diff --git "a/last-checkpoint/trainer_state.json" "b/last-checkpoint/trainer_state.json"
--- "a/last-checkpoint/trainer_state.json"
+++ "b/last-checkpoint/trainer_state.json"
@@ -1,9 +1,9 @@
 {
-  "best_metric": 13.5764,
-  "best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-58037",
-  "epoch": 15.99935392169531,
+  "best_metric": 12.1716,
+  "best_model_checkpoint": "/kaggle/tmp/amr-tst-indo/AMRBART-id/fine-tune/../outputs/mbart-en-id-smaller-fted/checkpoint-42562",
+  "epoch": 11.0,
   "eval_steps": 500,
-  "global_step": 61904,
+  "global_step": 42562,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
@@ -12891,5848 +12891,6 @@
       "eval_samples_per_second": 2.165,
       "eval_steps_per_second": 1.083,
       "step": 42562
-    },
-    {
-      "epoch": 11.0054270577594,
-      "learning_rate": 1.3131725657980034e-07,
-      "loss": 2.4441,
-      "step": 42580
-    },
-    {
-      "epoch": 11.010595684196925,
-      "learning_rate": 1.31284843770258e-07,
-      "loss": 2.5027,
-      "step": 42600
-    },
-    {
-      "epoch": 11.015764310634449,
-      "learning_rate": 1.3125243096071566e-07,
-      "loss": 2.4709,
-      "step": 42620
-    },
-    {
-      "epoch": 11.020932937071974,
-      "learning_rate": 1.3122001815117335e-07,
-      "loss": 2.4576,
-      "step": 42640
-    },
-    {
-      "epoch": 11.026101563509497,
-      "learning_rate": 1.31187605341631e-07,
-      "loss": 2.4395,
-      "step": 42660
-    },
-    {
-      "epoch": 11.031270189947021,
-      "learning_rate": 1.3115519253208867e-07,
-      "loss": 2.504,
-      "step": 42680
-    },
-    {
-      "epoch": 11.036438816384546,
-      "learning_rate": 1.3112277972254634e-07,
-      "loss": 2.5521,
-      "step": 42700
-    },
-    {
-      "epoch": 11.04160744282207,
-      "learning_rate": 1.31090366913004e-07,
-      "loss": 2.5074,
-      "step": 42720
-    },
-    {
-      "epoch": 11.046776069259595,
-      "learning_rate": 1.3105795410346169e-07,
-      "loss": 2.5079,
-      "step": 42740
-    },
-    {
-      "epoch": 11.051944695697118,
-      "learning_rate": 1.3102554129391935e-07,
-      "loss": 2.5337,
-      "step": 42760
-    },
-    {
-      "epoch": 11.057113322134644,
-      "learning_rate": 1.30993128484377e-07,
-      "loss": 2.4815,
-      "step": 42780
-    },
-    {
-      "epoch": 11.062281948572167,
-      "learning_rate": 1.309607156748347e-07,
-      "loss": 2.4897,
-      "step": 42800
-    },
-    {
-      "epoch": 11.06745057500969,
-      "learning_rate": 1.3092830286529236e-07,
-      "loss": 2.5184,
-      "step": 42820
-    },
-    {
-      "epoch": 11.072619201447216,
-      "learning_rate": 1.3089589005575002e-07,
-      "loss": 2.459,
-      "step": 42840
-    },
-    {
-      "epoch": 11.07778782788474,
-      "learning_rate": 1.308634772462077e-07,
-      "loss": 2.4623,
-      "step": 42860
-    },
-    {
-      "epoch": 11.082956454322265,
-      "learning_rate": 1.3083106443666535e-07,
-      "loss": 2.4422,
-      "step": 42880
-    },
-    {
-      "epoch": 11.088125080759788,
-      "learning_rate": 1.3079865162712303e-07,
-      "loss": 2.5212,
-      "step": 42900
-    },
-    {
-      "epoch": 11.093293707197311,
-      "learning_rate": 1.307662388175807e-07,
-      "loss": 2.4923,
-      "step": 42920
-    },
-    {
-      "epoch": 11.098462333634837,
-      "learning_rate": 1.3073382600803836e-07,
-      "loss": 2.5076,
-      "step": 42940
-    },
-    {
-      "epoch": 11.10363096007236,
-      "learning_rate": 1.3070141319849605e-07,
-      "loss": 2.5291,
-      "step": 42960
-    },
-    {
-      "epoch": 11.108799586509885,
-      "learning_rate": 1.306690003889537e-07,
-      "loss": 2.4459,
-      "step": 42980
-    },
-    {
-      "epoch": 11.113968212947409,
-      "learning_rate": 1.3063658757941137e-07,
-      "loss": 2.4522,
-      "step": 43000
-    },
-    {
-      "epoch": 11.119136839384934,
-      "learning_rate": 1.3060417476986906e-07,
-      "loss": 2.4752,
-      "step": 43020
-    },
-    {
-      "epoch": 11.124305465822458,
-      "learning_rate": 1.3057176196032672e-07,
-      "loss": 2.4877,
-      "step": 43040
-    },
-    {
-      "epoch": 11.129474092259981,
-      "learning_rate": 1.3053934915078438e-07,
-      "loss": 2.4969,
-      "step": 43060
-    },
-    {
-      "epoch": 11.134642718697506,
-      "learning_rate": 1.3050693634124207e-07,
-      "loss": 2.5331,
-      "step": 43080
-    },
-    {
-      "epoch": 11.13981134513503,
-      "learning_rate": 1.304745235316997e-07,
-      "loss": 2.4214,
-      "step": 43100
-    },
-    {
-      "epoch": 11.144979971572555,
-      "learning_rate": 1.304421107221574e-07,
-      "loss": 2.4795,
-      "step": 43120
-    },
-    {
-      "epoch": 11.150148598010079,
-      "learning_rate": 1.3040969791261506e-07,
-      "loss": 2.491,
-      "step": 43140
-    },
-    {
-      "epoch": 11.155317224447604,
-      "learning_rate": 1.3037728510307272e-07,
-      "loss": 2.4823,
-      "step": 43160
-    },
-    {
-      "epoch": 11.160485850885127,
-      "learning_rate": 1.303448722935304e-07,
-      "loss": 2.5129,
-      "step": 43180
-    },
-    {
-      "epoch": 11.16565447732265,
-      "learning_rate": 1.3031245948398807e-07,
-      "loss": 2.4543,
-      "step": 43200
-    },
-    {
-      "epoch": 11.170823103760176,
-      "learning_rate": 1.3028004667444573e-07,
-      "loss": 2.4735,
-      "step": 43220
-    },
-    {
-      "epoch": 11.1759917301977,
-      "learning_rate": 1.3024763386490342e-07,
-      "loss": 2.5127,
-      "step": 43240
-    },
-    {
-      "epoch": 11.181160356635225,
-      "learning_rate": 1.3021522105536105e-07,
-      "loss": 2.4045,
-      "step": 43260
-    },
-    {
-      "epoch": 11.186328983072748,
-      "learning_rate": 1.3018280824581874e-07,
-      "loss": 2.4599,
-      "step": 43280
-    },
-    {
-      "epoch": 11.191497609510273,
-      "learning_rate": 1.301503954362764e-07,
-      "loss": 2.4875,
-      "step": 43300
-    },
-    {
-      "epoch": 11.196666235947797,
-      "learning_rate": 1.3011798262673407e-07,
-      "loss": 2.4523,
-      "step": 43320
-    },
-    {
-      "epoch": 11.20183486238532,
-      "learning_rate": 1.3008556981719175e-07,
-      "loss": 2.47,
-      "step": 43340
-    },
-    {
-      "epoch": 11.207003488822846,
-      "learning_rate": 1.3005315700764942e-07,
-      "loss": 2.464,
-      "step": 43360
-    },
-    {
-      "epoch": 11.212172115260369,
-      "learning_rate": 1.3002074419810708e-07,
-      "loss": 2.5223,
-      "step": 43380
-    },
-    {
-      "epoch": 11.217340741697894,
-      "learning_rate": 1.2998833138856477e-07,
-      "loss": 2.4749,
-      "step": 43400
-    },
-    {
-      "epoch": 11.222509368135418,
-      "learning_rate": 1.2995591857902243e-07,
-      "loss": 2.5268,
-      "step": 43420
-    },
-    {
-      "epoch": 11.227677994572943,
-      "learning_rate": 1.299235057694801e-07,
-      "loss": 2.4837,
-      "step": 43440
-    },
-    {
-      "epoch": 11.232846621010466,
-      "learning_rate": 1.2989109295993778e-07,
-      "loss": 2.4285,
-      "step": 43460
-    },
-    {
-      "epoch": 11.23801524744799,
-      "learning_rate": 1.2985868015039541e-07,
-      "loss": 2.4918,
-      "step": 43480
-    },
-    {
-      "epoch": 11.243183873885515,
-      "learning_rate": 1.298262673408531e-07,
-      "loss": 2.4585,
-      "step": 43500
-    },
-    {
-      "epoch": 11.248352500323039,
-      "learning_rate": 1.2979385453131076e-07,
-      "loss": 2.5156,
-      "step": 43520
-    },
-    {
-      "epoch": 11.253521126760564,
-      "learning_rate": 1.2976144172176843e-07,
-      "loss": 2.5031,
-      "step": 43540
-    },
-    {
-      "epoch": 11.258689753198087,
-      "learning_rate": 1.2972902891222611e-07,
-      "loss": 2.456,
-      "step": 43560
-    },
-    {
-      "epoch": 11.263858379635613,
-      "learning_rate": 1.2969661610268378e-07,
-      "loss": 2.5083,
-      "step": 43580
-    },
-    {
-      "epoch": 11.269027006073136,
-      "learning_rate": 1.2966420329314144e-07,
-      "loss": 2.5241,
-      "step": 43600
-    },
-    {
-      "epoch": 11.27419563251066,
-      "learning_rate": 1.2963179048359913e-07,
-      "loss": 2.4751,
-      "step": 43620
-    },
-    {
-      "epoch": 11.279364258948185,
-      "learning_rate": 1.295993776740568e-07,
-      "loss": 2.4695,
-      "step": 43640
-    },
-    {
-      "epoch": 11.284532885385708,
-      "learning_rate": 1.2956696486451445e-07,
-      "loss": 2.5578,
-      "step": 43660
-    },
-    {
-      "epoch": 11.289701511823234,
-      "learning_rate": 1.2953455205497214e-07,
-      "loss": 2.4649,
-      "step": 43680
-    },
-    {
-      "epoch": 11.294870138260757,
-      "learning_rate": 1.2950213924542977e-07,
-      "loss": 2.4747,
-      "step": 43700
-    },
-    {
-      "epoch": 11.300038764698282,
-      "learning_rate": 1.2946972643588746e-07,
-      "loss": 2.4679,
-      "step": 43720
-    },
-    {
-      "epoch": 11.305207391135806,
-      "learning_rate": 1.2943731362634512e-07,
-      "loss": 2.4683,
-      "step": 43740
-    },
-    {
-      "epoch": 11.31037601757333,
-      "learning_rate": 1.2940490081680279e-07,
-      "loss": 2.4857,
-      "step": 43760
-    },
-    {
-      "epoch": 11.315544644010854,
-      "learning_rate": 1.2937248800726047e-07,
-      "loss": 2.426,
-      "step": 43780
-    },
-    {
-      "epoch": 11.320713270448378,
-      "learning_rate": 1.2934007519771814e-07,
-      "loss": 2.4733,
-      "step": 43800
-    },
-    {
-      "epoch": 11.325881896885903,
-      "learning_rate": 1.293076623881758e-07,
-      "loss": 2.4812,
-      "step": 43820
-    },
-    {
-      "epoch": 11.331050523323427,
-      "learning_rate": 1.2927524957863349e-07,
-      "loss": 2.4528,
-      "step": 43840
-    },
-    {
-      "epoch": 11.33621914976095,
-      "learning_rate": 1.2924283676909112e-07,
-      "loss": 2.494,
-      "step": 43860
-    },
-    {
-      "epoch": 11.341387776198475,
-      "learning_rate": 1.292104239595488e-07,
-      "loss": 2.4795,
-      "step": 43880
-    },
-    {
-      "epoch": 11.346556402635999,
-      "learning_rate": 1.2917801115000647e-07,
-      "loss": 2.4816,
-      "step": 43900
-    },
-    {
-      "epoch": 11.351725029073524,
-      "learning_rate": 1.2914559834046413e-07,
-      "loss": 2.5282,
-      "step": 43920
-    },
-    {
-      "epoch": 11.356893655511048,
-      "learning_rate": 1.2911318553092182e-07,
-      "loss": 2.4849,
-      "step": 43940
-    },
-    {
-      "epoch": 11.362062281948573,
-      "learning_rate": 1.2908077272137948e-07,
-      "loss": 2.4246,
-      "step": 43960
-    },
-    {
-      "epoch": 11.367230908386096,
-      "learning_rate": 1.2904835991183715e-07,
-      "loss": 2.5491,
-      "step": 43980
-    },
-    {
-      "epoch": 11.372399534823622,
-      "learning_rate": 1.2901594710229483e-07,
-      "loss": 2.4584,
-      "step": 44000
-    },
-    {
-      "epoch": 11.377568161261145,
-      "learning_rate": 1.289835342927525e-07,
-      "loss": 2.5351,
-      "step": 44020
-    },
-    {
-      "epoch": 11.382736787698668,
-      "learning_rate": 1.2895112148321016e-07,
-      "loss": 2.5108,
-      "step": 44040
-    },
-    {
-      "epoch": 11.387905414136194,
-      "learning_rate": 1.2891870867366782e-07,
-      "loss": 2.5328,
-      "step": 44060
-    },
-    {
-      "epoch": 11.393074040573717,
-      "learning_rate": 1.2888629586412548e-07,
-      "loss": 2.4501,
-      "step": 44080
-    },
-    {
-      "epoch": 11.398242667011242,
-      "learning_rate": 1.2885388305458317e-07,
-      "loss": 2.434,
-      "step": 44100
-    },
-    {
-      "epoch": 11.403411293448766,
-      "learning_rate": 1.2882147024504083e-07,
-      "loss": 2.4713,
-      "step": 44120
-    },
-    {
-      "epoch": 11.40857991988629,
-      "learning_rate": 1.287890574354985e-07,
-      "loss": 2.5093,
-      "step": 44140
-    },
-    {
-      "epoch": 11.413748546323815,
-      "learning_rate": 1.2875664462595618e-07,
-      "loss": 2.5031,
-      "step": 44160
-    },
-    {
-      "epoch": 11.418917172761338,
-      "learning_rate": 1.2872423181641384e-07,
-      "loss": 2.4712,
-      "step": 44180
-    },
-    {
-      "epoch": 11.424085799198863,
-      "learning_rate": 1.286918190068715e-07,
-      "loss": 2.4963,
-      "step": 44200
-    },
-    {
-      "epoch": 11.429254425636387,
-      "learning_rate": 1.2865940619732917e-07,
-      "loss": 2.4866,
-      "step": 44220
-    },
-    {
-      "epoch": 11.434423052073912,
-      "learning_rate": 1.2862699338778686e-07,
-      "loss": 2.4665,
-      "step": 44240
-    },
-    {
-      "epoch": 11.439591678511436,
-      "learning_rate": 1.2859458057824452e-07,
-      "loss": 2.4525,
-      "step": 44260
-    },
-    {
-      "epoch": 11.444760304948959,
-      "learning_rate": 1.2856216776870218e-07,
-      "loss": 2.5271,
-      "step": 44280
-    },
-    {
-      "epoch": 11.449928931386484,
-      "learning_rate": 1.2852975495915984e-07,
-      "loss": 2.4676,
-      "step": 44300
-    },
-    {
-      "epoch": 11.455097557824008,
-      "learning_rate": 1.2849734214961753e-07,
-      "loss": 2.4996,
-      "step": 44320
-    },
-    {
-      "epoch": 11.460266184261533,
-      "learning_rate": 1.284649293400752e-07,
-      "loss": 2.4901,
-      "step": 44340
-    },
-    {
-      "epoch": 11.465434810699056,
-      "learning_rate": 1.2843251653053285e-07,
-      "loss": 2.411,
-      "step": 44360
-    },
-    {
-      "epoch": 11.470603437136582,
-      "learning_rate": 1.2840010372099054e-07,
-      "loss": 2.4688,
-      "step": 44380
-    },
-    {
-      "epoch": 11.475772063574105,
-      "learning_rate": 1.283676909114482e-07,
-      "loss": 2.4458,
-      "step": 44400
-    },
-    {
-      "epoch": 11.480940690011629,
-      "learning_rate": 1.2833527810190587e-07,
-      "loss": 2.4871,
-      "step": 44420
-    },
-    {
-      "epoch": 11.486109316449154,
-      "learning_rate": 1.2830286529236353e-07,
-      "loss": 2.4476,
-      "step": 44440
-    },
-    {
-      "epoch": 11.491277942886677,
-      "learning_rate": 1.282704524828212e-07,
-      "loss": 2.4647,
-      "step": 44460
-    },
-    {
-      "epoch": 11.496446569324203,
-      "learning_rate": 1.2823803967327888e-07,
-      "loss": 2.4887,
-      "step": 44480
-    },
-    {
-      "epoch": 11.501615195761726,
-      "learning_rate": 1.2820562686373654e-07,
-      "loss": 2.4835,
-      "step": 44500
-    },
-    {
-      "epoch": 11.506783822199251,
-      "learning_rate": 1.281732140541942e-07,
-      "loss": 2.4435,
-      "step": 44520
-    },
-    {
-      "epoch": 11.511952448636775,
-      "learning_rate": 1.281408012446519e-07,
-      "loss": 2.4716,
-      "step": 44540
-    },
-    {
-      "epoch": 11.517121075074298,
-      "learning_rate": 1.2810838843510955e-07,
-      "loss": 2.457,
-      "step": 44560
-    },
-    {
-      "epoch": 11.522289701511824,
-      "learning_rate": 1.2807597562556721e-07,
