{
  "best_global_step": null,
  "best_metric": null,
  "best_model_checkpoint": null,
  "epoch": 0.9932104752667313,
  "eval_steps": 1024,
  "global_step": 21504,
  "is_hyper_param_search": false,
  "is_local_process_zero": true,
  "is_world_process_zero": true,
  "log_history": [
    {
      "epoch": 0.011823934229365849,
      "grad_norm": 0.8164597153663635,
      "learning_rate": 0.000498046875,
      "loss": 9.018741607666016,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 0.7249562740325928,
      "learning_rate": 0.000998046875,
      "loss": 4.788314342498779,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 1.273499608039856,
      "learning_rate": 0.000999640996023194,
      "loss": 3.9647223949432373,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.7055355310440063,
      "learning_rate": 0.0009985588674043958,
      "loss": 3.6936378479003906,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_bleu": 0.5077422295273598,
      "eval_ce_loss": 2.198924214600428,
      "eval_cos_loss": 0.4548168092557829,
      "eval_loss": 3.5576477192308262,
      "eval_mse_loss": 1.3587235043582306,
      "flow/cos_sim": 0.5451831924452629,
      "flow/improvement_ratio": 0.9661927943088148,
      "flow/mag_ratio_mean": 0.5132733742668204,
      "flow/mag_ratio_std": 0.23296540431236024,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_bleu": 0.5077422295273598,
      "eval_ce_loss": 2.198924214600428,
      "eval_cos_loss": 0.4548168092557829,
      "eval_loss": 3.5576477192308262,
      "eval_mse_loss": 1.3587235043582306,
      "eval_runtime": 114.937,
      "eval_samples_per_second": 243.551,
      "eval_steps_per_second": 3.811,
      "flow/cos_sim": 0.5451831924452629,
      "flow/improvement_ratio": 0.9661927943088148,
      "flow/mag_ratio_mean": 0.5132733742668204,
      "flow/mag_ratio_std": 0.23296540431236024,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 0.9717119336128235,
      "learning_rate": 0.0009967551747861387,
      "loss": 3.523725748062134,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 0.9234328866004944,
      "learning_rate": 0.000994232528651847,
      "loss": 3.442486047744751,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 0.9983293414115906,
      "learning_rate": 0.0009909945800260092,
      "loss": 3.3732779026031494,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.9987860918045044,
      "learning_rate": 0.0009870460151900522,
      "loss": 3.3335657119750977,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_bleu": 0.5583098662491817,
      "eval_ce_loss": 1.9428779066425481,
      "eval_cos_loss": 0.4089699534520711,
      "eval_loss": 3.2713746801358923,
      "eval_mse_loss": 1.3284967718603404,
      "flow/cos_sim": 0.5910300453231759,
      "flow/improvement_ratio": 0.9727095938436517,
      "flow/mag_ratio_mean": 0.5550923214791572,
      "flow/mag_ratio_std": 0.2607482071608713,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_bleu": 0.5583098662491817,
      "eval_ce_loss": 1.9428779066425481,
      "eval_cos_loss": 0.4089699534520711,
      "eval_loss": 3.2713746801358923,
      "eval_mse_loss": 1.3284967718603404,
      "eval_runtime": 112.8459,
      "eval_samples_per_second": 248.064,
      "eval_steps_per_second": 3.881,
      "flow/cos_sim": 0.5910300453231759,
      "flow/improvement_ratio": 0.9727095938436517,
      "flow/mag_ratio_mean": 0.5550923214791572,
      "flow/mag_ratio_std": 0.2607482071608713,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 0.8213550448417664,
      "learning_rate": 0.0009823925488998885,
      "loss": 3.3129796981811523,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.973803699016571,
      "learning_rate": 0.0009770409161149525,
      "loss": 3.2495813369750977,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 1.5047074556350708,
      "learning_rate": 0.0009709988622506973,
      "loss": 3.1962387561798096,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.9115265011787415,
      "learning_rate": 0.000964275131968659,
      "loss": 3.1230273246765137,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_bleu": 0.5612210129930764,
      "eval_ce_loss": 1.8150666712625931,
      "eval_cos_loss": 0.38208584085991393,
      "eval_loss": 3.1303975484142565,
      "eval_mse_loss": 1.3153308676258069,
      "flow/cos_sim": 0.6179141614535083,
