{
  "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": 1.9140732288360596,
      "learning_rate": 0.000498046875,
      "loss": 6.9310712814331055,
      "step": 256
    },
    {
      "epoch": 0.023647868458731697,
      "grad_norm": 0.966057538986206,
      "learning_rate": 0.000998046875,
      "loss": 1.017655372619629,
      "step": 512
    },
    {
      "epoch": 0.03547180268809755,
      "grad_norm": 0.6036155223846436,
      "learning_rate": 0.000999640996023194,
      "loss": 0.5299859046936035,
      "step": 768
    },
    {
      "epoch": 0.047295736917463395,
      "grad_norm": 0.6692569851875305,
      "learning_rate": 0.0009985588674043958,
      "loss": 0.43251049518585205,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_bleu": 0.6267380536600159,
      "eval_ce_loss": 1.563563620935292,
      "eval_cos_loss": 0.11390362637432198,
      "eval_loss": 0.40128198402113024,
      "eval_mse_loss": 0.2176408769061032,
      "eval_rec_loss": 0.015894381031734215,
      "flow/cos_sim": 0.8860963983078526,
      "flow/improvement_ratio": 0.9765386980146034,
      "flow/mag_ratio_mean": 0.8838755380889597,
      "flow/mag_ratio_std": 0.1646749794006892,
      "step": 1024
    },
    {
      "epoch": 0.047295736917463395,
      "eval_bleu": 0.6267380536600159,
      "eval_ce_loss": 1.563563620935292,
      "eval_cos_loss": 0.11390362637432198,
      "eval_loss": 0.40128198402113024,
      "eval_mse_loss": 0.2176408769061032,
      "eval_rec_loss": 0.015894381031734215,
      "eval_runtime": 145.0392,
      "eval_samples_per_second": 193.003,
      "eval_steps_per_second": 3.02,
      "flow/cos_sim": 0.8860963983078526,
      "flow/improvement_ratio": 0.9765386980146034,
      "flow/mag_ratio_mean": 0.8838755380889597,
      "flow/mag_ratio_std": 0.1646749794006892,
      "step": 1024
    },
    {
      "epoch": 0.05911967114682925,
      "grad_norm": 0.46750983595848083,
      "learning_rate": 0.0009967551747861387,
      "loss": 0.3926387429237366,
      "step": 1280
    },
    {
      "epoch": 0.0709436053761951,
      "grad_norm": 0.5673767328262329,
      "learning_rate": 0.000994232528651847,
      "loss": 0.3741423189640045,
      "step": 1536
    },
    {
      "epoch": 0.08276753960556095,
      "grad_norm": 0.5806387066841125,
      "learning_rate": 0.0009909945800260092,
      "loss": 0.35307449102401733,
      "step": 1792
    },
    {
      "epoch": 0.09459147383492679,
      "grad_norm": 0.6838217973709106,
      "learning_rate": 0.0009870460151900522,
      "loss": 0.3399304151535034,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_bleu": 0.6478349738418474,
      "eval_ce_loss": 1.3434572242982856,
      "eval_cos_loss": 0.09738318450306649,
      "eval_loss": 0.3373560005943525,
      "eval_mse_loss": 0.18800275102597938,
      "eval_rec_loss": 0.005269206022433341,
      "flow/cos_sim": 0.9026168323542973,
      "flow/improvement_ratio": 0.9768092384621433,
      "flow/mag_ratio_mean": 0.9078535956062682,
      "flow/mag_ratio_std": 0.12002966954419601,
      "step": 2048
    },
    {
      "epoch": 0.09459147383492679,
      "eval_bleu": 0.6478349738418474,
      "eval_ce_loss": 1.3434572242982856,
      "eval_cos_loss": 0.09738318450306649,
      "eval_loss": 0.3373560005943525,
      "eval_mse_loss": 0.18800275102597938,
      "eval_rec_loss": 0.005269206022433341,
      "eval_runtime": 143.5996,
      "eval_samples_per_second": 194.938,
      "eval_steps_per_second": 3.05,
      "flow/cos_sim": 0.9026168323542973,
      "flow/improvement_ratio": 0.9768092384621433,
      "flow/mag_ratio_mean": 0.9078535956062682,
      "flow/mag_ratio_std": 0.12002966954419601,
      "step": 2048
    },
    {
      "epoch": 0.10641540806429264,
      "grad_norm": 0.6462650299072266,
      "learning_rate": 0.0009823925488998885,
      "loss": 0.3380688428878784,
      "step": 2304
    },
    {
      "epoch": 0.1182393422936585,
      "grad_norm": 0.6545562744140625,
      "learning_rate": 0.0009770409161149525,
      "loss": 0.3214532732963562,
      "step": 2560
    },
    {
      "epoch": 0.13006327652302435,
      "grad_norm": 0.6667947173118591,
      "learning_rate": 0.0009709988622506973,
      "loss": 0.31182003021240234,
      "step": 2816
    },
    {
      "epoch": 0.1418872107523902,
      "grad_norm": 0.600439727306366,
      "learning_rate": 0.000964275131968659,
      "loss": 0.3082171380519867,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_bleu": 0.6794402302458825,
      "eval_ce_loss": 1.18399681810919,
      "eval_cos_loss": 0.08983433964453875,
      "eval_loss": 0.3044066195414491,
      "eval_mse_loss": 0.17457320809908652,
      "eval_rec_loss": 0.0024502936500277726,
      "flow/cos_sim": 0.9101656942062726,
      "flow/improvement_ratio": 0.9734727941691603,
      "flow/mag_ratio_mean": 0.9100713347463303,
      "flow/mag_ratio_std": 0.09228039255629392,
