{
  "best_global_step": 1200,
  "best_metric": 0.10889480262994766,
  "best_model_checkpoint": "./sft_output/checkpoint-1200",
  "epoch": 2.895489947473284,
  "eval_steps": 200,
  "global_step": 2000,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "entropy": 0.1292651420040056,
      "epoch": 0.07245064299945662,
      "grad_norm": 0.5799852013587952,
      "learning_rate": 0.00019527255185721178,
      "loss": 0.1073553466796875,
      "mean_token_accuracy": 0.9656643654406071,
      "num_tokens": 115293.0,
      "step": 50
    },
    {
      "entropy": 0.14149306831415742,
      "epoch": 0.14490128599891325,
      "grad_norm": 0.4752981960773468,
      "learning_rate": 0.00019044862518089725,
      "loss": 0.12214879989624024,
      "mean_token_accuracy": 0.9628654922544956,
      "num_tokens": 228079.0,
      "step": 100
    },
    {
      "entropy": 0.1316419944167137,
      "epoch": 0.21735192899836986,
      "grad_norm": 0.3439595699310303,
      "learning_rate": 0.00018562469850458274,
      "loss": 0.11439435958862304,
      "mean_token_accuracy": 0.9642146781086922,
      "num_tokens": 346451.0,
      "step": 150
    },
    {
      "entropy": 0.12753444796428084,
      "epoch": 0.2898025719978265,
      "grad_norm": 0.3129945695400238,
      "learning_rate": 0.0001808007718282682,
      "loss": 0.1101201343536377,
      "mean_token_accuracy": 0.9657260800898075,
      "num_tokens": 463838.0,
      "step": 200
    },
    {
      "epoch": 0.2898025719978265,
      "eval_entropy": 0.11204168354642802,
      "eval_loss": 0.1371564269065857,
      "eval_mean_token_accuracy": 0.9661400092059168,
      "eval_num_tokens": 463838.0,
      "eval_runtime": 135.07,
      "eval_samples_per_second": 5.108,
      "eval_steps_per_second": 0.644,
      "step": 200
    },
    {
      "entropy": 0.12514342403504997,
      "epoch": 0.3622532149972831,
      "grad_norm": 0.3354629874229431,
      "learning_rate": 0.0001759768451519537,
      "loss": 0.1064763069152832,
      "mean_token_accuracy": 0.9663380682468414,
      "num_tokens": 580179.0,
      "step": 250
    },
    {
      "entropy": 0.11715295719448476,
      "epoch": 0.4347038579967397,
      "grad_norm": 0.27767807245254517,
      "learning_rate": 0.00017115291847563918,
      "loss": 0.10067870140075684,
      "mean_token_accuracy": 0.9680409373342991,
      "num_tokens": 695580.0,
      "step": 300
    },
    {
      "entropy": 0.1172275464516133,
      "epoch": 0.5071545009961963,
      "grad_norm": 0.22333095967769623,
      "learning_rate": 0.00016632899179932467,
      "loss": 0.09892839431762696,
      "mean_token_accuracy": 0.968158980756998,
      "num_tokens": 813169.0,
      "step": 350
    },
    {
      "entropy": 0.13554750085109846,
      "epoch": 0.579605143995653,
      "grad_norm": 0.2716864049434662,
      "learning_rate": 0.00016150506512301014,
      "loss": 0.12139375686645508,
      "mean_token_accuracy": 0.9637367476522922,
      "num_tokens": 933136.0,
      "step": 400
    },
    {
      "epoch": 0.579605143995653,
      "eval_entropy": 0.12780291261686677,
      "eval_loss": 0.12689483165740967,
      "eval_mean_token_accuracy": 0.9676476744399674,
      "eval_num_tokens": 933136.0,
      "eval_runtime": 135.0083,
      "eval_samples_per_second": 5.111,
      "eval_steps_per_second": 0.644,
      "step": 400
    },
    {
      "entropy": 0.14845358284655957,
      "epoch": 0.6520557869951096,
      "grad_norm": 0.2887682616710663,
      "learning_rate": 0.0001566811384466956,
      "loss": 0.12956516265869142,
      "mean_token_accuracy": 0.961767661422491,
      "num_tokens": 1046558.0,
      "step": 450
    },
    {
      "entropy": 0.1415072442078963,
      "epoch": 0.7245064299945662,
      "grad_norm": 0.26643216609954834,
      "learning_rate": 0.0001518572117703811,
      "loss": 0.12267910957336425,
      "mean_token_accuracy": 0.9633507201075554,
      "num_tokens": 1163902.0,
      "step": 500
    },
    {
      "entropy": 0.1404953494342044,
      "epoch": 0.7969570729940229,
