{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 4.0,
  "eval_steps": 500,
  "global_step": 34368,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.00011638733705772812,
      "grad_norm": Infinity,
      "learning_rate": 0.0,
      "loss": 9.4531,
      "step": 1
    },
    {
      "epoch": 0.05819366852886406,
      "grad_norm": 6.4754815101623535,
      "learning_rate": 1.7390292166220463e-06,
      "loss": 6.7666,
      "step": 500
    },
    {
      "epoch": 0.05819366852886406,
      "eval_cosine_accuracy": 0.9417916536331177,
      "eval_loss": 4.736684322357178,
      "eval_runtime": 26.7399,
      "eval_samples_per_second": 354.003,
      "eval_steps_per_second": 1.384,
      "step": 500
    },
    {
      "epoch": 0.11638733705772812,
      "grad_norm": 5.042841911315918,
      "learning_rate": 3.4850424863228964e-06,
      "loss": 4.0612,
      "step": 1000
    },
    {
      "epoch": 0.11638733705772812,
      "eval_cosine_accuracy": 0.9421085715293884,
      "eval_loss": 4.456507682800293,
      "eval_runtime": 27.0721,
      "eval_samples_per_second": 349.659,
      "eval_steps_per_second": 1.367,
      "step": 1000
    },
    {
      "epoch": 0.17458100558659218,
      "grad_norm": 4.396075248718262,
      "learning_rate": 5.2310557560237455e-06,
      "loss": 3.8306,
      "step": 1500
    },
    {
      "epoch": 0.17458100558659218,
      "eval_cosine_accuracy": 0.9389393329620361,
      "eval_loss": 4.417118072509766,
      "eval_runtime": 29.5299,
      "eval_samples_per_second": 320.556,
      "eval_steps_per_second": 1.253,
      "step": 1500
    },
    {
      "epoch": 0.23277467411545624,
      "grad_norm": 4.9983229637146,
      "learning_rate": 6.9770690257245955e-06,
      "loss": 3.7166,
      "step": 2000
    },
    {
      "epoch": 0.23277467411545624,
      "eval_cosine_accuracy": 0.9376716613769531,
      "eval_loss": 4.384617805480957,
      "eval_runtime": 27.1288,
      "eval_samples_per_second": 348.928,
      "eval_steps_per_second": 1.364,
      "step": 2000
    },
    {
      "epoch": 0.2909683426443203,
      "grad_norm": 5.1397552490234375,
      "learning_rate": 8.723082295425445e-06,
      "loss": 3.5906,
      "step": 2500
    },
    {
      "epoch": 0.2909683426443203,
      "eval_cosine_accuracy": 0.929537296295166,
      "eval_loss": 4.382342338562012,
      "eval_runtime": 27.2377,
      "eval_samples_per_second": 347.534,
      "eval_steps_per_second": 1.358,
      "step": 2500
    },
    {
      "epoch": 0.34916201117318435,
      "grad_norm": 5.401367664337158,
      "learning_rate": 1.0469095565126295e-05,
      "loss": 3.504,
      "step": 3000
    },
    {
      "epoch": 0.34916201117318435,
      "eval_cosine_accuracy": 0.9236213564872742,
      "eval_loss": 4.398530006408691,
      "eval_runtime": 27.2096,
      "eval_samples_per_second": 347.891,
      "eval_steps_per_second": 1.36,
      "step": 3000
    },
    {
      "epoch": 0.4073556797020484,
      "grad_norm": 5.506592273712158,
      "learning_rate": 1.2215108834827144e-05,
      "loss": 3.4064,
      "step": 3500
    },
    {
      "epoch": 0.4073556797020484,
      "eval_cosine_accuracy": 0.9207690954208374,
      "eval_loss": 4.428633689880371,
      "eval_runtime": 27.2549,
      "eval_samples_per_second": 347.314,
      "eval_steps_per_second": 1.358,
      "step": 3500
    },
    {
      "epoch": 0.4655493482309125,
      "grad_norm": 6.144413471221924,
      "learning_rate": 1.3961122104527995e-05,
      "loss": 3.3219,
      "step": 4000
    },
    {
      "epoch": 0.4655493482309125,
      "eval_cosine_accuracy": 0.9240439534187317,
      "eval_loss": 4.413504600524902,
      "eval_runtime": 27.1733,
      "eval_samples_per_second": 348.357,
      "eval_steps_per_second": 1.362,
      "step": 4000
    },