      "flow/improvement_ratio": 0.9767382955986615,
      "flow/mag_ratio_mean": 0.5887199336807477,
      "flow/mag_ratio_std": 0.2682194593426299,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_bleu": 0.5612210129930764,
      "eval_ce_loss": 1.8150666712625931,
      "eval_cos_loss": 0.38208584085991393,
      "eval_loss": 3.1303975484142565,
      "eval_mse_loss": 1.3153308676258069,
      "eval_runtime": 112.4286,
      "eval_samples_per_second": 248.985,
      "eval_steps_per_second": 3.896,
      "flow/cos_sim": 0.6179141614535083,
      "flow/improvement_ratio": 0.9767382955986615,
      "flow/mag_ratio_mean": 0.5887199336807477,
      "flow/mag_ratio_std": 0.2682194593426299,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 1.048264980316162,
      "learning_rate": 0.0009568794565203123,
      "loss": 3.161437749862671,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.8199158906936646,
      "learning_rate": 0.0009488225396630347,
      "loss": 3.10720157623291,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 1.1004403829574585,
      "learning_rate": 0.0009401160421685646,
      "loss": 3.081758975982666,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.8042743802070618,
      "learning_rate": 0.0009307725649463714,
      "loss": 3.0546185970306396,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_bleu": 0.6482649669737099,
      "eval_ce_loss": 1.7232403153698195,
      "eval_cos_loss": 0.3577009423138344,
      "eval_loss": 3.022103031476339,
      "eval_mse_loss": 1.2988627180116905,
      "flow/cos_sim": 0.6422990591830859,
      "flow/improvement_ratio": 0.9803378113872929,
      "flow/mag_ratio_mean": 0.6228148394251523,
      "flow/mag_ratio_std": 0.2741846599399227,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_bleu": 0.6482649669737099,
      "eval_ce_loss": 1.7232403153698195,
      "eval_cos_loss": 0.3577009423138344,
      "eval_loss": 3.022103031476339,
      "eval_mse_loss": 1.2988627180116905,
      "eval_runtime": 114.314,
      "eval_samples_per_second": 244.878,
      "eval_steps_per_second": 3.832,
      "flow/cos_sim": 0.6422990591830859,
      "flow/improvement_ratio": 0.9803378113872929,
      "flow/mag_ratio_mean": 0.6228148394251523,
      "flow/mag_ratio_std": 0.2741846599399227,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 1.2677597999572754,
      "learning_rate": 0.0009208056308063659,
      "loss": 3.0482373237609863,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 1.2085199356079102,
      "learning_rate": 0.0009102296648873445,
      "loss": 3.047755241394043,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 1.0152058601379395,
      "learning_rate": 0.0008990599737794927,
      "loss": 2.990638494491577,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 1.1251702308654785,
      "learning_rate": 0.0008873127233711644,
      "loss": 3.0137414932250977,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_bleu": 0.5795940220557444,
      "eval_ce_loss": 1.7088007153985707,
      "eval_cos_loss": 0.3385225797080558,
      "eval_loss": 2.9921626473126346,
      "eval_mse_loss": 1.2833619313697293,
      "flow/cos_sim": 0.6614774217888645,
      "flow/improvement_ratio": 0.9821179653411587,
      "flow/mag_ratio_mean": 0.637896000794624,
      "flow/mag_ratio_std": 0.27109318641631025,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_bleu": 0.5795940220557444,
      "eval_ce_loss": 1.7088007153985707,
      "eval_cos_loss": 0.3385225797080558,
      "eval_loss": 2.9921626473126346,
      "eval_mse_loss": 1.2833619313697293,
      "eval_runtime": 117.7245,
      "eval_samples_per_second": 237.784,
      "eval_steps_per_second": 3.721,
      "flow/cos_sim": 0.6614774217888645,
      "flow/improvement_ratio": 0.9821179653411587,
      "flow/mag_ratio_mean": 0.637896000794624,
      "flow/mag_ratio_std": 0.27109318641631025,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 1.3458431959152222,
      "learning_rate": 0.0008750049154520011,
      "loss": 2.954760789871216,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.9947335124015808,
      "learning_rate": 0.0008621543631062487,
      "loss": 2.982330322265625,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.9248947501182556,
      "learning_rate": 0.0008487796649318904,