-      "loss": 2.4465,
-      "step": 44580
-    },
-    {
-      "epoch": 11.527458327949347,
-      "learning_rate": 1.2804356281602488e-07,
-      "loss": 2.4804,
-      "step": 44600
-    },
-    {
-      "epoch": 11.532626954386872,
-      "learning_rate": 1.2801115000648256e-07,
-      "loss": 2.5168,
-      "step": 44620
-    },
-    {
-      "epoch": 11.537795580824396,
-      "learning_rate": 1.2797873719694023e-07,
-      "loss": 2.4759,
-      "step": 44640
-    },
-    {
-      "epoch": 11.542964207261921,
-      "learning_rate": 1.279463243873979e-07,
-      "loss": 2.4929,
-      "step": 44660
-    },
-    {
-      "epoch": 11.548132833699444,
-      "learning_rate": 1.2791391157785555e-07,
-      "loss": 2.4136,
-      "step": 44680
-    },
-    {
-      "epoch": 11.553301460136968,
-      "learning_rate": 1.2788149876831324e-07,
-      "loss": 2.4497,
-      "step": 44700
-    },
-    {
-      "epoch": 11.558470086574493,
-      "learning_rate": 1.278490859587709e-07,
-      "loss": 2.4912,
-      "step": 44720
-    },
-    {
-      "epoch": 11.563638713012017,
-      "learning_rate": 1.2781667314922856e-07,
-      "loss": 2.5089,
-      "step": 44740
-    },
-    {
-      "epoch": 11.568807339449542,
-      "learning_rate": 1.2778426033968622e-07,
-      "loss": 2.4516,
-      "step": 44760
-    },
-    {
-      "epoch": 11.573975965887065,
-      "learning_rate": 1.277518475301439e-07,
-      "loss": 2.5023,
-      "step": 44780
-    },
-    {
-      "epoch": 11.579144592324589,
-      "learning_rate": 1.2771943472060157e-07,
-      "loss": 2.4883,
-      "step": 44800
-    },
-    {
-      "epoch": 11.584313218762114,
-      "learning_rate": 1.2768702191105924e-07,
-      "loss": 2.4885,
-      "step": 44820
-    },
-    {
-      "epoch": 11.589481845199638,
-      "learning_rate": 1.2765460910151692e-07,
-      "loss": 2.5421,
-      "step": 44840
-    },
-    {
-      "epoch": 11.594650471637163,
-      "learning_rate": 1.2762219629197459e-07,
-      "loss": 2.4938,
-      "step": 44860
-    },
-    {
-      "epoch": 11.599819098074686,
-      "learning_rate": 1.2758978348243225e-07,
-      "loss": 2.4712,
-      "step": 44880
-    },
-    {
-      "epoch": 11.604987724512211,
-      "learning_rate": 1.275573706728899e-07,
-      "loss": 2.5227,
-      "step": 44900
-    },
-    {
-      "epoch": 11.610156350949735,
-      "learning_rate": 1.275249578633476e-07,
-      "loss": 2.4463,
-      "step": 44920
-    },
-    {
-      "epoch": 11.61532497738726,
-      "learning_rate": 1.2749254505380526e-07,
-      "loss": 2.4526,
-      "step": 44940
-    },
-    {
-      "epoch": 11.620493603824784,
-      "learning_rate": 1.2746013224426292e-07,
-      "loss": 2.4763,
-      "step": 44960
-    },
-    {
-      "epoch": 11.625662230262307,
-      "learning_rate": 1.2742771943472058e-07,
-      "loss": 2.5575,
-      "step": 44980
-    },
-    {
-      "epoch": 11.630830856699832,
-      "learning_rate": 1.2739530662517827e-07,
-      "loss": 2.5079,
-      "step": 45000
-    },
-    {
-      "epoch": 11.635999483137356,
-      "learning_rate": 1.2736289381563593e-07,
-      "loss": 2.4868,
-      "step": 45020
-    },
-    {
-      "epoch": 11.641168109574881,
-      "learning_rate": 1.273304810060936e-07,
-      "loss": 2.4547,
-      "step": 45040
-    },
-    {
-      "epoch": 11.646336736012405,
-      "learning_rate": 1.2729806819655126e-07,
-      "loss": 2.4594,
-      "step": 45060
-    },
-    {
-      "epoch": 11.651505362449928,
-      "learning_rate": 1.2726565538700895e-07,
-      "loss": 2.495,
-      "step": 45080
-    },
-    {
-      "epoch": 11.656673988887453,
-      "learning_rate": 1.272332425774666e-07,
-      "loss": 2.5174,
-      "step": 45100
-    },
-    {
-      "epoch": 11.661842615324977,
-      "learning_rate": 1.2720082976792427e-07,
-      "loss": 2.4223,
-      "step": 45120
-    },
-    {
-      "epoch": 11.667011241762502,
-      "learning_rate": 1.2716841695838196e-07,
-      "loss": 2.4523,
-      "step": 45140
-    },
-    {
-      "epoch": 11.672179868200026,
-      "learning_rate": 1.2713600414883962e-07,
-      "loss": 2.4271,
-      "step": 45160
-    },
-    {
-      "epoch": 11.67734849463755,
-      "learning_rate": 1.2710359133929728e-07,
-      "loss": 2.5235,
-      "step": 45180
-    },
-    {
-      "epoch": 11.682517121075074,
-      "learning_rate": 1.2707117852975494e-07,
-      "loss": 2.4666,
-      "step": 45200
-    },
-    {
-      "epoch": 11.687685747512598,
-      "learning_rate": 1.2703876572021263e-07,
-      "loss": 2.4413,
-      "step": 45220
-    },
-    {
-      "epoch": 11.692854373950123,
-      "learning_rate": 1.270063529106703e-07,
-      "loss": 2.422,
-      "step": 45240
-    },
-    {
-      "epoch": 11.698023000387646,
-      "learning_rate": 1.2697394010112796e-07,
-      "loss": 2.5057,
-      "step": 45260
-    },
-    {
-      "epoch": 11.703191626825172,
-      "learning_rate": 1.2694152729158562e-07,
-      "loss": 2.5153,
-      "step": 45280
-    },
-    {
-      "epoch": 11.708360253262695,
-      "learning_rate": 1.269091144820433e-07,
-      "loss": 2.4894,
-      "step": 45300
-    },
-    {
-      "epoch": 11.71352887970022,
-      "learning_rate": 1.2687670167250097e-07,
-      "loss": 2.5211,
-      "step": 45320
-    },
-    {
-      "epoch": 11.718697506137744,
-      "learning_rate": 1.2684428886295863e-07,
-      "loss": 2.4449,
-      "step": 45340
-    },
-    {
-      "epoch": 11.723866132575267,
-      "learning_rate": 1.268118760534163e-07,
-      "loss": 2.4859,
-      "step": 45360
-    },
-    {
-      "epoch": 11.729034759012793,
-      "learning_rate": 1.2677946324387398e-07,
-      "loss": 2.4578,
-      "step": 45380
-    },
-    {
-      "epoch": 11.734203385450316,
-      "learning_rate": 1.2674705043433164e-07,
-      "loss": 2.4715,
-      "step": 45400
-    },
-    {
-      "epoch": 11.739372011887841,
-      "learning_rate": 1.267146376247893e-07,
-      "loss": 2.4258,
-      "step": 45420
-    },
-    {
-      "epoch": 11.744540638325365,
-      "learning_rate": 1.26682224815247e-07,
-      "loss": 2.4635,
-      "step": 45440
-    },
-    {
-      "epoch": 11.74970926476289,
-      "learning_rate": 1.2664981200570465e-07,
-      "loss": 2.4803,
-      "step": 45460
-    },
-    {
-      "epoch": 11.754877891200413,
-      "learning_rate": 1.2661739919616232e-07,
-      "loss": 2.4333,
-      "step": 45480
-    },
-    {
-      "epoch": 11.760046517637937,
-      "learning_rate": 1.2658498638661998e-07,
-      "loss": 2.5486,
-      "step": 45500
-    },
-    {
-      "epoch": 11.765215144075462,
-      "learning_rate": 1.2655257357707767e-07,
-      "loss": 2.4803,
-      "step": 45520
-    },
-    {
-      "epoch": 11.770383770512986,
-      "learning_rate": 1.2652016076753533e-07,
-      "loss": 2.4958,
-      "step": 45540
-    },
-    {
-      "epoch": 11.775552396950511,
-      "learning_rate": 1.26487747957993e-07,
-      "loss": 2.4918,
-      "step": 45560
-    },
-    {
-      "epoch": 11.780721023388034,
-      "learning_rate": 1.2645533514845065e-07,
-      "loss": 2.513,
-      "step": 45580
-    },
-    {
-      "epoch": 11.78588964982556,
-      "learning_rate": 1.2642292233890834e-07,
-      "loss": 2.4897,
-      "step": 45600
-    },
-    {
-      "epoch": 11.791058276263083,
-      "learning_rate": 1.26390509529366e-07,
-      "loss": 2.4468,
-      "step": 45620
-    },
-    {
-      "epoch": 11.796226902700607,
-      "learning_rate": 1.2635809671982366e-07,
-      "loss": 2.4928,
-      "step": 45640
-    },
-    {
-      "epoch": 11.801395529138132,
-      "learning_rate": 1.2632568391028132e-07,
-      "loss": 2.4793,
-      "step": 45660
-    },
-    {
-      "epoch": 11.806564155575655,
-      "learning_rate": 1.26293271100739e-07,
-      "loss": 2.4788,
-      "step": 45680
-    },
-    {
-      "epoch": 11.81173278201318,
-      "learning_rate": 1.2626085829119668e-07,
-      "loss": 2.4913,
-      "step": 45700
-    },
-    {
-      "epoch": 11.816901408450704,
-      "learning_rate": 1.2622844548165434e-07,
-      "loss": 2.4966,
-      "step": 45720
-    },
-    {
-      "epoch": 11.82207003488823,
-      "learning_rate": 1.2619603267211203e-07,
-      "loss": 2.4954,
-      "step": 45740
-    },
-    {
-      "epoch": 11.827238661325753,
-      "learning_rate": 1.261636198625697e-07,
-      "loss": 2.4608,
-      "step": 45760
-    },
-    {
-      "epoch": 11.832407287763276,
-      "learning_rate": 1.2613120705302735e-07,
-      "loss": 2.4901,
-      "step": 45780
-    },
-    {
-      "epoch": 11.837575914200801,
-      "learning_rate": 1.26098794243485e-07,
-      "loss": 2.5462,
-      "step": 45800
-    },
-    {
-      "epoch": 11.842744540638325,
-      "learning_rate": 1.260663814339427e-07,
-      "loss": 2.5217,
-      "step": 45820
-    },
-    {
-      "epoch": 11.84791316707585,
-      "learning_rate": 1.2603396862440036e-07,
-      "loss": 2.4889,
-      "step": 45840
-    },
-    {
-      "epoch": 11.853081793513374,
-      "learning_rate": 1.2600155581485802e-07,
-      "loss": 2.5021,
-      "step": 45860
-    },
-    {
-      "epoch": 11.858250419950899,
-      "learning_rate": 1.2596914300531568e-07,
-      "loss": 2.5027,
-      "step": 45880
-    },
-    {
-      "epoch": 11.863419046388422,
-      "learning_rate": 1.2593673019577337e-07,
-      "loss": 2.4681,
-      "step": 45900
-    },
-    {
-      "epoch": 11.868587672825946,
-      "learning_rate": 1.2590431738623104e-07,
-      "loss": 2.409,
-      "step": 45920
-    },
-    {
-      "epoch": 11.873756299263471,
-      "learning_rate": 1.258719045766887e-07,
-      "loss": 2.4426,
-      "step": 45940
-    },
-    {
-      "epoch": 11.878924925700995,
-      "learning_rate": 1.2583949176714636e-07,
-      "loss": 2.4285,
-      "step": 45960
-    },
-    {
-      "epoch": 11.88409355213852,
-      "learning_rate": 1.2580707895760405e-07,
-      "loss": 2.4452,
-      "step": 45980
-    },
-    {
-      "epoch": 11.889262178576043,
-      "learning_rate": 1.257746661480617e-07,
-      "loss": 2.4816,
-      "step": 46000
-    },
-    {
-      "epoch": 11.894430805013567,
-      "learning_rate": 1.2574225333851937e-07,
-      "loss": 2.4585,
-      "step": 46020
-    },
-    {
-      "epoch": 11.899599431451092,
-      "learning_rate": 1.2570984052897706e-07,
-      "loss": 2.4171,
-      "step": 46040
-    },
-    {
-      "epoch": 11.904768057888615,
-      "learning_rate": 1.2567742771943472e-07,
-      "loss": 2.4597,
-      "step": 46060
-    },
-    {
-      "epoch": 11.90993668432614,
-      "learning_rate": 1.2564501490989238e-07,
-      "loss": 2.4195,
-      "step": 46080
-    },
-    {
-      "epoch": 11.915105310763664,
-      "learning_rate": 1.2561260210035004e-07,
-      "loss": 2.5034,
-      "step": 46100
-    },
-    {
-      "epoch": 11.92027393720119,
-      "learning_rate": 1.2558018929080773e-07,
-      "loss": 2.4407,
-      "step": 46120
-    },
-    {
-      "epoch": 11.925442563638713,
-      "learning_rate": 1.255477764812654e-07,
-      "loss": 2.4813,
-      "step": 46140
-    },
-    {
-      "epoch": 11.930611190076238,
-      "learning_rate": 1.2551536367172306e-07,
-      "loss": 2.4599,
-      "step": 46160
-    },
-    {
-      "epoch": 11.935779816513762,
-      "learning_rate": 1.2548295086218072e-07,
-      "loss": 2.4641,
-      "step": 46180
-    },
-    {
-      "epoch": 11.940948442951285,
-      "learning_rate": 1.254505380526384e-07,
-      "loss": 2.5262,
-      "step": 46200
-    },
-    {
-      "epoch": 11.94611706938881,
-      "learning_rate": 1.2541812524309607e-07,
-      "loss": 2.4518,
-      "step": 46220
-    },
-    {
-      "epoch": 11.951285695826334,
-      "learning_rate": 1.2538571243355373e-07,
-      "loss": 2.5018,
-      "step": 46240
-    },
-    {
-      "epoch": 11.956454322263859,
-      "learning_rate": 1.253532996240114e-07,
-      "loss": 2.4634,
-      "step": 46260
-    },
-    {
-      "epoch": 11.961622948701383,
-      "learning_rate": 1.2532088681446908e-07,
-      "loss": 2.4889,
-      "step": 46280
-    },
-    {
-      "epoch": 11.966791575138906,
-      "learning_rate": 1.2528847400492674e-07,
-      "loss": 2.4577,
-      "step": 46300
-    },
-    {
-      "epoch": 11.971960201576431,
-      "learning_rate": 1.252560611953844e-07,
-      "loss": 2.4812,
-      "step": 46320
-    },
-    {
-      "epoch": 11.977128828013955,
-      "learning_rate": 1.252236483858421e-07,
-      "loss": 2.4448,
-      "step": 46340
-    },
-    {
-      "epoch": 11.98229745445148,
-      "learning_rate": 1.2519123557629975e-07,
-      "loss": 2.4594,
-      "step": 46360
-    },
-    {
-      "epoch": 11.987466080889003,
-      "learning_rate": 1.2515882276675742e-07,
-      "loss": 2.4473,
-      "step": 46380
-    },
-    {
-      "epoch": 11.992634707326529,
-      "learning_rate": 1.2512640995721508e-07,
-      "loss": 2.4831,
-      "step": 46400
-    },
-    {
-      "epoch": 11.997803333764052,
-      "learning_rate": 1.2509399714767277e-07,
-      "loss": 2.4706,
-      "step": 46420
-    },
-    {
-      "epoch": 11.999870784339063,
-      "eval_bleu": 12.51,
-      "eval_gen_len": 37.4146,
-      "eval_loss": 2.4478776454925537,
-      "eval_runtime": 807.7318,
-      "eval_samples_per_second": 2.132,
-      "eval_steps_per_second": 1.066,
-      "step": 46428
-    },
-    {
-      "epoch": 12.002971960201576,
-      "learning_rate": 1.2506158433813043e-07,
-      "loss": 2.4527,
-      "step": 46440
-    },
-    {
-      "epoch": 12.008140586639101,
-      "learning_rate": 1.250291715285881e-07,
-      "loss": 2.436,
-      "step": 46460
-    },
-    {
-      "epoch": 12.013309213076624,
-      "learning_rate": 1.2499675871904575e-07,
-      "loss": 2.4514,
-      "step": 46480
-    },
-    {
-      "epoch": 12.01847783951415,
-      "learning_rate": 1.2496434590950344e-07,
-      "loss": 2.4921,
-      "step": 46500
-    },
-    {
-      "epoch": 12.023646465951673,
-      "learning_rate": 1.249319330999611e-07,
-      "loss": 2.4637,
-      "step": 46520
-    },
-    {
-      "epoch": 12.028815092389198,
-      "learning_rate": 1.2489952029041876e-07,
-      "loss": 2.4924,
-      "step": 46540
-    },
-    {
-      "epoch": 12.033983718826722,
-      "learning_rate": 1.2486710748087643e-07,
-      "loss": 2.4788,
-      "step": 46560
-    },
-    {
-      "epoch": 12.039152345264245,
-      "learning_rate": 1.2483469467133411e-07,
-      "loss": 2.5291,
-      "step": 46580
-    },
-    {
-      "epoch": 12.04432097170177,
-      "learning_rate": 1.2480228186179178e-07,
-      "loss": 2.4419,
-      "step": 46600
-    },
-    {
-      "epoch": 12.049489598139294,
-      "learning_rate": 1.2476986905224944e-07,
-      "loss": 2.4768,
-      "step": 46620
-    },
-    {
-      "epoch": 12.05465822457682,
-      "learning_rate": 1.2473745624270713e-07,
-      "loss": 2.444,
-      "step": 46640
-    },
-    {
-      "epoch": 12.059826851014343,
