+    },
+    {
+      "epoch": 0.591743119266055,
+      "grad_norm": 0.1676388829946518,
+      "learning_rate": 3.929609611227817e-05,
+      "loss": 0.0263,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 258,
+      "tokens_per_second_per_gpu": 308.12
+    },
+    {
+      "epoch": 0.5940366972477065,
+      "grad_norm": 0.15992635488510132,
+      "learning_rate": 3.8926676258672866e-05,
+      "loss": 0.0271,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 259,
+      "tokens_per_second_per_gpu": 328.86
+    },
+    {
+      "epoch": 0.5963302752293578,
+      "grad_norm": 0.169114351272583,
+      "learning_rate": 3.855789097497794e-05,
+      "loss": 0.0306,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 260,
+      "tokens_per_second_per_gpu": 325.67
+    },
+    {
+      "epoch": 0.5986238532110092,
+      "grad_norm": 0.17272968590259552,
+      "learning_rate": 3.818976139487017e-05,
+      "loss": 0.0269,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 261,
+      "tokens_per_second_per_gpu": 414.66
+    },
+    {
+      "epoch": 0.6009174311926605,
+      "grad_norm": 0.17536921799182892,
+      "learning_rate": 3.7822308614450406e-05,
+      "loss": 0.0314,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 262,
+      "tokens_per_second_per_gpu": 369.24
+    },
+    {
+      "epoch": 0.6032110091743119,
+      "grad_norm": 0.18068668246269226,
+      "learning_rate": 3.745555369103471e-05,
+      "loss": 0.0313,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 263,
+      "tokens_per_second_per_gpu": 315.54
+    },
+    {
+      "epoch": 0.6055045871559633,
+      "grad_norm": 0.21385689079761505,
+      "learning_rate": 3.708951764194767e-05,
+      "loss": 0.0363,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 264,
+      "tokens_per_second_per_gpu": 327.61
+    },
+    {
+      "epoch": 0.6077981651376146,
+      "grad_norm": 0.15608981251716614,
+      "learning_rate": 3.6724221443317855e-05,
+      "loss": 0.0234,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 265,
+      "tokens_per_second_per_gpu": 475.2
+    },
+    {
+      "epoch": 0.6100917431192661,
+      "grad_norm": 0.16386674344539642,
+      "learning_rate": 3.635968602887585e-05,
+      "loss": 0.0336,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 266,
+      "tokens_per_second_per_gpu": 365.72
+    },
+    {
+      "epoch": 0.6123853211009175,
+      "grad_norm": 0.17866984009742737,
+      "learning_rate": 3.599593228875465e-05,
+      "loss": 0.0359,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 267,
+      "tokens_per_second_per_gpu": 391.2
+    },
+    {
+      "epoch": 0.6146788990825688,
+      "grad_norm": 0.1519947350025177,
+      "learning_rate": 3.563298106829244e-05,
+      "loss": 0.0262,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 268,
+      "tokens_per_second_per_gpu": 357.03
+    },
+    {
+      "epoch": 0.6169724770642202,
+      "grad_norm": 0.161304771900177,
+      "learning_rate": 3.527085316683805e-05,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 269,
+      "tokens_per_second_per_gpu": 416.5
+    },
+    {
+      "epoch": 0.6192660550458715,
+      "grad_norm": 0.1459219753742218,
+      "learning_rate": 3.490956933655909e-05,
+      "loss": 0.0209,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 270,
+      "tokens_per_second_per_gpu": 299.97
+    },
+    {
+      "epoch": 0.6215596330275229,
+      "grad_norm": 0.16339126229286194,
+      "learning_rate": 3.4549150281252636e-05,
+      "loss": 0.0317,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 271,
+      "tokens_per_second_per_gpu": 340.13
+    },
+    {
+      "epoch": 0.6238532110091743,
+      "grad_norm": 0.15448080003261566,
+      "learning_rate": 3.41896166551588e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 272,
+      "tokens_per_second_per_gpu": 401.64
+    },
+    {
+      "epoch": 0.6261467889908257,
+      "grad_norm": 0.17994728684425354,
+      "learning_rate": 3.383098906177719e-05,
+      "loss": 0.0312,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 273,
+      "tokens_per_second_per_gpu": 430.81
+    },
+    {
+      "epoch": 0.6284403669724771,
+      "grad_norm": 0.19077076017856598,
+      "learning_rate": 3.347328805268605e-05,
+      "loss": 0.0288,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 274,
+      "tokens_per_second_per_gpu": 308.99
+    },
+    {
+      "epoch": 0.6307339449541285,
+      "grad_norm": 0.18485191464424133,
+      "learning_rate": 3.3116534126364685e-05,
+      "loss": 0.0343,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 275,
+      "tokens_per_second_per_gpu": 294.35
+    },
+    {
+      "epoch": 0.6330275229357798,
+      "grad_norm": 0.1638234704732895,
+      "learning_rate": 3.2760747727018694e-05,
+      "loss": 0.0241,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 276,
+      "tokens_per_second_per_gpu": 294.45
+    },
+    {
+      "epoch": 0.6353211009174312,
+      "grad_norm": 0.17989729344844818,
+      "learning_rate": 3.240594924340835e-05,
+      "loss": 0.0273,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 277,
+      "tokens_per_second_per_gpu": 330.33
+    },
+    {
+      "epoch": 0.6376146788990825,
+      "grad_norm": 0.1611703485250473,
+      "learning_rate": 3.205215900768029e-05,
+      "loss": 0.0271,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 278,
+      "tokens_per_second_per_gpu": 359.1
+    },
+    {
+      "epoch": 0.6399082568807339,
+      "grad_norm": 0.18084922432899475,
+      "learning_rate": 3.169939729420233e-05,
+      "loss": 0.0275,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 279,
+      "tokens_per_second_per_gpu": 355.37
+    },
+    {
+      "epoch": 0.6422018348623854,
+      "grad_norm": 0.16204647719860077,
+      "learning_rate": 3.1347684318401536e-05,
+      "loss": 0.0238,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 280,