    {
      "epoch": 0.5237430167597765,
      "grad_norm": 6.7739691734313965,
      "learning_rate": 1.570364334768944e-05,
      "loss": 3.2308,
      "step": 4500
    },
    {
      "epoch": 0.5237430167597765,
      "eval_cosine_accuracy": 0.9183393120765686,
      "eval_loss": 4.584815979003906,
      "eval_runtime": 27.4916,
      "eval_samples_per_second": 344.324,
      "eval_steps_per_second": 1.346,
      "step": 4500
    },
    {
      "epoch": 0.5819366852886406,
      "grad_norm": 7.5280914306640625,
      "learning_rate": 1.7449656617390294e-05,
      "loss": 3.1167,
      "step": 5000
    },
    {
      "epoch": 0.5819366852886406,
      "eval_cosine_accuracy": 0.9231988191604614,
      "eval_loss": 4.731673240661621,
      "eval_runtime": 27.9733,
      "eval_samples_per_second": 338.394,
      "eval_steps_per_second": 1.323,
      "step": 5000
    },
    {
      "epoch": 0.6401303538175046,
      "grad_norm": 8.379385948181152,
      "learning_rate": 1.919217786055174e-05,
      "loss": 3.0155,
      "step": 5500
    },
    {
      "epoch": 0.6401303538175046,
      "eval_cosine_accuracy": 0.9202408790588379,
      "eval_loss": 4.590743064880371,
      "eval_runtime": 30.4387,
      "eval_samples_per_second": 310.985,
      "eval_steps_per_second": 1.216,
      "step": 5500
    },
    {
      "epoch": 0.6983240223463687,
      "grad_norm": 7.9107985496521,
      "learning_rate": 2.093819113025259e-05,
      "loss": 2.928,
      "step": 6000
    },
    {
      "epoch": 0.6983240223463687,
      "eval_cosine_accuracy": 0.9188675284385681,
      "eval_loss": 4.662985324859619,
      "eval_runtime": 27.6404,
      "eval_samples_per_second": 342.469,
      "eval_steps_per_second": 1.339,
      "step": 6000
    },
    {
      "epoch": 0.7565176908752328,
      "grad_norm": 8.108757972717285,
      "learning_rate": 2.268420439995344e-05,
      "loss": 2.8455,
      "step": 6500
    },
    {
      "epoch": 0.7565176908752328,
      "eval_cosine_accuracy": 0.9199239611625671,
      "eval_loss": 4.784646511077881,
      "eval_runtime": 27.3827,
      "eval_samples_per_second": 345.693,
      "eval_steps_per_second": 1.351,
      "step": 6500
    },
    {
      "epoch": 0.8147113594040968,
      "grad_norm": 8.092316627502441,
      "learning_rate": 2.443021766965429e-05,
      "loss": 2.7847,
      "step": 7000
    },
    {
      "epoch": 0.8147113594040968,
      "eval_cosine_accuracy": 0.9054510593414307,
      "eval_loss": 4.925645351409912,
      "eval_runtime": 27.6051,
      "eval_samples_per_second": 342.908,
      "eval_steps_per_second": 1.34,
      "step": 7000
    },
    {
      "epoch": 0.8729050279329609,
      "grad_norm": 8.676020622253418,
      "learning_rate": 2.617623093935514e-05,
      "loss": 2.7228,
      "step": 7500
    },
    {
      "epoch": 0.8729050279329609,
      "eval_cosine_accuracy": 0.9056623578071594,
      "eval_loss": 4.843690395355225,
      "eval_runtime": 27.3545,
      "eval_samples_per_second": 346.049,
      "eval_steps_per_second": 1.353,
      "step": 7500
    },
    {
      "epoch": 0.931098696461825,
      "grad_norm": 9.083893775939941,
      "learning_rate": 2.791875218251659e-05,
      "loss": 2.6738,
      "step": 8000
    },
    {
      "epoch": 0.931098696461825,
      "eval_cosine_accuracy": 0.907247006893158,
      "eval_loss": 4.833749771118164,
      "eval_runtime": 27.6952,
      "eval_samples_per_second": 341.791,
      "eval_steps_per_second": 1.336,
      "step": 8000
    },
    {
      "epoch": 0.9892923649906891,
      "grad_norm": 7.430295944213867,
      "learning_rate": 2.9664765452217438e-05,
      "loss": 2.6174,
      "step": 8500
    },
    {
      "epoch": 0.9892923649906891,
      "eval_cosine_accuracy": 0.9006972312927246,
      "eval_loss": 4.809013366699219,
