{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.6621403168444876,
  "eval_steps": 1024,
  "global_step": 14336,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.011823934229365849,
      "grad_norm": 5.310993194580078,
      "learning_rate": 0.000498046875,
      "loss": 22.2890625,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 7.661691188812256,
      "learning_rate": 0.000998046875,
      "loss": 12.93233871459961,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 2.532165050506592,
      "learning_rate": 0.000999640996023194,
      "loss": 9.235491752624512,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 3.695277214050293,
      "learning_rate": 0.0009985588674043958,
      "loss": 8.26991081237793,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_ce_loss": 6.770938226621445,
      "eval_cos_loss": 0.5271919946011887,
      "eval_loss": 8.123252781558799,
      "eval_mse_loss": 1.3523145494940074,
      "flow/cos_sim": 0.47280800812048457,
      "flow/improvement_ratio": 0.9578989227068478,
      "flow/mag_ratio_mean": 0.514818922025428,
      "flow/mag_ratio_std": 0.19883525225261575,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_ce_loss": 6.770938226621445,
      "eval_cos_loss": 0.5271919946011887,
      "eval_loss": 8.123252781558799,
      "eval_mse_loss": 1.3523145494940074,
      "eval_runtime": 64.4558,
      "eval_samples_per_second": 434.297,
      "eval_steps_per_second": 6.795,
      "flow/cos_sim": 0.47280800812048457,
      "flow/improvement_ratio": 0.9578989227068478,
      "flow/mag_ratio_mean": 0.514818922025428,
      "flow/mag_ratio_std": 0.19883525225261575,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 2.314648389816284,
      "learning_rate": 0.0009967551747861387,
      "loss": 8.022788047790527,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 1.662047266960144,
      "learning_rate": 0.000994232528651847,
      "loss": 7.858725070953369,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 1.137885570526123,
      "learning_rate": 0.0009909945800260092,
      "loss": 7.710772514343262,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 1.100211262702942,
      "learning_rate": 0.0009870460151900522,
      "loss": 7.598612308502197,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_ce_loss": 6.6824955722512716,
      "eval_cos_loss": 0.39051488293632525,
      "eval_loss": 7.547230463594063,
      "eval_mse_loss": 0.864734881272599,
      "flow/cos_sim": 0.6094851142739597,
      "flow/improvement_ratio": 0.9694367877152412,
      "flow/mag_ratio_mean": 0.6239872499960198,
      "flow/mag_ratio_std": 0.20368898976339053,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_ce_loss": 6.6824955722512716,
      "eval_cos_loss": 0.39051488293632525,
      "eval_loss": 7.547230463594063,
      "eval_mse_loss": 0.864734881272599,
      "eval_runtime": 64.6729,
      "eval_samples_per_second": 432.84,
      "eval_steps_per_second": 6.773,
      "flow/cos_sim": 0.6094851142739597,
      "flow/improvement_ratio": 0.9694367877152412,
      "flow/mag_ratio_mean": 0.6239872499960198,
      "flow/mag_ratio_std": 0.20368898976339053,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 1.3531696796417236,
      "learning_rate": 0.0009823925488998885,
      "loss": 7.5045647621154785,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.7993828654289246,
      "learning_rate": 0.0009770409161149525,
      "loss": 7.426115989685059,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 0.678104043006897,
      "learning_rate": 0.0009709988622506973,
      "loss": 7.355522632598877,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.5943172574043274,
      "learning_rate": 0.000964275131968659,
      "loss": 7.290462493896484,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_ce_loss": 6.662393192178038,
      "eval_cos_loss": 0.2890751386001774,
      "eval_loss": 7.268133973422116,
      "eval_mse_loss": 0.6057407744398945,
      "flow/cos_sim": 0.710924859188463,
      "flow/improvement_ratio": 0.9773364528401257,
      "flow/mag_ratio_mean": 0.724724873288037,
      "flow/mag_ratio_std": 0.22310954716788034,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_ce_loss": 6.662393192178038,
      "eval_cos_loss": 0.2890751386001774,
      "eval_loss": 7.268133973422116,
      "eval_mse_loss": 0.6057407744398945,