+      "tokens_per_second_per_gpu": 375.27
+    },
+    {
+      "epoch": 0.6444954128440367,
+      "grad_norm": 0.17873668670654297,
+      "learning_rate": 3.099704023560587e-05,
+      "loss": 0.0331,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 281,
+      "tokens_per_second_per_gpu": 344.67
+    },
+    {
+      "epoch": 0.6467889908256881,
+      "grad_norm": 0.1369648575782776,
+      "learning_rate": 3.0647485139889145e-05,
+      "loss": 0.0214,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 282,
+      "tokens_per_second_per_gpu": 289.83
+    },
+    {
+      "epoch": 0.6490825688073395,
+      "grad_norm": 0.16736926138401031,
+      "learning_rate": 3.0299039062919416e-05,
+      "loss": 0.0233,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 283,
+      "tokens_per_second_per_gpu": 391.42
+    },
+    {
+      "epoch": 0.6513761467889908,
+      "grad_norm": 0.1656859666109085,
+      "learning_rate": 2.995172197281113e-05,
+      "loss": 0.0279,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 284,
+      "tokens_per_second_per_gpu": 362.94
+    },
+    {
+      "epoch": 0.6536697247706422,
+      "grad_norm": 0.16381965577602386,
+      "learning_rate": 2.96055537729808e-05,
+      "loss": 0.0253,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 285,
+      "tokens_per_second_per_gpu": 396.6
+    },
+    {
+      "epoch": 0.6559633027522935,
+      "grad_norm": 0.13780774176120758,
+      "learning_rate": 2.926055430100647e-05,
+      "loss": 0.0224,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 286,
+      "tokens_per_second_per_gpu": 317.31
+    },
+    {
+      "epoch": 0.658256880733945,
+      "grad_norm": 0.14635130763053894,
+      "learning_rate": 2.8916743327490803e-05,
+      "loss": 0.0248,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 287,
+      "tokens_per_second_per_gpu": 345.66
+    },
+    {
+      "epoch": 0.6605504587155964,
+      "grad_norm": 0.1762857586145401,
+      "learning_rate": 2.8574140554928175e-05,
+      "loss": 0.032,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 288,
+      "tokens_per_second_per_gpu": 347.22
+    },
+    {
+      "epoch": 0.6628440366972477,
+      "grad_norm": 0.16524505615234375,
+      "learning_rate": 2.8232765616575563e-05,
+      "loss": 0.0295,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 289,
+      "tokens_per_second_per_gpu": 377.44
+    },
+    {
+      "epoch": 0.6651376146788991,
+      "grad_norm": 0.1379774957895279,
+      "learning_rate": 2.789263807532746e-05,
+      "loss": 0.0223,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 290,
+      "tokens_per_second_per_gpu": 387.67
+    },
+    {
+      "epoch": 0.6674311926605505,
+      "grad_norm": 0.1687348484992981,
+      "learning_rate": 2.7553777422594774e-05,
+      "loss": 0.0307,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 291,
+      "tokens_per_second_per_gpu": 349.05
+    },
+    {
+      "epoch": 0.6697247706422018,
+      "grad_norm": 0.1557503342628479,
+      "learning_rate": 2.721620307718793e-05,
+      "loss": 0.0229,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 292,
+      "tokens_per_second_per_gpu": 240.3
+    },
+    {
+      "epoch": 0.6720183486238532,
+      "grad_norm": 0.16598054766654968,
+      "learning_rate": 2.687993438420392e-05,
+      "loss": 0.0256,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 293,
+      "tokens_per_second_per_gpu": 340.81
+    },
+    {
+      "epoch": 0.6743119266055045,
+      "grad_norm": 0.17300239205360413,
+      "learning_rate": 2.65449906139178e-05,
+      "loss": 0.0315,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 294,
+      "tokens_per_second_per_gpu": 330.7
+    },
+    {
+      "epoch": 0.676605504587156,
+      "grad_norm": 0.18177662789821625,
+      "learning_rate": 2.6211390960678413e-05,
+      "loss": 0.0386,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 295,
+      "tokens_per_second_per_gpu": 303.69
+    },
+    {
+      "epoch": 0.6788990825688074,
+      "grad_norm": 0.1713094413280487,
+      "learning_rate": 2.5879154541808337e-05,
+      "loss": 0.0266,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 296,
+      "tokens_per_second_per_gpu": 421.27
+    },
+    {
+      "epoch": 0.6811926605504587,
+      "grad_norm": 0.15433934330940247,
+      "learning_rate": 2.554830039650834e-05,
+      "loss": 0.0237,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 297,
+      "tokens_per_second_per_gpu": 382.55
+    },
+    {
+      "epoch": 0.6834862385321101,
+      "grad_norm": 0.1905246078968048,
+      "learning_rate": 2.5218847484766495e-05,
+      "loss": 0.0296,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 298,
+      "tokens_per_second_per_gpu": 250.53
+    },
+    {
+      "epoch": 0.6857798165137615,
+      "grad_norm": 0.1752621829509735,
+      "learning_rate": 2.4890814686271448e-05,
+      "loss": 0.0254,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 299,
+      "tokens_per_second_per_gpu": 392.54
+    },
+    {
+      "epoch": 0.6880733944954128,
+      "grad_norm": 0.18215875327587128,
+      "learning_rate": 2.456422079933056e-05,
+      "loss": 0.0277,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 300,
+      "tokens_per_second_per_gpu": 400.99
+    },
+    {
+      "epoch": 0.6903669724770642,
+      "grad_norm": 0.15181544423103333,
+      "learning_rate": 2.4239084539792745e-05,
+      "loss": 0.0253,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 301,
+      "tokens_per_second_per_gpu": 364.7
+    },
+    {
+      "epoch": 0.6926605504587156,
+      "grad_norm": 0.1604120135307312,
+      "learning_rate": 2.391542453997578e-05,
+      "loss": 0.0234,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 302,
+      "tokens_per_second_per_gpu": 318.73
+    },
+    {
+      "epoch": 0.694954128440367,
+      "grad_norm": 0.14801405370235443,
+      "learning_rate": 2.3593259347598657e-05,