      "eval_runtime": 27.7002,
      "eval_samples_per_second": 341.731,
      "eval_steps_per_second": 1.336,
      "step": 8500
    },
    {
      "epoch": 1.047486033519553,
      "grad_norm": 8.466830253601074,
      "learning_rate": 2.998182094709628e-05,
      "loss": 2.5583,
      "step": 9000
    },
    {
      "epoch": 1.047486033519553,
      "eval_cosine_accuracy": 0.8939362168312073,
      "eval_loss": 4.823788166046143,
      "eval_runtime": 27.5804,
      "eval_samples_per_second": 343.214,
      "eval_steps_per_second": 1.342,
      "step": 9000
    },
    {
      "epoch": 1.105679702048417,
      "grad_norm": 7.473945617675781,
      "learning_rate": 2.9909051871166705e-05,
      "loss": 2.5004,
      "step": 9500
    },
    {
      "epoch": 1.105679702048417,
      "eval_cosine_accuracy": 0.9032326340675354,
      "eval_loss": 4.753988265991211,
      "eval_runtime": 27.8285,
      "eval_samples_per_second": 340.155,
      "eval_steps_per_second": 1.33,
      "step": 9500
    },
    {
      "epoch": 1.1638733705772812,
      "grad_norm": 7.371562480926514,
      "learning_rate": 2.9781247419540305e-05,
      "loss": 2.4574,
      "step": 10000
    },
    {
      "epoch": 1.1638733705772812,
      "eval_cosine_accuracy": 0.9001690149307251,
      "eval_loss": 4.775505065917969,
      "eval_runtime": 27.7857,
      "eval_samples_per_second": 340.679,
      "eval_steps_per_second": 1.332,
      "step": 10000
    },
    {
      "epoch": 1.2220670391061452,
      "grad_norm": 8.302326202392578,
      "learning_rate": 2.9598369954554405e-05,
      "loss": 2.44,
      "step": 10500
    },
    {
      "epoch": 1.2220670391061452,
      "eval_cosine_accuracy": 0.895415186882019,
      "eval_loss": 4.801342010498047,
      "eval_runtime": 27.4455,
      "eval_samples_per_second": 344.902,
      "eval_steps_per_second": 1.348,
      "step": 10500
    },
    {
      "epoch": 1.2802607076350094,
      "grad_norm": 7.892956733703613,
      "learning_rate": 2.936129926675645e-05,
      "loss": 2.3956,
      "step": 11000
    },
    {
      "epoch": 1.2802607076350094,
      "eval_cosine_accuracy": 0.9016479849815369,
      "eval_loss": 4.581014156341553,
      "eval_runtime": 27.82,
      "eval_samples_per_second": 340.259,
      "eval_steps_per_second": 1.33,
      "step": 11000
    },
    {
      "epoch": 1.3384543761638734,
      "grad_norm": 7.547521114349365,
      "learning_rate": 2.9070915428691924e-05,
      "loss": 2.3614,
      "step": 11500
    },
    {
      "epoch": 1.3384543761638734,
      "eval_cosine_accuracy": 0.8949925899505615,
      "eval_loss": 4.6524434089660645,
      "eval_runtime": 27.6717,
      "eval_samples_per_second": 342.082,
      "eval_steps_per_second": 1.337,
      "step": 11500
    },
    {
      "epoch": 1.3966480446927374,
      "grad_norm": 7.902958393096924,
      "learning_rate": 2.872829642619667e-05,
      "loss": 2.3639,
      "step": 12000
    },
    {
      "epoch": 1.3966480446927374,
      "eval_cosine_accuracy": 0.8828438520431519,
      "eval_loss": 4.835806369781494,
      "eval_runtime": 27.6782,
      "eval_samples_per_second": 342.002,
      "eval_steps_per_second": 1.337,
      "step": 12000
    },
    {
      "epoch": 1.4548417132216014,
      "grad_norm": 7.324765205383301,
      "learning_rate": 2.833555124233596e-05,
      "loss": 2.3222,
      "step": 12500
    },
    {
      "epoch": 1.4548417132216014,
      "eval_cosine_accuracy": 0.8947813510894775,
      "eval_loss": 4.768213272094727,
      "eval_runtime": 27.5948,
      "eval_samples_per_second": 343.035,
      "eval_steps_per_second": 1.341,
      "step": 12500
    },
    {
      "epoch": 1.5130353817504656,
      "grad_norm": 7.808898448944092,
      "learning_rate": 2.7892564215952825e-05,
      "loss": 2.2942,