      "eval_runtime": 64.5435,
      "eval_samples_per_second": 433.708,
      "eval_steps_per_second": 6.786,
      "flow/cos_sim": 0.710924859188463,
      "flow/improvement_ratio": 0.9773364528401257,
      "flow/mag_ratio_mean": 0.724724873288037,
      "flow/mag_ratio_std": 0.22310954716788034,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 1.2431687116622925,
      "learning_rate": 0.0009568794565203123,
      "loss": 7.288367748260498,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.9812341928482056,
      "learning_rate": 0.0009488225396630347,
      "loss": 7.193031311035156,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 0.46874845027923584,
      "learning_rate": 0.0009401160421685646,
      "loss": 7.153883457183838,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.47678303718566895,
      "learning_rate": 0.0009307725649463714,
      "loss": 7.12277889251709,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_ce_loss": 6.650159499416612,
      "eval_cos_loss": 0.22240412983720162,
      "eval_loss": 7.105518678551939,
      "eval_mse_loss": 0.4553591773662393,
      "flow/cos_sim": 0.7775958679854598,
      "flow/improvement_ratio": 0.9806619151814343,
      "flow/mag_ratio_mean": 0.7837155504858113,
      "flow/mag_ratio_std": 0.21672996865015595,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_ce_loss": 6.650159499416612,
      "eval_cos_loss": 0.22240412983720162,
      "eval_loss": 7.105518678551939,
      "eval_mse_loss": 0.4553591773662393,
      "eval_runtime": 63.9268,
      "eval_samples_per_second": 437.891,
      "eval_steps_per_second": 6.852,
      "flow/cos_sim": 0.7775958679854598,
      "flow/improvement_ratio": 0.9806619151814343,
      "flow/mag_ratio_mean": 0.7837155504858113,
      "flow/mag_ratio_std": 0.21672996865015595,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 0.6425520181655884,
      "learning_rate": 0.0009208056308063659,
      "loss": 7.0932536125183105,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 1.3422183990478516,
      "learning_rate": 0.0009102296648873445,
      "loss": 7.0613112449646,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 0.5829691290855408,
      "learning_rate": 0.0008990599737794927,
      "loss": 7.035637855529785,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.5854812860488892,
      "learning_rate": 0.0008873127233711644,
      "loss": 7.0129170417785645,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_ce_loss": 6.647287761784035,
      "eval_cos_loss": 0.1774802339267513,
      "eval_loss": 7.006625726342746,
      "eval_mse_loss": 0.3593379625174553,
      "flow/cos_sim": 0.8225197658691232,
      "flow/improvement_ratio": 0.9842402499560352,
      "flow/mag_ratio_mean": 0.8389692129609791,
      "flow/mag_ratio_std": 0.2145905472576346,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_ce_loss": 6.647287761784035,
      "eval_cos_loss": 0.1774802339267513,
      "eval_loss": 7.006625726342746,
      "eval_mse_loss": 0.3593379625174553,
      "eval_runtime": 65.6903,
      "eval_samples_per_second": 426.136,
      "eval_steps_per_second": 6.668,
      "flow/cos_sim": 0.8225197658691232,
      "flow/improvement_ratio": 0.9842402499560352,
      "flow/mag_ratio_mean": 0.8389692129609791,
      "flow/mag_ratio_std": 0.2145905472576346,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 0.7321183085441589,
      "learning_rate": 0.0008750049154520011,
      "loss": 6.990157604217529,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.5175312757492065,
      "learning_rate": 0.0008621543631062487,
      "loss": 6.97371244430542,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.7752848863601685,
      "learning_rate": 0.0008487796649318904,
      "loss": 6.963197708129883,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.5458380579948425,
      "learning_rate": 0.0008349001781229053,
      "loss": 6.956610679626465,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_ce_loss": 6.64574583806948,
      "eval_cos_loss": 0.14618989437410276,
      "eval_loss": 6.940428107840829,
      "eval_mse_loss": 0.2946822723569391,
      "flow/cos_sim": 0.8538101044691861,
      "flow/improvement_ratio": 0.9873136728321581,
      "flow/mag_ratio_mean": 0.8605127470678391,
      "flow/mag_ratio_std": 0.1925806637223997,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_ce_loss": 6.64574583806948,