+      "loss": 0.021,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 303,
+      "tokens_per_second_per_gpu": 403.51
+    },
+    {
+      "epoch": 0.6972477064220184,
+      "grad_norm": 0.1397543102502823,
+      "learning_rate": 2.3272607424718675e-05,
+      "loss": 0.0247,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 304,
+      "tokens_per_second_per_gpu": 342.14
+    },
+    {
+      "epoch": 0.6995412844036697,
+      "grad_norm": 0.14816102385520935,
+      "learning_rate": 2.29534871466734e-05,
+      "loss": 0.0298,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 305,
+      "tokens_per_second_per_gpu": 394.51
+    },
+    {
+      "epoch": 0.7018348623853211,
+      "grad_norm": 0.18773198127746582,
+      "learning_rate": 2.2635916801027706e-05,
+      "loss": 0.0382,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 306,
+      "tokens_per_second_per_gpu": 373.43
+    },
+    {
+      "epoch": 0.7041284403669725,
+      "grad_norm": 0.1440516710281372,
+      "learning_rate": 2.2319914586525777e-05,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 307,
+      "tokens_per_second_per_gpu": 391.41
+    },
+    {
+      "epoch": 0.7064220183486238,
+      "grad_norm": 0.1587175577878952,
+      "learning_rate": 2.2005498612048155e-05,
+      "loss": 0.0237,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 308,
+      "tokens_per_second_per_gpu": 316.38
+    },
+    {
+      "epoch": 0.7087155963302753,
+      "grad_norm": 0.16214604675769806,
+      "learning_rate": 2.1692686895574005e-05,
+      "loss": 0.0318,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 309,
+      "tokens_per_second_per_gpu": 412.91
+    },
+    {
+      "epoch": 0.7110091743119266,
+      "grad_norm": 0.15564551949501038,
+      "learning_rate": 2.1381497363148673e-05,
+      "loss": 0.0243,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 310,
+      "tokens_per_second_per_gpu": 351.83
+    },
+    {
+      "epoch": 0.713302752293578,
+      "grad_norm": 0.173199862241745,
+      "learning_rate": 2.1071947847856222e-05,
+      "loss": 0.0271,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 311,
+      "tokens_per_second_per_gpu": 354.94
+    },
+    {
+      "epoch": 0.7155963302752294,
+      "grad_norm": 0.1524014174938202,
+      "learning_rate": 2.0764056088797645e-05,
+      "loss": 0.0209,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 312,
+      "tokens_per_second_per_gpu": 334.1
+    },
+    {
+      "epoch": 0.7178899082568807,
+      "grad_norm": 0.16921748220920563,
+      "learning_rate": 2.045783973007429e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 313,
+      "tokens_per_second_per_gpu": 346.98
+    },
+    {
+      "epoch": 0.7201834862385321,
+      "grad_norm": 0.16978447139263153,
+      "learning_rate": 2.0153316319776662e-05,
+      "loss": 0.0264,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 314,
+      "tokens_per_second_per_gpu": 311.13
+    },
+    {
+      "epoch": 0.7224770642201835,
+      "grad_norm": 0.1572890430688858,
+      "learning_rate": 1.985050330897883e-05,
+      "loss": 0.0238,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 315,
+      "tokens_per_second_per_gpu": 343.83
+    },
+    {
+      "epoch": 0.7247706422018348,
+      "grad_norm": 0.20190350711345673,
+      "learning_rate": 1.954941805073848e-05,
+      "loss": 0.0361,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 316,
+      "tokens_per_second_per_gpu": 323.12
+    },
+    {
+      "epoch": 0.7270642201834863,
+      "grad_norm": 0.16654695570468903,
+      "learning_rate": 1.9250077799102322e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 317,
+      "tokens_per_second_per_gpu": 381.42
+    },
+    {
+      "epoch": 0.7293577981651376,
+      "grad_norm": 0.17696668207645416,
+      "learning_rate": 1.8952499708117432e-05,
+      "loss": 0.0243,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 318,
+      "tokens_per_second_per_gpu": 402.83
+    },
+    {
+      "epoch": 0.731651376146789,
+      "grad_norm": 0.15845169126987457,
+      "learning_rate": 1.8656700830848174e-05,
+      "loss": 0.0289,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 319,
+      "tokens_per_second_per_gpu": 390.76
+    },
+    {
+      "epoch": 0.7339449541284404,
+      "grad_norm": 0.15305587649345398,
+      "learning_rate": 1.8362698118398967e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 320,
+      "tokens_per_second_per_gpu": 334.05
+    },
+    {
+      "epoch": 0.7362385321100917,
+      "grad_norm": 0.17878518998622894,
+      "learning_rate": 1.8070508418942876e-05,
+      "loss": 0.0365,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 321,
+      "tokens_per_second_per_gpu": 332.38
+    },
+    {
+      "epoch": 0.7385321100917431,
+      "grad_norm": 0.16544903814792633,
+      "learning_rate": 1.7780148476756147e-05,
+      "loss": 0.0259,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 322,
+      "tokens_per_second_per_gpu": 435.22
+    },
+    {
+      "epoch": 0.7408256880733946,
+      "grad_norm": 0.1871042549610138,
+      "learning_rate": 1.7491634931258587e-05,
+      "loss": 0.0306,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 323,
+      "tokens_per_second_per_gpu": 340.78
+    },
+    {
+      "epoch": 0.7431192660550459,
+      "grad_norm": 0.13492454588413239,
+      "learning_rate": 1.7204984316060063e-05,
+      "loss": 0.0189,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 324,
+      "tokens_per_second_per_gpu": 353.84
+    },
+    {
+      "epoch": 0.7454128440366973,
+      "grad_norm": 0.16810919344425201,
+      "learning_rate": 1.6920213058013022e-05,
+      "loss": 0.0275,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 325,
+      "tokens_per_second_per_gpu": 404.89
+    },
+    {
+      "epoch": 0.7477064220183486,
+      "grad_norm": 0.16831310093402863,
+      "learning_rate": 1.6637337476271124e-05,
+      "loss": 0.0282,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 326,