-      "learning_rate": 1.247050434331648e-07,
-      "loss": 2.5193,
-      "step": 46660
-    },
-    {
-      "epoch": 12.064995477451868,
-      "learning_rate": 1.2467263062362245e-07,
-      "loss": 2.5019,
-      "step": 46680
-    },
-    {
-      "epoch": 12.070164103889391,
-      "learning_rate": 1.246402178140801e-07,
-      "loss": 2.465,
-      "step": 46700
-    },
-    {
-      "epoch": 12.075332730326915,
-      "learning_rate": 1.246078050045378e-07,
-      "loss": 2.4508,
-      "step": 46720
-    },
-    {
-      "epoch": 12.08050135676444,
-      "learning_rate": 1.2457539219499546e-07,
-      "loss": 2.4784,
-      "step": 46740
-    },
-    {
-      "epoch": 12.085669983201964,
-      "learning_rate": 1.2454297938545312e-07,
-      "loss": 2.5013,
-      "step": 46760
-    },
-    {
-      "epoch": 12.090838609639489,
-      "learning_rate": 1.2451056657591079e-07,
-      "loss": 2.3997,
-      "step": 46780
-    },
-    {
-      "epoch": 12.096007236077012,
-      "learning_rate": 1.2447815376636847e-07,
-      "loss": 2.4475,
-      "step": 46800
-    },
-    {
-      "epoch": 12.101175862514538,
-      "learning_rate": 1.2444574095682614e-07,
-      "loss": 2.4564,
-      "step": 46820
-    },
-    {
-      "epoch": 12.106344488952061,
-      "learning_rate": 1.244133281472838e-07,
-      "loss": 2.5028,
-      "step": 46840
-    },
-    {
-      "epoch": 12.111513115389585,
-      "learning_rate": 1.2438091533774146e-07,
-      "loss": 2.4969,
-      "step": 46860
-    },
-    {
-      "epoch": 12.11668174182711,
-      "learning_rate": 1.2434850252819915e-07,
-      "loss": 2.4488,
-      "step": 46880
-    },
-    {
-      "epoch": 12.121850368264633,
-      "learning_rate": 1.243160897186568e-07,
-      "loss": 2.4779,
-      "step": 46900
-    },
-    {
-      "epoch": 12.127018994702158,
-      "learning_rate": 1.2428367690911447e-07,
-      "loss": 2.4605,
-      "step": 46920
-    },
-    {
-      "epoch": 12.132187621139682,
-      "learning_rate": 1.2425126409957216e-07,
-      "loss": 2.4592,
-      "step": 46940
-    },
-    {
-      "epoch": 12.137356247577207,
-      "learning_rate": 1.2421885129002982e-07,
-      "loss": 2.5115,
-      "step": 46960
-    },
-    {
-      "epoch": 12.14252487401473,
-      "learning_rate": 1.2418643848048748e-07,
-      "loss": 2.494,
-      "step": 46980
-    },
-    {
-      "epoch": 12.147693500452254,
-      "learning_rate": 1.2415402567094515e-07,
-      "loss": 2.4545,
-      "step": 47000
-    },
-    {
-      "epoch": 12.15286212688978,
-      "learning_rate": 1.2412161286140283e-07,
-      "loss": 2.5211,
-      "step": 47020
-    },
-    {
-      "epoch": 12.158030753327303,
-      "learning_rate": 1.240892000518605e-07,
-      "loss": 2.4661,
-      "step": 47040
-    },
-    {
-      "epoch": 12.163199379764828,
-      "learning_rate": 1.2405678724231816e-07,
-      "loss": 2.4593,
-      "step": 47060
-    },
-    {
-      "epoch": 12.168368006202352,
-      "learning_rate": 1.2402437443277582e-07,
-      "loss": 2.555,
-      "step": 47080
-    },
-    {
-      "epoch": 12.173536632639877,
-      "learning_rate": 1.239919616232335e-07,
-      "loss": 2.5006,
-      "step": 47100
-    },
-    {
-      "epoch": 12.1787052590774,
-      "learning_rate": 1.2395954881369117e-07,
-      "loss": 2.4607,
-      "step": 47120
-    },
-    {
-      "epoch": 12.183873885514924,
-      "learning_rate": 1.2392713600414883e-07,
-      "loss": 2.4595,
-      "step": 47140
-    },
-    {
-      "epoch": 12.189042511952449,
-      "learning_rate": 1.238947231946065e-07,
-      "loss": 2.444,
-      "step": 47160
-    },
-    {
-      "epoch": 12.194211138389973,
-      "learning_rate": 1.2386231038506418e-07,
-      "loss": 2.4657,
-      "step": 47180
-    },
-    {
-      "epoch": 12.199379764827498,
-      "learning_rate": 1.2382989757552184e-07,
-      "loss": 2.454,
-      "step": 47200
-    },
-    {
-      "epoch": 12.204548391265021,
-      "learning_rate": 1.237974847659795e-07,
-      "loss": 2.4939,
-      "step": 47220
-    },
-    {
-      "epoch": 12.209717017702545,
-      "learning_rate": 1.237650719564372e-07,
-      "loss": 2.4465,
-      "step": 47240
-    },
-    {
-      "epoch": 12.21488564414007,
-      "learning_rate": 1.2373265914689486e-07,
-      "loss": 2.5464,
-      "step": 47260
-    },
-    {
-      "epoch": 12.220054270577593,
-      "learning_rate": 1.2370024633735252e-07,
-      "loss": 2.4521,
-      "step": 47280
-    },
-    {
-      "epoch": 12.225222897015119,
-      "learning_rate": 1.2366783352781018e-07,
-      "loss": 2.5132,
-      "step": 47300
-    },
-    {
-      "epoch": 12.230391523452642,
-      "learning_rate": 1.2363542071826787e-07,
-      "loss": 2.43,
-      "step": 47320
-    },
-    {
-      "epoch": 12.235560149890167,
-      "learning_rate": 1.2360300790872553e-07,
-      "loss": 2.4295,
-      "step": 47340
-    },
-    {
-      "epoch": 12.24072877632769,
-      "learning_rate": 1.235705950991832e-07,
-      "loss": 2.5134,
-      "step": 47360
-    },
-    {
-      "epoch": 12.245897402765214,
-      "learning_rate": 1.2353818228964085e-07,
-      "loss": 2.461,
-      "step": 47380
-    },
-    {
-      "epoch": 12.25106602920274,
-      "learning_rate": 1.2350576948009854e-07,
-      "loss": 2.4942,
-      "step": 47400
-    },
-    {
-      "epoch": 12.256234655640263,
-      "learning_rate": 1.234733566705562e-07,
-      "loss": 2.4607,
-      "step": 47420
-    },
-    {
-      "epoch": 12.261403282077788,
-      "learning_rate": 1.2344094386101387e-07,
-      "loss": 2.4634,
-      "step": 47440
-    },
-    {
-      "epoch": 12.266571908515312,
-      "learning_rate": 1.2340853105147153e-07,
-      "loss": 2.4448,
-      "step": 47460
-    },
-    {
-      "epoch": 12.271740534952837,
-      "learning_rate": 1.2337611824192922e-07,
-      "loss": 2.4803,
-      "step": 47480
-    },
-    {
-      "epoch": 12.27690916139036,
-      "learning_rate": 1.2334370543238688e-07,
-      "loss": 2.4154,
-      "step": 47500
-    },
-    {
-      "epoch": 12.282077787827884,
-      "learning_rate": 1.2331129262284454e-07,
-      "loss": 2.5181,
-      "step": 47520
-    },
-    {
-      "epoch": 12.28724641426541,
-      "learning_rate": 1.2327887981330223e-07,
-      "loss": 2.5086,
-      "step": 47540
-    },
-    {
-      "epoch": 12.292415040702933,
-      "learning_rate": 1.232464670037599e-07,
-      "loss": 2.4487,
-      "step": 47560
-    },
-    {
-      "epoch": 12.297583667140458,
-      "learning_rate": 1.2321405419421755e-07,
-      "loss": 2.4109,
-      "step": 47580
-    },
-    {
-      "epoch": 12.302752293577981,
-      "learning_rate": 1.2318164138467521e-07,
-      "loss": 2.4675,
-      "step": 47600
-    },
-    {
-      "epoch": 12.307920920015507,
-      "learning_rate": 1.2314922857513288e-07,
-      "loss": 2.4326,
-      "step": 47620
-    },
-    {
-      "epoch": 12.31308954645303,
-      "learning_rate": 1.2311681576559056e-07,
-      "loss": 2.441,
-      "step": 47640
-    },
-    {
-      "epoch": 12.318258172890554,
-      "learning_rate": 1.2308440295604823e-07,
-      "loss": 2.4154,
-      "step": 47660
-    },
-    {
-      "epoch": 12.323426799328079,
-      "learning_rate": 1.230519901465059e-07,
-      "loss": 2.4491,
-      "step": 47680
-    },
-    {
-      "epoch": 12.328595425765602,
-      "learning_rate": 1.2301957733696358e-07,
-      "loss": 2.4764,
-      "step": 47700
-    },
-    {
-      "epoch": 12.333764052203128,
-      "learning_rate": 1.2298716452742124e-07,
-      "loss": 2.4786,
-      "step": 47720
-    },
-    {
-      "epoch": 12.338932678640651,
-      "learning_rate": 1.229547517178789e-07,
-      "loss": 2.4033,
-      "step": 47740
-    },
-    {
-      "epoch": 12.344101305078176,
-      "learning_rate": 1.2292233890833656e-07,
-      "loss": 2.4742,
-      "step": 47760
-    },
-    {
-      "epoch": 12.3492699315157,
-      "learning_rate": 1.2288992609879422e-07,
-      "loss": 2.5316,
-      "step": 47780
-    },
-    {
-      "epoch": 12.354438557953223,
-      "learning_rate": 1.228575132892519e-07,
-      "loss": 2.4717,
-      "step": 47800
-    },
-    {
-      "epoch": 12.359607184390748,
-      "learning_rate": 1.2282510047970957e-07,
-      "loss": 2.4518,
-      "step": 47820
-    },
-    {
-      "epoch": 12.364775810828272,
-      "learning_rate": 1.2279268767016724e-07,
-      "loss": 2.4747,
-      "step": 47840
-    },
-    {
-      "epoch": 12.369944437265797,
-      "learning_rate": 1.2276027486062492e-07,
-      "loss": 2.4391,
-      "step": 47860
-    },
-    {
-      "epoch": 12.37511306370332,
-      "learning_rate": 1.2272786205108259e-07,
-      "loss": 2.4076,
-      "step": 47880
-    },
-    {
-      "epoch": 12.380281690140846,
-      "learning_rate": 1.2269544924154025e-07,
-      "loss": 2.4275,
-      "step": 47900
-    },
-    {
-      "epoch": 12.38545031657837,
-      "learning_rate": 1.2266303643199794e-07,
-      "loss": 2.4593,
-      "step": 47920
-    },
-    {
-      "epoch": 12.390618943015893,
-      "learning_rate": 1.226306236224556e-07,
-      "loss": 2.4666,
-      "step": 47940
-    },
-    {
-      "epoch": 12.395787569453418,
-      "learning_rate": 1.2259821081291326e-07,
-      "loss": 2.4939,
-      "step": 47960
-    },
-    {
-      "epoch": 12.400956195890942,
-      "learning_rate": 1.2256579800337092e-07,
-      "loss": 2.4933,
-      "step": 47980
-    },
-    {
-      "epoch": 12.406124822328467,
-      "learning_rate": 1.2253338519382858e-07,
-      "loss": 2.4298,
-      "step": 48000
-    },
-    {
-      "epoch": 12.41129344876599,
-      "learning_rate": 1.2250097238428627e-07,
-      "loss": 2.4923,
-      "step": 48020
-    },
-    {
-      "epoch": 12.416462075203516,
-      "learning_rate": 1.2246855957474393e-07,
-      "loss": 2.4368,
-      "step": 48040
-    },
-    {
-      "epoch": 12.421630701641039,
-      "learning_rate": 1.224361467652016e-07,
-      "loss": 2.4882,
-      "step": 48060
-    },
-    {
-      "epoch": 12.426799328078562,
-      "learning_rate": 1.2240373395565928e-07,
-      "loss": 2.4417,
-      "step": 48080
-    },
-    {
-      "epoch": 12.431967954516088,
-      "learning_rate": 1.2237132114611695e-07,
-      "loss": 2.4881,
-      "step": 48100
-    },
-    {
-      "epoch": 12.437136580953611,
-      "learning_rate": 1.223389083365746e-07,
-      "loss": 2.4633,
-      "step": 48120
-    },
-    {
-      "epoch": 12.442305207391136,
-      "learning_rate": 1.223064955270323e-07,
-      "loss": 2.4914,
-      "step": 48140
-    },
-    {
-      "epoch": 12.44747383382866,
-      "learning_rate": 1.2227408271748993e-07,
-      "loss": 2.4236,
-      "step": 48160
-    },
-    {
-      "epoch": 12.452642460266183,
-      "learning_rate": 1.2224166990794762e-07,
-      "loss": 2.4603,
-      "step": 48180
-    },
-    {
-      "epoch": 12.457811086703709,
-      "learning_rate": 1.2220925709840528e-07,
-      "loss": 2.5019,
-      "step": 48200
-    },
-    {
-      "epoch": 12.462979713141232,
-      "learning_rate": 1.2217684428886294e-07,
-      "loss": 2.4368,
-      "step": 48220
-    },
-    {
-      "epoch": 12.468148339578757,
-      "learning_rate": 1.2214443147932063e-07,
-      "loss": 2.5019,
-      "step": 48240
-    },
-    {
-      "epoch": 12.47331696601628,
-      "learning_rate": 1.221120186697783e-07,
-      "loss": 2.4857,
-      "step": 48260
-    },
-    {
-      "epoch": 12.478485592453806,
-      "learning_rate": 1.2207960586023596e-07,
-      "loss": 2.4496,
-      "step": 48280
-    },
-    {
-      "epoch": 12.48365421889133,
-      "learning_rate": 1.2204719305069364e-07,
-      "loss": 2.5045,
-      "step": 48300
-    },
-    {
-      "epoch": 12.488822845328853,
-      "learning_rate": 1.2201478024115128e-07,
-      "loss": 2.4589,
-      "step": 48320
-    },
-    {
-      "epoch": 12.493991471766378,
-      "learning_rate": 1.2198236743160897e-07,
-      "loss": 2.4778,
-      "step": 48340
-    },
-    {
-      "epoch": 12.499160098203902,
-      "learning_rate": 1.2194995462206663e-07,
-      "loss": 2.4297,
-      "step": 48360
-    },
-    {
-      "epoch": 12.504328724641427,
-      "learning_rate": 1.219175418125243e-07,
-      "loss": 2.395,
-      "step": 48380
-    },
-    {
-      "epoch": 12.50949735107895,
-      "learning_rate": 1.2188512900298198e-07,
-      "loss": 2.4515,
-      "step": 48400
-    },
-    {
-      "epoch": 12.514665977516476,
-      "learning_rate": 1.2185271619343964e-07,
-      "loss": 2.4912,
-      "step": 48420
-    },
-    {
-      "epoch": 12.519834603954,
-      "learning_rate": 1.218203033838973e-07,
-      "loss": 2.4681,
-      "step": 48440
-    },
-    {
-      "epoch": 12.525003230391523,
-      "learning_rate": 1.21787890574355e-07,
-      "loss": 2.4598,
-      "step": 48460
-    },
-    {
-      "epoch": 12.530171856829048,
-      "learning_rate": 1.2175547776481265e-07,
-      "loss": 2.4792,
-      "step": 48480
-    },
-    {
-      "epoch": 12.535340483266571,
-      "learning_rate": 1.2172306495527032e-07,
-      "loss": 2.4331,
-      "step": 48500
-    },
-    {
-      "epoch": 12.540509109704097,
-      "learning_rate": 1.21690652145728e-07,
-      "loss": 2.4537,
-      "step": 48520
-    },
-    {
-      "epoch": 12.54567773614162,
-      "learning_rate": 1.2165823933618564e-07,
-      "loss": 2.5384,
-      "step": 48540
-    },
-    {
-      "epoch": 12.550846362579145,
-      "learning_rate": 1.2162582652664333e-07,
-      "loss": 2.5251,
-      "step": 48560
-    },
-    {
-      "epoch": 12.556014989016669,
-      "learning_rate": 1.21593413717101e-07,
-      "loss": 2.4403,
-      "step": 48580
-    },
-    {
-      "epoch": 12.561183615454192,
-      "learning_rate": 1.2156100090755865e-07,
-      "loss": 2.4652,
-      "step": 48600
-    },
-    {
-      "epoch": 12.566352241891718,
-      "learning_rate": 1.2152858809801634e-07,
-      "loss": 2.5051,
-      "step": 48620
-    },
-    {
-      "epoch": 12.571520868329241,
-      "learning_rate": 1.21496175288474e-07,
-      "loss": 2.4572,
-      "step": 48640
-    },
-    {
-      "epoch": 12.576689494766766,
-      "learning_rate": 1.2146376247893166e-07,
-      "loss": 2.4891,
-      "step": 48660
-    },
-    {
-      "epoch": 12.58185812120429,
-      "learning_rate": 1.2143134966938935e-07,
-      "loss": 2.5186,
-      "step": 48680
-    },
-    {
-      "epoch": 12.587026747641815,
-      "learning_rate": 1.21398936859847e-07,
-      "loss": 2.4751,
-      "step": 48700
-    },
-    {
-      "epoch": 12.592195374079338,
-      "learning_rate": 1.2136652405030468e-07,
-      "loss": 2.4716,
-      "step": 48720
-    },
-    {
-      "epoch": 12.597364000516862,
-      "learning_rate": 1.2133411124076234e-07,
-      "loss": 2.4849,
-      "step": 48740
-    },
-    {
-      "epoch": 12.602532626954387,
-      "learning_rate": 1.2130169843122e-07,
-      "loss": 2.4891,
-      "step": 48760
-    },
-    {
-      "epoch": 12.60770125339191,