      "eval_cos_loss": 0.14618989437410276,
      "eval_loss": 6.940428107840829,
      "eval_mse_loss": 0.2946822723569391,
      "eval_runtime": 65.5471,
      "eval_samples_per_second": 427.067,
      "eval_steps_per_second": 6.682,
      "flow/cos_sim": 0.8538101044691861,
      "flow/improvement_ratio": 0.9873136728321581,
      "flow/mag_ratio_mean": 0.8605127470678391,
      "flow/mag_ratio_std": 0.1925806637223997,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 0.437857985496521,
      "learning_rate": 0.0008205359904536107,
      "loss": 6.932642936706543,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 0.6009621024131775,
      "learning_rate": 0.0008057078912056363,
      "loss": 6.932676315307617,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.4097861051559448,
      "learning_rate": 0.0007904373410796086,
      "loss": 6.922003746032715,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.54083251953125,
      "learning_rate": 0.0007747464411350876,
      "loss": 6.903475284576416,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_ce_loss": 6.644999326636258,
      "eval_cos_loss": 0.12948155590252244,
      "eval_loss": 6.905599483071941,
      "eval_mse_loss": 0.26060015364596834,
      "flow/cos_sim": 0.87051844270262,
      "flow/improvement_ratio": 0.9893111555271497,
      "flow/mag_ratio_mean": 0.8712390643548747,
      "flow/mag_ratio_std": 0.19501862627322272,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_ce_loss": 6.644999326636258,
      "eval_cos_loss": 0.12948155590252244,
      "eval_loss": 6.905599483071941,
      "eval_mse_loss": 0.26060015364596834,
      "eval_runtime": 64.508,
      "eval_samples_per_second": 433.946,
      "eval_steps_per_second": 6.79,
      "flow/cos_sim": 0.87051844270262,
      "flow/improvement_ratio": 0.9893111555271497,
      "flow/mag_ratio_mean": 0.8712390643548747,
      "flow/mag_ratio_std": 0.19501862627322272,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 0.5095380544662476,
      "learning_rate": 0.000758657900803716,
      "loss": 6.9005961418151855,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 0.341415673494339,
      "learning_rate": 0.000742195005021869,
      "loss": 6.8952412605285645,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.2902733385562897,
      "learning_rate": 0.0007253815805303786,
      "loss": 6.8849945068359375,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.5425611138343811,
      "learning_rate": 0.0007082419613901028,
      "loss": 6.8753275871276855,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_ce_loss": 6.643324745300154,
      "eval_cos_loss": 0.11695763337761844,
      "eval_loss": 6.877702439756698,
      "eval_mse_loss": 0.23437770422054752,
      "flow/cos_sim": 0.8830423678981659,
      "flow/improvement_ratio": 0.9898773733883688,
      "flow/mag_ratio_mean": 0.8962584248144333,
      "flow/mag_ratio_std": 0.18067260168186605,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_ce_loss": 6.643324745300154,
      "eval_cos_loss": 0.11695763337761844,
      "eval_loss": 6.877702439756698,
      "eval_mse_loss": 0.23437770422054752,
      "eval_runtime": 64.3325,
      "eval_samples_per_second": 435.13,
      "eval_steps_per_second": 6.808,
      "flow/cos_sim": 0.8830423678981659,
      "flow/improvement_ratio": 0.9898773733883688,
      "flow/mag_ratio_mean": 0.8962584248144333,
      "flow/mag_ratio_std": 0.18067260168186605,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 0.7520615458488464,
      "learning_rate": 0.0006908009537632514,
      "loss": 6.879398345947266,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.5208486914634705,
      "learning_rate": 0.0006730838000114403,
      "loss": 6.873098373413086,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 0.43452462553977966,
      "learning_rate": 0.0006551161421624341,
      "loss": 6.862982749938965,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.45747169852256775,
      "learning_rate": 0.0006369239847984517,
      "loss": 6.855817794799805,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_ce_loss": 6.643397864685755,
      "eval_cos_loss": 0.1082146630099375,
      "eval_loss": 6.860041695642689,
      "eval_mse_loss": 0.21664383299818865,
      "flow/cos_sim": 0.8917853384529619,
      "flow/improvement_ratio": 0.990278246740228,
