{
  "best_global_step": 2500,
  "best_metric": 0.14181293547153473,
  "best_model_checkpoint": "Output/LightOnOCR-2-ft-Tsongkhapa_v1_lines/checkpoint-2500",
  "epoch": 9.765625,
  "eval_steps": 50,
  "global_step": 2500,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.1953125,
      "grad_norm": 0.5390625,
      "learning_rate": 5.969491525423729e-05,
      "loss": 0.9187405395507813,
      "step": 50
    },
    {
      "epoch": 0.1953125,
      "eval_loss": 0.6442809700965881,
      "eval_runtime": 46.8811,
      "eval_samples_per_second": 21.842,
      "eval_steps_per_second": 3.648,
      "step": 50
    },
    {
      "epoch": 0.390625,
      "grad_norm": 0.73828125,
      "learning_rate": 5.9303780964797915e-05,
      "loss": 0.5725218582153321,
      "step": 100
    },
    {
      "epoch": 0.390625,
      "eval_loss": 0.5042728185653687,
      "eval_runtime": 47.3174,
      "eval_samples_per_second": 21.641,
      "eval_steps_per_second": 3.614,
      "step": 100
    },
    {
      "epoch": 0.5859375,
      "grad_norm": 0.6484375,
      "learning_rate": 5.891264667535854e-05,
      "loss": 0.47095882415771484,
      "step": 150
    },
    {
      "epoch": 0.5859375,
      "eval_loss": 0.43541908264160156,
      "eval_runtime": 47.1075,
      "eval_samples_per_second": 21.737,
      "eval_steps_per_second": 3.63,
      "step": 150
    },
    {
      "epoch": 0.78125,
      "grad_norm": 0.73828125,
      "learning_rate": 5.8521512385919166e-05,
      "loss": 0.4152865219116211,
      "step": 200
    },
    {
      "epoch": 0.78125,
      "eval_loss": 0.37952595949172974,
      "eval_runtime": 47.0492,
      "eval_samples_per_second": 21.764,
      "eval_steps_per_second": 3.634,
      "step": 200
    },
    {
      "epoch": 0.9765625,
      "grad_norm": 0.7265625,
      "learning_rate": 5.813037809647979e-05,
      "loss": 0.363801155090332,
      "step": 250
    },
    {
      "epoch": 0.9765625,
      "eval_loss": 0.35072338581085205,
      "eval_runtime": 46.9489,
      "eval_samples_per_second": 21.811,
      "eval_steps_per_second": 3.642,
      "step": 250
    },
    {
      "epoch": 1.171875,
      "grad_norm": 0.7890625,
      "learning_rate": 5.773924380704042e-05,
      "loss": 0.33179210662841796,
      "step": 300
    },
    {
      "epoch": 1.171875,
      "eval_loss": 0.3240472674369812,
      "eval_runtime": 46.9989,
      "eval_samples_per_second": 21.788,
      "eval_steps_per_second": 3.638,
      "step": 300
    },
    {
      "epoch": 1.3671875,
      "grad_norm": 0.65234375,
      "learning_rate": 5.734810951760104e-05,
      "loss": 0.30721635818481446,
      "step": 350
    },
    {
      "epoch": 1.3671875,
      "eval_loss": 0.3084252178668976,
      "eval_runtime": 47.0141,
      "eval_samples_per_second": 21.781,
      "eval_steps_per_second": 3.637,
      "step": 350
    },
    {
      "epoch": 1.5625,
      "grad_norm": 0.353515625,
      "learning_rate": 5.695697522816167e-05,
      "loss": 0.28962944030761717,
      "step": 400
    },
    {
      "epoch": 1.5625,
      "eval_loss": 0.2849891781806946,
      "eval_runtime": 47.0271,
      "eval_samples_per_second": 21.775,
      "eval_steps_per_second": 3.636,
      "step": 400
    },
    {
      "epoch": 1.7578125,
      "grad_norm": 0.47265625,
      "learning_rate": 5.6565840938722294e-05,
      "loss": 0.2774631881713867,
      "step": 450
    },
    {
      "epoch": 1.7578125,
      "eval_loss": 0.2731478810310364,
      "eval_runtime": 47.0547,
      "eval_samples_per_second": 21.762,
      "eval_steps_per_second": 3.634,
      "step": 450
    },
    {
      "epoch": 1.953125,
      "grad_norm": 0.3203125,
      "learning_rate": 5.617470664928292e-05,
      "loss": 0.25862661361694333,
      "step": 500
    },
    {
      "epoch": 1.953125,
      "eval_loss": 0.2546536326408386,