-      "learning_rate": 1.212692856216777e-07,
-      "loss": 2.4776,
-      "step": 48780
-    },
-    {
-      "epoch": 12.612869879829436,
-      "learning_rate": 1.2123687281213535e-07,
-      "loss": 2.5003,
-      "step": 48800
-    },
-    {
-      "epoch": 12.61803850626696,
-      "learning_rate": 1.21204460002593e-07,
-      "loss": 2.4541,
-      "step": 48820
-    },
-    {
-      "epoch": 12.623207132704485,
-      "learning_rate": 1.211720471930507e-07,
-      "loss": 2.3995,
-      "step": 48840
-    },
-    {
-      "epoch": 12.628375759142008,
-      "learning_rate": 1.2113963438350836e-07,
-      "loss": 2.4991,
-      "step": 48860
-    },
-    {
-      "epoch": 12.633544385579532,
-      "learning_rate": 1.2110722157396602e-07,
-      "loss": 2.4393,
-      "step": 48880
-    },
-    {
-      "epoch": 12.638713012017057,
-      "learning_rate": 1.2107480876442369e-07,
-      "loss": 2.4266,
-      "step": 48900
-    },
-    {
-      "epoch": 12.64388163845458,
-      "learning_rate": 1.2104239595488135e-07,
-      "loss": 2.4504,
-      "step": 48920
-    },
-    {
-      "epoch": 12.649050264892105,
-      "learning_rate": 1.2100998314533904e-07,
-      "loss": 2.4426,
-      "step": 48940
-    },
-    {
-      "epoch": 12.654218891329629,
-      "learning_rate": 1.209775703357967e-07,
-      "loss": 2.4279,
-      "step": 48960
-    },
-    {
-      "epoch": 12.659387517767154,
-      "learning_rate": 1.2094515752625436e-07,
-      "loss": 2.4973,
-      "step": 48980
-    },
-    {
-      "epoch": 12.664556144204678,
-      "learning_rate": 1.2091274471671205e-07,
-      "loss": 2.4447,
-      "step": 49000
-    },
-    {
-      "epoch": 12.669724770642201,
-      "learning_rate": 1.208803319071697e-07,
-      "loss": 2.497,
-      "step": 49020
-    },
-    {
-      "epoch": 12.674893397079726,
-      "learning_rate": 1.2084791909762737e-07,
-      "loss": 2.4223,
-      "step": 49040
-    },
-    {
-      "epoch": 12.68006202351725,
-      "learning_rate": 1.2081550628808503e-07,
-      "loss": 2.447,
-      "step": 49060
-    },
-    {
-      "epoch": 12.685230649954775,
-      "learning_rate": 1.2078309347854272e-07,
-      "loss": 2.4652,
-      "step": 49080
-    },
-    {
-      "epoch": 12.690399276392299,
-      "learning_rate": 1.2075068066900038e-07,
-      "loss": 2.4167,
-      "step": 49100
-    },
-    {
-      "epoch": 12.695567902829822,
-      "learning_rate": 1.2071826785945805e-07,
-      "loss": 2.4103,
-      "step": 49120
-    },
-    {
-      "epoch": 12.700736529267347,
-      "learning_rate": 1.206858550499157e-07,
-      "loss": 2.4225,
-      "step": 49140
-    },
-    {
-      "epoch": 12.70590515570487,
-      "learning_rate": 1.206534422403734e-07,
-      "loss": 2.4668,
-      "step": 49160
-    },
-    {
-      "epoch": 12.711073782142396,
-      "learning_rate": 1.2062102943083106e-07,
-      "loss": 2.4861,
-      "step": 49180
-    },
-    {
-      "epoch": 12.71624240857992,
-      "learning_rate": 1.2058861662128872e-07,
-      "loss": 2.433,
-      "step": 49200
-    },
-    {
-      "epoch": 12.721411035017445,
-      "learning_rate": 1.205562038117464e-07,
-      "loss": 2.4395,
-      "step": 49220
-    },
-    {
-      "epoch": 12.726579661454968,
-      "learning_rate": 1.2052379100220407e-07,
-      "loss": 2.4702,
-      "step": 49240
-    },
-    {
-      "epoch": 12.731748287892493,
-      "learning_rate": 1.2049137819266173e-07,
-      "loss": 2.4585,
-      "step": 49260
-    },
-    {
-      "epoch": 12.736916914330017,
-      "learning_rate": 1.204589653831194e-07,
-      "loss": 2.414,
-      "step": 49280
-    },
-    {
-      "epoch": 12.74208554076754,
-      "learning_rate": 1.2042655257357706e-07,
-      "loss": 2.4743,
-      "step": 49300
-    },
-    {
-      "epoch": 12.747254167205066,
-      "learning_rate": 1.2039413976403474e-07,
-      "loss": 2.4963,
-      "step": 49320
-    },
-    {
-      "epoch": 12.75242279364259,
-      "learning_rate": 1.203617269544924e-07,
-      "loss": 2.4157,
-      "step": 49340
-    },
-    {
-      "epoch": 12.757591420080114,
-      "learning_rate": 1.2032931414495007e-07,
-      "loss": 2.4355,
-      "step": 49360
-    },
-    {
-      "epoch": 12.762760046517638,
-      "learning_rate": 1.2029690133540776e-07,
-      "loss": 2.4174,
-      "step": 49380
-    },
-    {
-      "epoch": 12.767928672955161,
-      "learning_rate": 1.2026448852586542e-07,
-      "loss": 2.4903,
-      "step": 49400
-    },
-    {
-      "epoch": 12.773097299392687,
-      "learning_rate": 1.2023207571632308e-07,
-      "loss": 2.4285,
-      "step": 49420
-    },
-    {
-      "epoch": 12.77826592583021,
-      "learning_rate": 1.2019966290678074e-07,
-      "loss": 2.5139,
-      "step": 49440
-    },
-    {
-      "epoch": 12.783434552267735,
-      "learning_rate": 1.2016725009723843e-07,
-      "loss": 2.4306,
-      "step": 49460
-    },
-    {
-      "epoch": 12.788603178705259,
-      "learning_rate": 1.201348372876961e-07,
-      "loss": 2.4208,
-      "step": 49480
-    },
-    {
-      "epoch": 12.793771805142784,
-      "learning_rate": 1.2010242447815375e-07,
-      "loss": 2.4154,
-      "step": 49500
-    },
-    {
-      "epoch": 12.798940431580307,
-      "learning_rate": 1.2007001166861142e-07,
-      "loss": 2.4862,
-      "step": 49520
-    },
-    {
-      "epoch": 12.804109058017831,
-      "learning_rate": 1.200375988590691e-07,
-      "loss": 2.4184,
-      "step": 49540
-    },
-    {
-      "epoch": 12.809277684455356,
-      "learning_rate": 1.2000518604952677e-07,
-      "loss": 2.374,
-      "step": 49560
-    },
-    {
-      "epoch": 12.81444631089288,
-      "learning_rate": 1.1997277323998443e-07,
-      "loss": 2.4885,
-      "step": 49580
-    },
-    {
-      "epoch": 12.819614937330405,
-      "learning_rate": 1.199403604304421e-07,
-      "loss": 2.4429,
-      "step": 49600
-    },
-    {
-      "epoch": 12.824783563767928,
-      "learning_rate": 1.1990794762089978e-07,
-      "loss": 2.3834,
-      "step": 49620
-    },
-    {
-      "epoch": 12.829952190205454,
-      "learning_rate": 1.1987553481135744e-07,
-      "loss": 2.415,
-      "step": 49640
-    },
-    {
-      "epoch": 12.835120816642977,
-      "learning_rate": 1.198431220018151e-07,
-      "loss": 2.5206,
-      "step": 49660
-    },
-    {
-      "epoch": 12.8402894430805,
-      "learning_rate": 1.198107091922728e-07,
-      "loss": 2.4883,
-      "step": 49680
-    },
-    {
-      "epoch": 12.845458069518026,
-      "learning_rate": 1.1977829638273045e-07,
-      "loss": 2.4729,
-      "step": 49700
-    },
-    {
-      "epoch": 12.85062669595555,
-      "learning_rate": 1.1974588357318811e-07,
-      "loss": 2.3815,
-      "step": 49720
-    },
-    {
-      "epoch": 12.855795322393075,
-      "learning_rate": 1.1971347076364578e-07,
-      "loss": 2.453,
-      "step": 49740
-    },
-    {
-      "epoch": 12.860963948830598,
-      "learning_rate": 1.1968105795410346e-07,
-      "loss": 2.4414,
-      "step": 49760
-    },
-    {
-      "epoch": 12.866132575268123,
-      "learning_rate": 1.1964864514456113e-07,
-      "loss": 2.5048,
-      "step": 49780
-    },
-    {
-      "epoch": 12.871301201705647,
-      "learning_rate": 1.196162323350188e-07,
-      "loss": 2.4836,
-      "step": 49800
-    },
-    {
-      "epoch": 12.87646982814317,
-      "learning_rate": 1.1958381952547645e-07,
-      "loss": 2.4018,
-      "step": 49820
-    },
-    {
-      "epoch": 12.881638454580695,
-      "learning_rate": 1.1955140671593414e-07,
-      "loss": 2.4562,
-      "step": 49840
-    },
-    {
-      "epoch": 12.886807081018219,
-      "learning_rate": 1.195189939063918e-07,
-      "loss": 2.4431,
-      "step": 49860
-    },
-    {
-      "epoch": 12.891975707455744,
-      "learning_rate": 1.1948658109684946e-07,
-      "loss": 2.4554,
-      "step": 49880
-    },
-    {
-      "epoch": 12.897144333893268,
-      "learning_rate": 1.1945416828730712e-07,
-      "loss": 2.4341,
-      "step": 49900
-    },
-    {
-      "epoch": 12.902312960330793,
-      "learning_rate": 1.194217554777648e-07,
-      "loss": 2.4189,
-      "step": 49920
-    },
-    {
-      "epoch": 12.907481586768316,
-      "learning_rate": 1.1938934266822247e-07,
-      "loss": 2.4739,
-      "step": 49940
-    },
-    {
-      "epoch": 12.91265021320584,
-      "learning_rate": 1.1935692985868014e-07,
-      "loss": 2.3941,
-      "step": 49960
-    },
-    {
-      "epoch": 12.917818839643365,
-      "learning_rate": 1.1932451704913782e-07,
-      "loss": 2.4357,
-      "step": 49980
-    },
-    {
-      "epoch": 12.922987466080889,
-      "learning_rate": 1.1929210423959549e-07,
-      "loss": 2.4676,
-      "step": 50000
-    },
-    {
-      "epoch": 12.928156092518414,
-      "learning_rate": 1.1925969143005315e-07,
-      "loss": 2.4551,
-      "step": 50020
-    },
-    {
-      "epoch": 12.933324718955937,
-      "learning_rate": 1.192272786205108e-07,
-      "loss": 2.4811,
-      "step": 50040
-    },
-    {
-      "epoch": 12.938493345393463,
-      "learning_rate": 1.191948658109685e-07,
-      "loss": 2.4513,
-      "step": 50060
-    },
-    {
-      "epoch": 12.943661971830986,
-      "learning_rate": 1.1916245300142616e-07,
-      "loss": 2.4607,
-      "step": 50080
-    },
-    {
-      "epoch": 12.94883059826851,
-      "learning_rate": 1.1913004019188382e-07,
-      "loss": 2.482,
-      "step": 50100
-    },
-    {
-      "epoch": 12.953999224706035,
-      "learning_rate": 1.1909762738234151e-07,
-      "loss": 2.4465,
-      "step": 50120
-    },
-    {
-      "epoch": 12.959167851143558,
-      "learning_rate": 1.1906521457279917e-07,
-      "loss": 2.4672,
-      "step": 50140
-    },
-    {
-      "epoch": 12.964336477581083,
-      "learning_rate": 1.1903280176325683e-07,
-      "loss": 2.421,
-      "step": 50160
-    },
-    {
-      "epoch": 12.969505104018607,
-      "learning_rate": 1.190003889537145e-07,
-      "loss": 2.4271,
-      "step": 50180
-    },
-    {
-      "epoch": 12.974673730456132,
-      "learning_rate": 1.1896797614417218e-07,
-      "loss": 2.4226,
-      "step": 50200
-    },
-    {
-      "epoch": 12.979842356893656,
-      "learning_rate": 1.1893556333462985e-07,
-      "loss": 2.4162,
-      "step": 50220
-    },
-    {
-      "epoch": 12.985010983331179,
-      "learning_rate": 1.1890315052508751e-07,
-      "loss": 2.509,
-      "step": 50240
-    },
-    {
-      "epoch": 12.990179609768704,
-      "learning_rate": 1.1887073771554518e-07,
-      "loss": 2.4553,
-      "step": 50260
-    },
-    {
-      "epoch": 12.995348236206228,
-      "learning_rate": 1.1883832490600286e-07,
-      "loss": 2.4558,
-      "step": 50280
-    },
-    {
-      "epoch": 13.0,
-      "eval_bleu": 12.8144,
-      "eval_gen_len": 37.2979,
-      "eval_loss": 2.433004140853882,
-      "eval_runtime": 806.1399,
-      "eval_samples_per_second": 2.136,
-      "eval_steps_per_second": 1.068,
-      "step": 50298
-    },
-    {
-      "epoch": 13.000516862643753,
-      "learning_rate": 1.1880591209646052e-07,
-      "loss": 2.4458,
-      "step": 50300
-    },
-    {
-      "epoch": 13.005685489081277,
-      "learning_rate": 1.1877349928691818e-07,
-      "loss": 2.4137,
-      "step": 50320
-    },
-    {
-      "epoch": 13.0108541155188,
-      "learning_rate": 1.1874108647737586e-07,
-      "loss": 2.4151,
-      "step": 50340
-    },
-    {
-      "epoch": 13.016022741956325,
-      "learning_rate": 1.1870867366783353e-07,
-      "loss": 2.4124,
-      "step": 50360
-    },
-    {
-      "epoch": 13.021191368393849,
-      "learning_rate": 1.1867626085829119e-07,
-      "loss": 2.4259,
-      "step": 50380
-    },
-    {
-      "epoch": 13.026359994831374,
-      "learning_rate": 1.1864384804874886e-07,
-      "loss": 2.4933,
-      "step": 50400
-    },
-    {
-      "epoch": 13.031528621268897,
-      "learning_rate": 1.1861143523920653e-07,
-      "loss": 2.4579,
-      "step": 50420
-    },
-    {
-      "epoch": 13.036697247706423,
-      "learning_rate": 1.185790224296642e-07,
-      "loss": 2.4766,
-      "step": 50440
-    },
-    {
-      "epoch": 13.041865874143946,
-      "learning_rate": 1.1854660962012187e-07,
-      "loss": 2.4438,
-      "step": 50460
-    },
-    {
-      "epoch": 13.04703450058147,
-      "learning_rate": 1.1851419681057953e-07,
-      "loss": 2.4372,
-      "step": 50480
-    },
-    {
-      "epoch": 13.052203127018995,
-      "learning_rate": 1.184817840010372e-07,
-      "loss": 2.4097,
-      "step": 50500
-    },
-    {
-      "epoch": 13.057371753456518,
-      "learning_rate": 1.1844937119149488e-07,
-      "loss": 2.4199,
-      "step": 50520
-    },
-    {
-      "epoch": 13.062540379894044,
-      "learning_rate": 1.1841695838195254e-07,
-      "loss": 2.477,
-      "step": 50540
-    },
-    {
-      "epoch": 13.067709006331567,
-      "learning_rate": 1.183845455724102e-07,
-      "loss": 2.3863,
-      "step": 50560
-    },
-    {
-      "epoch": 13.072877632769092,
-      "learning_rate": 1.1835213276286788e-07,
-      "loss": 2.4797,
-      "step": 50580
-    },
-    {
-      "epoch": 13.078046259206616,
-      "learning_rate": 1.1831971995332555e-07,
-      "loss": 2.4302,
-      "step": 50600
-    },
-    {
-      "epoch": 13.08321488564414,
-      "learning_rate": 1.1828730714378322e-07,
-      "loss": 2.4693,
-      "step": 50620
-    },
-    {
-      "epoch": 13.088383512081665,
-      "learning_rate": 1.1825489433424088e-07,
-      "loss": 2.5014,
-      "step": 50640
-    },
-    {
-      "epoch": 13.093552138519188,
-      "learning_rate": 1.1822248152469855e-07,
-      "loss": 2.4204,
-      "step": 50660
-    },
-    {
-      "epoch": 13.098720764956713,
-      "learning_rate": 1.1819006871515623e-07,
-      "loss": 2.4063,
-      "step": 50680
-    },
-    {
-      "epoch": 13.103889391394237,
-      "learning_rate": 1.1815765590561389e-07,
-      "loss": 2.4893,
-      "step": 50700
-    },
-    {
-      "epoch": 13.109058017831762,
-      "learning_rate": 1.1812524309607155e-07,
-      "loss": 2.4443,
-      "step": 50720
-    },
-    {
-      "epoch": 13.114226644269285,
-      "learning_rate": 1.1809283028652923e-07,
-      "loss": 2.4188,
-      "step": 50740
-    },
-    {
-      "epoch": 13.119395270706809,
-      "learning_rate": 1.180604174769869e-07,
-      "loss": 2.4566,
-      "step": 50760
-    },
-    {
-      "epoch": 13.124563897144334,
-      "learning_rate": 1.1802800466744456e-07,
-      "loss": 2.4187,
-      "step": 50780
-    },
-    {
-      "epoch": 13.129732523581858,
-      "learning_rate": 1.1799559185790222e-07,
-      "loss": 2.4418,
-      "step": 50800
-    },
-    {
-      "epoch": 13.134901150019383,
-      "learning_rate": 1.179631790483599e-07,
-      "loss": 2.4705,
-      "step": 50820
-    },
-    {
-      "epoch": 13.140069776456906,
-      "learning_rate": 1.1793076623881758e-07,
-      "loss": 2.4966,
-      "step": 50840
-    },
-    {
-      "epoch": 13.145238402894432,