      "step": 13000
    },
    {
      "epoch": 1.5130353817504656,
      "eval_cosine_accuracy": 0.8974223732948303,
      "eval_loss": 4.828193664550781,
      "eval_runtime": 27.5674,
      "eval_samples_per_second": 343.377,
      "eval_steps_per_second": 1.342,
      "step": 13000
    },
    {
      "epoch": 1.5712290502793296,
      "grad_norm": 6.751107692718506,
      "learning_rate": 2.7401716393648702e-05,
      "loss": 2.2723,
      "step": 13500
    },
    {
      "epoch": 1.5712290502793296,
      "eval_cosine_accuracy": 0.8914008140563965,
      "eval_loss": 4.6300153732299805,
      "eval_runtime": 27.5296,
      "eval_samples_per_second": 343.848,
      "eval_steps_per_second": 1.344,
      "step": 13500
    },
    {
      "epoch": 1.6294227188081938,
      "grad_norm": 7.299578666687012,
      "learning_rate": 2.6864829939506563e-05,
      "loss": 2.2473,
      "step": 14000
    },
    {
      "epoch": 1.6294227188081938,
      "eval_cosine_accuracy": 0.892034649848938,
      "eval_loss": 4.730409145355225,
      "eval_runtime": 27.5527,
      "eval_samples_per_second": 343.56,
      "eval_steps_per_second": 1.343,
      "step": 14000
    },
    {
      "epoch": 1.6876163873370578,
      "grad_norm": 6.653262138366699,
      "learning_rate": 2.6283897925857392e-05,
      "loss": 2.236,
      "step": 14500
    },
    {
      "epoch": 1.6876163873370578,
      "eval_cosine_accuracy": 0.9077752232551575,
      "eval_loss": 4.478672981262207,
      "eval_runtime": 27.5484,
      "eval_samples_per_second": 343.613,
      "eval_steps_per_second": 1.343,
      "step": 14500
    },
    {
      "epoch": 1.7458100558659218,
      "grad_norm": 7.958475112915039,
      "learning_rate": 2.566107693443959e-05,
      "loss": 2.2211,
      "step": 15000
    },
    {
      "epoch": 1.7458100558659218,
      "eval_cosine_accuracy": 0.8841115832328796,
      "eval_loss": 4.705593109130859,
      "eval_runtime": 27.838,
      "eval_samples_per_second": 340.039,
      "eval_steps_per_second": 1.329,
      "step": 15000
    },
    {
      "epoch": 1.8040037243947857,
      "grad_norm": 6.931536674499512,
      "learning_rate": 2.499867905056577e-05,
      "loss": 2.2015,
      "step": 15500
    },
    {
      "epoch": 1.8040037243947857,
      "eval_cosine_accuracy": 0.8960490226745605,
      "eval_loss": 4.561342716217041,
      "eval_runtime": 27.6406,
      "eval_samples_per_second": 342.467,
      "eval_steps_per_second": 1.339,
      "step": 15500
    },
    {
      "epoch": 1.86219739292365,
      "grad_norm": 6.317222595214844,
      "learning_rate": 2.4300597651196786e-05,
      "loss": 2.1922,
      "step": 16000
    },
    {
      "epoch": 1.86219739292365,
      "eval_cosine_accuracy": 0.9019649028778076,
      "eval_loss": 4.534769535064697,
      "eval_runtime": 27.5599,
      "eval_samples_per_second": 343.47,
      "eval_steps_per_second": 1.343,
      "step": 16000
    },
    {
      "epoch": 1.920391061452514,
      "grad_norm": 5.440205097198486,
      "learning_rate": 2.3566627155448065e-05,
      "loss": 2.1537,
      "step": 16500
    },
    {
      "epoch": 1.920391061452514,
      "eval_cosine_accuracy": 0.913057267665863,
      "eval_loss": 4.725633144378662,
      "eval_runtime": 27.6819,
      "eval_samples_per_second": 341.957,
      "eval_steps_per_second": 1.337,
      "step": 16500
    },
    {
      "epoch": 1.9785847299813781,
      "grad_norm": 6.925086498260498,
      "learning_rate": 2.2800854949259868e-05,
      "loss": 2.1493,
      "step": 17000
    },
    {
      "epoch": 1.9785847299813781,
      "eval_cosine_accuracy": 0.8949925899505615,
      "eval_loss": 4.73307991027832,
      "eval_runtime": 27.5609,
      "eval_samples_per_second": 343.457,
      "eval_steps_per_second": 1.342,
      "step": 17000
    },