      "flow/mag_ratio_mean": 0.8969191525080432,
      "flow/mag_ratio_std": 0.17551493791004294,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_ce_loss": 6.643397864685755,
      "eval_cos_loss": 0.1082146630099375,
      "eval_loss": 6.860041695642689,
      "eval_mse_loss": 0.21664383299818865,
      "eval_runtime": 64.562,
      "eval_samples_per_second": 433.583,
      "eval_steps_per_second": 6.784,
      "flow/cos_sim": 0.8917853384529619,
      "flow/improvement_ratio": 0.990278246740228,
      "flow/mag_ratio_mean": 0.8969191525080432,
      "flow/mag_ratio_std": 0.17551493791004294,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 0.529183030128479,
      "learning_rate": 0.0006185336574197479,
      "loss": 6.849204063415527,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 0.2831822633743286,
      "learning_rate": 0.0005999717763379407,
      "loss": 6.8596038818359375,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 1.0217152833938599,
      "learning_rate": 0.0005812652061542363,
      "loss": 6.847334861755371,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.44654422998428345,
      "learning_rate": 0.0005624410208783071,
      "loss": 6.846036911010742,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_ce_loss": 6.641612795389951,
      "eval_cos_loss": 0.10159792694939326,
      "eval_loss": 6.845016151802725,
      "eval_mse_loss": 0.20340334783950353,
      "flow/cos_sim": 0.8984020736119519,
      "flow/improvement_ratio": 0.9897379034186062,
      "flow/mag_ratio_mean": 0.9109135144921743,
      "flow/mag_ratio_std": 0.16760862588065945,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_ce_loss": 6.641612795389951,
      "eval_cos_loss": 0.10159792694939326,
      "eval_loss": 6.845016151802725,
      "eval_mse_loss": 0.20340334783950353,
      "eval_runtime": 63.9023,
      "eval_samples_per_second": 438.059,
      "eval_steps_per_second": 6.854,
      "flow/cos_sim": 0.8984020736119519,
      "flow/improvement_ratio": 0.9897379034186062,
      "flow/mag_ratio_mean": 0.9109135144921743,
      "flow/mag_ratio_std": 0.16760862588065945,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 0.31733208894729614,
      "learning_rate": 0.0005435264647440881,
      "loss": 6.841334342956543,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 0.7933322191238403,
      "learning_rate": 0.000524548912779213,
      "loss": 6.833290100097656,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 0.273179829120636,
      "learning_rate": 0.0005055358311851499,
      "loss": 6.844526290893555,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 0.3748645782470703,
      "learning_rate": 0.0004865147375853812,
      "loss": 6.834465026855469,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_ce_loss": 6.6411590695925495,
      "eval_cos_loss": 0.09674425246236532,
      "eval_loss": 6.83495632480813,
      "eval_mse_loss": 0.19379725269803175,
      "flow/cos_sim": 0.9032557489665132,
      "flow/improvement_ratio": 0.9913779068483065,
      "flow/mag_ratio_mean": 0.916442300631031,
      "flow/mag_ratio_std": 0.16842460159433487,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_ce_loss": 6.6411590695925495,
      "eval_cos_loss": 0.09674425246236532,
      "eval_loss": 6.83495632480813,
      "eval_mse_loss": 0.19379725269803175,
      "eval_runtime": 63.6545,
      "eval_samples_per_second": 439.765,
      "eval_steps_per_second": 6.881,
      "flow/cos_sim": 0.9032557489665132,
      "flow/improvement_ratio": 0.9913779068483065,
      "flow/mag_ratio_mean": 0.916442300631031,
      "flow/mag_ratio_std": 0.16842460159433487,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 0.3057026267051697,
      "learning_rate": 0.0004675131611991607,
      "loss": 6.823686599731445,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 0.6613107919692993,
      "learning_rate": 0.0004485586029984899,
      "loss": 6.826196670532227,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 0.7032549381256104,
      "learning_rate": 0.00042967849590597266,
      "loss": 6.824607849121094,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.3788175880908966,
      "learning_rate": 0.0004109001650911621,
      "loss": 6.826193332672119,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_ce_loss": 6.639694696147692,
      "eval_cos_loss": 0.0918180431371138,