      "eval_runtime": 47.0859,
      "eval_samples_per_second": 21.747,
      "eval_steps_per_second": 3.632,
      "step": 500
    },
    {
      "epoch": 2.1484375,
      "grad_norm": 0.58203125,
      "learning_rate": 5.5783572359843546e-05,
      "loss": 0.23830808639526369,
      "step": 550
    },
    {
      "epoch": 2.1484375,
      "eval_loss": 0.25039613246917725,
      "eval_runtime": 47.0985,
      "eval_samples_per_second": 21.742,
      "eval_steps_per_second": 3.631,
      "step": 550
    },
    {
      "epoch": 2.34375,
      "grad_norm": 0.42578125,
      "learning_rate": 5.539243807040417e-05,
      "loss": 0.23750856399536133,
      "step": 600
    },
    {
      "epoch": 2.34375,
      "eval_loss": 0.238111674785614,
      "eval_runtime": 47.0814,
      "eval_samples_per_second": 21.75,
      "eval_steps_per_second": 3.632,
      "step": 600
    },
    {
      "epoch": 2.5390625,
      "grad_norm": 0.5078125,
      "learning_rate": 5.5001303780964804e-05,
      "loss": 0.22395227432250978,
      "step": 650
    },
    {
      "epoch": 2.5390625,
      "eval_loss": 0.23060572147369385,
      "eval_runtime": 47.296,
      "eval_samples_per_second": 21.651,
      "eval_steps_per_second": 3.616,
      "step": 650
    },
    {
      "epoch": 2.734375,
      "grad_norm": 0.37109375,
      "learning_rate": 5.461016949152542e-05,
      "loss": 0.20927860260009765,
      "step": 700
    },
    {
      "epoch": 2.734375,
      "eval_loss": 0.21895363926887512,
      "eval_runtime": 47.2492,
      "eval_samples_per_second": 21.672,
      "eval_steps_per_second": 3.619,
      "step": 700
    },
    {
      "epoch": 2.9296875,
      "grad_norm": 0.5546875,
      "learning_rate": 5.421903520208605e-05,
      "loss": 0.21066415786743165,
      "step": 750
    },
    {
      "epoch": 2.9296875,
      "eval_loss": 0.21317130327224731,
      "eval_runtime": 47.1154,
      "eval_samples_per_second": 21.734,
      "eval_steps_per_second": 3.629,
      "step": 750
    },
    {
      "epoch": 3.125,
      "grad_norm": 0.61328125,
      "learning_rate": 5.382790091264668e-05,
      "loss": 0.19511587142944337,
      "step": 800
    },
    {
      "epoch": 3.125,
      "eval_loss": 0.20500582456588745,
      "eval_runtime": 47.3019,
      "eval_samples_per_second": 21.648,
      "eval_steps_per_second": 3.615,
      "step": 800
    },
    {
      "epoch": 3.3203125,
      "grad_norm": 0.37109375,
      "learning_rate": 5.34367666232073e-05,
      "loss": 0.18393211364746093,
      "step": 850
    },
    {
      "epoch": 3.3203125,
      "eval_loss": 0.20138591527938843,
      "eval_runtime": 47.0993,
      "eval_samples_per_second": 21.741,
      "eval_steps_per_second": 3.631,
      "step": 850
    },
    {
      "epoch": 3.515625,
      "grad_norm": 0.353515625,
      "learning_rate": 5.3045632333767925e-05,
      "loss": 0.18533748626708985,
      "step": 900
    },
    {
      "epoch": 3.515625,
      "eval_loss": 0.19779549539089203,
      "eval_runtime": 47.1339,
      "eval_samples_per_second": 21.725,
      "eval_steps_per_second": 3.628,
      "step": 900
    },
    {
      "epoch": 3.7109375,
      "grad_norm": 0.73046875,
      "learning_rate": 5.265449804432856e-05,
      "loss": 0.18759201049804688,
      "step": 950
    },
    {
      "epoch": 3.7109375,
      "eval_loss": 0.19392603635787964,
      "eval_runtime": 47.1618,
      "eval_samples_per_second": 21.712,
      "eval_steps_per_second": 3.626,
      "step": 950
    },
    {
      "epoch": 3.90625,
      "grad_norm": 0.56640625,
      "learning_rate": 5.226336375488918e-05,
      "loss": 0.17709009170532228,
      "step": 1000
    },
    {
      "epoch": 3.90625,
      "eval_loss": 0.1904962658882141,
      "eval_runtime": 47.1385,
      "eval_samples_per_second": 21.723,
      "eval_steps_per_second": 3.628,