    {
      "epoch": 2.036778398510242,
      "grad_norm": 9.615991592407227,
      "learning_rate": 2.200612379274325e-05,
      "loss": 2.12,
      "step": 17500
    },
    {
      "epoch": 2.036778398510242,
      "eval_cosine_accuracy": 0.8875977396965027,
      "eval_loss": 4.846031188964844,
      "eval_runtime": 27.5584,
      "eval_samples_per_second": 343.488,
      "eval_steps_per_second": 1.343,
      "step": 17500
    },
    {
      "epoch": 2.094972067039106,
      "grad_norm": 6.895238399505615,
      "learning_rate": 2.1185383949713617e-05,
      "loss": 2.0944,
      "step": 18000
    },
    {
      "epoch": 2.094972067039106,
      "eval_cosine_accuracy": 0.8969997763633728,
      "eval_loss": 4.649144649505615,
      "eval_runtime": 30.2381,
      "eval_samples_per_second": 313.049,
      "eval_steps_per_second": 1.224,
      "step": 18000
    },
    {
      "epoch": 2.15316573556797,
      "grad_norm": 7.440585136413574,
      "learning_rate": 2.03416822354882e-05,
      "loss": 2.093,
      "step": 18500
    },
    {
      "epoch": 2.15316573556797,
      "eval_cosine_accuracy": 0.8948869705200195,
      "eval_loss": 4.538448810577393,
      "eval_runtime": 27.8445,
      "eval_samples_per_second": 339.96,
      "eval_steps_per_second": 1.329,
      "step": 18500
    },
    {
      "epoch": 2.211359404096834,
      "grad_norm": 7.8712639808654785,
      "learning_rate": 1.9479895439841827e-05,
      "loss": 2.0801,
      "step": 19000
    },
    {
      "epoch": 2.211359404096834,
      "eval_cosine_accuracy": 0.9011197686195374,
      "eval_loss": 4.58160400390625,
      "eval_runtime": 27.5957,
      "eval_samples_per_second": 343.024,
      "eval_steps_per_second": 1.341,
      "step": 19000
    },
    {
      "epoch": 2.2695530726256985,
      "grad_norm": 9.599085807800293,
      "learning_rate": 1.859976977160929e-05,
      "loss": 2.081,
      "step": 19500
    },
    {
      "epoch": 2.2695530726256985,
      "eval_cosine_accuracy": 0.9015423655509949,
      "eval_loss": 4.580519199371338,
      "eval_runtime": 27.8327,
      "eval_samples_per_second": 340.103,
      "eval_steps_per_second": 1.329,
      "step": 19500
    },
    {
      "epoch": 2.3277467411545625,
      "grad_norm": 7.2574238777160645,
      "learning_rate": 1.770628075350192e-05,
      "loss": 2.0818,
      "step": 20000
    },
    {
      "epoch": 2.3277467411545625,
      "eval_cosine_accuracy": 0.8975279927253723,
      "eval_loss": 4.5440673828125,
      "eval_runtime": 27.5922,
      "eval_samples_per_second": 343.068,
      "eval_steps_per_second": 1.341,
      "step": 20000
    },
    {
      "epoch": 2.3859404096834265,
      "grad_norm": 7.82363748550415,
      "learning_rate": 1.6802745266074818e-05,
      "loss": 2.0571,
      "step": 20500
    },
    {
      "epoch": 2.3859404096834265,
      "eval_cosine_accuracy": 0.8922459483146667,
      "eval_loss": 4.727934837341309,
      "eval_runtime": 27.6474,
      "eval_samples_per_second": 342.383,
      "eval_steps_per_second": 1.338,
      "step": 20500
    },
    {
      "epoch": 2.4441340782122905,
      "grad_norm": 6.985836982727051,
      "learning_rate": 1.5892517485180574e-05,
      "loss": 2.0463,
      "step": 21000
    },
    {
      "epoch": 2.4441340782122905,
      "eval_cosine_accuracy": 0.9073526263237,
      "eval_loss": 4.589358806610107,
      "eval_runtime": 27.4789,
      "eval_samples_per_second": 344.483,
      "eval_steps_per_second": 1.346,
      "step": 21000
    },
    {
      "epoch": 2.5023277467411544,
      "grad_norm": 7.737358093261719,
      "learning_rate": 1.4978976430332556e-05,
      "loss": 2.0326,
      "step": 21500
    },
    {
      "epoch": 2.5023277467411544,
      "eval_cosine_accuracy": 0.8977392911911011,
      "eval_loss": 4.702400207519531,