      "loss": 2.9677340984344482,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 1.0004583597183228,
      "learning_rate": 0.0008349001781229053,
      "loss": 2.931910753250122,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_bleu": 0.5966938923470444,
      "eval_ce_loss": 1.6562061356082898,
      "eval_cos_loss": 0.32249008207560675,
      "eval_loss": 2.9227684172865462,
      "eval_mse_loss": 1.2665622860329335,
      "flow/cos_sim": 0.6775099191491463,
      "flow/improvement_ratio": 0.9837169728866996,
      "flow/mag_ratio_mean": 0.643968923587233,
      "flow/mag_ratio_std": 0.2556537515632638,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_bleu": 0.5966938923470444,
      "eval_ce_loss": 1.6562061356082898,
      "eval_cos_loss": 0.32249008207560675,
      "eval_loss": 2.9227684172865462,
      "eval_mse_loss": 1.2665622860329335,
      "eval_runtime": 117.1936,
      "eval_samples_per_second": 238.861,
      "eval_steps_per_second": 3.737,
      "flow/cos_sim": 0.6775099191491463,
      "flow/improvement_ratio": 0.9837169728866996,
      "flow/mag_ratio_mean": 0.643968923587233,
      "flow/mag_ratio_std": 0.2556537515632638,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 1.0061293840408325,
      "learning_rate": 0.0008205359904536107,
      "loss": 2.9485950469970703,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 1.0945794582366943,
      "learning_rate": 0.0008057078912056363,
      "loss": 2.9102377891540527,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.8923175930976868,
      "learning_rate": 0.0007904373410796086,
      "loss": 2.9341015815734863,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 1.009055733680725,
      "learning_rate": 0.0007747464411350876,
      "loss": 2.9085001945495605,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_bleu": 0.5952239839816952,
      "eval_ce_loss": 1.5963479481875624,
      "eval_cos_loss": 0.30743498025147337,
      "eval_loss": 2.836080116224071,
      "eval_mse_loss": 1.2397321745685246,
      "flow/cos_sim": 0.6925650225382417,
      "flow/improvement_ratio": 0.9844940459891541,
      "flow/mag_ratio_mean": 0.6763553567672973,
      "flow/mag_ratio_std": 0.26531474469187055,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_bleu": 0.5952239839816952,
      "eval_ce_loss": 1.5963479481875624,
      "eval_cos_loss": 0.30743498025147337,
      "eval_loss": 2.836080116224071,
      "eval_mse_loss": 1.2397321745685246,
      "eval_runtime": 114.8272,
      "eval_samples_per_second": 243.784,
      "eval_steps_per_second": 3.814,
      "flow/cos_sim": 0.6925650225382417,
      "flow/improvement_ratio": 0.9844940459891541,
      "flow/mag_ratio_mean": 0.6763553567672973,
      "flow/mag_ratio_std": 0.26531474469187055,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 1.058870792388916,
      "learning_rate": 0.000758657900803716,
      "loss": 2.8502509593963623,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 1.6279115676879883,
      "learning_rate": 0.000742195005021869,
      "loss": 2.8677358627319336,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.719463586807251,
      "learning_rate": 0.0007253815805303786,
      "loss": 2.8696165084838867,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 1.2832756042480469,
      "learning_rate": 0.0007082419613901028,
      "loss": 2.8660786151885986,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_bleu": 0.5924882946880239,
      "eval_ce_loss": 1.5985924006869259,
      "eval_cos_loss": 0.298040063073646,
      "eval_loss": 2.8347644716092986,
      "eval_mse_loss": 1.2361720814008146,
      "flow/cos_sim": 0.7019599421655751,
      "flow/improvement_ratio": 0.985667058049816,
      "flow/mag_ratio_mean": 0.676283470844025,
      "flow/mag_ratio_std": 0.2555381012697742,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_bleu": 0.5924882946880239,
      "eval_ce_loss": 1.5985924006869259,
      "eval_cos_loss": 0.298040063073646,
      "eval_loss": 2.8347644716092986,
      "eval_mse_loss": 1.2361720814008146,
      "eval_runtime": 111.5666,
      "eval_samples_per_second": 250.908,
      "eval_steps_per_second": 3.926,
      "flow/cos_sim": 0.7019599421655751,
      "flow/improvement_ratio": 0.985667058049816,
      "flow/mag_ratio_mean": 0.676283470844025,
      "flow/mag_ratio_std": 0.2555381012697742,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 1.0193252563476562,