      "step": 3072
    },
    {
      "epoch": 0.1418872107523902,
      "eval_bleu": 0.6794402302458825,
      "eval_ce_loss": 1.18399681810919,
      "eval_cos_loss": 0.08983433964453875,
      "eval_loss": 0.3044066195414491,
      "eval_mse_loss": 0.17457320809908652,
      "eval_rec_loss": 0.0024502936500277726,
      "eval_runtime": 142.616,
      "eval_samples_per_second": 196.282,
      "eval_steps_per_second": 3.071,
      "flow/cos_sim": 0.9101656942062726,
      "flow/improvement_ratio": 0.9734727941691603,
      "flow/mag_ratio_mean": 0.9100713347463303,
      "flow/mag_ratio_std": 0.09228039255629392,
      "step": 3072
    },
    {
      "epoch": 0.15371114498175603,
      "grad_norm": 0.5188813209533691,
      "learning_rate": 0.0009568794565203123,
      "loss": 0.3006798326969147,
      "step": 3328
    },
    {
      "epoch": 0.1655350792111219,
      "grad_norm": 0.3702296316623688,
      "learning_rate": 0.0009488225396630347,
      "loss": 0.29647037386894226,
      "step": 3584
    },
    {
      "epoch": 0.17735901344048774,
      "grad_norm": 0.49310481548309326,
      "learning_rate": 0.0009401160421685646,
      "loss": 0.3130263388156891,
      "step": 3840
    },
    {
      "epoch": 0.18918294766985358,
      "grad_norm": 0.5644649267196655,
      "learning_rate": 0.0009307725649463714,
      "loss": 0.2926834523677826,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_bleu": 0.6839715035857402,
      "eval_ce_loss": 1.1360810732460458,
      "eval_cos_loss": 0.08503438343536364,
      "eval_loss": 0.2889118402788084,
      "eval_mse_loss": 0.16517186124030858,
      "eval_rec_loss": 0.0016284317239192974,
      "flow/cos_sim": 0.9149656362457362,
      "flow/improvement_ratio": 0.9750959768415042,
      "flow/mag_ratio_mean": 0.9128544130553938,
      "flow/mag_ratio_std": 0.09354445787325297,
      "step": 4096
    },
    {
      "epoch": 0.18918294766985358,
      "eval_bleu": 0.6839715035857402,
      "eval_ce_loss": 1.1360810732460458,
      "eval_cos_loss": 0.08503438343536364,
      "eval_loss": 0.2889118402788084,
      "eval_mse_loss": 0.16517186124030858,
      "eval_rec_loss": 0.0016284317239192974,
      "eval_runtime": 143.3522,
      "eval_samples_per_second": 195.274,
      "eval_steps_per_second": 3.055,
      "flow/cos_sim": 0.9149656362457362,
      "flow/improvement_ratio": 0.9750959768415042,
      "flow/mag_ratio_mean": 0.9128544130553938,
      "flow/mag_ratio_std": 0.09354445787325297,
      "step": 4096
    },
    {
      "epoch": 0.20100688189921945,
      "grad_norm": 0.3648587763309479,
      "learning_rate": 0.0009208056308063659,
      "loss": 0.28915518522262573,
      "step": 4352
    },
    {
      "epoch": 0.2128308161285853,
      "grad_norm": 0.5004331469535828,
      "learning_rate": 0.0009102296648873445,
      "loss": 0.28424444794654846,
      "step": 4608
    },
    {
      "epoch": 0.22465475035795113,
      "grad_norm": 0.27222171425819397,
      "learning_rate": 0.0008990599737794927,
      "loss": 0.2838270366191864,
      "step": 4864
    },
    {
      "epoch": 0.236478684587317,
      "grad_norm": 0.39166298508644104,
      "learning_rate": 0.0008873127233711644,
      "loss": 0.2802566885948181,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_bleu": 0.6966429437264656,
      "eval_ce_loss": 1.0860335096784923,
      "eval_cos_loss": 0.08069509963534738,
      "eval_loss": 0.27587767567944854,
      "eval_mse_loss": 0.15802378180228413,
      "eval_rec_loss": 0.0011810324893420196,
      "flow/cos_sim": 0.9193049165756191,
      "flow/improvement_ratio": 0.9735453040360316,
      "flow/mag_ratio_mean": 0.9165048223652251,
      "flow/mag_ratio_std": 0.08227625193252955,
      "step": 5120
    },
    {
      "epoch": 0.236478684587317,
      "eval_bleu": 0.6966429437264656,
      "eval_ce_loss": 1.0860335096784923,
      "eval_cos_loss": 0.08069509963534738,
      "eval_loss": 0.27587767567944854,
      "eval_mse_loss": 0.15802378180228413,
      "eval_rec_loss": 0.0011810324893420196,
      "eval_runtime": 144.8851,
      "eval_samples_per_second": 193.208,
      "eval_steps_per_second": 3.023,
      "flow/cos_sim": 0.9193049165756191,
      "flow/improvement_ratio": 0.9735453040360316,
      "flow/mag_ratio_mean": 0.9165048223652251,
      "flow/mag_ratio_std": 0.08227625193252955,
      "step": 5120
    },
    {
      "epoch": 0.24830261881668284,
      "grad_norm": 0.4552461802959442,
      "learning_rate": 0.0008750049154520011,
      "loss": 0.27747780084609985,
      "step": 5376
    },
    {
      "epoch": 0.2601265530460487,
      "grad_norm": 0.3430848717689514,
      "learning_rate": 0.0008621543631062487,
      "loss": 0.2891384959220886,
      "step": 5632
    },
    {
      "epoch": 0.27195048727541454,
      "grad_norm": 0.4337022304534912,
      "learning_rate": 0.0008487796649318904,