      "step": 1000
    },
    {
      "epoch": 4.1015625,
      "grad_norm": 0.494140625,
      "learning_rate": 5.18722294654498e-05,
      "loss": 0.17133207321166993,
      "step": 1050
    },
    {
      "epoch": 4.1015625,
      "eval_loss": 0.18719176948070526,
      "eval_runtime": 47.1539,
      "eval_samples_per_second": 21.716,
      "eval_steps_per_second": 3.626,
      "step": 1050
    },
    {
      "epoch": 4.296875,
      "grad_norm": 0.29296875,
      "learning_rate": 5.1481095176010435e-05,
      "loss": 0.1665992546081543,
      "step": 1100
    },
    {
      "epoch": 4.296875,
      "eval_loss": 0.18399563431739807,
      "eval_runtime": 47.1505,
      "eval_samples_per_second": 21.718,
      "eval_steps_per_second": 3.627,
      "step": 1100
    },
    {
      "epoch": 4.4921875,
      "grad_norm": 0.375,
      "learning_rate": 5.1089960886571054e-05,
      "loss": 0.16016250610351562,
      "step": 1150
    },
    {
      "epoch": 4.4921875,
      "eval_loss": 0.18017655611038208,
      "eval_runtime": 47.1593,
      "eval_samples_per_second": 21.714,
      "eval_steps_per_second": 3.626,
      "step": 1150
    },
    {
      "epoch": 4.6875,
      "grad_norm": 0.298828125,
      "learning_rate": 5.069882659713168e-05,
      "loss": 0.15881264686584473,
      "step": 1200
    },
    {
      "epoch": 4.6875,
      "eval_loss": 0.1789371222257614,
      "eval_runtime": 47.1606,
      "eval_samples_per_second": 21.713,
      "eval_steps_per_second": 3.626,
      "step": 1200
    },
    {
      "epoch": 4.8828125,
      "grad_norm": 1.0234375,
      "learning_rate": 5.030769230769231e-05,
      "loss": 0.16435245513916016,
      "step": 1250
    },
    {
      "epoch": 4.8828125,
      "eval_loss": 0.17935042083263397,
      "eval_runtime": 47.1859,
      "eval_samples_per_second": 21.701,
      "eval_steps_per_second": 3.624,
      "step": 1250
    },
    {
      "epoch": 5.078125,
      "grad_norm": 0.31640625,
      "learning_rate": 4.991655801825294e-05,
      "loss": 0.16247941970825194,
      "step": 1300
    },
    {
      "epoch": 5.078125,
      "eval_loss": 0.1745760142803192,
      "eval_runtime": 47.3117,
      "eval_samples_per_second": 21.644,
      "eval_steps_per_second": 3.614,
      "step": 1300
    },
    {
      "epoch": 5.2734375,
      "grad_norm": 0.435546875,
      "learning_rate": 4.9525423728813556e-05,
      "loss": 0.14758566856384278,
      "step": 1350
    },
    {
      "epoch": 5.2734375,
      "eval_loss": 0.1702026128768921,
      "eval_runtime": 47.1167,
      "eval_samples_per_second": 21.733,
      "eval_steps_per_second": 3.629,
      "step": 1350
    },
    {
      "epoch": 5.46875,
      "grad_norm": 0.62890625,
      "learning_rate": 4.913428943937419e-05,
      "loss": 0.1457064628601074,
      "step": 1400
    },
    {
      "epoch": 5.46875,
      "eval_loss": 0.16571737825870514,
      "eval_runtime": 47.0795,
      "eval_samples_per_second": 21.75,
      "eval_steps_per_second": 3.632,
      "step": 1400
    },
    {
      "epoch": 5.6640625,
      "grad_norm": 0.9609375,
      "learning_rate": 4.8743155149934814e-05,
      "loss": 0.135118408203125,
      "step": 1450
    },
    {
      "epoch": 5.6640625,
      "eval_loss": 0.16433748602867126,
      "eval_runtime": 47.0384,
      "eval_samples_per_second": 21.769,
      "eval_steps_per_second": 3.635,
      "step": 1450
    },
    {
      "epoch": 5.859375,
      "grad_norm": 0.2890625,
      "learning_rate": 4.835202086049543e-05,
      "loss": 0.14427947044372558,
      "step": 1500
    },
    {
      "epoch": 5.859375,
      "eval_loss": 0.16130518913269043,
      "eval_runtime": 47.0567,
      "eval_samples_per_second": 21.761,
      "eval_steps_per_second": 3.634,
      "step": 1500
    },
    {
      "epoch": 6.0546875,
      "grad_norm": 0.439453125,