      "eval_runtime": 27.6834,
      "eval_samples_per_second": 341.937,
      "eval_steps_per_second": 1.337,
      "step": 21500
    },
    {
      "epoch": 2.560521415270019,
      "grad_norm": 6.782015800476074,
      "learning_rate": 1.4065513420841569e-05,
      "loss": 2.0332,
      "step": 22000
    },
    {
      "epoch": 2.560521415270019,
      "eval_cosine_accuracy": 0.9057680368423462,
      "eval_loss": 4.601651668548584,
      "eval_runtime": 27.6133,
      "eval_samples_per_second": 342.806,
      "eval_steps_per_second": 1.34,
      "step": 22000
    },
    {
      "epoch": 2.618715083798883,
      "grad_norm": 7.085362911224365,
      "learning_rate": 1.3155519486292301e-05,
      "loss": 2.0155,
      "step": 22500
    },
    {
      "epoch": 2.618715083798883,
      "eval_cosine_accuracy": 0.9074582457542419,
      "eval_loss": 4.483080863952637,
      "eval_runtime": 27.8124,
      "eval_samples_per_second": 340.352,
      "eval_steps_per_second": 1.33,
      "step": 22500
    },
    {
      "epoch": 2.676908752327747,
      "grad_norm": 6.544814109802246,
      "learning_rate": 1.2254170004046857e-05,
      "loss": 2.0111,
      "step": 23000
    },
    {
      "epoch": 2.676908752327747,
      "eval_cosine_accuracy": 0.8939362168312073,
      "eval_loss": 4.641151428222656,
      "eval_runtime": 27.6114,
      "eval_samples_per_second": 342.829,
      "eval_steps_per_second": 1.34,
      "step": 23000
    },
    {
      "epoch": 2.735102420856611,
      "grad_norm": 6.633194446563721,
      "learning_rate": 1.1361199531960475e-05,
      "loss": 2.0176,
      "step": 23500
    },
    {
      "epoch": 2.735102420856611,
      "eval_cosine_accuracy": 0.8949925899505615,
      "eval_loss": 4.60220193862915,
      "eval_runtime": 27.6158,
      "eval_samples_per_second": 342.775,
      "eval_steps_per_second": 1.34,
      "step": 23500
    },
    {
      "epoch": 2.793296089385475,
      "grad_norm": 8.512932777404785,
      "learning_rate": 1.0481737302585223e-05,
      "loss": 2.0028,
      "step": 24000
    },
    {
      "epoch": 2.793296089385475,
      "eval_cosine_accuracy": 0.8974223732948303,
      "eval_loss": 4.529559135437012,
      "eval_runtime": 27.968,
      "eval_samples_per_second": 338.459,
      "eval_steps_per_second": 1.323,
      "step": 24000
    },
    {
      "epoch": 2.851489757914339,
      "grad_norm": 7.090054988861084,
      "learning_rate": 9.619048125122396e-06,
      "loss": 1.9965,
      "step": 24500
    },
    {
      "epoch": 2.851489757914339,
      "eval_cosine_accuracy": 0.8923515677452087,
      "eval_loss": 4.752791881561279,
      "eval_runtime": 27.5418,
      "eval_samples_per_second": 343.696,
      "eval_steps_per_second": 1.343,
      "step": 24500
    },
    {
      "epoch": 2.9096834264432028,
      "grad_norm": 8.278507232666016,
      "learning_rate": 8.777997941190181e-06,
      "loss": 2.0011,
      "step": 25000
    },
    {
      "epoch": 2.9096834264432028,
      "eval_cosine_accuracy": 0.8986900448799133,
      "eval_loss": 4.698167324066162,
      "eval_runtime": 27.6911,
      "eval_samples_per_second": 341.842,
      "eval_steps_per_second": 1.336,
      "step": 25000
    },
    {
      "epoch": 2.967877094972067,
      "grad_norm": 7.455854892730713,
      "learning_rate": 7.958339066434103e-06,
      "loss": 1.9851,
      "step": 25500
    },
    {
      "epoch": 2.967877094972067,
      "eval_cosine_accuracy": 0.8960490226745605,
      "eval_loss": 4.680306434631348,
      "eval_runtime": 27.9245,
      "eval_samples_per_second": 338.986,
      "eval_steps_per_second": 1.325,
      "step": 25500
    },
    {
      "epoch": 3.026070763500931,
      "grad_norm": 6.495429039001465,
      "learning_rate": 7.164820801974087e-06,
      "loss": 1.9658,