+      "tokens_per_second_per_gpu": 337.09
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 0.1542624533176422,
+      "learning_rate": 1.6356373781354058e-05,
+      "loss": 0.0268,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 327,
+      "tokens_per_second_per_gpu": 319.71
+    },
+    {
+      "epoch": 0.7522935779816514,
+      "grad_norm": 0.16454322636127472,
+      "learning_rate": 1.6077338074218596e-05,
+      "loss": 0.0307,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 328,
+      "tokens_per_second_per_gpu": 363.92
+    },
+    {
+      "epoch": 0.7545871559633027,
+      "grad_norm": 0.15118440985679626,
+      "learning_rate": 1.580024634533587e-05,
+      "loss": 0.0285,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 329,
+      "tokens_per_second_per_gpu": 381.21
+    },
+    {
+      "epoch": 0.7568807339449541,
+      "grad_norm": 0.19343721866607666,
+      "learning_rate": 1.5525114473775014e-05,
+      "loss": 0.0467,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 330,
+      "tokens_per_second_per_gpu": 381.71
+    },
+    {
+      "epoch": 0.7591743119266054,
+      "grad_norm": 0.2053116112947464,
+      "learning_rate": 1.5251958226293306e-05,
+      "loss": 0.0364,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 331,
+      "tokens_per_second_per_gpu": 314.33
+    },
+    {
+      "epoch": 0.7614678899082569,
+      "grad_norm": 0.1531161516904831,
+      "learning_rate": 1.4980793256432474e-05,
+      "loss": 0.0244,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 332,
+      "tokens_per_second_per_gpu": 305.61
+    },
+    {
+      "epoch": 0.7637614678899083,
+      "grad_norm": 0.16694028675556183,
+      "learning_rate": 1.4711635103621719e-05,
+      "loss": 0.0246,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 333,
+      "tokens_per_second_per_gpu": 349.73
+    },
+    {
+      "epoch": 0.7660550458715596,
+      "grad_norm": 0.1601162850856781,
+      "learning_rate": 1.4444499192287275e-05,
+      "loss": 0.0248,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 334,
+      "tokens_per_second_per_gpu": 323.25
+    },
+    {
+      "epoch": 0.768348623853211,
+      "grad_norm": 0.18338176608085632,
+      "learning_rate": 1.4179400830968415e-05,
+      "loss": 0.0357,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 335,
+      "tokens_per_second_per_gpu": 273.45
+    },
+    {
+      "epoch": 0.7706422018348624,
+      "grad_norm": 0.15707750618457794,
+      "learning_rate": 1.3916355211440164e-05,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 336,
+      "tokens_per_second_per_gpu": 314.74
+    },
+    {
+      "epoch": 0.7729357798165137,
+      "grad_norm": 0.1581530123949051,
+      "learning_rate": 1.3655377407842812e-05,
+      "loss": 0.024,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 337,
+      "tokens_per_second_per_gpu": 396.34
+    },
+    {
+      "epoch": 0.7752293577981652,
+      "grad_norm": 0.17221379280090332,
+      "learning_rate": 1.3396482375817975e-05,
+      "loss": 0.0312,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 338,
+      "tokens_per_second_per_gpu": 397.25
+    },
+    {
+      "epoch": 0.7775229357798165,
+      "grad_norm": 0.14726734161376953,
+      "learning_rate": 1.3139684951651588e-05,
+      "loss": 0.0252,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 339,
+      "tokens_per_second_per_gpu": 346.27
+    },
+    {
+      "epoch": 0.7798165137614679,
+      "grad_norm": 0.1696820855140686,
+      "learning_rate": 1.2884999851423673e-05,
+      "loss": 0.0263,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 340,
+      "tokens_per_second_per_gpu": 366.61
+    },
+    {
+      "epoch": 0.7821100917431193,
+      "grad_norm": 0.15241795778274536,
+      "learning_rate": 1.2632441670165056e-05,
+      "loss": 0.0258,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 341,
+      "tokens_per_second_per_gpu": 384.41
+    },
+    {
+      "epoch": 0.7844036697247706,
+      "grad_norm": 0.14132796227931976,
+      "learning_rate": 1.2382024881020937e-05,
+      "loss": 0.0245,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 342,
+      "tokens_per_second_per_gpu": 425.45
+    },
+    {
+      "epoch": 0.786697247706422,
+      "grad_norm": 0.17556384205818176,
+      "learning_rate": 1.213376383442153e-05,
+      "loss": 0.0296,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 343,
+      "tokens_per_second_per_gpu": 341.95
+    },
+    {
+      "epoch": 0.7889908256880734,
+      "grad_norm": 0.19094258546829224,
+      "learning_rate": 1.188767275725966e-05,
+      "loss": 0.0342,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 344,
+      "tokens_per_second_per_gpu": 370.51
+    },
+    {
+      "epoch": 0.7912844036697247,
+      "grad_norm": 0.18251046538352966,
+      "learning_rate": 1.164376575207547e-05,
+      "loss": 0.0298,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 345,
+      "tokens_per_second_per_gpu": 247.25
+    },
+    {
+      "epoch": 0.7935779816513762,
+      "grad_norm": 0.14971286058425903,
+      "learning_rate": 1.140205679624834e-05,
+      "loss": 0.0225,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 346,
+      "tokens_per_second_per_gpu": 306.05
+    },
+    {
+      "epoch": 0.7958715596330275,
+      "grad_norm": 0.1707334965467453,
+      "learning_rate": 1.1162559741195733e-05,
+      "loss": 0.0269,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 347,
+      "tokens_per_second_per_gpu": 330.86
+    },
+    {
+      "epoch": 0.7981651376146789,
+      "grad_norm": 0.15573203563690186,
+      "learning_rate": 1.092528831157959e-05,
+      "loss": 0.0222,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 348,
+      "tokens_per_second_per_gpu": 410.05
+    },
+    {
+      "epoch": 0.8004587155963303,
+      "grad_norm": 0.155721515417099,
+      "learning_rate": 1.0690256104519764e-05,