      "learning_rate": 0.0006908009537632514,
      "loss": 2.831209897994995,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.9136300086975098,
      "learning_rate": 0.0006730838000114403,
      "loss": 2.838761329650879,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 1.4517526626586914,
      "learning_rate": 0.0006551161421624341,
      "loss": 2.8446762561798096,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 1.3003419637680054,
      "learning_rate": 0.0006369239847984517,
      "loss": 2.815347909927368,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_bleu": 0.5888005150326592,
      "eval_ce_loss": 1.5748343447299853,
      "eval_cos_loss": 0.28894776742071865,
      "eval_loss": 2.7988699277242026,
      "eval_mse_loss": 1.2240355853076395,
      "flow/cos_sim": 0.7110522344504318,
      "flow/improvement_ratio": 0.9867022692068527,
      "flow/mag_ratio_mean": 0.6932177976386188,
      "flow/mag_ratio_std": 0.25633022032644104,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_bleu": 0.5888005150326592,
      "eval_ce_loss": 1.5748343447299853,
      "eval_cos_loss": 0.28894776742071865,
      "eval_loss": 2.7988699277242026,
      "eval_mse_loss": 1.2240355853076395,
      "eval_runtime": 116.9194,
      "eval_samples_per_second": 239.421,
      "eval_steps_per_second": 3.746,
      "flow/cos_sim": 0.7110522344504318,
      "flow/improvement_ratio": 0.9867022692068527,
      "flow/mag_ratio_mean": 0.6932177976386188,
      "flow/mag_ratio_std": 0.25633022032644104,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 1.0110423564910889,
      "learning_rate": 0.0006185336574197479,
      "loss": 2.7822299003601074,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 0.8770432472229004,
      "learning_rate": 0.0005999717763379407,
      "loss": 2.7894957065582275,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 1.0478971004486084,
      "learning_rate": 0.0005812652061542363,
      "loss": 2.7769644260406494,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.9663735032081604,
      "learning_rate": 0.0005624410208783071,
      "loss": 2.770944595336914,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_bleu": 0.615012894123598,
      "eval_ce_loss": 1.5484308441207832,
      "eval_cos_loss": 0.28064475538523775,
      "eval_loss": 2.7590326413716357,
      "eval_mse_loss": 1.2106017988838562,
      "flow/cos_sim": 0.7193552492416069,
      "flow/improvement_ratio": 0.9875411418474973,
      "flow/mag_ratio_mean": 0.7026482544261027,
      "flow/mag_ratio_std": 0.24951966559369815,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_bleu": 0.615012894123598,
      "eval_ce_loss": 1.5484308441207832,
      "eval_cos_loss": 0.28064475538523775,
      "eval_loss": 2.7590326413716357,
      "eval_mse_loss": 1.2106017988838562,
      "eval_runtime": 116.4465,
      "eval_samples_per_second": 240.394,
      "eval_steps_per_second": 3.761,
      "flow/cos_sim": 0.7193552492416069,
      "flow/improvement_ratio": 0.9875411418474973,
      "flow/mag_ratio_mean": 0.7026482544261027,
      "flow/mag_ratio_std": 0.24951966559369815,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 1.2616784572601318,
      "learning_rate": 0.0005435264647440881,
      "loss": 2.7979249954223633,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 0.906895101070404,
      "learning_rate": 0.000524548912779213,
      "loss": 2.790898561477661,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 1.108110785484314,
      "learning_rate": 0.0005055358311851499,
      "loss": 2.7871291637420654,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 1.0619992017745972,
      "learning_rate": 0.0004865147375853812,
      "loss": 2.774048089981079,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_bleu": 0.6019788667659534,
      "eval_ce_loss": 1.5569582646840239,
      "eval_cos_loss": 0.2751268859119176,
      "eval_loss": 2.763672876575766,
      "eval_mse_loss": 1.2067146095783199,
      "flow/cos_sim": 0.7248731139860197,
      "flow/improvement_ratio": 0.9875338717138387,
      "flow/mag_ratio_mean": 0.7146132612881595,
      "flow/mag_ratio_std": 0.2529917611447099,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_bleu": 0.6019788667659534,
      "eval_ce_loss": 1.5569582646840239,