      "step": 26000
    },
    {
      "epoch": 3.026070763500931,
      "eval_cosine_accuracy": 0.8993238806724548,
      "eval_loss": 4.683339595794678,
      "eval_runtime": 27.8555,
      "eval_samples_per_second": 339.825,
      "eval_steps_per_second": 1.328,
      "step": 26000
    },
    {
      "epoch": 3.084264432029795,
      "grad_norm": 8.352217674255371,
      "learning_rate": 6.400388909031767e-06,
      "loss": 1.9651,
      "step": 26500
    },
    {
      "epoch": 3.084264432029795,
      "eval_cosine_accuracy": 0.896682858467102,
      "eval_loss": 4.631614685058594,
      "eval_runtime": 27.6286,
      "eval_samples_per_second": 342.616,
      "eval_steps_per_second": 1.339,
      "step": 26500
    },
    {
      "epoch": 3.142458100558659,
      "grad_norm": 7.33491325378418,
      "learning_rate": 5.669312499098547e-06,
      "loss": 1.9609,
      "step": 27000
    },
    {
      "epoch": 3.142458100558659,
      "eval_cosine_accuracy": 0.8975279927253723,
      "eval_loss": 4.696747779846191,
      "eval_runtime": 27.5958,
      "eval_samples_per_second": 343.023,
      "eval_steps_per_second": 1.341,
      "step": 27000
    },
    {
      "epoch": 3.200651769087523,
      "grad_norm": 7.817756652832031,
      "learning_rate": 4.9727358335650156e-06,
      "loss": 1.9588,
      "step": 27500
    },
    {
      "epoch": 3.200651769087523,
      "eval_cosine_accuracy": 0.896682858467102,
      "eval_loss": 4.715923309326172,
      "eval_runtime": 27.6817,
      "eval_samples_per_second": 341.958,
      "eval_steps_per_second": 1.337,
      "step": 27500
    },
    {
      "epoch": 3.2588454376163876,
      "grad_norm": 6.541388034820557,
      "learning_rate": 4.3119517050406135e-06,
      "loss": 1.958,
      "step": 28000
    },
    {
      "epoch": 3.2588454376163876,
      "eval_cosine_accuracy": 0.9014367461204529,
      "eval_loss": 4.600804328918457,
      "eval_runtime": 27.8473,
      "eval_samples_per_second": 339.925,
      "eval_steps_per_second": 1.329,
      "step": 28000
    },
    {
      "epoch": 3.3170391061452515,
      "grad_norm": 7.466308116912842,
      "learning_rate": 3.6908445941885822e-06,
      "loss": 1.9494,
      "step": 28500
    },
    {
      "epoch": 3.3170391061452515,
      "eval_cosine_accuracy": 0.9018592834472656,
      "eval_loss": 4.584753036499023,
      "eval_runtime": 27.8346,
      "eval_samples_per_second": 340.08,
      "eval_steps_per_second": 1.329,
      "step": 28500
    },
    {
      "epoch": 3.3752327746741155,
      "grad_norm": 7.9546990394592285,
      "learning_rate": 3.1117202239186415e-06,
      "loss": 1.9541,
      "step": 29000
    },
    {
      "epoch": 3.3752327746741155,
      "eval_cosine_accuracy": 0.9040777683258057,
      "eval_loss": 4.553245544433594,
      "eval_runtime": 27.7545,
      "eval_samples_per_second": 341.062,
      "eval_steps_per_second": 1.333,
      "step": 29000
    },
    {
      "epoch": 3.4334264432029795,
      "grad_norm": 6.409430027008057,
      "learning_rate": 2.5767284654925557e-06,
      "loss": 1.9563,
      "step": 29500
    },
    {
      "epoch": 3.4334264432029795,
      "eval_cosine_accuracy": 0.9052398204803467,
      "eval_loss": 4.594734191894531,
      "eval_runtime": 27.8799,
      "eval_samples_per_second": 339.528,
      "eval_steps_per_second": 1.327,
      "step": 29500
    },
    {
      "epoch": 3.4916201117318435,
      "grad_norm": 6.676296234130859,
      "learning_rate": 2.087855357602124e-06,
      "loss": 1.9593,
      "step": 30000
    },
    {
      "epoch": 3.4916201117318435,
      "eval_cosine_accuracy": 0.9071413278579712,
      "eval_loss": 4.592144012451172,
      "eval_runtime": 27.8717,
      "eval_samples_per_second": 339.628,
      "eval_steps_per_second": 1.328,
      "step": 30000