      "learning_rate": 4.7960886571056066e-05,
      "loss": 0.14234776496887208,
      "step": 1550
    },
    {
      "epoch": 6.0546875,
      "eval_loss": 0.15913954377174377,
      "eval_runtime": 47.2404,
      "eval_samples_per_second": 21.676,
      "eval_steps_per_second": 3.62,
      "step": 1550
    },
    {
      "epoch": 6.25,
      "grad_norm": 0.3515625,
      "learning_rate": 4.756975228161669e-05,
      "loss": 0.12651596069335938,
      "step": 1600
    },
    {
      "epoch": 6.25,
      "eval_loss": 0.16049708425998688,
      "eval_runtime": 47.1173,
      "eval_samples_per_second": 21.733,
      "eval_steps_per_second": 3.629,
      "step": 1600
    },
    {
      "epoch": 6.4453125,
      "grad_norm": 0.3125,
      "learning_rate": 4.717861799217731e-05,
      "loss": 0.13076963424682617,
      "step": 1650
    },
    {
      "epoch": 6.4453125,
      "eval_loss": 0.158030703663826,
      "eval_runtime": 47.1458,
      "eval_samples_per_second": 21.72,
      "eval_steps_per_second": 3.627,
      "step": 1650
    },
    {
      "epoch": 6.640625,
      "grad_norm": 1.0390625,
      "learning_rate": 4.678748370273794e-05,
      "loss": 0.13330907821655275,
      "step": 1700
    },
    {
      "epoch": 6.640625,
      "eval_loss": 0.15628746151924133,
      "eval_runtime": 47.2327,
      "eval_samples_per_second": 21.68,
      "eval_steps_per_second": 3.62,
      "step": 1700
    },
    {
      "epoch": 6.8359375,
      "grad_norm": 0.38671875,
      "learning_rate": 4.639634941329857e-05,
      "loss": 0.1278824996948242,
      "step": 1750
    },
    {
      "epoch": 6.8359375,
      "eval_loss": 0.15447524189949036,
      "eval_runtime": 47.1896,
      "eval_samples_per_second": 21.7,
      "eval_steps_per_second": 3.624,
      "step": 1750
    },
    {
      "epoch": 7.03125,
      "grad_norm": 0.97265625,
      "learning_rate": 4.600521512385919e-05,
      "loss": 0.12293228149414062,
      "step": 1800
    },
    {
      "epoch": 7.03125,
      "eval_loss": 0.15221962332725525,
      "eval_runtime": 47.1928,
      "eval_samples_per_second": 21.698,
      "eval_steps_per_second": 3.623,
      "step": 1800
    },
    {
      "epoch": 7.2265625,
      "grad_norm": 0.328125,
      "learning_rate": 4.561408083441982e-05,
      "loss": 0.12022055625915527,
      "step": 1850
    },
    {
      "epoch": 7.2265625,
      "eval_loss": 0.15250065922737122,
      "eval_runtime": 47.2688,
      "eval_samples_per_second": 21.663,
      "eval_steps_per_second": 3.618,
      "step": 1850
    },
    {
      "epoch": 7.421875,
      "grad_norm": 0.30078125,
      "learning_rate": 4.5222946544980445e-05,
      "loss": 0.11987478256225587,
      "step": 1900
    },
    {
      "epoch": 7.421875,
      "eval_loss": 0.150522381067276,
      "eval_runtime": 47.2684,
      "eval_samples_per_second": 21.664,
      "eval_steps_per_second": 3.618,
      "step": 1900
    },
    {
      "epoch": 7.6171875,
      "grad_norm": 0.361328125,
      "learning_rate": 4.4831812255541064e-05,
      "loss": 0.11880167007446289,
      "step": 1950
    },
    {
      "epoch": 7.6171875,
      "eval_loss": 0.14916279911994934,
      "eval_runtime": 47.3178,
      "eval_samples_per_second": 21.641,
      "eval_steps_per_second": 3.614,
      "step": 1950
    },
    {
      "epoch": 7.8125,
      "grad_norm": 0.31640625,
      "learning_rate": 4.44406779661017e-05,
      "loss": 0.11192596435546875,
      "step": 2000
    },
    {
      "epoch": 7.8125,
      "eval_loss": 0.1477399468421936,
      "eval_runtime": 47.3373,
      "eval_samples_per_second": 21.632,
      "eval_steps_per_second": 3.612,
      "step": 2000
    },
    {
      "epoch": 8.0078125,
      "grad_norm": 0.404296875,
      "learning_rate": 4.404954367666232e-05,
      "loss": 0.11280095100402832,
      "step": 2050