      "eval_cos_loss": 0.2751268859119176,
      "eval_loss": 2.763672876575766,
      "eval_mse_loss": 1.2067146095783199,
      "eval_runtime": 118.9502,
      "eval_samples_per_second": 235.334,
      "eval_steps_per_second": 3.682,
      "flow/cos_sim": 0.7248731139860197,
      "flow/improvement_ratio": 0.9875338717138387,
      "flow/mag_ratio_mean": 0.7146132612881595,
      "flow/mag_ratio_std": 0.2529917611447099,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 1.5396530628204346,
      "learning_rate": 0.0004675131611991607,
      "loss": 2.753552198410034,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 0.9192457795143127,
      "learning_rate": 0.0004485586029984899,
      "loss": 2.7478880882263184,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 1.0996524095535278,
      "learning_rate": 0.00042967849590597266,
      "loss": 2.73809552192688,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.9251135587692261,
      "learning_rate": 0.0004109001650911621,
      "loss": 2.757793664932251,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_bleu": 0.6123695589769809,
      "eval_ce_loss": 1.540999772069661,
      "eval_cos_loss": 0.27151700275945884,
      "eval_loss": 2.744947803891413,
      "eval_mse_loss": 1.203948032910421,
      "flow/cos_sim": 0.7284829948590771,
      "flow/improvement_ratio": 0.9874945781002306,
      "flow/mag_ratio_mean": 0.7196056571997465,
      "flow/mag_ratio_std": 0.25078553958162325,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_bleu": 0.6123695589769809,
      "eval_ce_loss": 1.540999772069661,
      "eval_cos_loss": 0.27151700275945884,
      "eval_loss": 2.744947803891413,
      "eval_mse_loss": 1.203948032910421,
      "eval_runtime": 116.6151,
      "eval_samples_per_second": 240.046,
      "eval_steps_per_second": 3.756,
      "flow/cos_sim": 0.7284829948590771,
      "flow/improvement_ratio": 0.9874945781002306,
      "flow/mag_ratio_mean": 0.7196056571997465,
      "flow/mag_ratio_std": 0.25078553958162325,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 0.7880428433418274,
      "learning_rate": 0.0003922507884228551,
      "loss": 2.7514264583587646,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 1.0263415575027466,
      "learning_rate": 0.00037375735713457723,
      "loss": 2.728660821914673,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 1.3691045045852661,
      "learning_rate": 0.00035544663676018276,
      "loss": 2.7708632946014404,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 1.1120959520339966,
      "learning_rate": 0.00033734512839611255,
      "loss": 2.7620890140533447,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_bleu": 0.6120219445702544,
      "eval_ce_loss": 1.4990101543735697,
      "eval_cos_loss": 0.26497433351599464,
      "eval_loss": 2.6835129650216123,
      "eval_mse_loss": 1.1845028109202103,
      "flow/cos_sim": 0.7350256655314197,
      "flow/improvement_ratio": 0.9874911498805704,
      "flow/mag_ratio_mean": 0.7202309272060655,
      "flow/mag_ratio_std": 0.25011629386715695,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_bleu": 0.6120219445702544,
      "eval_ce_loss": 1.4990101543735697,
      "eval_cos_loss": 0.26497433351599464,
      "eval_loss": 2.6835129650216123,
      "eval_mse_loss": 1.1845028109202103,
      "eval_runtime": 115.7143,
      "eval_samples_per_second": 241.915,
      "eval_steps_per_second": 3.785,
      "flow/cos_sim": 0.7350256655314197,
      "flow/improvement_ratio": 0.9874911498805704,
      "flow/mag_ratio_mean": 0.7202309272060655,
      "flow/mag_ratio_std": 0.25011629386715695,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 1.2354153394699097,
      "learning_rate": 0.0003194790303463687,
      "loss": 2.736755847930908,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 1.015051245689392,
      "learning_rate": 0.00030187420020572406,
      "loss": 2.7210988998413086,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 0.7805538773536682,
      "learning_rate": 0.00028455611743603626,
      "loss": 2.7070095539093018,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 1.3120993375778198,
      "learning_rate": 0.0002675498464898373,
      "loss": 2.7322652339935303,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_bleu": 0.5479981354239509,
      "eval_ce_loss": 1.5251544244485358,