+      "loss": 0.0217,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 349,
+      "tokens_per_second_per_gpu": 369.28
+    },
+    {
+      "epoch": 0.8027522935779816,
+      "grad_norm": 0.15273097157478333,
+      "learning_rate": 1.0457476588814774e-05,
+      "loss": 0.0259,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 350,
+      "tokens_per_second_per_gpu": 336.84
+    },
+    {
+      "epoch": 0.805045871559633,
+      "grad_norm": 0.16290222108364105,
+      "learning_rate": 1.0226963104170002e-05,
+      "loss": 0.031,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 351,
+      "tokens_per_second_per_gpu": 376.86
+    },
+    {
+      "epoch": 0.8073394495412844,
+      "grad_norm": 0.15131717920303345,
+      "learning_rate": 9.998728860433276e-06,
+      "loss": 0.027,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 352,
+      "tokens_per_second_per_gpu": 341.03
+    },
+    {
+      "epoch": 0.8096330275229358,
+      "grad_norm": 0.18805041909217834,
+      "learning_rate": 9.772786936837785e-06,
+      "loss": 0.03,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 353,
+      "tokens_per_second_per_gpu": 345.52
+    },
+    {
+      "epoch": 0.8119266055045872,
+      "grad_norm": 0.1632698029279709,
+      "learning_rate": 9.549150281252633e-06,
+      "loss": 0.0261,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 354,
+      "tokens_per_second_per_gpu": 389.37
+    },
+    {
+      "epoch": 0.8142201834862385,
+      "grad_norm": 0.15074226260185242,
+      "learning_rate": 9.327831709440792e-06,
+      "loss": 0.0255,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 355,
+      "tokens_per_second_per_gpu": 331.29
+    },
+    {
+      "epoch": 0.8165137614678899,
+      "grad_norm": 0.14309869706630707,
+      "learning_rate": 9.108843904324715e-06,
+      "loss": 0.0253,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 356,
+      "tokens_per_second_per_gpu": 362.07
+    },
+    {
+      "epoch": 0.8188073394495413,
+      "grad_norm": 0.17357462644577026,
+      "learning_rate": 8.8921994152595e-06,
+      "loss": 0.0246,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 357,
+      "tokens_per_second_per_gpu": 250.82
+    },
+    {
+      "epoch": 0.8211009174311926,
+      "grad_norm": 0.16879694163799286,
+      "learning_rate": 8.677910657313782e-06,
+      "loss": 0.0288,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 358,
+      "tokens_per_second_per_gpu": 398.32
+    },
+    {
+      "epoch": 0.823394495412844,
+      "grad_norm": 0.17209598422050476,
+      "learning_rate": 8.465989910558209e-06,
+      "loss": 0.0311,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 359,
+      "tokens_per_second_per_gpu": 320.4
+    },
+    {
+      "epoch": 0.8256880733944955,
+      "grad_norm": 0.15807099640369415,
+      "learning_rate": 8.256449319361748e-06,
+      "loss": 0.0261,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 360,
+      "tokens_per_second_per_gpu": 333.25
+    },
+    {
+      "epoch": 0.8279816513761468,
+      "grad_norm": 0.20308910310268402,
+      "learning_rate": 8.049300891695744e-06,
+      "loss": 0.0349,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 361,
+      "tokens_per_second_per_gpu": 277.97
+    },
+    {
+      "epoch": 0.8302752293577982,
+      "grad_norm": 0.1718447357416153,
+      "learning_rate": 7.844556498445788e-06,
+      "loss": 0.0268,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 362,
+      "tokens_per_second_per_gpu": 382.17
+    },
+    {
+      "epoch": 0.8325688073394495,
+      "grad_norm": 0.17684946954250336,
+      "learning_rate": 7.642227872731417e-06,
+      "loss": 0.025,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 363,
+      "tokens_per_second_per_gpu": 287.92
+    },
+    {
+      "epoch": 0.8348623853211009,
+      "grad_norm": 0.17828437685966492,
+      "learning_rate": 7.4423266092337855e-06,
+      "loss": 0.0313,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 364,
+      "tokens_per_second_per_gpu": 316.84
+    },
+    {
+      "epoch": 0.8371559633027523,
+      "grad_norm": 0.18794777989387512,
+      "learning_rate": 7.244864163531162e-06,
+      "loss": 0.0315,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 365,
+      "tokens_per_second_per_gpu": 319.23
+    },
+    {
+      "epoch": 0.8394495412844036,
+      "grad_norm": 0.1593928337097168,
+      "learning_rate": 7.049851851442468e-06,
+      "loss": 0.0292,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 366,
+      "tokens_per_second_per_gpu": 412.94
+    },
+    {
+      "epoch": 0.841743119266055,
+      "grad_norm": 0.18850290775299072,
+      "learning_rate": 6.857300848378856e-06,
+      "loss": 0.0357,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 367,
+      "tokens_per_second_per_gpu": 357.3
+    },
+    {
+      "epoch": 0.8440366972477065,
+      "grad_norm": 0.13592322170734406,
+      "learning_rate": 6.667222188703226e-06,
+      "loss": 0.0241,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 368,
+      "tokens_per_second_per_gpu": 382.98
+    },
+    {
+      "epoch": 0.8463302752293578,
+      "grad_norm": 0.16348065435886383,
+      "learning_rate": 6.479626765097918e-06,
+      "loss": 0.0299,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 369,
+      "tokens_per_second_per_gpu": 400.18
+    },
+    {
+      "epoch": 0.8486238532110092,
+      "grad_norm": 0.17598161101341248,
+      "learning_rate": 6.294525327940515e-06,
+      "loss": 0.0304,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 370,
+      "tokens_per_second_per_gpu": 342.87
+    },
+    {
+      "epoch": 0.8509174311926605,
+      "grad_norm": 0.14915841817855835,
+      "learning_rate": 6.111928484687723e-06,
+      "loss": 0.0253,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 371,
+      "tokens_per_second_per_gpu": 361.06
+    },
+    {
+      "epoch": 0.8532110091743119,
+      "grad_norm": 0.1544603556394577,