      "loss": 0.274080365896225,
      "step": 5888
    },
    {
      "epoch": 0.2837744215047804,
      "grad_norm": 0.40342026948928833,
      "learning_rate": 0.0008349001781229053,
      "loss": 0.2737179100513458,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_bleu": 0.7005510372485733,
      "eval_ce_loss": 1.067978764981984,
      "eval_cos_loss": 0.0791517910994079,
      "eval_loss": 0.2710464134744313,
      "eval_mse_loss": 0.15530247285485813,
      "eval_rec_loss": 0.00103088276002016,
      "flow/cos_sim": 0.9208482295682986,
      "flow/improvement_ratio": 0.9731569932476026,
      "flow/mag_ratio_mean": 0.9284612283314744,
      "flow/mag_ratio_std": 0.07813496222574962,
      "step": 6144
    },
    {
      "epoch": 0.2837744215047804,
      "eval_bleu": 0.7005510372485733,
      "eval_ce_loss": 1.067978764981984,
      "eval_cos_loss": 0.0791517910994079,
      "eval_loss": 0.2710464134744313,
      "eval_mse_loss": 0.15530247285485813,
      "eval_rec_loss": 0.00103088276002016,
      "eval_runtime": 142.9519,
      "eval_samples_per_second": 195.821,
      "eval_steps_per_second": 3.064,
      "flow/cos_sim": 0.9208482295682986,
      "flow/improvement_ratio": 0.9731569932476026,
      "flow/mag_ratio_mean": 0.9284612283314744,
      "flow/mag_ratio_std": 0.07813496222574962,
      "step": 6144
    },
    {
      "epoch": 0.2955983557341462,
      "grad_norm": 0.36669906973838806,
      "learning_rate": 0.0008205359904536107,
      "loss": 0.2686825394630432,
      "step": 6400
    },
    {
      "epoch": 0.30742228996351206,
      "grad_norm": 0.2844366729259491,
      "learning_rate": 0.0008057078912056363,
      "loss": 0.26841098070144653,
      "step": 6656
    },
    {
      "epoch": 0.3192462241928779,
      "grad_norm": 0.341339647769928,
      "learning_rate": 0.0007904373410796086,
      "loss": 0.2661970555782318,
      "step": 6912
    },
    {
      "epoch": 0.3310701584222438,
      "grad_norm": 0.35041218996047974,
      "learning_rate": 0.0007747464411350876,
      "loss": 0.26556745171546936,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_bleu": 0.710653856755971,
      "eval_ce_loss": 1.0037036097213015,
      "eval_cos_loss": 0.0774884378733156,
      "eval_loss": 0.2617515635817018,
      "eval_mse_loss": 0.15282062886784611,
      "eval_rec_loss": 0.0008117282769037305,
      "flow/cos_sim": 0.9225115811443765,
      "flow/improvement_ratio": 0.9752856880834658,
      "flow/mag_ratio_mean": 0.9207794453999768,
      "flow/mag_ratio_std": 0.08132727078447059,
      "step": 7168
    },
    {
      "epoch": 0.3310701584222438,
      "eval_bleu": 0.710653856755971,
      "eval_ce_loss": 1.0037036097213015,
      "eval_cos_loss": 0.0774884378733156,
      "eval_loss": 0.2617515635817018,
      "eval_mse_loss": 0.15282062886784611,
      "eval_rec_loss": 0.0008117282769037305,
      "eval_runtime": 142.6116,
      "eval_samples_per_second": 196.288,
      "eval_steps_per_second": 3.071,
      "flow/cos_sim": 0.9225115811443765,
      "flow/improvement_ratio": 0.9752856880834658,
      "flow/mag_ratio_mean": 0.9207794453999768,
      "flow/mag_ratio_std": 0.08132727078447059,
      "step": 7168
    },
    {
      "epoch": 0.34289409265160964,
      "grad_norm": 0.328535258769989,
      "learning_rate": 0.000758657900803716,
      "loss": 0.27869582176208496,
      "step": 7424
    },
    {
      "epoch": 0.3547180268809755,
      "grad_norm": 0.27767524123191833,
      "learning_rate": 0.000742195005021869,
      "loss": 0.26837044954299927,
      "step": 7680
    },
    {
      "epoch": 0.3665419611103413,
      "grad_norm": 0.40343427658081055,
      "learning_rate": 0.0007253815805303786,
      "loss": 0.26287224888801575,
      "step": 7936
    },
    {
      "epoch": 0.37836589533970716,
      "grad_norm": 0.3494727611541748,
      "learning_rate": 0.0007082419613901028,
      "loss": 0.2608710527420044,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_bleu": 0.7258561394806508,
      "eval_ce_loss": 0.954990051584701,
      "eval_cos_loss": 0.07798272708099183,
      "eval_loss": 0.2584831827239359,
      "eval_mse_loss": 0.15443320324023566,
      "eval_rec_loss": 0.0007526992644982298,
      "flow/cos_sim": 0.9220172950666244,
      "flow/improvement_ratio": 0.9747015749482804,
      "flow/mag_ratio_mean": 0.9233354187175019,
      "flow/mag_ratio_std": 0.07788482179108276,
      "step": 8192
    },
    {
      "epoch": 0.37836589533970716,
      "eval_bleu": 0.7258561394806508,
      "eval_ce_loss": 0.954990051584701,
      "eval_cos_loss": 0.07798272708099183,
      "eval_loss": 0.2584831827239359,
      "eval_mse_loss": 0.15443320324023566,
      "eval_rec_loss": 0.0007526992644982298,
      "eval_runtime": 144.1923,
      "eval_samples_per_second": 194.137,
      "eval_steps_per_second": 3.038,
      "flow/cos_sim": 0.9220172950666244,