      "grad_norm": 0.2829984128475189,
      "learning_rate": 0.00014703328509406657,
      "loss": 0.12197186470031739,
      "mean_token_accuracy": 0.9628035886585713,
      "num_tokens": 1276260.0,
      "step": 550
    },
    {
      "entropy": 0.13520956406835466,
      "epoch": 0.8694077159934794,
      "grad_norm": 0.2343612164258957,
      "learning_rate": 0.00014220935841775204,
      "loss": 0.11497893333435058,
      "mean_token_accuracy": 0.964199658036232,
      "num_tokens": 1388505.0,
      "step": 600
    },
    {
      "epoch": 0.8694077159934794,
      "eval_entropy": 0.11246106024274881,
      "eval_loss": 0.11639242619276047,
      "eval_mean_token_accuracy": 0.9700561001383025,
      "eval_num_tokens": 1388505.0,
      "eval_runtime": 134.8927,
      "eval_samples_per_second": 5.115,
      "eval_steps_per_second": 0.645,
      "step": 600
    },
    {
      "entropy": 0.1313706226553768,
      "epoch": 0.941858358992936,
      "grad_norm": 0.2563211917877197,
      "learning_rate": 0.00013738543174143754,
      "loss": 0.1108535385131836,
      "mean_token_accuracy": 0.965393750667572,
      "num_tokens": 1506542.0,
      "step": 650
    },
    {
      "entropy": 0.12733429687870215,
      "epoch": 1.0130411157399022,
      "grad_norm": 0.20515012741088867,
      "learning_rate": 0.000132561505065123,
      "loss": 0.10263432502746582,
      "mean_token_accuracy": 0.9680144694318602,
      "num_tokens": 1625554.0,
      "step": 700
    },
    {
      "entropy": 0.09378954898566008,
      "epoch": 1.0854917587393589,
      "grad_norm": 0.1810425966978073,
      "learning_rate": 0.0001277375783888085,
      "loss": 0.07341335773468018,
      "mean_token_accuracy": 0.9737974387407303,
      "num_tokens": 1742352.0,
      "step": 750
    },
    {
      "entropy": 0.08947869850322604,
      "epoch": 1.1579424017388154,
      "grad_norm": 0.2107703685760498,
      "learning_rate": 0.00012291365171249397,
      "loss": 0.07443580627441407,
      "mean_token_accuracy": 0.974107117652893,
      "num_tokens": 1859066.0,
      "step": 800
    },
    {
      "epoch": 1.1579424017388154,
      "eval_entropy": 0.08398804492477713,
      "eval_loss": 0.11319845169782639,
      "eval_mean_token_accuracy": 0.9715756554713194,
      "eval_num_tokens": 1859066.0,
      "eval_runtime": 134.9594,
      "eval_samples_per_second": 5.113,
      "eval_steps_per_second": 0.645,
      "step": 800
    },
    {
      "entropy": 0.09298701203661039,
      "epoch": 1.230393044738272,
      "grad_norm": 0.2118096500635147,
      "learning_rate": 0.00011808972503617945,
      "loss": 0.07880696296691894,
      "mean_token_accuracy": 0.9729460710287094,
      "num_tokens": 1974287.0,
      "step": 850
    },
    {
      "entropy": 0.09084601077716797,
      "epoch": 1.3028436877377287,
      "grad_norm": 0.14472977817058563,
      "learning_rate": 0.00011326579835986493,
      "loss": 0.07501977443695068,
      "mean_token_accuracy": 0.973245297074318,
      "num_tokens": 2089089.0,
      "step": 900
    },
    {
      "entropy": 0.09346557535696774,
      "epoch": 1.3752943307371852,
      "grad_norm": 0.5793337225914001,
      "learning_rate": 0.00010844187168355043,
      "loss": 0.07840880870819092,
      "mean_token_accuracy": 0.9730692146718503,
      "num_tokens": 2203786.0,
      "step": 950
    },
    {
      "entropy": 0.09212197210872546,
      "epoch": 1.447744973736642,
      "grad_norm": 0.22646810114383698,
      "learning_rate": 0.0001036179450072359,
      "loss": 0.07762556552886962,
      "mean_token_accuracy": 0.9733995048701763,
      "num_tokens": 2318554.0,
      "step": 1000
    },
    {
      "epoch": 1.447744973736642,
      "eval_entropy": 0.08479796379975889,
      "eval_loss": 0.11282521486282349,
      "eval_mean_token_accuracy": 0.9713243239227383,
      "eval_num_tokens": 2318554.0,
      "eval_runtime": 135.0253,
      "eval_samples_per_second": 5.11,
      "eval_steps_per_second": 0.644,
      "step": 1000
    },
    {
      "entropy": 0.09136153725441545,