-      "learning_rate": 1.1789835342927524e-07,
-      "loss": 2.4615,
-      "step": 50860
-    },
-    {
-      "epoch": 13.150407029331955,
-      "learning_rate": 1.1786594061973291e-07,
-      "loss": 2.4513,
-      "step": 50880
-    },
-    {
-      "epoch": 13.155575655769479,
-      "learning_rate": 1.1783352781019059e-07,
-      "loss": 2.4163,
-      "step": 50900
-    },
-    {
-      "epoch": 13.160744282207004,
-      "learning_rate": 1.1780111500064825e-07,
-      "loss": 2.457,
-      "step": 50920
-    },
-    {
-      "epoch": 13.165912908644527,
-      "learning_rate": 1.1776870219110591e-07,
-      "loss": 2.4637,
-      "step": 50940
-    },
-    {
-      "epoch": 13.171081535082052,
-      "learning_rate": 1.1773628938156359e-07,
-      "loss": 2.4972,
-      "step": 50960
-    },
-    {
-      "epoch": 13.176250161519576,
-      "learning_rate": 1.1770387657202126e-07,
-      "loss": 2.4628,
-      "step": 50980
-    },
-    {
-      "epoch": 13.181418787957101,
-      "learning_rate": 1.1767146376247892e-07,
-      "loss": 2.404,
-      "step": 51000
-    },
-    {
-      "epoch": 13.186587414394625,
-      "learning_rate": 1.1763905095293658e-07,
-      "loss": 2.4146,
-      "step": 51020
-    },
-    {
-      "epoch": 13.191756040832148,
-      "learning_rate": 1.1760663814339426e-07,
-      "loss": 2.4528,
-      "step": 51040
-    },
-    {
-      "epoch": 13.196924667269673,
-      "learning_rate": 1.1757422533385193e-07,
-      "loss": 2.4843,
-      "step": 51060
-    },
-    {
-      "epoch": 13.202093293707197,
-      "learning_rate": 1.175418125243096e-07,
-      "loss": 2.4556,
-      "step": 51080
-    },
-    {
-      "epoch": 13.207261920144722,
-      "learning_rate": 1.1750939971476726e-07,
-      "loss": 2.4307,
-      "step": 51100
-    },
-    {
-      "epoch": 13.212430546582246,
-      "learning_rate": 1.1747698690522493e-07,
-      "loss": 2.4733,
-      "step": 51120
-    },
-    {
-      "epoch": 13.21759917301977,
-      "learning_rate": 1.1744457409568261e-07,
-      "loss": 2.4003,
-      "step": 51140
-    },
-    {
-      "epoch": 13.222767799457294,
-      "learning_rate": 1.1741216128614027e-07,
-      "loss": 2.4786,
-      "step": 51160
-    },
-    {
-      "epoch": 13.227936425894818,
-      "learning_rate": 1.1737974847659795e-07,
-      "loss": 2.4754,
-      "step": 51180
-    },
-    {
-      "epoch": 13.233105052332343,
-      "learning_rate": 1.1734733566705561e-07,
-      "loss": 2.4504,
-      "step": 51200
-    },
-    {
-      "epoch": 13.238273678769866,
-      "learning_rate": 1.1731492285751328e-07,
-      "loss": 2.3622,
-      "step": 51220
-    },
-    {
-      "epoch": 13.243442305207392,
-      "learning_rate": 1.1728251004797094e-07,
-      "loss": 2.4765,
-      "step": 51240
-    },
-    {
-      "epoch": 13.248610931644915,
-      "learning_rate": 1.1725009723842862e-07,
-      "loss": 2.5202,
-      "step": 51260
-    },
-    {
-      "epoch": 13.253779558082439,
-      "learning_rate": 1.1721768442888628e-07,
-      "loss": 2.4383,
-      "step": 51280
-    },
-    {
-      "epoch": 13.258948184519964,
-      "learning_rate": 1.1718527161934396e-07,
-      "loss": 2.4267,
-      "step": 51300
-    },
-    {
-      "epoch": 13.264116810957487,
-      "learning_rate": 1.1715285880980162e-07,
-      "loss": 2.4442,
-      "step": 51320
-    },
-    {
-      "epoch": 13.269285437395013,
-      "learning_rate": 1.171204460002593e-07,
-      "loss": 2.4723,
-      "step": 51340
-    },
-    {
-      "epoch": 13.274454063832536,
-      "learning_rate": 1.1708803319071696e-07,
-      "loss": 2.4552,
-      "step": 51360
-    },
-    {
-      "epoch": 13.279622690270061,
-      "learning_rate": 1.1705562038117463e-07,
-      "loss": 2.4261,
-      "step": 51380
-    },
-    {
-      "epoch": 13.284791316707585,
-      "learning_rate": 1.1702320757163229e-07,
-      "loss": 2.3728,
-      "step": 51400
-    },
-    {
-      "epoch": 13.28995994314511,
-      "learning_rate": 1.1699079476208997e-07,
-      "loss": 2.4434,
-      "step": 51420
-    },
-    {
-      "epoch": 13.295128569582634,
-      "learning_rate": 1.1695838195254763e-07,
-      "loss": 2.433,
-      "step": 51440
-    },
-    {
-      "epoch": 13.300297196020157,
-      "learning_rate": 1.169259691430053e-07,
-      "loss": 2.4091,
-      "step": 51460
-    },
-    {
-      "epoch": 13.305465822457682,
-      "learning_rate": 1.1689355633346298e-07,
-      "loss": 2.418,
-      "step": 51480
-    },
-    {
-      "epoch": 13.310634448895206,
-      "learning_rate": 1.1686114352392064e-07,
-      "loss": 2.4226,
-      "step": 51500
-    },
-    {
-      "epoch": 13.315803075332731,
-      "learning_rate": 1.168287307143783e-07,
-      "loss": 2.4603,
-      "step": 51520
-    },
-    {
-      "epoch": 13.320971701770254,
-      "learning_rate": 1.1679631790483598e-07,
-      "loss": 2.4939,
-      "step": 51540
-    },
-    {
-      "epoch": 13.326140328207778,
-      "learning_rate": 1.1676390509529365e-07,
-      "loss": 2.444,
-      "step": 51560
-    },
-    {
-      "epoch": 13.331308954645303,
-      "learning_rate": 1.1673149228575132e-07,
-      "loss": 2.5078,
-      "step": 51580
-    },
-    {
-      "epoch": 13.336477581082827,
-      "learning_rate": 1.1669907947620898e-07,
-      "loss": 2.5158,
-      "step": 51600
-    },
-    {
-      "epoch": 13.341646207520352,
-      "learning_rate": 1.1666666666666665e-07,
-      "loss": 2.5183,
-      "step": 51620
-    },
-    {
-      "epoch": 13.346814833957875,
-      "learning_rate": 1.1663425385712433e-07,
-      "loss": 2.3932,
-      "step": 51640
-    },
-    {
-      "epoch": 13.3519834603954,
-      "learning_rate": 1.1660184104758199e-07,
-      "loss": 2.413,
-      "step": 51660
-    },
-    {
-      "epoch": 13.357152086832924,
-      "learning_rate": 1.1656942823803965e-07,
-      "loss": 2.4563,
-      "step": 51680
-    },
-    {
-      "epoch": 13.362320713270448,
-      "learning_rate": 1.1653701542849733e-07,
-      "loss": 2.4141,
-      "step": 51700
-    },
-    {
-      "epoch": 13.367489339707973,
-      "learning_rate": 1.16504602618955e-07,
-      "loss": 2.486,
-      "step": 51720
-    },
-    {
-      "epoch": 13.372657966145496,
-      "learning_rate": 1.1647218980941266e-07,
-      "loss": 2.444,
-      "step": 51740
-    },
-    {
-      "epoch": 13.377826592583022,
-      "learning_rate": 1.1643977699987033e-07,
-      "loss": 2.4608,
-      "step": 51760
-    },
-    {
-      "epoch": 13.382995219020545,
-      "learning_rate": 1.1640736419032801e-07,
-      "loss": 2.4173,
-      "step": 51780
-    },
-    {
-      "epoch": 13.38816384545807,
-      "learning_rate": 1.1637495138078568e-07,
-      "loss": 2.4247,
-      "step": 51800
-    },
-    {
-      "epoch": 13.393332471895594,
-      "learning_rate": 1.1634253857124334e-07,
-      "loss": 2.4383,
-      "step": 51820
-    },
-    {
-      "epoch": 13.398501098333117,
-      "learning_rate": 1.16310125761701e-07,
-      "loss": 2.4455,
-      "step": 51840
-    },
-    {
-      "epoch": 13.403669724770642,
-      "learning_rate": 1.1627771295215869e-07,
-      "loss": 2.4771,
-      "step": 51860
-    },
-    {
-      "epoch": 13.408838351208166,
-      "learning_rate": 1.1624530014261635e-07,
-      "loss": 2.4686,
-      "step": 51880
-    },
-    {
-      "epoch": 13.414006977645691,
-      "learning_rate": 1.1621288733307401e-07,
-      "loss": 2.4417,
-      "step": 51900
-    },
-    {
-      "epoch": 13.419175604083215,
-      "learning_rate": 1.1618047452353167e-07,
-      "loss": 2.4765,
-      "step": 51920
-    },
-    {
-      "epoch": 13.42434423052074,
-      "learning_rate": 1.1614806171398936e-07,
-      "loss": 2.4501,
-      "step": 51940
-    },
-    {
-      "epoch": 13.429512856958263,
-      "learning_rate": 1.1611564890444702e-07,
-      "loss": 2.467,
-      "step": 51960
-    },
-    {
-      "epoch": 13.434681483395787,
-      "learning_rate": 1.1608323609490469e-07,
-      "loss": 2.4404,
-      "step": 51980
-    },
-    {
-      "epoch": 13.439850109833312,
-      "learning_rate": 1.1605082328536235e-07,
-      "loss": 2.4413,
-      "step": 52000
-    },
-    {
-      "epoch": 13.445018736270836,
-      "learning_rate": 1.1601841047582004e-07,
-      "loss": 2.4529,
-      "step": 52020
-    },
-    {
-      "epoch": 13.45018736270836,
-      "learning_rate": 1.159859976662777e-07,
-      "loss": 2.4719,
-      "step": 52040
-    },
-    {
-      "epoch": 13.455355989145884,
-      "learning_rate": 1.1595358485673536e-07,
-      "loss": 2.3729,
-      "step": 52060
-    },
-    {
-      "epoch": 13.46052461558341,
-      "learning_rate": 1.1592117204719305e-07,
-      "loss": 2.4296,
-      "step": 52080
-    },
-    {
-      "epoch": 13.465693242020933,
-      "learning_rate": 1.1588875923765071e-07,
-      "loss": 2.4939,
-      "step": 52100
-    },
-    {
-      "epoch": 13.470861868458456,
-      "learning_rate": 1.1585634642810837e-07,
-      "loss": 2.4214,
-      "step": 52120
-    },
-    {
-      "epoch": 13.476030494895982,
-      "learning_rate": 1.1582393361856603e-07,
-      "loss": 2.4435,
-      "step": 52140
-    },
-    {
-      "epoch": 13.481199121333505,
-      "learning_rate": 1.1579152080902372e-07,
-      "loss": 2.4724,
-      "step": 52160
-    },
-    {
-      "epoch": 13.48636774777103,
-      "learning_rate": 1.1575910799948138e-07,
-      "loss": 2.4219,
-      "step": 52180
-    },
-    {
-      "epoch": 13.491536374208554,
-      "learning_rate": 1.1572669518993905e-07,
-      "loss": 2.4166,
-      "step": 52200
-    },
-    {
-      "epoch": 13.49670500064608,
-      "learning_rate": 1.1569428238039671e-07,
-      "loss": 2.456,
-      "step": 52220
-    },
-    {
-      "epoch": 13.501873627083603,
-      "learning_rate": 1.156618695708544e-07,
-      "loss": 2.4335,
-      "step": 52240
-    },
-    {
-      "epoch": 13.507042253521126,
-      "learning_rate": 1.1562945676131206e-07,
-      "loss": 2.4549,
-      "step": 52260
-    },
-    {
-      "epoch": 13.512210879958651,
-      "learning_rate": 1.1559704395176972e-07,
-      "loss": 2.4274,
-      "step": 52280
-    },
-    {
-      "epoch": 13.517379506396175,
-      "learning_rate": 1.1556463114222738e-07,
-      "loss": 2.4509,
-      "step": 52300
-    },
-    {
-      "epoch": 13.5225481328337,
-      "learning_rate": 1.1553221833268507e-07,
-      "loss": 2.4466,
-      "step": 52320
-    },
-    {
-      "epoch": 13.527716759271224,
-      "learning_rate": 1.1549980552314273e-07,
-      "loss": 2.486,
-      "step": 52340
-    },
-    {
-      "epoch": 13.532885385708749,
-      "learning_rate": 1.1546739271360039e-07,
-      "loss": 2.4423,
-      "step": 52360
-    },
-    {
-      "epoch": 13.538054012146272,
-      "learning_rate": 1.1543497990405808e-07,
-      "loss": 2.3907,
-      "step": 52380
-    },
-    {
-      "epoch": 13.543222638583796,
-      "learning_rate": 1.1540256709451574e-07,
-      "loss": 2.4323,
-      "step": 52400
-    },
-    {
-      "epoch": 13.548391265021321,
-      "learning_rate": 1.153701542849734e-07,
-      "loss": 2.4375,
-      "step": 52420
-    },
-    {
-      "epoch": 13.553559891458844,
-      "learning_rate": 1.1533774147543107e-07,
-      "loss": 2.4197,
-      "step": 52440
-    },
-    {
-      "epoch": 13.55872851789637,
-      "learning_rate": 1.1530532866588876e-07,
-      "loss": 2.4902,
-      "step": 52460
-    },
-    {
-      "epoch": 13.563897144333893,
-      "learning_rate": 1.1527291585634642e-07,
-      "loss": 2.4065,
-      "step": 52480
-    },
-    {
-      "epoch": 13.569065770771417,
-      "learning_rate": 1.1524050304680408e-07,
-      "loss": 2.4606,
-      "step": 52500
-    },
-    {
-      "epoch": 13.574234397208942,
-      "learning_rate": 1.1520809023726174e-07,
-      "loss": 2.4576,
-      "step": 52520
-    },
-    {
-      "epoch": 13.579403023646465,
-      "learning_rate": 1.1517567742771943e-07,
-      "loss": 2.4596,
-      "step": 52540
-    },
-    {
-      "epoch": 13.58457165008399,
-      "learning_rate": 1.1514326461817709e-07,
-      "loss": 2.4678,
-      "step": 52560
-    },
-    {
-      "epoch": 13.589740276521514,
-      "learning_rate": 1.1511085180863475e-07,
-      "loss": 2.4333,
-      "step": 52580
-    },
-    {
-      "epoch": 13.59490890295904,
-      "learning_rate": 1.1507843899909241e-07,
-      "loss": 2.4255,
-      "step": 52600
-    },
-    {
-      "epoch": 13.600077529396563,
-      "learning_rate": 1.150460261895501e-07,
-      "loss": 2.414,
-      "step": 52620
-    },
-    {
-      "epoch": 13.605246155834088,
-      "learning_rate": 1.1501361338000777e-07,
-      "loss": 2.4667,
-      "step": 52640
-    },
-    {
-      "epoch": 13.610414782271612,
-      "learning_rate": 1.1498120057046543e-07,
-      "loss": 2.4574,
-      "step": 52660
-    },
-    {
-      "epoch": 13.615583408709135,
-      "learning_rate": 1.1494878776092312e-07,
-      "loss": 2.4472,
-      "step": 52680
-    },
-    {
-      "epoch": 13.62075203514666,
-      "learning_rate": 1.1491637495138078e-07,
-      "loss": 2.4838,
-      "step": 52700
-    },
-    {
-      "epoch": 13.625920661584184,
-      "learning_rate": 1.1488396214183844e-07,
-      "loss": 2.4244,
-      "step": 52720
-    },
-    {
-      "epoch": 13.631089288021709,
-      "learning_rate": 1.148515493322961e-07,
-      "loss": 2.4486,
-      "step": 52740
-    },
-    {
-      "epoch": 13.636257914459232,
-      "learning_rate": 1.1481913652275379e-07,
-      "loss": 2.4595,
-      "step": 52760
-    },
-    {
-      "epoch": 13.641426540896756,
-      "learning_rate": 1.1478672371321145e-07,
-      "loss": 2.4267,
-      "step": 52780
-    },
-    {
-      "epoch": 13.646595167334281,
-      "learning_rate": 1.1475431090366911e-07,
-      "loss": 2.4348,
-      "step": 52800
-    },
-    {
-      "epoch": 13.651763793771805,
-      "learning_rate": 1.1472189809412677e-07,
-      "loss": 2.4963,
-      "step": 52820
-    },
-    {
-      "epoch": 13.65693242020933,
-      "learning_rate": 1.1468948528458446e-07,
-      "loss": 2.3726,
-      "step": 52840
-    },
-    {
-      "epoch": 13.662101046646853,
-      "learning_rate": 1.1465707247504213e-07,
-      "loss": 2.3944,
-      "step": 52860
-    },
-    {
-      "epoch": 13.667269673084379,
-      "learning_rate": 1.1462465966549979e-07,
-      "loss": 2.5007,
-      "step": 52880
-    },
-    {
-      "epoch": 13.672438299521902,
-      "learning_rate": 1.1459224685595745e-07,
-      "loss": 2.4834,
-      "step": 52900
-    },
-    {
-      "epoch": 13.677606925959426,
-      "learning_rate": 1.1455983404641514e-07,
-      "loss": 2.3963,
-      "step": 52920
-    },
-    {
-      "epoch": 13.68277555239695,
-      "learning_rate": 1.1452742123687283e-07,
-      "loss": 2.4304,
-      "step": 52940
-    },
-    {
-      "epoch": 13.687944178834474,
-      "learning_rate": 1.1449500842733046e-07,
-      "loss": 2.4219,
-      "step": 52960
-    },
-    {
-      "epoch": 13.693112805272,