      "flow/improvement_ratio": 0.9747015749482804,
      "flow/mag_ratio_mean": 0.9233354187175019,
      "flow/mag_ratio_std": 0.07788482179108276,
      "step": 8192
    },
    {
      "epoch": 0.390189829569073,
      "grad_norm": 0.4279082417488098,
      "learning_rate": 0.0006908009537632514,
      "loss": 0.2599097788333893,
      "step": 8448
    },
    {
      "epoch": 0.4020137637984389,
      "grad_norm": 0.43606042861938477,
      "learning_rate": 0.0006730838000114403,
      "loss": 0.26111704111099243,
      "step": 8704
    },
    {
      "epoch": 0.41383769802780473,
      "grad_norm": 0.21560809016227722,
      "learning_rate": 0.0006551161421624341,
      "loss": 0.268413245677948,
      "step": 8960
    },
    {
      "epoch": 0.4256616322571706,
      "grad_norm": 0.5756855607032776,
      "learning_rate": 0.0006369239847984517,
      "loss": 0.2562841475009918,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_bleu": 0.7095258647762687,
      "eval_ce_loss": 1.0158189947474492,
      "eval_cos_loss": 0.0742807937743457,
      "eval_loss": 0.256784121705789,
      "eval_mse_loss": 0.14713338680871546,
      "eval_rec_loss": 0.0006407552403591884,
      "flow/cos_sim": 0.9257192251072627,
      "flow/improvement_ratio": 0.9743275895510635,
      "flow/mag_ratio_mean": 0.9202466026016566,
      "flow/mag_ratio_std": 0.0764387545588354,
      "step": 9216
    },
    {
      "epoch": 0.4256616322571706,
      "eval_bleu": 0.7095258647762687,
      "eval_ce_loss": 1.0158189947474492,
      "eval_cos_loss": 0.0742807937743457,
      "eval_loss": 0.256784121705789,
      "eval_mse_loss": 0.14713338680871546,
      "eval_rec_loss": 0.0006407552403591884,
      "eval_runtime": 145.076,
      "eval_samples_per_second": 192.954,
      "eval_steps_per_second": 3.019,
      "flow/cos_sim": 0.9257192251072627,
      "flow/improvement_ratio": 0.9743275895510635,
      "flow/mag_ratio_mean": 0.9202466026016566,
      "flow/mag_ratio_std": 0.0764387545588354,
      "step": 9216
    },
    {
      "epoch": 0.4374855664865364,
      "grad_norm": 0.376579225063324,
      "learning_rate": 0.0006185336574197479,
      "loss": 0.25477761030197144,
      "step": 9472
    },
    {
      "epoch": 0.44930950071590225,
      "grad_norm": 0.4893060624599457,
      "learning_rate": 0.0005999717763379407,
      "loss": 0.2554967701435089,
      "step": 9728
    },
    {
      "epoch": 0.4611334349452681,
      "grad_norm": 0.5277368426322937,
      "learning_rate": 0.0005812652061542363,
      "loss": 0.2537921667098999,
      "step": 9984
    },
    {
      "epoch": 0.472957369174634,
      "grad_norm": 0.3493780791759491,
      "learning_rate": 0.0005624410208783071,
      "loss": 0.2518165707588196,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_bleu": 0.7175600139987302,
      "eval_ce_loss": 0.9684542848231041,
      "eval_cos_loss": 0.07330035061902923,
      "eval_loss": 0.25065779049782994,
      "eval_mse_loss": 0.1458930650292194,
      "eval_rec_loss": 0.000589260474122746,
      "flow/cos_sim": 0.9266996725236989,
      "flow/improvement_ratio": 0.9765301047394809,
      "flow/mag_ratio_mean": 0.9194769399351181,
      "flow/mag_ratio_std": 0.07280441308946914,
      "step": 10240
    },
    {
      "epoch": 0.472957369174634,
      "eval_bleu": 0.7175600139987302,
      "eval_ce_loss": 0.9684542848231041,
      "eval_cos_loss": 0.07330035061902923,
      "eval_loss": 0.25065779049782994,
      "eval_mse_loss": 0.1458930650292194,
      "eval_rec_loss": 0.000589260474122746,
      "eval_runtime": 144.907,
      "eval_samples_per_second": 193.179,
      "eval_steps_per_second": 3.023,
      "flow/cos_sim": 0.9266996725236989,
      "flow/improvement_ratio": 0.9765301047394809,
      "flow/mag_ratio_mean": 0.9194769399351181,
      "flow/mag_ratio_std": 0.07280441308946914,
      "step": 10240
    },
    {
      "epoch": 0.48478130340399983,
      "grad_norm": 0.5341387391090393,
      "learning_rate": 0.0005435264647440881,
      "loss": 0.2537766695022583,
      "step": 10496
    },
    {
      "epoch": 0.49660523763336567,
      "grad_norm": 0.37015679478645325,
      "learning_rate": 0.000524548912779213,
      "loss": 0.2499093860387802,
      "step": 10752
    },
    {
      "epoch": 0.5084291718627315,
      "grad_norm": 0.6591759920120239,
      "learning_rate": 0.0005055358311851499,
      "loss": 0.25186970829963684,
      "step": 11008
    },
    {
      "epoch": 0.5202531060920974,
      "grad_norm": 0.31277748942375183,
      "learning_rate": 0.0004865147375853812,
      "loss": 0.25062766671180725,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_bleu": 0.7260803484693739,
      "eval_ce_loss": 0.9400546782092961,
      "eval_cos_loss": 0.07318596699228298,
      "eval_loss": 0.2483225218169221,
      "eval_mse_loss": 0.14649179712210073,
      "eval_rec_loss": 0.0005066586955975142,