      "eval_cos_loss": 0.2640067928185746,
      "eval_loss": 2.7153855798995656,
      "eval_mse_loss": 1.1902311546345279,
      "flow/cos_sim": 0.7359932074535928,
      "flow/improvement_ratio": 0.9878783013722668,
      "flow/mag_ratio_mean": 0.7280373928481585,
      "flow/mag_ratio_std": 0.25223626611439603,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_bleu": 0.5479981354239509,
      "eval_ce_loss": 1.5251544244485358,
      "eval_cos_loss": 0.2640067928185746,
      "eval_loss": 2.7153855798995656,
      "eval_mse_loss": 1.1902311546345279,
      "eval_runtime": 119.0169,
      "eval_samples_per_second": 235.202,
      "eval_steps_per_second": 3.68,
      "flow/cos_sim": 0.7359932074535928,
      "flow/improvement_ratio": 0.9878783013722668,
      "flow/mag_ratio_mean": 0.7280373928481585,
      "flow/mag_ratio_std": 0.25223626611439603,
      "step": 14336
    },
    {
      "epoch": 0.6739642510738534,
      "grad_norm": 1.2459907531738281,
      "learning_rate": 0.0002508800005345623,
      "loss": 2.746056079864502,
      "step": 14592
    },
    {
      "epoch": 0.6857881853032193,
      "grad_norm": 1.0967656373977661,
      "learning_rate": 0.00023457070582992562,
      "loss": 2.6974754333496094,
      "step": 14848
    },
    {
      "epoch": 0.6976121195325851,
      "grad_norm": 1.05544114112854,
      "learning_rate": 0.00021864556680999692,
      "loss": 2.7050912380218506,
      "step": 15104
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 1.2396007776260376,
      "learning_rate": 0.0002031276319205152,
      "loss": 2.699166774749756,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_bleu": 0.60911048337298,
      "eval_ce_loss": 1.4985592939537955,
      "eval_cos_loss": 0.26051041347795423,
      "eval_loss": 2.6781036116216823,
      "eval_mse_loss": 1.1795443144018791,
      "flow/cos_sim": 0.7394895870663803,
      "flow/improvement_ratio": 0.9873927095437158,
      "flow/mag_ratio_mean": 0.7277950643132266,
      "flow/mag_ratio_std": 0.24724318032667517,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_bleu": 0.60911048337298,
      "eval_ce_loss": 1.4985592939537955,
      "eval_cos_loss": 0.26051041347795423,
      "eval_loss": 2.6781036116216823,
      "eval_mse_loss": 1.1795443144018791,
      "eval_runtime": 116.4844,
      "eval_samples_per_second": 240.315,
      "eval_steps_per_second": 3.76,
      "flow/cos_sim": 0.7394895870663803,
      "flow/improvement_ratio": 0.9873927095437158,
      "flow/mag_ratio_mean": 0.7277950643132266,
      "flow/mag_ratio_std": 0.24724318032667517,
      "step": 15360
    },
    {
      "epoch": 0.7212599879913169,
      "grad_norm": 1.060723900794983,
      "learning_rate": 0.00018803936026088542,
      "loss": 2.698255777359009,
      "step": 15616
    },
    {
      "epoch": 0.7330839222206826,
      "grad_norm": 1.1871165037155151,
      "learning_rate": 0.00017340258907913464,
      "loss": 2.6915080547332764,
      "step": 15872
    },
    {
      "epoch": 0.7449078564500485,
      "grad_norm": 0.9529109597206116,
      "learning_rate": 0.0001592385021668743,
      "loss": 2.6958580017089844,
      "step": 16128
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 1.0608938932418823,
      "learning_rate": 0.0001455675992000087,
      "loss": 2.685572862625122,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_bleu": 0.6033715700217472,
      "eval_ce_loss": 1.5173689544745232,
      "eval_cos_loss": 0.26027777413391084,
      "eval_loss": 2.701112314990666,
      "eval_mse_loss": 1.1837433556171313,
      "flow/cos_sim": 0.7397222264444447,
      "flow/improvement_ratio": 0.9886682670410365,
      "flow/mag_ratio_mean": 0.7254251331499179,
      "flow/mag_ratio_std": 0.24604364749911714,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_bleu": 0.6033715700217472,
      "eval_ce_loss": 1.5173689544745232,
      "eval_cos_loss": 0.26027777413391084,
      "eval_loss": 2.701112314990666,
      "eval_mse_loss": 1.1837433556171313,
      "eval_runtime": 116.0079,
      "eval_samples_per_second": 241.302,
      "eval_steps_per_second": 3.776,
      "flow/cos_sim": 0.7397222264444447,
      "flow/improvement_ratio": 0.9886682670410365,
      "flow/mag_ratio_mean": 0.7254251331499179,
      "flow/mag_ratio_std": 0.24604364749911714,
      "step": 16384
    },
    {