-      "learning_rate": 1.1446259561778815e-07,
-      "loss": 2.5008,
-      "step": 52980
-    },
-    {
-      "epoch": 13.698281431709523,
-      "learning_rate": 1.1443018280824581e-07,
-      "loss": 2.4243,
-      "step": 53000
-    },
-    {
-      "epoch": 13.703450058147048,
-      "learning_rate": 1.143977699987035e-07,
-      "loss": 2.462,
-      "step": 53020
-    },
-    {
-      "epoch": 13.708618684584572,
-      "learning_rate": 1.1436535718916113e-07,
-      "loss": 2.4512,
-      "step": 53040
-    },
-    {
-      "epoch": 13.713787311022095,
-      "learning_rate": 1.1433294437961882e-07,
-      "loss": 2.508,
-      "step": 53060
-    },
-    {
-      "epoch": 13.71895593745962,
-      "learning_rate": 1.1430053157007649e-07,
-      "loss": 2.4558,
-      "step": 53080
-    },
-    {
-      "epoch": 13.724124563897144,
-      "learning_rate": 1.1426811876053417e-07,
-      "loss": 2.4658,
-      "step": 53100
-    },
-    {
-      "epoch": 13.729293190334669,
-      "learning_rate": 1.1423570595099181e-07,
-      "loss": 2.3885,
-      "step": 53120
-    },
-    {
-      "epoch": 13.734461816772193,
-      "learning_rate": 1.142032931414495e-07,
-      "loss": 2.44,
-      "step": 53140
-    },
-    {
-      "epoch": 13.739630443209718,
-      "learning_rate": 1.1417088033190716e-07,
-      "loss": 2.4517,
-      "step": 53160
-    },
-    {
-      "epoch": 13.744799069647241,
-      "learning_rate": 1.1413846752236485e-07,
-      "loss": 2.4766,
-      "step": 53180
-    },
-    {
-      "epoch": 13.749967696084765,
-      "learning_rate": 1.1410605471282248e-07,
-      "loss": 2.447,
-      "step": 53200
-    },
-    {
-      "epoch": 13.75513632252229,
-      "learning_rate": 1.1407364190328017e-07,
-      "loss": 2.4432,
-      "step": 53220
-    },
-    {
-      "epoch": 13.760304948959813,
-      "learning_rate": 1.1404122909373786e-07,
-      "loss": 2.4773,
-      "step": 53240
-    },
-    {
-      "epoch": 13.765473575397339,
-      "learning_rate": 1.140088162841955e-07,
-      "loss": 2.455,
-      "step": 53260
-    },
-    {
-      "epoch": 13.770642201834862,
-      "learning_rate": 1.1397640347465318e-07,
-      "loss": 2.4651,
-      "step": 53280
-    },
-    {
-      "epoch": 13.775810828272387,
-      "learning_rate": 1.1394399066511085e-07,
-      "loss": 2.4309,
-      "step": 53300
-    },
-    {
-      "epoch": 13.780979454709911,
-      "learning_rate": 1.1391157785556853e-07,
-      "loss": 2.4539,
-      "step": 53320
-    },
-    {
-      "epoch": 13.786148081147434,
-      "learning_rate": 1.1387916504602617e-07,
-      "loss": 2.4259,
-      "step": 53340
-    },
-    {
-      "epoch": 13.79131670758496,
-      "learning_rate": 1.1384675223648386e-07,
-      "loss": 2.357,
-      "step": 53360
-    },
-    {
-      "epoch": 13.796485334022483,
-      "learning_rate": 1.1381433942694152e-07,
-      "loss": 2.4122,
-      "step": 53380
-    },
-    {
-      "epoch": 13.801653960460008,
-      "learning_rate": 1.1378192661739921e-07,
-      "loss": 2.435,
-      "step": 53400
-    },
-    {
-      "epoch": 13.806822586897532,
-      "learning_rate": 1.1374951380785684e-07,
-      "loss": 2.4501,
-      "step": 53420
-    },
-    {
-      "epoch": 13.811991213335055,
-      "learning_rate": 1.1371710099831453e-07,
-      "loss": 2.4191,
-      "step": 53440
-    },
-    {
-      "epoch": 13.81715983977258,
-      "learning_rate": 1.1368468818877219e-07,
-      "loss": 2.4168,
-      "step": 53460
-    },
-    {
-      "epoch": 13.822328466210104,
-      "learning_rate": 1.1365227537922988e-07,
-      "loss": 2.4114,
-      "step": 53480
-    },
-    {
-      "epoch": 13.82749709264763,
-      "learning_rate": 1.1361986256968752e-07,
-      "loss": 2.4437,
-      "step": 53500
-    },
-    {
-      "epoch": 13.832665719085153,
-      "learning_rate": 1.135874497601452e-07,
-      "loss": 2.5263,
-      "step": 53520
-    },
-    {
-      "epoch": 13.837834345522678,
-      "learning_rate": 1.1355503695060289e-07,
-      "loss": 2.3909,
-      "step": 53540
-    },
-    {
-      "epoch": 13.843002971960201,
-      "learning_rate": 1.1352262414106056e-07,
-      "loss": 2.4661,
-      "step": 53560
-    },
-    {
-      "epoch": 13.848171598397727,
-      "learning_rate": 1.1349021133151822e-07,
-      "loss": 2.3978,
-      "step": 53580
-    },
-    {
-      "epoch": 13.85334022483525,
-      "learning_rate": 1.1345779852197588e-07,
-      "loss": 2.4506,
-      "step": 53600
-    },
-    {
-      "epoch": 13.858508851272774,
-      "learning_rate": 1.1342538571243357e-07,
-      "loss": 2.4424,
-      "step": 53620
-    },
-    {
-      "epoch": 13.863677477710299,
-      "learning_rate": 1.1339297290289123e-07,
-      "loss": 2.4247,
-      "step": 53640
-    },
-    {
-      "epoch": 13.868846104147822,
-      "learning_rate": 1.1336056009334889e-07,
-      "loss": 2.4441,
-      "step": 53660
-    },
-    {
-      "epoch": 13.874014730585348,
-      "learning_rate": 1.1332814728380655e-07,
-      "loss": 2.4742,
-      "step": 53680
-    },
-    {
-      "epoch": 13.879183357022871,
-      "learning_rate": 1.1329573447426424e-07,
-      "loss": 2.4507,
-      "step": 53700
-    },
-    {
-      "epoch": 13.884351983460395,
-      "learning_rate": 1.132633216647219e-07,
-      "loss": 2.4422,
-      "step": 53720
-    },
-    {
-      "epoch": 13.88952060989792,
-      "learning_rate": 1.1323090885517956e-07,
-      "loss": 2.5032,
-      "step": 53740
-    },
-    {
-      "epoch": 13.894689236335443,
-      "learning_rate": 1.1319849604563723e-07,
-      "loss": 2.4601,
-      "step": 53760
-    },
-    {
-      "epoch": 13.899857862772969,
-      "learning_rate": 1.1316608323609492e-07,
-      "loss": 2.3983,
-      "step": 53780
-    },
-    {
-      "epoch": 13.905026489210492,
-      "learning_rate": 1.1313367042655258e-07,
-      "loss": 2.3913,
-      "step": 53800
-    },
-    {
-      "epoch": 13.910195115648017,
-      "learning_rate": 1.1310125761701024e-07,
-      "loss": 2.4417,
-      "step": 53820
-    },
-    {
-      "epoch": 13.91536374208554,
-      "learning_rate": 1.1306884480746793e-07,
-      "loss": 2.422,
-      "step": 53840
-    },
-    {
-      "epoch": 13.920532368523064,
-      "learning_rate": 1.1303643199792559e-07,
-      "loss": 2.4629,
-      "step": 53860
-    },
-    {
-      "epoch": 13.92570099496059,
-      "learning_rate": 1.1300401918838325e-07,
-      "loss": 2.4489,
-      "step": 53880
-    },
-    {
-      "epoch": 13.930869621398113,
-      "learning_rate": 1.1297160637884091e-07,
-      "loss": 2.4243,
-      "step": 53900
-    },
-    {
-      "epoch": 13.936038247835638,
-      "learning_rate": 1.129391935692986e-07,
-      "loss": 2.4241,
-      "step": 53920
-    },
-    {
-      "epoch": 13.941206874273162,
-      "learning_rate": 1.1290678075975626e-07,
-      "loss": 2.4386,
-      "step": 53940
-    },
-    {
-      "epoch": 13.946375500710687,
-      "learning_rate": 1.1287436795021392e-07,
-      "loss": 2.4117,
-      "step": 53960
-    },
-    {
-      "epoch": 13.95154412714821,
-      "learning_rate": 1.1284195514067159e-07,
-      "loss": 2.4356,
-      "step": 53980
-    },
-    {
-      "epoch": 13.956712753585734,
-      "learning_rate": 1.1280954233112928e-07,
-      "loss": 2.4823,
-      "step": 54000
-    },
-    {
-      "epoch": 13.961881380023259,
-      "learning_rate": 1.1277712952158694e-07,
-      "loss": 2.4368,
-      "step": 54020
-    },
-    {
-      "epoch": 13.967050006460783,
-      "learning_rate": 1.127447167120446e-07,
-      "loss": 2.4806,
-      "step": 54040
-    },
-    {
-      "epoch": 13.972218632898308,
-      "learning_rate": 1.1271230390250226e-07,
-      "loss": 2.3614,
-      "step": 54060
-    },
-    {
-      "epoch": 13.977387259335831,
-      "learning_rate": 1.1267989109295995e-07,
-      "loss": 2.4549,
-      "step": 54080
-    },
-    {
-      "epoch": 13.982555885773357,
-      "learning_rate": 1.1264747828341761e-07,
-      "loss": 2.4648,
-      "step": 54100
-    },
-    {
-      "epoch": 13.98772451221088,
-      "learning_rate": 1.1261506547387527e-07,
-      "loss": 2.482,
-      "step": 54120
-    },
-    {
-      "epoch": 13.992893138648403,
-      "learning_rate": 1.1258265266433296e-07,
-      "loss": 2.4405,
-      "step": 54140
-    },
-    {
-      "epoch": 13.998061765085929,
-      "learning_rate": 1.1255023985479062e-07,
-      "loss": 2.4125,
-      "step": 54160
-    },
-    {
-      "epoch": 13.999870784339063,
-      "eval_bleu": 13.0772,
-      "eval_gen_len": 37.0436,
-      "eval_loss": 2.4198501110076904,
-      "eval_runtime": 802.9059,
-      "eval_samples_per_second": 2.145,
-      "eval_steps_per_second": 1.072,
-      "step": 54167
-    },
-    {
-      "epoch": 14.003230391523452,
-      "learning_rate": 1.1251782704524828e-07,
-      "loss": 2.497,
-      "step": 54180
-    },
-    {
-      "epoch": 14.008399017960977,
-      "learning_rate": 1.1248541423570595e-07,
-      "loss": 2.4448,
-      "step": 54200
-    },
-    {
-      "epoch": 14.013567644398501,
-      "learning_rate": 1.1245300142616364e-07,
-      "loss": 2.4748,
-      "step": 54220
-    },
-    {
-      "epoch": 14.018736270836026,
-      "learning_rate": 1.124205886166213e-07,
-      "loss": 2.5235,
-      "step": 54240
-    },
-    {
-      "epoch": 14.02390489727355,
-      "learning_rate": 1.1238817580707896e-07,
-      "loss": 2.4309,
-      "step": 54260
-    },
-    {
-      "epoch": 14.029073523711073,
-      "learning_rate": 1.1235576299753662e-07,
-      "loss": 2.4434,
-      "step": 54280
-    },
-    {
-      "epoch": 14.034242150148598,
-      "learning_rate": 1.1232335018799431e-07,
-      "loss": 2.4613,
-      "step": 54300
-    },
-    {
-      "epoch": 14.039410776586122,
-      "learning_rate": 1.1229093737845197e-07,
-      "loss": 2.4494,
-      "step": 54320
-    },
-    {
-      "epoch": 14.044579403023647,
-      "learning_rate": 1.1225852456890963e-07,
-      "loss": 2.4625,
-      "step": 54340
-    },
-    {
-      "epoch": 14.04974802946117,
-      "learning_rate": 1.122261117593673e-07,
-      "loss": 2.4266,
-      "step": 54360
-    },
-    {
-      "epoch": 14.054916655898696,
-      "learning_rate": 1.1219369894982498e-07,
-      "loss": 2.4216,
-      "step": 54380
-    },
-    {
-      "epoch": 14.06008528233622,
-      "learning_rate": 1.1216128614028264e-07,
-      "loss": 2.4619,
-      "step": 54400
-    },
-    {
-      "epoch": 14.065253908773743,
-      "learning_rate": 1.1212887333074031e-07,
-      "loss": 2.4161,
-      "step": 54420
-    },
-    {
-      "epoch": 14.070422535211268,
-      "learning_rate": 1.12096460521198e-07,
-      "loss": 2.4824,
-      "step": 54440
-    },
-    {
-      "epoch": 14.075591161648791,
-      "learning_rate": 1.1206404771165566e-07,
-      "loss": 2.3989,
-      "step": 54460
-    },
-    {
-      "epoch": 14.080759788086317,
-      "learning_rate": 1.1203163490211332e-07,
-      "loss": 2.4347,
-      "step": 54480
-    },
-    {
-      "epoch": 14.08592841452384,
-      "learning_rate": 1.1199922209257098e-07,
-      "loss": 2.4447,
-      "step": 54500
-    },
-    {
-      "epoch": 14.091097040961365,
-      "learning_rate": 1.1196680928302867e-07,
-      "loss": 2.4,
-      "step": 54520
-    },
-    {
-      "epoch": 14.096265667398889,
-      "learning_rate": 1.1193439647348633e-07,
-      "loss": 2.421,
-      "step": 54540
-    },
-    {
-      "epoch": 14.101434293836412,
-      "learning_rate": 1.1190198366394399e-07,
-      "loss": 2.4029,
-      "step": 54560
-    },
-    {
-      "epoch": 14.106602920273938,
-      "learning_rate": 1.1186957085440165e-07,
-      "loss": 2.4102,
-      "step": 54580
-    },
-    {
-      "epoch": 14.111771546711461,
-      "learning_rate": 1.1183715804485934e-07,
-      "loss": 2.391,
-      "step": 54600
-    },
-    {
-      "epoch": 14.116940173148986,
-      "learning_rate": 1.11804745235317e-07,
-      "loss": 2.4074,
-      "step": 54620
-    },
-    {
-      "epoch": 14.12210879958651,
-      "learning_rate": 1.1177233242577467e-07,
-      "loss": 2.4133,
-      "step": 54640
-    },
-    {
-      "epoch": 14.127277426024033,
-      "learning_rate": 1.1173991961623233e-07,
-      "loss": 2.4479,
-      "step": 54660
-    },
-    {
-      "epoch": 14.132446052461558,
-      "learning_rate": 1.1170750680669002e-07,
-      "loss": 2.4476,
-      "step": 54680
-    },
-    {
-      "epoch": 14.137614678899082,
-      "learning_rate": 1.1167509399714768e-07,
-      "loss": 2.4295,
-      "step": 54700
-    },
-    {
-      "epoch": 14.142783305336607,
-      "learning_rate": 1.1164268118760534e-07,
-      "loss": 2.4715,
-      "step": 54720
-    },
-    {
-      "epoch": 14.14795193177413,
-      "learning_rate": 1.1161026837806302e-07,
-      "loss": 2.4425,
-      "step": 54740
-    },
-    {
-      "epoch": 14.153120558211656,
-      "learning_rate": 1.1157785556852069e-07,
-      "loss": 2.4664,
-      "step": 54760
-    },
-    {
-      "epoch": 14.15828918464918,
-      "learning_rate": 1.1154544275897835e-07,
-      "loss": 2.424,
-      "step": 54780
-    },
-    {
-      "epoch": 14.163457811086703,
-      "learning_rate": 1.1151302994943601e-07,
-      "loss": 2.4141,
-      "step": 54800
-    },
-    {
-      "epoch": 14.168626437524228,
-      "learning_rate": 1.1148061713989369e-07,
-      "loss": 2.4566,
-      "step": 54820
-    },
-    {
-      "epoch": 14.173795063961752,
-      "learning_rate": 1.1144820433035136e-07,
-      "loss": 2.4483,
-      "step": 54840
-    },
-    {
-      "epoch": 14.178963690399277,
-      "learning_rate": 1.1141579152080903e-07,
-      "loss": 2.4681,
-      "step": 54860
-    },
-    {
-      "epoch": 14.1841323168368,
-      "learning_rate": 1.1138337871126669e-07,
-      "loss": 2.4122,
-      "step": 54880
-    },
-    {
-      "epoch": 14.189300943274326,
-      "learning_rate": 1.1135096590172436e-07,
-      "loss": 2.4219,
-      "step": 54900
-    },
-    {
-      "epoch": 14.194469569711849,
-      "learning_rate": 1.1131855309218204e-07,
-      "loss": 2.398,
-      "step": 54920
-    },
-    {
-      "epoch": 14.199638196149373,
-      "learning_rate": 1.112861402826397e-07,
-      "loss": 2.4346,
-      "step": 54940
-    },
-    {
-      "epoch": 14.204806822586898,
-      "learning_rate": 1.1125372747309736e-07,
-      "loss": 2.477,
-      "step": 54960
-    },
-    {
-      "epoch": 14.209975449024421,
-      "learning_rate": 1.1122131466355504e-07,
-      "loss": 2.412,
-      "step": 54980
-    },
-    {
-      "epoch": 14.215144075461946,
-      "learning_rate": 1.1118890185401271e-07,
-      "loss": 2.4304,
-      "step": 55000
-    },
-    {
-      "epoch": 14.22031270189947,
-      "learning_rate": 1.1115648904447037e-07,
-      "loss": 2.4396,
-      "step": 55020
-    },
-    {
-      "epoch": 14.225481328336995,
-      "learning_rate": 1.1112407623492805e-07,
-      "loss": 2.4407,
-      "step": 55040
-    },
-    {