      "flow/cos_sim": 0.9268140568308634,
      "flow/improvement_ratio": 0.9762515691045213,
      "flow/mag_ratio_mean": 0.9157658136326429,
      "flow/mag_ratio_std": 0.07639177793373256,
      "step": 11264
    },
    {
      "epoch": 0.5202531060920974,
      "eval_bleu": 0.7260803484693739,
      "eval_ce_loss": 0.9400546782092961,
      "eval_cos_loss": 0.07318596699228298,
      "eval_loss": 0.2483225218169221,
      "eval_mse_loss": 0.14649179712210073,
      "eval_rec_loss": 0.0005066586955975142,
      "eval_runtime": 143.4087,
      "eval_samples_per_second": 195.197,
      "eval_steps_per_second": 3.054,
      "flow/cos_sim": 0.9268140568308634,
      "flow/improvement_ratio": 0.9762515691045213,
      "flow/mag_ratio_mean": 0.9157658136326429,
      "flow/mag_ratio_std": 0.07639177793373256,
      "step": 11264
    },
    {
      "epoch": 0.5320770403214632,
      "grad_norm": 0.7305059432983398,
      "learning_rate": 0.0004675131611991607,
      "loss": 0.2499980330467224,
      "step": 11520
    },
    {
      "epoch": 0.5439009745508291,
      "grad_norm": 0.2771352231502533,
      "learning_rate": 0.0004485586029984899,
      "loss": 0.24800576269626617,
      "step": 11776
    },
    {
      "epoch": 0.5557249087801949,
      "grad_norm": 0.2686781883239746,
      "learning_rate": 0.00042967849590597266,
      "loss": 0.24677802622318268,
      "step": 12032
    },
    {
      "epoch": 0.5675488430095608,
      "grad_norm": 0.37717506289482117,
      "learning_rate": 0.0004109001650911621,
      "loss": 0.24666211009025574,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_bleu": 0.724567852062454,
      "eval_ce_loss": 0.9509265820185343,
      "eval_cos_loss": 0.07338333341663014,
      "eval_loss": 0.24947158049911125,
      "eval_mse_loss": 0.1465663725518745,
      "eval_rec_loss": 0.00047421499268288965,
      "flow/cos_sim": 0.9266166863920482,
      "flow/improvement_ratio": 0.9748387396607769,
      "flow/mag_ratio_mean": 0.9328112143631939,
      "flow/mag_ratio_std": 0.07196474395113993,
      "step": 12288
    },
    {
      "epoch": 0.5675488430095608,
      "eval_bleu": 0.724567852062454,
      "eval_ce_loss": 0.9509265820185343,
      "eval_cos_loss": 0.07338333341663014,
      "eval_loss": 0.24947158049911125,
      "eval_mse_loss": 0.1465663725518745,
      "eval_rec_loss": 0.00047421499268288965,
      "eval_runtime": 144.4458,
      "eval_samples_per_second": 193.796,
      "eval_steps_per_second": 3.032,
      "flow/cos_sim": 0.9266166863920482,
      "flow/improvement_ratio": 0.9748387396607769,
      "flow/mag_ratio_mean": 0.9328112143631939,
      "flow/mag_ratio_std": 0.07196474395113993,
      "step": 12288
    },
    {
      "epoch": 0.5793727772389267,
      "grad_norm": 0.7573108673095703,
      "learning_rate": 0.0003922507884228551,
      "loss": 0.24700023233890533,
      "step": 12544
    },
    {
      "epoch": 0.5911967114682924,
      "grad_norm": 0.2122378647327423,
      "learning_rate": 0.00037375735713457723,
      "loss": 0.2433316856622696,
      "step": 12800
    },
    {
      "epoch": 0.6030206456976583,
      "grad_norm": 0.22209292650222778,
      "learning_rate": 0.00035544663676018276,
      "loss": 0.246763676404953,
      "step": 13056
    },
    {
      "epoch": 0.6148445799270241,
      "grad_norm": 0.309989333152771,
      "learning_rate": 0.00033734512839611255,
      "loss": 0.24262374639511108,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_bleu": 0.7300156788258321,
      "eval_ce_loss": 0.933146977737614,
      "eval_cos_loss": 0.07192521648626077,
      "eval_loss": 0.24540512995224564,
      "eval_mse_loss": 0.14446957482526837,
      "eval_rec_loss": 0.00042833433703537476,
      "flow/cos_sim": 0.9280747922315989,
      "flow/improvement_ratio": 0.9743193072271129,
      "flow/mag_ratio_mean": 0.9250430587219866,
      "flow/mag_ratio_std": 0.07064938510355612,
      "step": 13312
    },
    {
      "epoch": 0.6148445799270241,
      "eval_bleu": 0.7300156788258321,
      "eval_ce_loss": 0.933146977737614,
      "eval_cos_loss": 0.07192521648626077,
      "eval_loss": 0.24540512995224564,
      "eval_mse_loss": 0.14446957482526837,
      "eval_rec_loss": 0.00042833433703537476,
      "eval_runtime": 144.2347,
      "eval_samples_per_second": 194.079,
      "eval_steps_per_second": 3.037,
      "flow/cos_sim": 0.9280747922315989,
      "flow/improvement_ratio": 0.9743193072271129,
      "flow/mag_ratio_mean": 0.9250430587219866,
      "flow/mag_ratio_std": 0.07064938510355612,
      "step": 13312
    },
    {
      "epoch": 0.62666851415639,
      "grad_norm": 0.2954619824886322,
      "learning_rate": 0.0003194790303463687,
      "loss": 0.24288520216941833,
      "step": 13568
    },
    {
      "epoch": 0.6384924483857558,
      "grad_norm": 0.6633331775665283,
      "learning_rate": 0.00030187420020572406,