      "epoch": 0.7685557249087802,
      "grad_norm": 0.7993431687355042,
      "learning_rate": 0.000132409666069565,
      "loss": 2.687924861907959,
      "step": 16640
    },
    {
      "epoch": 0.780379659138146,
      "grad_norm": 0.7456022500991821,
      "learning_rate": 0.0001197837462455823,
      "loss": 2.69299054145813,
      "step": 16896
    },
    {
      "epoch": 0.7922035933675119,
      "grad_norm": 0.9615710377693176,
      "learning_rate": 0.00010770811321550749,
      "loss": 2.700143337249756,
      "step": 17152
    },
    {
      "epoch": 0.8040275275968778,
      "grad_norm": 0.9432339072227478,
      "learning_rate": 9.620024403698591e-05,
      "loss": 2.7163219451904297,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_bleu": 0.548308169220463,
      "eval_ce_loss": 1.5058127282416984,
      "eval_cos_loss": 0.25919875846197615,
      "eval_loss": 2.6880075305564217,
      "eval_mse_loss": 1.1821947968713769,
      "flow/cos_sim": 0.7408012430689651,
      "flow/improvement_ratio": 0.9881188005345053,
      "flow/mag_ratio_mean": 0.7311118910573933,
      "flow/mag_ratio_std": 0.24796967136941545,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_bleu": 0.548308169220463,
      "eval_ce_loss": 1.5058127282416984,
      "eval_cos_loss": 0.25919875846197615,
      "eval_loss": 2.6880075305564217,
      "eval_mse_loss": 1.1821947968713769,
      "eval_runtime": 117.7488,
      "eval_samples_per_second": 237.735,
      "eval_steps_per_second": 3.72,
      "flow/cos_sim": 0.7408012430689651,
      "flow/improvement_ratio": 0.9881188005345053,
      "flow/mag_ratio_mean": 0.7311118910573933,
      "flow/mag_ratio_std": 0.24796967136941545,
      "step": 17408
    },
    {
      "epoch": 0.8158514618262436,
      "grad_norm": 0.8737886548042297,
      "learning_rate": 8.527679404332429e-05,
      "loss": 2.6902272701263428,
      "step": 17664
    },
    {
      "epoch": 0.8276753960556095,
      "grad_norm": 0.8381933569908142,
      "learning_rate": 7.495357273823544e-05,
      "loss": 2.678302764892578,
      "step": 17920
    },
    {
      "epoch": 0.8394993302849753,
      "grad_norm": 1.1766976118087769,
      "learning_rate": 6.524552091475183e-05,
      "loss": 2.6962482929229736,
      "step": 18176
    },
    {
      "epoch": 0.8513232645143411,
      "grad_norm": 1.1689873933792114,
      "learning_rate": 5.6166689031422024e-05,
      "loss": 2.6863315105438232,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_bleu": 0.545809467957221,
      "eval_ce_loss": 1.5024473004417331,
      "eval_cos_loss": 0.2577548390722166,
      "eval_loss": 2.6791613673510617,
      "eval_mse_loss": 1.1767140667732447,
      "flow/cos_sim": 0.742245161097888,
      "flow/improvement_ratio": 0.9881984092057023,
      "flow/mag_ratio_mean": 0.7322580480956595,
      "flow/mag_ratio_std": 0.24721906929527787,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_bleu": 0.545809467957221,
      "eval_ce_loss": 1.5024473004417331,
      "eval_cos_loss": 0.2577548390722166,
      "eval_loss": 2.6791613673510617,
      "eval_mse_loss": 1.1767140667732447,
      "eval_runtime": 117.9266,
      "eval_samples_per_second": 237.377,
      "eval_steps_per_second": 3.714,
      "flow/cos_sim": 0.742245161097888,
      "flow/improvement_ratio": 0.9881984092057023,
      "flow/mag_ratio_mean": 0.7322580480956595,
      "flow/mag_ratio_std": 0.24721906929527787,
      "step": 18432
    },
    {
      "epoch": 0.8631471987437069,
      "grad_norm": 1.2247194051742554,
      "learning_rate": 4.773021687709067e-05,
      "loss": 2.6804656982421875,
      "step": 18688
    },
    {
      "epoch": 0.8749711329730728,
      "grad_norm": 0.9656364917755127,
      "learning_rate": 3.994831455368719e-05,
      "loss": 2.6799473762512207,
      "step": 18944
    },
    {
      "epoch": 0.8867950672024387,
      "grad_norm": 1.1167409420013428,
      "learning_rate": 3.283224480455282e-05,
      "loss": 2.6692895889282227,
      "step": 19200
    },
    {
      "epoch": 0.8986190014318045,
      "grad_norm": 0.8527817726135254,
      "learning_rate": 2.639230671387627e-05,
      "loss": 2.6880292892456055,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_bleu": 0.5496708764507602,
      "eval_ce_loss": 1.4893978382354458,
      "eval_cos_loss": 0.2560059982075539,
      "eval_loss": 2.6582647809699247,