    },
    {
      "epoch": 8.0078125,
      "eval_loss": 0.14646019041538239,
      "eval_runtime": 47.3114,
      "eval_samples_per_second": 21.644,
      "eval_steps_per_second": 3.614,
      "step": 2050
    },
    {
      "epoch": 8.203125,
      "grad_norm": 0.3359375,
      "learning_rate": 4.365840938722295e-05,
      "loss": 0.11032476425170898,
      "step": 2100
    },
    {
      "epoch": 8.203125,
      "eval_loss": 0.14793652296066284,
      "eval_runtime": 47.1893,
      "eval_samples_per_second": 21.7,
      "eval_steps_per_second": 3.624,
      "step": 2100
    },
    {
      "epoch": 8.3984375,
      "grad_norm": 0.388671875,
      "learning_rate": 4.3267275097783574e-05,
      "loss": 0.11091269493103027,
      "step": 2150
    },
    {
      "epoch": 8.3984375,
      "eval_loss": 0.14682111144065857,
      "eval_runtime": 47.1132,
      "eval_samples_per_second": 21.735,
      "eval_steps_per_second": 3.63,
      "step": 2150
    },
    {
      "epoch": 8.59375,
      "grad_norm": 0.265625,
      "learning_rate": 4.28761408083442e-05,
      "loss": 0.10571674346923828,
      "step": 2200
    },
    {
      "epoch": 8.59375,
      "eval_loss": 0.14549781382083893,
      "eval_runtime": 47.0828,
      "eval_samples_per_second": 21.749,
      "eval_steps_per_second": 3.632,
      "step": 2200
    },
    {
      "epoch": 8.7890625,
      "grad_norm": 0.3203125,
      "learning_rate": 4.2485006518904825e-05,
      "loss": 0.112736177444458,
      "step": 2250
    },
    {
      "epoch": 8.7890625,
      "eval_loss": 0.1448877900838852,
      "eval_runtime": 47.1721,
      "eval_samples_per_second": 21.708,
      "eval_steps_per_second": 3.625,
      "step": 2250
    },
    {
      "epoch": 8.984375,
      "grad_norm": 0.3203125,
      "learning_rate": 4.209387222946545e-05,
      "loss": 0.10553077697753906,
      "step": 2300
    },
    {
      "epoch": 8.984375,
      "eval_loss": 0.14488515257835388,
      "eval_runtime": 47.1566,
      "eval_samples_per_second": 21.715,
      "eval_steps_per_second": 3.626,
      "step": 2300
    },
    {
      "epoch": 9.1796875,
      "grad_norm": 0.265625,
      "learning_rate": 4.1702737940026076e-05,
      "loss": 0.10591485023498536,
      "step": 2350
    },
    {
      "epoch": 9.1796875,
      "eval_loss": 0.1439216583967209,
      "eval_runtime": 47.3149,
      "eval_samples_per_second": 21.642,
      "eval_steps_per_second": 3.614,
      "step": 2350
    },
    {
      "epoch": 9.375,
      "grad_norm": 0.3046875,
      "learning_rate": 4.13116036505867e-05,
      "loss": 0.10057873725891113,
      "step": 2400
    },
    {
      "epoch": 9.375,
      "eval_loss": 0.1432727873325348,
      "eval_runtime": 47.1444,
      "eval_samples_per_second": 21.72,
      "eval_steps_per_second": 3.627,
      "step": 2400
    },
    {
      "epoch": 9.5703125,
      "grad_norm": 0.2197265625,
      "learning_rate": 4.092046936114733e-05,
      "loss": 0.09765372276306153,
      "step": 2450
    },
    {
      "epoch": 9.5703125,
      "eval_loss": 0.14208123087882996,
      "eval_runtime": 47.0846,
      "eval_samples_per_second": 21.748,
      "eval_steps_per_second": 3.632,
      "step": 2450
    },
    {
      "epoch": 9.765625,
      "grad_norm": 0.2578125,
      "learning_rate": 4.0529335071707953e-05,
      "loss": 0.09726298332214356,
      "step": 2500
    },
    {
      "epoch": 9.765625,
      "eval_loss": 0.14181293547153473,
      "eval_runtime": 47.1351,
      "eval_samples_per_second": 21.725,
      "eval_steps_per_second": 3.628,
      "step": 2500
    }
  ],
  "logging_steps": 50,
  "max_steps": 7680,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 30,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 4.0241545572335616e+17,
  "train_batch_size": 4,
  "trial_name": null,
  "trial_params": null
}