      "loss": 0.24463221430778503,
      "step": 13824
    },
    {
      "epoch": 0.6503163826151217,
      "grad_norm": 0.2086828500032425,
      "learning_rate": 0.00028455611743603626,
      "loss": 0.2416921854019165,
      "step": 14080
    },
    {
      "epoch": 0.6621403168444876,
      "grad_norm": 0.19425669312477112,
      "learning_rate": 0.0002675498464898373,
      "loss": 0.2426968514919281,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_bleu": 0.7358716031963392,
      "eval_ce_loss": 0.8970746389126669,
      "eval_cos_loss": 0.07039773773793216,
      "eval_loss": 0.23921880884665878,
      "eval_mse_loss": 0.14204559988899318,
      "eval_rec_loss": 0.0004259694892114143,
      "flow/cos_sim": 0.9296022757003296,
      "flow/improvement_ratio": 0.9752817895586632,
      "flow/mag_ratio_mean": 0.9346142843978046,
      "flow/mag_ratio_std": 0.06967301141963975,
      "step": 14336
    },
    {
      "epoch": 0.6621403168444876,
      "eval_bleu": 0.7358716031963392,
      "eval_ce_loss": 0.8970746389126669,
      "eval_cos_loss": 0.07039773773793216,
      "eval_loss": 0.23921880884665878,
      "eval_mse_loss": 0.14204559988899318,
      "eval_rec_loss": 0.0004259694892114143,
      "eval_runtime": 146.9688,
      "eval_samples_per_second": 190.469,
      "eval_steps_per_second": 2.98,
      "flow/cos_sim": 0.9296022757003296,
      "flow/improvement_ratio": 0.9752817895586632,
      "flow/mag_ratio_mean": 0.9346142843978046,
      "flow/mag_ratio_std": 0.06967301141963975,
      "step": 14336
    },
    {
      "epoch": 0.6739642510738534,
      "grad_norm": 0.5798227787017822,
      "learning_rate": 0.0002508800005345623,
      "loss": 0.24217651784420013,
      "step": 14592
    },
    {
      "epoch": 0.6857881853032193,
      "grad_norm": 0.8549785017967224,
      "learning_rate": 0.00023457070582992562,
      "loss": 0.24345579743385315,
      "step": 14848
    },
    {
      "epoch": 0.6976121195325851,
      "grad_norm": 0.4544009566307068,
      "learning_rate": 0.00021864556680999692,
      "loss": 0.23991872370243073,
      "step": 15104
    },
    {
      "epoch": 0.709436053761951,
      "grad_norm": 0.5695450305938721,
      "learning_rate": 0.0002031276319205152,
      "loss": 0.2396196871995926,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_bleu": 0.7367086437042604,
      "eval_ce_loss": 0.8948814071476732,
      "eval_cos_loss": 0.07030742877422402,
      "eval_loss": 0.23915054540247677,
      "eval_mse_loss": 0.14226854150289814,
      "eval_rec_loss": 0.00036311902771741107,
      "flow/cos_sim": 0.9296925847661005,
      "flow/improvement_ratio": 0.9746487660495113,
      "flow/mag_ratio_mean": 0.9314249843767245,
      "flow/mag_ratio_std": 0.0696038415476884,
      "step": 15360
    },
    {
      "epoch": 0.709436053761951,
      "eval_bleu": 0.7367086437042604,
      "eval_ce_loss": 0.8948814071476732,
      "eval_cos_loss": 0.07030742877422402,
      "eval_loss": 0.23915054540247677,
      "eval_mse_loss": 0.14226854150289814,
      "eval_rec_loss": 0.00036311902771741107,
      "eval_runtime": 143.8829,
      "eval_samples_per_second": 194.554,
      "eval_steps_per_second": 3.044,
      "flow/cos_sim": 0.9296925847661005,
      "flow/improvement_ratio": 0.9746487660495113,
      "flow/mag_ratio_mean": 0.9314249843767245,
      "flow/mag_ratio_std": 0.0696038415476884,
      "step": 15360
    },
    {
      "epoch": 0.7212599879913169,
      "grad_norm": 0.24349141120910645,
      "learning_rate": 0.00018803936026088542,
      "loss": 0.23957179486751556,
      "step": 15616
    },
    {
      "epoch": 0.7330839222206826,
      "grad_norm": 0.2847910225391388,
      "learning_rate": 0.00017340258907913464,
      "loss": 0.23970942199230194,
      "step": 15872
    },
    {
      "epoch": 0.7449078564500485,
      "grad_norm": 0.5366371870040894,
      "learning_rate": 0.0001592385021668743,
      "loss": 0.23960672318935394,
      "step": 16128
    },
    {
      "epoch": 0.7567317906794143,
      "grad_norm": 0.16883717477321625,
      "learning_rate": 0.0001455675992000087,
      "loss": 0.238803431391716,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_bleu": 0.734623877372401,
      "eval_ce_loss": 0.8963056962647938,
      "eval_cos_loss": 0.06929709047045066,
      "eval_loss": 0.23715336164927373,
      "eval_mse_loss": 0.14023235147674334,
      "eval_rec_loss": 0.00036073007059260167,
      "flow/cos_sim": 0.9307029282665689,
      "flow/improvement_ratio": 0.9735523003693585,
      "flow/mag_ratio_mean": 0.9296198963574623,
      "flow/mag_ratio_std": 0.07116366107203917,
      "step": 16384
    },
    {
      "epoch": 0.7567317906794143,
      "eval_bleu": 0.734623877372401,
      "eval_ce_loss": 0.8963056962647938,
      "eval_cos_loss": 0.06929709047045066,
      "eval_loss": 0.23715336164927373,