+      "learning_rate": 5.931846699267557e-06,
+      "loss": 0.0287,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 372,
+      "tokens_per_second_per_gpu": 441.28
+    },
+    {
+      "epoch": 0.8555045871559633,
+      "grad_norm": 0.14758428931236267,
+      "learning_rate": 5.7542902914796745e-06,
+      "loss": 0.0217,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 373,
+      "tokens_per_second_per_gpu": 484.61
+    },
+    {
+      "epoch": 0.8577981651376146,
+      "grad_norm": 0.15728335082530975,
+      "learning_rate": 5.579269436403967e-06,
+      "loss": 0.0239,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 374,
+      "tokens_per_second_per_gpu": 335.35
+    },
+    {
+      "epoch": 0.8600917431192661,
+      "grad_norm": 0.18564024567604065,
+      "learning_rate": 5.4067941638174806e-06,
+      "loss": 0.0302,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 375,
+      "tokens_per_second_per_gpu": 305.66
+    },
+    {
+      "epoch": 0.8623853211009175,
+      "grad_norm": 0.15752846002578735,
+      "learning_rate": 5.2368743576196536e-06,
+      "loss": 0.0232,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 376,
+      "tokens_per_second_per_gpu": 286.15
+    },
+    {
+      "epoch": 0.8646788990825688,
+      "grad_norm": 0.14457005262374878,
+      "learning_rate": 5.0695197552659e-06,
+      "loss": 0.0262,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 377,
+      "tokens_per_second_per_gpu": 379.69
+    },
+    {
+      "epoch": 0.8669724770642202,
+      "grad_norm": 0.15968698263168335,
+      "learning_rate": 4.9047399472095746e-06,
+      "loss": 0.0279,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 378,
+      "tokens_per_second_per_gpu": 300.1
+    },
+    {
+      "epoch": 0.8692660550458715,
+      "grad_norm": 0.16619980335235596,
+      "learning_rate": 4.742544376352443e-06,
+      "loss": 0.0261,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 379,
+      "tokens_per_second_per_gpu": 379.48
+    },
+    {
+      "epoch": 0.8715596330275229,
+      "grad_norm": 0.15459509193897247,
+      "learning_rate": 4.582942337503465e-06,
+      "loss": 0.0248,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 380,
+      "tokens_per_second_per_gpu": 391.55
+    },
+    {
+      "epoch": 0.8738532110091743,
+      "grad_norm": 0.17279919981956482,
+      "learning_rate": 4.425942976846187e-06,
+      "loss": 0.0334,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 381,
+      "tokens_per_second_per_gpu": 288.48
+    },
+    {
+      "epoch": 0.8761467889908257,
+      "grad_norm": 0.18266154825687408,
+      "learning_rate": 4.271555291414636e-06,
+      "loss": 0.032,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 382,
+      "tokens_per_second_per_gpu": 287.87
+    },
+    {
+      "epoch": 0.8784403669724771,
+      "grad_norm": 0.18472951650619507,
+      "learning_rate": 4.119788128577667e-06,
+      "loss": 0.0321,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 383,
+      "tokens_per_second_per_gpu": 387.69
+    },
+    {
+      "epoch": 0.8807339449541285,
+      "grad_norm": 0.1689075380563736,
+      "learning_rate": 3.9706501855319765e-06,
+      "loss": 0.0284,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 384,
+      "tokens_per_second_per_gpu": 385.7
+    },
+    {
+      "epoch": 0.8830275229357798,
+      "grad_norm": 0.1800360083580017,
+      "learning_rate": 3.824150008803767e-06,
+      "loss": 0.0353,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 385,
+      "tokens_per_second_per_gpu": 298.16
+    },
+    {
+      "epoch": 0.8853211009174312,
+      "grad_norm": 0.1640298217535019,
+      "learning_rate": 3.680295993758881e-06,
+      "loss": 0.0295,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 386,
+      "tokens_per_second_per_gpu": 342.93
+    },
+    {
+      "epoch": 0.8876146788990825,
+      "grad_norm": 0.1996641308069229,
+      "learning_rate": 3.539096384121743e-06,
+      "loss": 0.0421,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 387,
+      "tokens_per_second_per_gpu": 335.85
+    },
+    {
+      "epoch": 0.8899082568807339,
+      "grad_norm": 0.16044816374778748,
+      "learning_rate": 3.40055927150294e-06,
+      "loss": 0.0296,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 388,
+      "tokens_per_second_per_gpu": 356.79
+    },
+    {
+      "epoch": 0.8922018348623854,
+      "grad_norm": 0.14716744422912598,
+      "learning_rate": 3.2646925949355312e-06,
+      "loss": 0.0249,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 389,
+      "tokens_per_second_per_gpu": 390.05
+    },
+    {
+      "epoch": 0.8944954128440367,
+      "grad_norm": 0.16763879358768463,
+      "learning_rate": 3.1315041404200663e-06,
+      "loss": 0.0263,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 390,
+      "tokens_per_second_per_gpu": 364.38
+    },
+    {
+      "epoch": 0.8967889908256881,
+      "grad_norm": 0.16172398626804352,
+      "learning_rate": 3.00100154047841e-06,
+      "loss": 0.0283,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 391,
+      "tokens_per_second_per_gpu": 404.12
+    },
+    {
+      "epoch": 0.8990825688073395,
+      "grad_norm": 0.1692630648612976,
+      "learning_rate": 2.8731922737163685e-06,
+      "loss": 0.0307,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 392,
+      "tokens_per_second_per_gpu": 411.48
+    },
+    {
+      "epoch": 0.9013761467889908,
+      "grad_norm": 0.15489362180233002,
+      "learning_rate": 2.7480836643950956e-06,
+      "loss": 0.0215,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 393,
+      "tokens_per_second_per_gpu": 390.38
+    },
+    {
+      "epoch": 0.9036697247706422,
+      "grad_norm": 0.16410787403583527,
+      "learning_rate": 2.6256828820113766e-06,
+      "loss": 0.0272,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 394,
+      "tokens_per_second_per_gpu": 284.58