      "eval_mse_loss": 1.1688669447757338,
      "flow/cos_sim": 0.7439940057388724,
      "flow/improvement_ratio": 0.9888677304466021,
      "flow/mag_ratio_mean": 0.7347901796913583,
      "flow/mag_ratio_std": 0.2447861627288605,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_bleu": 0.5496708764507602,
      "eval_ce_loss": 1.4893978382354458,
      "eval_cos_loss": 0.2560059982075539,
      "eval_loss": 2.6582647809699247,
      "eval_mse_loss": 1.1688669447757338,
      "eval_runtime": 117.3021,
      "eval_samples_per_second": 238.64,
      "eval_steps_per_second": 3.734,
      "flow/cos_sim": 0.7439940057388724,
      "flow/improvement_ratio": 0.9888677304466021,
      "flow/mag_ratio_mean": 0.7347901796913583,
      "flow/mag_ratio_std": 0.2447861627288605,
      "step": 19456
    },
    {
      "epoch": 0.9104429356611704,
      "grad_norm": 0.5750219225883484,
      "learning_rate": 2.063782080083576e-05,
      "loss": 2.677205801010132,
      "step": 19712
    },
    {
      "epoch": 0.9222668698905362,
      "grad_norm": 1.096099615097046,
      "learning_rate": 1.557711553001523e-05,
      "loss": 2.6789114475250244,
      "step": 19968
    },
    {
      "epoch": 0.9340908041199021,
      "grad_norm": 0.9802102446556091,
      "learning_rate": 1.1217515257622269e-05,
      "loss": 2.662231922149658,
      "step": 20224
    },
    {
      "epoch": 0.945914738349268,
      "grad_norm": 0.7547997236251831,
      "learning_rate": 7.565329630950746e-06,
      "loss": 2.676410436630249,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_bleu": 0.5464821752660199,
      "eval_ce_loss": 1.4990644795165213,
      "eval_cos_loss": 0.25722625508019914,
      "eval_loss": 2.67388042249636,
      "eval_mse_loss": 1.1748159417550859,
      "flow/cos_sim": 0.7427737461105329,
      "flow/improvement_ratio": 0.9887178831448838,
      "flow/mag_ratio_mean": 0.7328161248877713,
      "flow/mag_ratio_std": 0.2466802823965408,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_bleu": 0.5464821752660199,
      "eval_ce_loss": 1.4990644795165213,
      "eval_cos_loss": 0.25722625508019914,
      "eval_loss": 2.67388042249636,
      "eval_mse_loss": 1.1748159417550859,
      "eval_runtime": 116.8427,
      "eval_samples_per_second": 239.578,
      "eval_steps_per_second": 3.749,
      "flow/cos_sim": 0.7427737461105329,
      "flow/improvement_ratio": 0.9887178831448838,
      "flow/mag_ratio_mean": 0.7328161248877713,
      "flow/mag_ratio_std": 0.2466802823965408,
      "step": 20480
    },
    {
      "epoch": 0.9577386725786338,
      "grad_norm": 0.9208499193191528,
      "learning_rate": 4.62584445643166e-06,
      "loss": 2.667086362838745,
      "step": 20736
    },
    {
      "epoch": 0.9695626068079997,
      "grad_norm": 0.6824806332588196,
      "learning_rate": 2.40331404948807e-06,
      "loss": 2.655881404876709,
      "step": 20992
    },
    {
      "epoch": 0.9813865410373654,
      "grad_norm": 0.7420426607131958,
      "learning_rate": 9.009550772663965e-07,
      "loss": 2.674077033996582,
      "step": 21248
    },
    {
      "epoch": 0.9932104752667313,
      "grad_norm": 0.6535339951515198,
      "learning_rate": 1.2094190315575791e-07,
      "loss": 2.6826326847076416,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_bleu": 0.5500433951110925,
      "eval_ce_loss": 1.4851820075620799,
      "eval_cos_loss": 0.25606720632614066,
      "eval_loss": 2.6555024687558006,
      "eval_mse_loss": 1.170320453709119,
      "flow/cos_sim": 0.7439327969398672,
      "flow/improvement_ratio": 0.9888444078839533,
      "flow/mag_ratio_mean": 0.7340902883440392,
      "flow/mag_ratio_std": 0.24632065518669885,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_bleu": 0.5500433951110925,
      "eval_ce_loss": 1.4851820075620799,
      "eval_cos_loss": 0.25606720632614066,
      "eval_loss": 2.6555024687558006,
      "eval_mse_loss": 1.170320453709119,
      "eval_runtime": 115.8505,
      "eval_samples_per_second": 241.63,
      "eval_steps_per_second": 3.781,
      "flow/cos_sim": 0.7439327969398672,
      "flow/improvement_ratio": 0.9888444078839533,
      "flow/mag_ratio_mean": 0.7340902883440392,
      "flow/mag_ratio_std": 0.24632065518669885,
      "step": 21504
    }
  ],
  "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
}