      "eval_mse_loss": 0.14023235147674334,
      "eval_rec_loss": 0.00036073007059260167,
      "eval_runtime": 144.8569,
      "eval_samples_per_second": 193.246,
      "eval_steps_per_second": 3.024,
      "flow/cos_sim": 0.9307029282665689,
      "flow/improvement_ratio": 0.9735523003693585,
      "flow/mag_ratio_mean": 0.9296198963574623,
      "flow/mag_ratio_std": 0.07116366107203917,
      "step": 16384
    },
    {
      "epoch": 0.7685557249087802,
      "grad_norm": 0.18049044907093048,
      "learning_rate": 0.000132409666069565,
      "loss": 0.23929160833358765,
      "step": 16640
    },
    {
      "epoch": 0.780379659138146,
      "grad_norm": 0.42043963074684143,
      "learning_rate": 0.0001197837462455823,
      "loss": 0.23627738654613495,
      "step": 16896
    },
    {
      "epoch": 0.7922035933675119,
      "grad_norm": 0.19926320016384125,
      "learning_rate": 0.00010770811321550749,
      "loss": 0.23734550178050995,
      "step": 17152
    },
    {
      "epoch": 0.8040275275968778,
      "grad_norm": 0.2398398518562317,
      "learning_rate": 9.620024403698591e-05,
      "loss": 0.23703321814537048,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_bleu": 0.7436939325306295,
      "eval_ce_loss": 0.8676520106999297,
      "eval_cos_loss": 0.06962325224933559,
      "eval_loss": 0.23564744887014502,
      "eval_mse_loss": 0.14157898780213643,
      "eval_rec_loss": 0.00034093371078227947,
      "flow/cos_sim": 0.9303767629410034,
      "flow/improvement_ratio": 0.9734724662075304,
      "flow/mag_ratio_mean": 0.9335192646065803,
      "flow/mag_ratio_std": 0.06777094987531503,
      "step": 17408
    },
    {
      "epoch": 0.8040275275968778,
      "eval_bleu": 0.7436939325306295,
      "eval_ce_loss": 0.8676520106999297,
      "eval_cos_loss": 0.06962325224933559,
      "eval_loss": 0.23564744887014502,
      "eval_mse_loss": 0.14157898780213643,
      "eval_rec_loss": 0.00034093371078227947,
      "eval_runtime": 144.4529,
      "eval_samples_per_second": 193.786,
      "eval_steps_per_second": 3.032,
      "flow/cos_sim": 0.9303767629410034,
      "flow/improvement_ratio": 0.9734724662075304,
      "flow/mag_ratio_mean": 0.9335192646065803,
      "flow/mag_ratio_std": 0.06777094987531503,
      "step": 17408
    },
    {
      "epoch": 0.8158514618262436,
      "grad_norm": 0.781293511390686,
      "learning_rate": 8.527679404332429e-05,
      "loss": 0.23650652170181274,
      "step": 17664
    },
    {
      "epoch": 0.8276753960556095,
      "grad_norm": 0.14831966161727905,
      "learning_rate": 7.495357273823544e-05,
      "loss": 0.2379007786512375,
      "step": 17920
    },
    {
      "epoch": 0.8394993302849753,
      "grad_norm": 0.15485990047454834,
      "learning_rate": 6.524552091475183e-05,
      "loss": 0.2373773455619812,
      "step": 18176
    },
    {
      "epoch": 0.8513232645143411,
      "grad_norm": 0.13972342014312744,
      "learning_rate": 5.6166689031422024e-05,
      "loss": 0.23547400534152985,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_bleu": 0.7404975173841,
      "eval_ce_loss": 0.8763441998664647,
      "eval_cos_loss": 0.06848628710121869,
      "eval_loss": 0.2341211859698165,
      "eval_mse_loss": 0.1393070275040522,
      "eval_rec_loss": 0.00033110864400998134,
      "flow/cos_sim": 0.9315137378701336,
      "flow/improvement_ratio": 0.9746690928663837,
      "flow/mag_ratio_mean": 0.9297302695986343,
      "flow/mag_ratio_std": 0.06708622893922406,
      "step": 18432
    },
    {
      "epoch": 0.8513232645143411,
      "eval_bleu": 0.7404975173841,
      "eval_ce_loss": 0.8763441998664647,
      "eval_cos_loss": 0.06848628710121869,
      "eval_loss": 0.2341211859698165,
      "eval_mse_loss": 0.1393070275040522,
      "eval_rec_loss": 0.00033110864400998134,
      "eval_runtime": 144.9916,
      "eval_samples_per_second": 193.066,
      "eval_steps_per_second": 3.021,
      "flow/cos_sim": 0.9315137378701336,
      "flow/improvement_ratio": 0.9746690928663837,
      "flow/mag_ratio_mean": 0.9297302695986343,
      "flow/mag_ratio_std": 0.06708622893922406,
      "step": 18432
    },
    {
      "epoch": 0.8631471987437069,
      "grad_norm": 0.13306647539138794,
      "learning_rate": 4.773021687709067e-05,
      "loss": 0.23608988523483276,
      "step": 18688
    },
    {
      "epoch": 0.8749711329730728,
      "grad_norm": 0.16064585745334625,
      "learning_rate": 3.994831455368719e-05,
      "loss": 0.23599746823310852,
      "step": 18944
    },
    {
      "epoch": 0.8867950672024387,
      "grad_norm": 0.19070403277873993,
      "learning_rate": 3.283224480455282e-05,
      "loss": 0.23425938189029694,
      "step": 19200
    },
    {
      "epoch": 0.8986190014318045,
      "grad_norm": 1.2540472745895386,
      "learning_rate": 2.639230671387627e-05,