+    },
+    {
+      "epoch": 0.9059633027522935,
+      "grad_norm": 0.14239473640918732,
+      "learning_rate": 2.5059969408867843e-06,
+      "loss": 0.0232,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 395,
+      "tokens_per_second_per_gpu": 356.47
+    },
+    {
+      "epoch": 0.908256880733945,
+      "grad_norm": 0.15464159846305847,
+      "learning_rate": 2.3890326997656975e-06,
+      "loss": 0.0263,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 396,
+      "tokens_per_second_per_gpu": 332.14
+    },
+    {
+      "epoch": 0.9105504587155964,
+      "grad_norm": 0.19348366558551788,
+      "learning_rate": 2.274796861422246e-06,
+      "loss": 0.0286,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 397,
+      "tokens_per_second_per_gpu": 391.97
+    },
+    {
+      "epoch": 0.9128440366972477,
+      "grad_norm": 0.14645299315452576,
+      "learning_rate": 2.163295972276219e-06,
+      "loss": 0.0194,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 398,
+      "tokens_per_second_per_gpu": 333.01
+    },
+    {
+      "epoch": 0.9151376146788991,
+      "grad_norm": 0.190754696726799,
+      "learning_rate": 2.054536422017922e-06,
+      "loss": 0.0351,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 399,
+      "tokens_per_second_per_gpu": 295.61
+    },
+    {
+      "epoch": 0.9174311926605505,
+      "grad_norm": 0.165267676115036,
+      "learning_rate": 1.9485244432419667e-06,
+      "loss": 0.0267,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 400,
+      "tokens_per_second_per_gpu": 323.64
+    },
+    {
+      "epoch": 0.9197247706422018,
+      "grad_norm": 0.13596034049987793,
+      "learning_rate": 1.8452661110901715e-06,
+      "loss": 0.0215,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 401,
+      "tokens_per_second_per_gpu": 413.82
+    },
+    {
+      "epoch": 0.9220183486238532,
+      "grad_norm": 0.39827218651771545,
+      "learning_rate": 1.7447673429033362e-06,
+      "loss": 0.0297,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 402,
+      "tokens_per_second_per_gpu": 281.37
+    },
+    {
+      "epoch": 0.9243119266055045,
+      "grad_norm": 0.1586989313364029,
+      "learning_rate": 1.6470338978822108e-06,
+      "loss": 0.029,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 403,
+      "tokens_per_second_per_gpu": 353.92
+    },
+    {
+      "epoch": 0.926605504587156,
+      "grad_norm": 0.18811003863811493,
+      "learning_rate": 1.5520713767574246e-06,
+      "loss": 0.0355,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 404,
+      "tokens_per_second_per_gpu": 326.13
+    },
+    {
+      "epoch": 0.9288990825688074,
+      "grad_norm": 0.18732599914073944,
+      "learning_rate": 1.4598852214685488e-06,
+      "loss": 0.024,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 405,
+      "tokens_per_second_per_gpu": 364.56
+    },
+    {
+      "epoch": 0.9311926605504587,
+      "grad_norm": 0.17649473249912262,
+      "learning_rate": 1.3704807148521903e-06,
+      "loss": 0.0272,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 406,
+      "tokens_per_second_per_gpu": 340.97
+    },
+    {
+      "epoch": 0.9334862385321101,
+      "grad_norm": 0.15469086170196533,
+      "learning_rate": 1.2838629803393342e-06,
+      "loss": 0.0251,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 407,
+      "tokens_per_second_per_gpu": 319.7
+    },
+    {
+      "epoch": 0.9357798165137615,
+      "grad_norm": 0.17929796874523163,
+      "learning_rate": 1.2000369816616674e-06,
+      "loss": 0.028,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 408,
+      "tokens_per_second_per_gpu": 426.66
+    },
+    {
+      "epoch": 0.9380733944954128,
+      "grad_norm": 0.1820838302373886,
+      "learning_rate": 1.119007522567167e-06,
+      "loss": 0.0372,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 409,
+      "tokens_per_second_per_gpu": 385.56
+    },
+    {
+      "epoch": 0.9403669724770642,
+      "grad_norm": 0.1381448209285736,
+      "learning_rate": 1.0407792465447986e-06,
+      "loss": 0.0227,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 410,
+      "tokens_per_second_per_gpu": 443.85
+    },
+    {
+      "epoch": 0.9426605504587156,
+      "grad_norm": 0.16802094876766205,
+      "learning_rate": 9.653566365584176e-07,
+      "loss": 0.0287,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 411,
+      "tokens_per_second_per_gpu": 354.89
+    },
+    {
+      "epoch": 0.944954128440367,
+      "grad_norm": 0.19988204538822174,
+      "learning_rate": 8.927440147898702e-07,
+      "loss": 0.0337,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 412,
+      "tokens_per_second_per_gpu": 267.46
+    },
+    {
+      "epoch": 0.9472477064220184,
+      "grad_norm": 0.17694270610809326,
+      "learning_rate": 8.229455423913013e-07,
+      "loss": 0.0303,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.08,
+      "memory/max_allocated (GiB)": 54.08,
+      "step": 413,
+      "tokens_per_second_per_gpu": 287.7
+    },
+    {
+      "epoch": 0.9495412844036697,
+      "grad_norm": 0.1923583745956421,
+      "learning_rate": 7.559652192467126e-07,
+      "loss": 0.0374,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 414,
+      "tokens_per_second_per_gpu": 333.67
+    },
+    {
+      "epoch": 0.9518348623853211,
+      "grad_norm": 0.17904430627822876,
+      "learning_rate": 6.918068837427128e-07,
+      "loss": 0.0241,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 415,
+      "tokens_per_second_per_gpu": 355.41
+    },
+    {
+      "epoch": 0.9541284403669725,
+      "grad_norm": 0.14790676534175873,
+      "learning_rate": 6.304742125485874e-07,
+      "loss": 0.0227,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 416,
+      "tokens_per_second_per_gpu": 336.07
+    },
+    {
+      "epoch": 0.9564220183486238,
+      "grad_norm": 0.1487906128168106,
+      "learning_rate": 5.719707204055735e-07,