      "epoch": 1.5201956167360984,
      "grad_norm": 0.24745002388954163,
      "learning_rate": 9.879401833092138e-05,
      "loss": 0.0756511640548706,
      "mean_token_accuracy": 0.9727346667647362,
      "num_tokens": 2434178.0,
      "step": 1050
    },
    {
      "entropy": 0.08874542655656115,
      "epoch": 1.592646259735555,
      "grad_norm": 0.16458512842655182,
      "learning_rate": 9.397009165460685e-05,
      "loss": 0.0733512830734253,
      "mean_token_accuracy": 0.9746969762444496,
      "num_tokens": 2551600.0,
      "step": 1100
    },
    {
      "entropy": 0.08679104059003294,
      "epoch": 1.6650969027350118,
      "grad_norm": 0.2363063395023346,
      "learning_rate": 8.914616497829233e-05,
      "loss": 0.07282801628112794,
      "mean_token_accuracy": 0.9748657821118831,
      "num_tokens": 2670568.0,
      "step": 1150
    },
    {
      "entropy": 0.08530640854965896,
      "epoch": 1.7375475457344685,
      "grad_norm": 0.3205758035182953,
      "learning_rate": 8.432223830197782e-05,
      "loss": 0.07208527088165283,
      "mean_token_accuracy": 0.9753711603581905,
      "num_tokens": 2790395.0,
      "step": 1200
    },
    {
      "epoch": 1.7375475457344685,
      "eval_entropy": 0.08271760874609838,
      "eval_loss": 0.10889480262994766,
      "eval_mean_token_accuracy": 0.9722844334854477,
      "eval_num_tokens": 2790395.0,
      "eval_runtime": 135.0173,
      "eval_samples_per_second": 5.11,
      "eval_steps_per_second": 0.644,
      "step": 1200
    },
    {
      "entropy": 0.08853163382969796,
      "epoch": 1.8099981887339252,
      "grad_norm": 0.2860631048679352,
      "learning_rate": 7.949831162566329e-05,
      "loss": 0.07203876972198486,
      "mean_token_accuracy": 0.9745123912394047,
      "num_tokens": 2906250.0,
      "step": 1250
    },
    {
      "entropy": 0.09159044577972963,
      "epoch": 1.8824488317333816,
      "grad_norm": 0.22217783331871033,
      "learning_rate": 7.467438494934877e-05,
      "loss": 0.0765086030960083,
      "mean_token_accuracy": 0.9737336048483849,
      "num_tokens": 3020888.0,
      "step": 1300
    },
    {
      "entropy": 0.08859600902302191,
      "epoch": 1.954899474732838,
      "grad_norm": 0.35960590839385986,
      "learning_rate": 6.985045827303426e-05,
      "loss": 0.07379885196685791,
      "mean_token_accuracy": 0.9746835605800152,
      "num_tokens": 3134684.0,
      "step": 1350
    },
    {
      "entropy": 0.08423688769122496,
      "epoch": 2.0260822314798044,
      "grad_norm": 0.1249847561120987,
      "learning_rate": 6.502653159671973e-05,
      "loss": 0.06818146228790284,
      "mean_token_accuracy": 0.9766376161089987,
      "num_tokens": 3245618.0,
      "step": 1400
    },
    {
      "epoch": 2.0260822314798044,
      "eval_entropy": 0.07047315447152346,
      "eval_loss": 0.11015637964010239,
      "eval_mean_token_accuracy": 0.9731149563844177,
      "eval_num_tokens": 3245618.0,
      "eval_runtime": 134.9477,
      "eval_samples_per_second": 5.113,
      "eval_steps_per_second": 0.645,
      "step": 1400
    },
    {
      "entropy": 0.06371946840547026,
      "epoch": 2.098532874479261,
      "grad_norm": 0.16305267810821533,
      "learning_rate": 6.0202604920405214e-05,
      "loss": 0.05102997303009033,
      "mean_token_accuracy": 0.9815227979421616,
      "num_tokens": 3362965.0,
      "step": 1450
    },
    {
      "entropy": 0.0659942712332122,
      "epoch": 2.1709835174787178,
      "grad_norm": 0.1745850145816803,
      "learning_rate": 5.537867824409069e-05,
      "loss": 0.05241162776947021,
      "mean_token_accuracy": 0.9807419489324093,
      "num_tokens": 3475907.0,
      "step": 1500
    },
    {
      "entropy": 0.06338218597928062,
      "epoch": 2.2434341604781745,
      "grad_norm": 0.19584746658802032,
      "learning_rate": 5.055475156777617e-05,
      "loss": 0.051362857818603516,
      "mean_token_accuracy": 0.9814340886473656,
      "num_tokens": 3593538.0,
      "step": 1550
    },
    {