    },
    {
      "epoch": 3.5498137802607075,
      "grad_norm": 6.646908283233643,
      "learning_rate": 1.6469157336393776e-06,
      "loss": 1.9416,
      "step": 30500
    },
    {
      "epoch": 3.5498137802607075,
      "eval_cosine_accuracy": 0.9049229025840759,
      "eval_loss": 4.61123514175415,
      "eval_runtime": 28.163,
      "eval_samples_per_second": 336.114,
      "eval_steps_per_second": 1.314,
      "step": 30500
    },
    {
      "epoch": 3.6080074487895715,
      "grad_norm": 6.432226657867432,
      "learning_rate": 1.2555464845286114e-06,
      "loss": 1.9653,
      "step": 31000
    },
    {
      "epoch": 3.6080074487895715,
      "eval_cosine_accuracy": 0.9083033800125122,
      "eval_loss": 4.575016498565674,
      "eval_runtime": 27.7872,
      "eval_samples_per_second": 340.661,
      "eval_steps_per_second": 1.332,
      "step": 31000
    },
    {
      "epoch": 3.666201117318436,
      "grad_norm": 6.147933006286621,
      "learning_rate": 9.15829381125019e-07,
      "loss": 1.9563,
      "step": 31500
    },
    {
      "epoch": 3.666201117318436,
      "eval_cosine_accuracy": 0.9043946862220764,
      "eval_loss": 4.594618320465088,
      "eval_runtime": 27.8051,
      "eval_samples_per_second": 340.441,
      "eval_steps_per_second": 1.331,
      "step": 31500
    },
    {
      "epoch": 3.7243947858473,
      "grad_norm": 6.538316249847412,
      "learning_rate": 6.276643787308245e-07,
      "loss": 1.9507,
      "step": 32000
    },
    {
      "epoch": 3.7243947858473,
      "eval_cosine_accuracy": 0.9060849547386169,
      "eval_loss": 4.5688347816467285,
      "eval_runtime": 27.7627,
      "eval_samples_per_second": 340.961,
      "eval_steps_per_second": 1.333,
      "step": 32000
    },
    {
      "epoch": 3.782588454376164,
      "grad_norm": 7.255917549133301,
      "learning_rate": 3.928534966307357e-07,
      "loss": 1.9378,
      "step": 32500
    },
    {
      "epoch": 3.782588454376164,
      "eval_cosine_accuracy": 0.9062961935997009,
      "eval_loss": 4.608202934265137,
      "eval_runtime": 27.7075,
      "eval_samples_per_second": 341.641,
      "eval_steps_per_second": 1.335,
      "step": 32500
    },
    {
      "epoch": 3.840782122905028,
      "grad_norm": 7.074450969696045,
      "learning_rate": 2.1257499177363416e-07,
      "loss": 1.9494,
      "step": 33000
    },
    {
      "epoch": 3.840782122905028,
      "eval_cosine_accuracy": 0.9059792757034302,
      "eval_loss": 4.594712734222412,
      "eval_runtime": 27.8802,
      "eval_samples_per_second": 339.524,
      "eval_steps_per_second": 1.327,
      "step": 33000
    },
    {
      "epoch": 3.898975791433892,
      "grad_norm": 6.569212436676025,
      "learning_rate": 8.677577319543827e-08,
      "loss": 1.9424,
      "step": 33500
    },
    {
      "epoch": 3.898975791433892,
      "eval_cosine_accuracy": 0.9069300889968872,
      "eval_loss": 4.600465297698975,
      "eval_runtime": 27.843,
      "eval_samples_per_second": 339.977,
      "eval_steps_per_second": 1.329,
      "step": 33500
    },
    {
      "epoch": 3.9571694599627563,
      "grad_norm": 6.682833194732666,
      "learning_rate": 1.6338604345084162e-08,
      "loss": 1.9399,
      "step": 34000
    },
    {
      "epoch": 3.9571694599627563,
      "eval_cosine_accuracy": 0.9064018726348877,
      "eval_loss": 4.5989670753479,
      "eval_runtime": 28.0362,
      "eval_samples_per_second": 337.635,
      "eval_steps_per_second": 1.32,
      "step": 34000
    }
  ],
  "logging_steps": 500,
  "max_steps": 34368,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 4,
  "save_steps": 500,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": true
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 256,
  "trial_name": null,
  "trial_params": null
}