      "eval_loss": 6.82323066937869,
      "eval_mse_loss": 0.18353597942280442,
      "flow/cos_sim": 0.9081819589551725,
      "flow/improvement_ratio": 0.9904580299984919,
      "flow/mag_ratio_mean": 0.9139516707424704,
      "flow/mag_ratio_std": 0.15966295250202422,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_ce_loss": 6.639694696147692,
      "eval_cos_loss": 0.0918180431371138,
      "eval_loss": 6.82323066937869,
      "eval_mse_loss": 0.18353597942280442,
      "eval_runtime": 63.9479,
      "eval_samples_per_second": 437.747,
      "eval_steps_per_second": 6.849,
      "flow/cos_sim": 0.9081819589551725,
      "flow/improvement_ratio": 0.9904580299984919,
      "flow/mag_ratio_mean": 0.9139516707424704,
      "flow/mag_ratio_std": 0.15966295250202422,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 0.3380263149738312,
      "learning_rate": 0.0003922507884228551,
      "loss": 6.8268632888793945,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 0.3507392704486847,
      "learning_rate": 0.00037375735713457723,
      "loss": 6.814322471618652,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 0.34482619166374207,
      "learning_rate": 0.00035544663676018276,
      "loss": 6.817571640014648,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 0.3506380617618561,
      "learning_rate": 0.00033734512839611255,
      "loss": 6.815121173858643,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_ce_loss": 6.639219018422305,
      "eval_cos_loss": 0.08840062753318652,
      "eval_loss": 6.816446893291387,
      "eval_mse_loss": 0.17722788187738966,
      "flow/cos_sim": 0.9115993724838239,
      "flow/improvement_ratio": 0.9913340507576999,
      "flow/mag_ratio_mean": 0.9105819005127911,
      "flow/mag_ratio_std": 0.15417975639779818,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_ce_loss": 6.639219018422305,
      "eval_cos_loss": 0.08840062753318652,
      "eval_loss": 6.816446893291387,
      "eval_mse_loss": 0.17722788187738966,
      "eval_runtime": 64.1691,
      "eval_samples_per_second": 436.238,
      "eval_steps_per_second": 6.826,
      "flow/cos_sim": 0.9115993724838239,
      "flow/improvement_ratio": 0.9913340507576999,
      "flow/mag_ratio_mean": 0.9105819005127911,
      "flow/mag_ratio_std": 0.15417975639779818,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 0.39763471484184265,
      "learning_rate": 0.0003194790303463687,
      "loss": 6.808745861053467,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 0.2740715444087982,
      "learning_rate": 0.00030187420020572406,
      "loss": 6.8159685134887695,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 0.36115965247154236,
      "learning_rate": 0.00028455611743603626,
      "loss": 6.8075947761535645,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 0.33763498067855835,
      "learning_rate": 0.0002675498464898373,
      "loss": 6.80622673034668,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_ce_loss": 6.63860485216254,
      "eval_cos_loss": 0.08730801206813556,
      "eval_loss": 6.812814682041673,
      "eval_mse_loss": 0.17420982753169045,
      "flow/cos_sim": 0.9126919905344645,
      "flow/improvement_ratio": 0.9905721886789418,
      "flow/mag_ratio_mean": 0.920888199517716,
      "flow/mag_ratio_std": 0.1555312630044271,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_ce_loss": 6.63860485216254,
      "eval_cos_loss": 0.08730801206813556,
      "eval_loss": 6.812814682041673,
      "eval_mse_loss": 0.17420982753169045,
      "eval_runtime": 64.0106,
      "eval_samples_per_second": 437.318,
      "eval_steps_per_second": 6.843,
      "flow/cos_sim": 0.9126919905344645,
      "flow/improvement_ratio": 0.9905721886789418,
      "flow/mag_ratio_mean": 0.920888199517716,
      "flow/mag_ratio_std": 0.1555312630044271,
      "step": 14336
    }
  ],
  "logging_steps": 256,
  "max_steps": 21651,
  "num_input_tokens_seen": 0,
  "num_train_epochs": 1,
  "save_steps": 1024,
  "stateful_callbacks": {
    "TrainerControl": {
      "args": {
        "should_epoch_stop": false,
        "should_evaluate": false,
        "should_log": false,
        "should_save": true,
        "should_training_stop": false
      },
      "attributes": {}
    }
  },
  "total_flos": 0.0,
  "train_batch_size": 64,
  "trial_name": null,
  "trial_params": null
}