      "entropy": 0.06354268383933231,
      "epoch": 2.3158848034776307,
      "grad_norm": 0.19582241773605347,
      "learning_rate": 4.573082489146165e-05,
      "loss": 0.052291646003723144,
      "mean_token_accuracy": 0.981376932412386,
      "num_tokens": 3707445.0,
      "step": 1600
    },
    {
      "epoch": 2.3158848034776307,
      "eval_entropy": 0.06323016272194085,
      "eval_loss": 0.11573585867881775,
      "eval_mean_token_accuracy": 0.973355067872453,
      "eval_num_tokens": 3707445.0,
      "eval_runtime": 135.1558,
      "eval_samples_per_second": 5.105,
      "eval_steps_per_second": 0.644,
      "step": 1600
    },
    {
      "entropy": 0.06490354960318655,
      "epoch": 2.3883354464770874,
      "grad_norm": 0.17022456228733063,
      "learning_rate": 4.0906898215147135e-05,
      "loss": 0.054670772552490234,
      "mean_token_accuracy": 0.980761057138443,
      "num_tokens": 3820475.0,
      "step": 1650
    },
    {
      "entropy": 0.06598329161293805,
      "epoch": 2.460786089476544,
      "grad_norm": 0.20498181879520416,
      "learning_rate": 3.608297153883261e-05,
      "loss": 0.05304833889007568,
      "mean_token_accuracy": 0.9809189030528068,
      "num_tokens": 3933729.0,
      "step": 1700
    },
    {
      "entropy": 0.06734917678404599,
      "epoch": 2.533236732476001,
      "grad_norm": 0.1891096830368042,
      "learning_rate": 3.1259044862518086e-05,
      "loss": 0.0533873462677002,
      "mean_token_accuracy": 0.980241654664278,
      "num_tokens": 4047398.0,
      "step": 1750
    },
    {
      "entropy": 0.05935755535727367,
      "epoch": 2.6056873754754575,
      "grad_norm": 0.17223551869392395,
      "learning_rate": 2.643511818620357e-05,
      "loss": 0.04701524257659912,
      "mean_token_accuracy": 0.9830030533671379,
      "num_tokens": 4168983.0,
      "step": 1800
    },
    {
      "epoch": 2.6056873754754575,
      "eval_entropy": 0.06277676872994709,
      "eval_loss": 0.11459118127822876,
      "eval_mean_token_accuracy": 0.9735316360133818,
      "eval_num_tokens": 4168983.0,
      "eval_runtime": 135.0545,
      "eval_samples_per_second": 5.109,
      "eval_steps_per_second": 0.644,
      "step": 1800
    },
    {
      "entropy": 0.06294791254680604,
      "epoch": 2.6781380184749137,
      "grad_norm": 0.21268633008003235,
      "learning_rate": 2.161119150988905e-05,
      "loss": 0.05056027889251709,
      "mean_token_accuracy": 0.9813353289663792,
      "num_tokens": 4286132.0,
      "step": 1850
    },
    {
      "entropy": 0.06273156431736425,
      "epoch": 2.7505886614743704,
      "grad_norm": 0.166688933968544,
      "learning_rate": 1.678726483357453e-05,
      "loss": 0.05039352893829346,
      "mean_token_accuracy": 0.9812411935627461,
      "num_tokens": 4403199.0,
      "step": 1900
    },
    {
      "entropy": 0.06210695546586067,
      "epoch": 2.823039304473827,
      "grad_norm": 0.20107661187648773,
      "learning_rate": 1.1963338157260012e-05,
      "loss": 0.050140600204467776,
      "mean_token_accuracy": 0.9815662078559398,
      "num_tokens": 4519243.0,
      "step": 1950
    },
    {
      "entropy": 0.06227384155383334,
      "epoch": 2.895489947473284,
      "grad_norm": 0.1902814358472824,
      "learning_rate": 7.13941148094549e-06,
      "loss": 0.048079490661621094,
      "mean_token_accuracy": 0.9821359276771545,
      "num_tokens": 4636178.0,
      "step": 2000
    },
    {
      "epoch": 2.895489947473284,
      "eval_entropy": 0.06232287272297103,
      "eval_loss": 0.11387787759304047,
      "eval_mean_token_accuracy": 0.9737936319976017,
      "eval_num_tokens": 4636178.0,
      "eval_runtime": 135.2272,
      "eval_samples_per_second": 5.103,
      "eval_steps_per_second": 0.643,
      "step": 2000
    }
  ],
  "logging_steps": 50,
  "max_steps": 2073,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 3,
  "save_steps": 200,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 8752228753593600.0,
  "train_batch_size": 1,
  "trial_name": null,
  "trial_params": null
}