-      "epoch": 14.230649954774519,
-      "learning_rate": 1.1109166342538571e-07,
-      "loss": 2.3734,
-      "step": 55060
-    },
-    {
-      "epoch": 14.235818581212042,
-      "learning_rate": 1.1105925061584339e-07,
-      "loss": 2.4121,
-      "step": 55080
-    },
-    {
-      "epoch": 14.240987207649567,
-      "learning_rate": 1.1102683780630105e-07,
-      "loss": 2.4242,
-      "step": 55100
-    },
-    {
-      "epoch": 14.24615583408709,
-      "learning_rate": 1.1099442499675872e-07,
-      "loss": 2.446,
-      "step": 55120
-    },
-    {
-      "epoch": 14.251324460524616,
-      "learning_rate": 1.1096201218721639e-07,
-      "loss": 2.3918,
-      "step": 55140
-    },
-    {
-      "epoch": 14.25649308696214,
-      "learning_rate": 1.1092959937767406e-07,
-      "loss": 2.4142,
-      "step": 55160
-    },
-    {
-      "epoch": 14.261661713399665,
-      "learning_rate": 1.1089718656813172e-07,
-      "loss": 2.4401,
-      "step": 55180
-    },
-    {
-      "epoch": 14.266830339837188,
-      "learning_rate": 1.108647737585894e-07,
-      "loss": 2.4477,
-      "step": 55200
-    },
-    {
-      "epoch": 14.271998966274712,
-      "learning_rate": 1.1083236094904706e-07,
-      "loss": 2.434,
-      "step": 55220
-    },
-    {
-      "epoch": 14.277167592712237,
-      "learning_rate": 1.1079994813950473e-07,
-      "loss": 2.4021,
-      "step": 55240
-    },
-    {
-      "epoch": 14.28233621914976,
-      "learning_rate": 1.107675353299624e-07,
-      "loss": 2.4173,
-      "step": 55260
-    },
-    {
-      "epoch": 14.287504845587286,
-      "learning_rate": 1.1073512252042007e-07,
-      "loss": 2.4248,
-      "step": 55280
-    },
-    {
-      "epoch": 14.29267347202481,
-      "learning_rate": 1.1070270971087773e-07,
-      "loss": 2.449,
-      "step": 55300
-    },
-    {
-      "epoch": 14.297842098462334,
-      "learning_rate": 1.1067029690133541e-07,
-      "loss": 2.5167,
-      "step": 55320
-    },
-    {
-      "epoch": 14.303010724899858,
-      "learning_rate": 1.1063788409179308e-07,
-      "loss": 2.439,
-      "step": 55340
-    },
-    {
-      "epoch": 14.308179351337381,
-      "learning_rate": 1.1060547128225075e-07,
-      "loss": 2.4282,
-      "step": 55360
-    },
-    {
-      "epoch": 14.313347977774907,
-      "learning_rate": 1.1057305847270841e-07,
-      "loss": 2.4502,
-      "step": 55380
-    },
-    {
-      "epoch": 14.31851660421243,
-      "learning_rate": 1.1054064566316608e-07,
-      "loss": 2.4245,
-      "step": 55400
-    },
-    {
-      "epoch": 14.323685230649955,
-      "learning_rate": 1.1050823285362376e-07,
-      "loss": 2.4192,
-      "step": 55420
-    },
-    {
-      "epoch": 14.328853857087479,
-      "learning_rate": 1.1047582004408142e-07,
-      "loss": 2.4461,
-      "step": 55440
-    },
-    {
-      "epoch": 14.334022483525004,
-      "learning_rate": 1.1044340723453908e-07,
-      "loss": 2.3875,
-      "step": 55460
-    },
-    {
-      "epoch": 14.339191109962528,
-      "learning_rate": 1.1041099442499676e-07,
-      "loss": 2.3687,
-      "step": 55480
-    },
-    {
-      "epoch": 14.344359736400051,
-      "learning_rate": 1.1037858161545443e-07,
-      "loss": 2.5006,
-      "step": 55500
-    },
-    {
-      "epoch": 14.349528362837576,
-      "learning_rate": 1.1034616880591209e-07,
-      "loss": 2.4207,
-      "step": 55520
-    },
-    {
-      "epoch": 14.3546969892751,
-      "learning_rate": 1.1031375599636977e-07,
-      "loss": 2.4137,
-      "step": 55540
-    },
-    {
-      "epoch": 14.359865615712625,
-      "learning_rate": 1.1028134318682743e-07,
-      "loss": 2.4323,
-      "step": 55560
-    },
-    {
-      "epoch": 14.365034242150148,
-      "learning_rate": 1.102489303772851e-07,
-      "loss": 2.3947,
-      "step": 55580
-    },
-    {
-      "epoch": 14.370202868587672,
-      "learning_rate": 1.1021651756774277e-07,
-      "loss": 2.4286,
-      "step": 55600
-    },
-    {
-      "epoch": 14.375371495025197,
-      "learning_rate": 1.1018410475820044e-07,
-      "loss": 2.3842,
-      "step": 55620
-    },
-    {
-      "epoch": 14.38054012146272,
-      "learning_rate": 1.101516919486581e-07,
-      "loss": 2.4287,
-      "step": 55640
-    },
-    {
-      "epoch": 14.385708747900246,
-      "learning_rate": 1.1011927913911578e-07,
-      "loss": 2.441,
-      "step": 55660
-    },
-    {
-      "epoch": 14.39087737433777,
-      "learning_rate": 1.1008686632957344e-07,
-      "loss": 2.467,
-      "step": 55680
-    },
-    {
-      "epoch": 14.396046000775295,
-      "learning_rate": 1.1005445352003112e-07,
-      "loss": 2.4228,
-      "step": 55700
-    },
-    {
-      "epoch": 14.401214627212818,
-      "learning_rate": 1.1002204071048878e-07,
-      "loss": 2.4338,
-      "step": 55720
-    },
-    {
-      "epoch": 14.406383253650343,
-      "learning_rate": 1.0998962790094645e-07,
-      "loss": 2.4264,
-      "step": 55740
-    },
-    {
-      "epoch": 14.411551880087867,
-      "learning_rate": 1.0995721509140412e-07,
-      "loss": 2.3694,
-      "step": 55760
-    },
-    {
-      "epoch": 14.41672050652539,
-      "learning_rate": 1.0992480228186179e-07,
-      "loss": 2.4507,
-      "step": 55780
-    },
-    {
-      "epoch": 14.421889132962916,
-      "learning_rate": 1.0989238947231945e-07,
-      "loss": 2.4463,
-      "step": 55800
-    },
-    {
-      "epoch": 14.427057759400439,
-      "learning_rate": 1.0985997666277713e-07,
-      "loss": 2.4241,
-      "step": 55820
-    },
-    {
-      "epoch": 14.432226385837964,
-      "learning_rate": 1.0982756385323479e-07,
-      "loss": 2.4437,
-      "step": 55840
-    },
-    {
-      "epoch": 14.437395012275488,
-      "learning_rate": 1.0979515104369246e-07,
-      "loss": 2.4293,
-      "step": 55860
-    },
-    {
-      "epoch": 14.442563638713011,
-      "learning_rate": 1.0976273823415014e-07,
-      "loss": 2.4539,
-      "step": 55880
-    },
-    {
-      "epoch": 14.447732265150536,
-      "learning_rate": 1.097303254246078e-07,
-      "loss": 2.444,
-      "step": 55900
-    },
-    {
-      "epoch": 14.45290089158806,
-      "learning_rate": 1.0969791261506546e-07,
-      "loss": 2.4366,
-      "step": 55920
-    },
-    {
-      "epoch": 14.458069518025585,
-      "learning_rate": 1.0966549980552314e-07,
-      "loss": 2.3765,
-      "step": 55940
-    },
-    {
-      "epoch": 14.463238144463109,
-      "learning_rate": 1.0963308699598081e-07,
-      "loss": 2.4172,
-      "step": 55960
-    },
-    {
-      "epoch": 14.468406770900634,
-      "learning_rate": 1.0960067418643847e-07,
-      "loss": 2.4072,
-      "step": 55980
-    },
-    {
-      "epoch": 14.473575397338157,
-      "learning_rate": 1.0956826137689614e-07,
-      "loss": 2.4183,
-      "step": 56000
-    },
-    {
-      "epoch": 14.47874402377568,
-      "learning_rate": 1.0953584856735381e-07,
-      "loss": 2.4149,
-      "step": 56020
-    },
-    {
-      "epoch": 14.483912650213206,
-      "learning_rate": 1.0950343575781149e-07,
-      "loss": 2.4087,
-      "step": 56040
-    },
-    {
-      "epoch": 14.48908127665073,
-      "learning_rate": 1.0947102294826915e-07,
-      "loss": 2.4551,
-      "step": 56060
-    },
-    {
-      "epoch": 14.494249903088255,
-      "learning_rate": 1.0943861013872681e-07,
-      "loss": 2.431,
-      "step": 56080
-    },
-    {
-      "epoch": 14.499418529525778,
-      "learning_rate": 1.0940619732918449e-07,
-      "loss": 2.4769,
-      "step": 56100
-    },
-    {
-      "epoch": 14.504587155963304,
-      "learning_rate": 1.0937378451964216e-07,
-      "loss": 2.4141,
-      "step": 56120
-    },
-    {
-      "epoch": 14.509755782400827,
-      "learning_rate": 1.0934137171009982e-07,
-      "loss": 2.4434,
-      "step": 56140
-    },
-    {
-      "epoch": 14.51492440883835,
-      "learning_rate": 1.0930895890055748e-07,
-      "loss": 2.4858,
-      "step": 56160
-    },
-    {
-      "epoch": 14.520093035275876,
-      "learning_rate": 1.0927654609101516e-07,
-      "loss": 2.4497,
-      "step": 56180
-    },
-    {
-      "epoch": 14.5252616617134,
-      "learning_rate": 1.0924413328147283e-07,
-      "loss": 2.3597,
-      "step": 56200
-    },
-    {
-      "epoch": 14.530430288150924,
-      "learning_rate": 1.092117204719305e-07,
-      "loss": 2.4058,
-      "step": 56220
-    },
-    {
-      "epoch": 14.535598914588448,
-      "learning_rate": 1.0917930766238816e-07,
-      "loss": 2.4425,
-      "step": 56240
-    },
-    {
-      "epoch": 14.540767541025973,
-      "learning_rate": 1.0914689485284583e-07,
-      "loss": 2.4314,
-      "step": 56260
-    },
-    {
-      "epoch": 14.545936167463497,
-      "learning_rate": 1.0911448204330351e-07,
-      "loss": 2.4842,
-      "step": 56280
-    },
-    {
-      "epoch": 14.55110479390102,
-      "learning_rate": 1.0908206923376117e-07,
-      "loss": 2.4116,
-      "step": 56300
-    },
-    {
-      "epoch": 14.556273420338545,
-      "learning_rate": 1.0904965642421883e-07,
-      "loss": 2.4123,
-      "step": 56320
-    },
-    {
-      "epoch": 14.561442046776069,
-      "learning_rate": 1.0901724361467651e-07,
-      "loss": 2.4266,
-      "step": 56340
-    },
-    {
-      "epoch": 14.566610673213594,
-      "learning_rate": 1.0898483080513418e-07,
-      "loss": 2.4135,
-      "step": 56360
-    },
-    {
-      "epoch": 14.571779299651118,
-      "learning_rate": 1.0895241799559184e-07,
-      "loss": 2.4508,
-      "step": 56380
-    },
-    {
-      "epoch": 14.576947926088643,
-      "learning_rate": 1.089200051860495e-07,
-      "loss": 2.4001,
-      "step": 56400
-    },
-    {
-      "epoch": 14.582116552526166,
-      "learning_rate": 1.0888759237650718e-07,
-      "loss": 2.3891,
-      "step": 56420
-    },
-    {
-      "epoch": 14.58728517896369,
-      "learning_rate": 1.0885517956696486e-07,
-      "loss": 2.4768,
-      "step": 56440
-    },
-    {
-      "epoch": 14.592453805401215,
-      "learning_rate": 1.0882276675742252e-07,
-      "loss": 2.4198,
-      "step": 56460
-    },
-    {
-      "epoch": 14.597622431838738,
-      "learning_rate": 1.0879035394788018e-07,
-      "loss": 2.404,
-      "step": 56480
-    },
-    {
-      "epoch": 14.602791058276264,
-      "learning_rate": 1.0875794113833786e-07,
-      "loss": 2.4938,
-      "step": 56500
-    },
-    {
-      "epoch": 14.607959684713787,
-      "learning_rate": 1.0872552832879553e-07,
-      "loss": 2.4509,
-      "step": 56520
-    },
-    {
-      "epoch": 14.61312831115131,
-      "learning_rate": 1.0869311551925319e-07,
-      "loss": 2.4481,
-      "step": 56540
-    },
-    {
-      "epoch": 14.618296937588836,
-      "learning_rate": 1.0866070270971085e-07,
-      "loss": 2.4265,
-      "step": 56560
-    },
-    {
-      "epoch": 14.62346556402636,
-      "learning_rate": 1.0862828990016854e-07,
-      "loss": 2.4469,
-      "step": 56580
-    },
-    {
-      "epoch": 14.628634190463885,
-      "learning_rate": 1.085958770906262e-07,
-      "loss": 2.4133,
-      "step": 56600
-    },
-    {
-      "epoch": 14.633802816901408,
-      "learning_rate": 1.0856346428108387e-07,
-      "loss": 2.4522,
-      "step": 56620
-    },
-    {
-      "epoch": 14.638971443338933,
-      "learning_rate": 1.0853105147154153e-07,
-      "loss": 2.4041,
-      "step": 56640
-    },
-    {
-      "epoch": 14.644140069776457,
-      "learning_rate": 1.0849863866199922e-07,
-      "loss": 2.4375,
-      "step": 56660
-    },
-    {
-      "epoch": 14.649308696213982,
-      "learning_rate": 1.0846622585245688e-07,
-      "loss": 2.4063,
-      "step": 56680
-    },
-    {
-      "epoch": 14.654477322651505,
-      "learning_rate": 1.0843381304291454e-07,
-      "loss": 2.4281,
-      "step": 56700
-    },
-    {
-      "epoch": 14.659645949089029,
-      "learning_rate": 1.084014002333722e-07,
-      "loss": 2.3715,
-      "step": 56720
-    },
-    {
-      "epoch": 14.664814575526554,
-      "learning_rate": 1.0836898742382989e-07,
-      "loss": 2.4447,
-      "step": 56740
-    },
-    {
-      "epoch": 14.669983201964078,
-      "learning_rate": 1.0833657461428755e-07,
-      "loss": 2.4509,
-      "step": 56760
-    },
-    {
-      "epoch": 14.675151828401603,
-      "learning_rate": 1.0830416180474521e-07,
-      "loss": 2.4565,
-      "step": 56780
-    },
-    {
-      "epoch": 14.680320454839126,
-      "learning_rate": 1.0827174899520288e-07,
-      "loss": 2.4184,
-      "step": 56800
-    },
-    {
-      "epoch": 14.68548908127665,
-      "learning_rate": 1.0823933618566056e-07,
-      "loss": 2.3964,
-      "step": 56820
-    },
-    {
-      "epoch": 14.690657707714175,
-      "learning_rate": 1.0820692337611823e-07,
-      "loss": 2.3597,
-      "step": 56840
-    },
-    {
-      "epoch": 14.695826334151699,
-      "learning_rate": 1.0817451056657589e-07,
-      "loss": 2.4697,
-      "step": 56860
-    },
-    {
-      "epoch": 14.700994960589224,
-      "learning_rate": 1.0814209775703358e-07,
-      "loss": 2.4048,
-      "step": 56880
-    },
-    {
-      "epoch": 14.706163587026747,
-      "learning_rate": 1.0810968494749124e-07,
-      "loss": 2.3896,
-      "step": 56900
-    },
-    {
-      "epoch": 14.711332213464273,
-      "learning_rate": 1.080772721379489e-07,
-      "loss": 2.4108,
-      "step": 56920
-    },
-    {
-      "epoch": 14.716500839901796,
-      "learning_rate": 1.0804485932840656e-07,
-      "loss": 2.3934,
-      "step": 56940
-    },
-    {
-      "epoch": 14.72166946633932,
-      "learning_rate": 1.0801244651886425e-07,
-      "loss": 2.3797,
-      "step": 56960
-    },
-    {
-      "epoch": 14.726838092776845,
-      "learning_rate": 1.0798003370932191e-07,
-      "loss": 2.3676,
-      "step": 56980
-    },
-    {
-      "epoch": 14.732006719214368,
-      "learning_rate": 1.0794762089977957e-07,
-      "loss": 2.426,
-      "step": 57000
-    },
-    {
-      "epoch": 14.737175345651893,
-      "learning_rate": 1.0791520809023724e-07,
-      "loss": 2.4503,
-      "step": 57020
-    },
-    {
-      "epoch": 14.742343972089417,
-      "learning_rate": 1.0788279528069492e-07,
-      "loss": 2.449,
-      "step": 57040
-    },
-    {
-      "epoch": 14.747512598526942,
-      "learning_rate": 1.0785038247115259e-07,
-      "loss": 2.4417,
-      "step": 57060
-    },
-    {
-      "epoch": 14.752681224964466,
-      "learning_rate": 1.0781796966161025e-07,
-      "loss": 2.4536,
-      "step": 57080
-    },
-    {
-      "epoch": 14.75784985140199,
-      "learning_rate": 1.0778555685206791e-07,
-      "loss": 2.4133,
-      "step": 57100
-    },
-    {
-      "epoch": 14.763018477839514,
-      "learning_rate": 1.077531440425256e-07,
-      "loss": 2.4921,
-      "step": 57120
-    },
-    {
-      "epoch": 14.768187104277038,
-      "learning_rate": 1.0772073123298326e-07,
-      "loss": 2.4226,
-      "step": 57140
-    },
-    {
-      "epoch": 14.773355730714563,
-      "learning_rate": 1.0768831842344092e-07,
-      "loss": 2.4188,
-      "step": 57160