      "loss": 0.23733244836330414,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_bleu": 0.7411587696578733,
      "eval_ce_loss": 0.8768332646862013,
      "eval_cos_loss": 0.06816527606349558,
      "eval_loss": 0.23368886810595588,
      "eval_mse_loss": 0.138866622085985,
      "eval_rec_loss": 0.00032239026382972687,
      "flow/cos_sim": 0.9318347360989819,
      "flow/improvement_ratio": 0.974329279574085,
      "flow/mag_ratio_mean": 0.9297914792141414,
      "flow/mag_ratio_std": 0.06777094615852997,
      "step": 19456
    },
    {
      "epoch": 0.8986190014318045,
      "eval_bleu": 0.7411587696578733,
      "eval_ce_loss": 0.8768332646862013,
      "eval_cos_loss": 0.06816527606349558,
      "eval_loss": 0.23368886810595588,
      "eval_mse_loss": 0.138866622085985,
      "eval_rec_loss": 0.00032239026382972687,
      "eval_runtime": 144.7333,
      "eval_samples_per_second": 193.411,
      "eval_steps_per_second": 3.026,
      "flow/cos_sim": 0.9318347360989819,
      "flow/improvement_ratio": 0.974329279574085,
      "flow/mag_ratio_mean": 0.9297914792141414,
      "flow/mag_ratio_std": 0.06777094615852997,
      "step": 19456
    },
    {
      "epoch": 0.9104429356611704,
      "grad_norm": 0.1263592690229416,
      "learning_rate": 2.063782080083576e-05,
      "loss": 0.23605787754058838,
      "step": 19712
    },
    {
      "epoch": 0.9222668698905362,
      "grad_norm": 0.12157848477363586,
      "learning_rate": 1.557711553001523e-05,
      "loss": 0.23632152378559113,
      "step": 19968
    },
    {
      "epoch": 0.9340908041199021,
      "grad_norm": 0.1504155695438385,
      "learning_rate": 1.1217515257622269e-05,
      "loss": 0.2339780181646347,
      "step": 20224
    },
    {
      "epoch": 0.945914738349268,
      "grad_norm": 0.22383330762386322,
      "learning_rate": 7.565329630950746e-06,
      "loss": 0.23464307188987732,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_bleu": 0.7380424224705183,
      "eval_ce_loss": 0.887897579365125,
      "eval_cos_loss": 0.06765534596934439,
      "eval_loss": 0.23374415763981266,
      "eval_mse_loss": 0.13786488870099256,
      "eval_rec_loss": 0.0003239751909095337,
      "flow/cos_sim": 0.9323446643134775,
      "flow/improvement_ratio": 0.9756706645499625,
      "flow/mag_ratio_mean": 0.9308563253106592,
      "flow/mag_ratio_std": 0.06712424952417748,
      "step": 20480
    },
    {
      "epoch": 0.945914738349268,
      "eval_bleu": 0.7380424224705183,
      "eval_ce_loss": 0.887897579365125,
      "eval_cos_loss": 0.06765534596934439,
      "eval_loss": 0.23374415763981266,
      "eval_mse_loss": 0.13786488870099256,
      "eval_rec_loss": 0.0003239751909095337,
      "eval_runtime": 145.0128,
      "eval_samples_per_second": 193.038,
      "eval_steps_per_second": 3.02,
      "flow/cos_sim": 0.9323446643134775,
      "flow/improvement_ratio": 0.9756706645499625,
      "flow/mag_ratio_mean": 0.9308563253106592,
      "flow/mag_ratio_std": 0.06712424952417748,
      "step": 20480
    },
    {
      "epoch": 0.9577386725786338,
      "grad_norm": 1.0795046091079712,
      "learning_rate": 4.62584445643166e-06,
      "loss": 0.23464855551719666,
      "step": 20736
    },
    {
      "epoch": 0.9695626068079997,
      "grad_norm": 0.14688174426555634,
      "learning_rate": 2.40331404948807e-06,
      "loss": 0.23495320975780487,
      "step": 20992
    },
    {
      "epoch": 0.9813865410373654,
      "grad_norm": 0.5085982084274292,
      "learning_rate": 9.009550772663965e-07,
      "loss": 0.23527467250823975,
      "step": 21248
    },
    {
      "epoch": 0.9932104752667313,
      "grad_norm": 0.1799926906824112,
      "learning_rate": 1.2094190315575791e-07,
      "loss": 0.23269928991794586,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_bleu": 0.7394941421899379,
      "eval_ce_loss": 0.8877883761032531,
      "eval_cos_loss": 0.06756893683173885,
      "eval_loss": 0.23357277302301094,
      "eval_mse_loss": 0.13771423821647963,
      "eval_rec_loss": 0.00032280229498863754,
      "flow/cos_sim": 0.932431082475131,
      "flow/improvement_ratio": 0.97383240123862,
      "flow/mag_ratio_mean": 0.9317929653544405,
      "flow/mag_ratio_std": 0.06708553743008609,
      "step": 21504
    },
    {
      "epoch": 0.9932104752667313,
      "eval_bleu": 0.7394941421899379,
      "eval_ce_loss": 0.8877883761032531,
      "eval_cos_loss": 0.06756893683173885,
      "eval_loss": 0.23357277302301094,
      "eval_mse_loss": 0.13771423821647963,
      "eval_rec_loss": 0.00032280229498863754,
      "eval_runtime": 146.0954,
      "eval_samples_per_second": 191.608,
      "eval_steps_per_second": 2.998,
      "flow/cos_sim": 0.932431082475131,
      "flow/improvement_ratio": 0.97383240123862,
      "flow/mag_ratio_mean": 0.9317929653544405,
      "flow/mag_ratio_std": 0.06708553743008609,
      "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
}