+      "loss": 0.0221,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 417,
+      "tokens_per_second_per_gpu": 357.03
+    },
+    {
+      "epoch": 0.9587155963302753,
+      "grad_norm": 0.13644175231456757,
+      "learning_rate": 5.162997599254704e-07,
+      "loss": 0.0211,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.39,
+      "memory/max_allocated (GiB)": 54.39,
+      "step": 418,
+      "tokens_per_second_per_gpu": 414.62
+    },
+    {
+      "epoch": 0.9610091743119266,
+      "grad_norm": 0.15555576980113983,
+      "learning_rate": 4.634645213984934e-07,
+      "loss": 0.0274,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 419,
+      "tokens_per_second_per_gpu": 383.45
+    },
+    {
+      "epoch": 0.963302752293578,
+      "grad_norm": 0.16330547630786896,
+      "learning_rate": 4.134680326104645e-07,
+      "loss": 0.0315,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 420,
+      "tokens_per_second_per_gpu": 417.47
+    },
+    {
+      "epoch": 0.9655963302752294,
+      "grad_norm": 0.16092361509799957,
+      "learning_rate": 3.663131586692792e-07,
+      "loss": 0.024,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 421,
+      "tokens_per_second_per_gpu": 285.66
+    },
+    {
+      "epoch": 0.9678899082568807,
+      "grad_norm": 0.16476912796497345,
+      "learning_rate": 3.2200260184075406e-07,
+      "loss": 0.0264,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 422,
+      "tokens_per_second_per_gpu": 368.99
+    },
+    {
+      "epoch": 0.9701834862385321,
+      "grad_norm": 0.13190625607967377,
+      "learning_rate": 2.805389013937454e-07,
+      "loss": 0.0181,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 423,
+      "tokens_per_second_per_gpu": 349.07
+    },
+    {
+      "epoch": 0.9724770642201835,
+      "grad_norm": 0.1542016565799713,
+      "learning_rate": 2.419244334546267e-07,
+      "loss": 0.0208,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.04,
+      "memory/max_allocated (GiB)": 54.04,
+      "step": 424,
+      "tokens_per_second_per_gpu": 291.68
+    },
+    {
+      "epoch": 0.9747706422018348,
+      "grad_norm": 0.20206497609615326,
+      "learning_rate": 2.061614108711474e-07,
+      "loss": 0.0423,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 425,
+      "tokens_per_second_per_gpu": 323.72
+    },
+    {
+      "epoch": 0.9770642201834863,
+      "grad_norm": 0.14892390370368958,
+      "learning_rate": 1.732518830856067e-07,
+      "loss": 0.0226,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 426,
+      "tokens_per_second_per_gpu": 375.55
+    },
+    {
+      "epoch": 0.9793577981651376,
+      "grad_norm": 0.1744110882282257,
+      "learning_rate": 1.431977360173975e-07,
+      "loss": 0.0261,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 427,
+      "tokens_per_second_per_gpu": 380.25
+    },
+    {
+      "epoch": 0.981651376146789,
+      "grad_norm": 0.16802337765693665,
+      "learning_rate": 1.16000691954965e-07,
+      "loss": 0.0312,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.32,
+      "memory/max_allocated (GiB)": 54.32,
+      "step": 428,
+      "tokens_per_second_per_gpu": 414.47
+    },
+    {
+      "epoch": 0.9839449541284404,
+      "grad_norm": 0.15291042625904083,
+      "learning_rate": 9.1662309457069e-08,
+      "loss": 0.026,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.28,
+      "memory/max_allocated (GiB)": 54.28,
+      "step": 429,
+      "tokens_per_second_per_gpu": 363.62
+    },
+    {
+      "epoch": 0.9862385321100917,
+      "grad_norm": 0.15010935068130493,
+      "learning_rate": 7.018398326350539e-08,
+      "loss": 0.0228,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.16,
+      "memory/max_allocated (GiB)": 54.16,
+      "step": 430,
+      "tokens_per_second_per_gpu": 285.9
+    },
+    {
+      "epoch": 0.9885321100917431,
+      "grad_norm": 0.17680975794792175,
+      "learning_rate": 5.15669442151423e-08,
+      "loss": 0.0307,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.12,
+      "memory/max_allocated (GiB)": 54.12,
+      "step": 431,
+      "tokens_per_second_per_gpu": 235.14
+    },
+    {
+      "epoch": 0.9908256880733946,
+      "grad_norm": 0.17848782241344452,
+      "learning_rate": 3.581225918342646e-08,
+      "loss": 0.0309,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.24,
+      "memory/max_allocated (GiB)": 54.24,
+      "step": 432,
+      "tokens_per_second_per_gpu": 330.64
+    },
+    {
+      "epoch": 0.9931192660550459,
+      "grad_norm": 0.14756126701831818,
+      "learning_rate": 2.292083100920994e-08,
+      "loss": 0.0261,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 433,
+      "tokens_per_second_per_gpu": 370.84
+    },
+    {
+      "epoch": 0.9954128440366973,
+      "grad_norm": 0.17790447175502777,
+      "learning_rate": 1.2893398451024886e-08,
+      "loss": 0.0329,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.35,
+      "memory/max_allocated (GiB)": 54.35,
+      "step": 434,
+      "tokens_per_second_per_gpu": 403.79
+    },
+    {
+      "epoch": 0.9977064220183486,
+      "grad_norm": 0.20343028008937836,
+      "learning_rate": 5.730536142745102e-09,
+      "loss": 0.0368,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 435,
+      "tokens_per_second_per_gpu": 361.41
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.16160258650779724,
+      "learning_rate": 1.432654560679092e-09,
+      "loss": 0.0258,
+      "memory/device_reserved (GiB)": 63.06,
+      "memory/max_active (GiB)": 54.2,
+      "memory/max_allocated (GiB)": 54.2,
+      "step": 436,
+      "tokens_per_second_per_gpu": 297.23
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 436,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 60,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5.772961672511619e+18,
+  "train_batch_size": 16,
+  "trial_name": null,
+  "trial_params": null
+}