-    },
-    {
-      "epoch": 14.778524357152087,
-      "learning_rate": 1.0765590561389861e-07,
-      "loss": 2.4282,
-      "step": 57180
-    },
-    {
-      "epoch": 14.783692983589612,
-      "learning_rate": 1.0762349280435627e-07,
-      "loss": 2.4692,
-      "step": 57200
-    },
-    {
-      "epoch": 14.788861610027135,
-      "learning_rate": 1.0759107999481393e-07,
-      "loss": 2.394,
-      "step": 57220
-    },
-    {
-      "epoch": 14.794030236464659,
-      "learning_rate": 1.075586671852716e-07,
-      "loss": 2.4628,
-      "step": 57240
-    },
-    {
-      "epoch": 14.799198862902184,
-      "learning_rate": 1.0752625437572928e-07,
-      "loss": 2.4294,
-      "step": 57260
-    },
-    {
-      "epoch": 14.804367489339707,
-      "learning_rate": 1.0749384156618695e-07,
-      "loss": 2.4667,
-      "step": 57280
-    },
-    {
-      "epoch": 14.809536115777233,
-      "learning_rate": 1.0746142875664461e-07,
-      "loss": 2.4368,
-      "step": 57300
-    },
-    {
-      "epoch": 14.814704742214756,
-      "learning_rate": 1.0742901594710227e-07,
-      "loss": 2.3656,
-      "step": 57320
-    },
-    {
-      "epoch": 14.819873368652281,
-      "learning_rate": 1.0739660313755996e-07,
-      "loss": 2.4152,
-      "step": 57340
-    },
-    {
-      "epoch": 14.825041995089805,
-      "learning_rate": 1.0736419032801762e-07,
-      "loss": 2.4619,
-      "step": 57360
-    },
-    {
-      "epoch": 14.830210621527328,
-      "learning_rate": 1.0733177751847528e-07,
-      "loss": 2.4696,
-      "step": 57380
-    },
-    {
-      "epoch": 14.835379247964854,
-      "learning_rate": 1.0729936470893294e-07,
-      "loss": 2.4359,
-      "step": 57400
-    },
-    {
-      "epoch": 14.840547874402377,
-      "learning_rate": 1.0726695189939063e-07,
-      "loss": 2.4534,
-      "step": 57420
-    },
-    {
-      "epoch": 14.845716500839902,
-      "learning_rate": 1.072345390898483e-07,
-      "loss": 2.4296,
-      "step": 57440
-    },
-    {
-      "epoch": 14.850885127277426,
-      "learning_rate": 1.0720212628030596e-07,
-      "loss": 2.4417,
-      "step": 57460
-    },
-    {
-      "epoch": 14.856053753714951,
-      "learning_rate": 1.0716971347076364e-07,
-      "loss": 2.4688,
-      "step": 57480
-    },
-    {
-      "epoch": 14.861222380152475,
-      "learning_rate": 1.071373006612213e-07,
-      "loss": 2.4566,
-      "step": 57500
-    },
-    {
-      "epoch": 14.866391006589998,
-      "learning_rate": 1.0710488785167897e-07,
-      "loss": 2.4179,
-      "step": 57520
-    },
-    {
-      "epoch": 14.871559633027523,
-      "learning_rate": 1.0707247504213663e-07,
-      "loss": 2.4653,
-      "step": 57540
-    },
-    {
-      "epoch": 14.876728259465047,
-      "learning_rate": 1.0704006223259432e-07,
-      "loss": 2.4089,
-      "step": 57560
-    },
-    {
-      "epoch": 14.881896885902572,
-      "learning_rate": 1.0700764942305198e-07,
-      "loss": 2.4813,
-      "step": 57580
-    },
-    {
-      "epoch": 14.887065512340095,
-      "learning_rate": 1.0697523661350964e-07,
-      "loss": 2.3872,
-      "step": 57600
-    },
-    {
-      "epoch": 14.89223413877762,
-      "learning_rate": 1.069428238039673e-07,
-      "loss": 2.4628,
-      "step": 57620
-    },
-    {
-      "epoch": 14.897402765215144,
-      "learning_rate": 1.0691041099442499e-07,
-      "loss": 2.3651,
-      "step": 57640
-    },
-    {
-      "epoch": 14.902571391652668,
-      "learning_rate": 1.0687799818488265e-07,
-      "loss": 2.4236,
-      "step": 57660
-    },
-    {
-      "epoch": 14.907740018090193,
-      "learning_rate": 1.0684558537534032e-07,
-      "loss": 2.4099,
-      "step": 57680
-    },
-    {
-      "epoch": 14.912908644527716,
-      "learning_rate": 1.0681317256579798e-07,
-      "loss": 2.4265,
-      "step": 57700
-    },
-    {
-      "epoch": 14.918077270965242,
-      "learning_rate": 1.0678075975625567e-07,
-      "loss": 2.4424,
-      "step": 57720
-    },
-    {
-      "epoch": 14.923245897402765,
-      "learning_rate": 1.0674834694671335e-07,
-      "loss": 2.3981,
-      "step": 57740
-    },
-    {
-      "epoch": 14.928414523840289,
-      "learning_rate": 1.0671593413717099e-07,
-      "loss": 2.4498,
-      "step": 57760
-    },
-    {
-      "epoch": 14.933583150277814,
-      "learning_rate": 1.0668352132762868e-07,
-      "loss": 2.4153,
-      "step": 57780
-    },
-    {
-      "epoch": 14.938751776715337,
-      "learning_rate": 1.0665110851808634e-07,
-      "loss": 2.4706,
-      "step": 57800
-    },
-    {
-      "epoch": 14.943920403152863,
-      "learning_rate": 1.0661869570854403e-07,
-      "loss": 2.4511,
-      "step": 57820
-    },
-    {
-      "epoch": 14.949089029590386,
-      "learning_rate": 1.0658628289900166e-07,
-      "loss": 2.4389,
-      "step": 57840
-    },
-    {
-      "epoch": 14.954257656027911,
-      "learning_rate": 1.0655387008945935e-07,
-      "loss": 2.4287,
-      "step": 57860
-    },
-    {
-      "epoch": 14.959426282465435,
-      "learning_rate": 1.0652145727991701e-07,
-      "loss": 2.4765,
-      "step": 57880
-    },
-    {
-      "epoch": 14.96459490890296,
-      "learning_rate": 1.0648904447037468e-07,
-      "loss": 2.4132,
-      "step": 57900
-    },
-    {
-      "epoch": 14.969763535340483,
-      "learning_rate": 1.0645663166083234e-07,
-      "loss": 2.4539,
-      "step": 57920
-    },
-    {
-      "epoch": 14.974932161778007,
-      "learning_rate": 1.0642421885129003e-07,
-      "loss": 2.4428,
-      "step": 57940
-    },
-    {
-      "epoch": 14.980100788215532,
-      "learning_rate": 1.0639180604174769e-07,
-      "loss": 2.3815,
-      "step": 57960
-    },
-    {
-      "epoch": 14.985269414653056,
-      "learning_rate": 1.0635939323220535e-07,
-      "loss": 2.4527,
-      "step": 57980
-    },
-    {
-      "epoch": 14.99043804109058,
-      "learning_rate": 1.0632698042266301e-07,
-      "loss": 2.4376,
-      "step": 58000
-    },
-    {
-      "epoch": 14.995606667528104,
-      "learning_rate": 1.062945676131207e-07,
-      "loss": 2.4053,
-      "step": 58020
-    },
-    {
-      "epoch": 15.0,
-      "eval_bleu": 13.5764,
-      "eval_gen_len": 36.1492,
-      "eval_loss": 2.4080686569213867,
-      "eval_runtime": 787.1528,
-      "eval_samples_per_second": 2.188,
-      "eval_steps_per_second": 1.094,
-      "step": 58037
-    },
-    {
-      "epoch": 15.000775293965628,
-      "learning_rate": 1.0626215480357839e-07,
-      "loss": 2.3623,
-      "step": 58040
-    },
-    {
-      "epoch": 15.005943920403153,
-      "learning_rate": 1.0622974199403602e-07,
-      "loss": 2.4089,
-      "step": 58060
-    },
-    {
-      "epoch": 15.011112546840677,
-      "learning_rate": 1.0619732918449371e-07,
-      "loss": 2.5134,
-      "step": 58080
-    },
-    {
-      "epoch": 15.016281173278202,
-      "learning_rate": 1.0616491637495137e-07,
-      "loss": 2.4379,
-      "step": 58100
-    },
-    {
-      "epoch": 15.021449799715725,
-      "learning_rate": 1.0613250356540906e-07,
-      "loss": 2.4265,
-      "step": 58120
-    },
-    {
-      "epoch": 15.02661842615325,
-      "learning_rate": 1.061000907558667e-07,
-      "loss": 2.438,
-      "step": 58140
-    },
-    {
-      "epoch": 15.031787052590774,
-      "learning_rate": 1.0606767794632439e-07,
-      "loss": 2.4204,
-      "step": 58160
-    },
-    {
-      "epoch": 15.036955679028297,
-      "learning_rate": 1.0603526513678205e-07,
-      "loss": 2.4318,
-      "step": 58180
-    },
-    {
-      "epoch": 15.042124305465823,
-      "learning_rate": 1.0600285232723974e-07,
-      "loss": 2.4512,
-      "step": 58200
-    },
-    {
-      "epoch": 15.047292931903346,
-      "learning_rate": 1.0597043951769737e-07,
-      "loss": 2.4247,
-      "step": 58220
-    },
-    {
-      "epoch": 15.052461558340871,
-      "learning_rate": 1.0593802670815506e-07,
-      "loss": 2.4195,
-      "step": 58240
-    },
-    {
-      "epoch": 15.057630184778395,
-      "learning_rate": 1.0590561389861272e-07,
-      "loss": 2.4264,
-      "step": 58260
-    },
-    {
-      "epoch": 15.06279881121592,
-      "learning_rate": 1.0587320108907041e-07,
-      "loss": 2.4425,
-      "step": 58280
-    },
-    {
-      "epoch": 15.067967437653444,
-      "learning_rate": 1.0584078827952805e-07,
-      "loss": 2.3938,
-      "step": 58300
-    },
-    {
-      "epoch": 15.073136064090967,
-      "learning_rate": 1.0580837546998573e-07,
-      "loss": 2.4447,
-      "step": 58320
-    },
-    {
-      "epoch": 15.078304690528492,
-      "learning_rate": 1.0577596266044342e-07,
-      "loss": 2.3923,
-      "step": 58340
-    },
-    {
-      "epoch": 15.083473316966016,
-      "learning_rate": 1.0574354985090108e-07,
-      "loss": 2.4065,
-      "step": 58360
-    },
-    {
-      "epoch": 15.088641943403541,
-      "learning_rate": 1.0571113704135875e-07,
-      "loss": 2.4171,
-      "step": 58380
-    },
-    {
-      "epoch": 15.093810569841065,
-      "learning_rate": 1.0567872423181641e-07,
-      "loss": 2.4148,
-      "step": 58400
-    },
-    {
-      "epoch": 15.09897919627859,
-      "learning_rate": 1.056463114222741e-07,
-      "loss": 2.4317,
-      "step": 58420
-    },
-    {
-      "epoch": 15.104147822716113,
-      "learning_rate": 1.0561389861273176e-07,
-      "loss": 2.3749,
-      "step": 58440
-    },
-    {
-      "epoch": 15.109316449153637,
-      "learning_rate": 1.0558148580318942e-07,
-      "loss": 2.3947,
-      "step": 58460
-    },
-    {
-      "epoch": 15.114485075591162,
-      "learning_rate": 1.0554907299364708e-07,
-      "loss": 2.3908,
-      "step": 58480
-    },
-    {
-      "epoch": 15.119653702028685,
-      "learning_rate": 1.0551666018410477e-07,
-      "loss": 2.3965,
-      "step": 58500
-    },
-    {
-      "epoch": 15.12482232846621,
-      "learning_rate": 1.0548424737456243e-07,
-      "loss": 2.4536,
-      "step": 58520
-    },
-    {
-      "epoch": 15.129990954903734,
-      "learning_rate": 1.054518345650201e-07,
-      "loss": 2.4602,
-      "step": 58540
-    },
-    {
-      "epoch": 15.13515958134126,
-      "learning_rate": 1.0541942175547776e-07,
-      "loss": 2.4214,
-      "step": 58560
-    },
-    {
-      "epoch": 15.140328207778783,
-      "learning_rate": 1.0538700894593544e-07,
-      "loss": 2.4254,
-      "step": 58580
-    },
-    {
-      "epoch": 15.145496834216306,
-      "learning_rate": 1.053545961363931e-07,
-      "loss": 2.446,
-      "step": 58600
-    },
-    {
-      "epoch": 15.150665460653832,
-      "learning_rate": 1.0532218332685077e-07,
-      "loss": 2.4541,
-      "step": 58620
-    },
-    {
-      "epoch": 15.155834087091355,
-      "learning_rate": 1.0528977051730846e-07,
-      "loss": 2.4487,
-      "step": 58640
-    },
-    {
-      "epoch": 15.16100271352888,
-      "learning_rate": 1.0525735770776612e-07,
-      "loss": 2.3776,
-      "step": 58660
-    },
-    {
-      "epoch": 15.166171339966404,
-      "learning_rate": 1.0522494489822378e-07,
-      "loss": 2.3837,
-      "step": 58680
-    },
-    {
-      "epoch": 15.171339966403929,
-      "learning_rate": 1.0519253208868144e-07,
-      "loss": 2.3505,
-      "step": 58700
-    },
-    {
-      "epoch": 15.176508592841452,
-      "learning_rate": 1.0516011927913913e-07,
-      "loss": 2.4518,
-      "step": 58720
-    },
-    {
-      "epoch": 15.181677219278976,
-      "learning_rate": 1.0512770646959679e-07,
-      "loss": 2.3862,
-      "step": 58740
-    },
-    {
-      "epoch": 15.186845845716501,
-      "learning_rate": 1.0509529366005445e-07,
-      "loss": 2.3564,
-      "step": 58760
-    },
-    {
-      "epoch": 15.192014472154025,
-      "learning_rate": 1.0506288085051212e-07,
-      "loss": 2.3832,
-      "step": 58780
-    },
-    {
-      "epoch": 15.19718309859155,
-      "learning_rate": 1.050304680409698e-07,
-      "loss": 2.4699,
-      "step": 58800
-    },
-    {
-      "epoch": 15.202351725029073,
-      "learning_rate": 1.0499805523142747e-07,
-      "loss": 2.4436,
-      "step": 58820
-    },
-    {
-      "epoch": 15.207520351466599,
-      "learning_rate": 1.0496564242188513e-07,
-      "loss": 2.3911,
-      "step": 58840
-    },
-    {
-      "epoch": 15.212688977904122,
-      "learning_rate": 1.0493322961234279e-07,
-      "loss": 2.3914,
-      "step": 58860
-    },
-    {
-      "epoch": 15.217857604341646,
-      "learning_rate": 1.0490081680280048e-07,
-      "loss": 2.454,
-      "step": 58880
-    },
-    {
-      "epoch": 15.22302623077917,
-      "learning_rate": 1.0486840399325814e-07,
-      "loss": 2.424,
-      "step": 58900
-    },
-    {
-      "epoch": 15.228194857216694,
-      "learning_rate": 1.048359911837158e-07,
-      "loss": 2.4053,
-      "step": 58920
-    },
-    {
-      "epoch": 15.23336348365422,
-      "learning_rate": 1.0480357837417349e-07,
-      "loss": 2.41,
-      "step": 58940
-    },
-    {
-      "epoch": 15.238532110091743,
-      "learning_rate": 1.0477116556463115e-07,
-      "loss": 2.4504,
-      "step": 58960
-    },
-    {
-      "epoch": 15.243700736529266,
-      "learning_rate": 1.0473875275508881e-07,
-      "loss": 2.4132,
-      "step": 58980
-    },
-    {
-      "epoch": 15.248869362966792,
-      "learning_rate": 1.0470633994554648e-07,
-      "loss": 2.4664,
-      "step": 59000
-    },
-    {
-      "epoch": 15.254037989404315,
-      "learning_rate": 1.0467392713600416e-07,
-      "loss": 2.4222,
-      "step": 59020
-    },
-    {
-      "epoch": 15.25920661584184,
-      "learning_rate": 1.0464151432646183e-07,
-      "loss": 2.4471,
-      "step": 59040
-    },
-    {
-      "epoch": 15.264375242279364,
-      "learning_rate": 1.0460910151691949e-07,
-      "loss": 2.402,
-      "step": 59060
-    },
-    {
-      "epoch": 15.26954386871689,
-      "learning_rate": 1.0457668870737715e-07,
-      "loss": 2.3928,
-      "step": 59080
-    },
-    {
-      "epoch": 15.274712495154413,
-      "learning_rate": 1.0454427589783484e-07,
-      "loss": 2.3765,
-      "step": 59100
-    },
-    {
-      "epoch": 15.279881121591936,
-      "learning_rate": 1.045118630882925e-07,
-      "loss": 2.3731,
-      "step": 59120
-    },
-    {
-      "epoch": 15.285049748029461,
-      "learning_rate": 1.0447945027875016e-07,
-      "loss": 2.4251,
-      "step": 59140
-    },
-    {
-      "epoch": 15.290218374466985,
-      "learning_rate": 1.0444703746920782e-07,
-      "loss": 2.4078,
-      "step": 59160
-    },
-    {
-      "epoch": 15.29538700090451,
-      "learning_rate": 1.0441462465966551e-07,
-      "loss": 2.3711,
-      "step": 59180
-    },
-    {
-      "epoch": 15.300555627342034,
-      "learning_rate": 1.0438221185012317e-07,
-      "loss": 2.4432,
-      "step": 59200
-    },
-    {
-      "epoch": 15.305724253779559,
-      "learning_rate": 1.0434979904058084e-07,
-      "loss": 2.4697,
-      "step": 59220
-    },
-    {
-      "epoch": 15.310892880217082,
-      "learning_rate": 1.0431738623103852e-07,
-      "loss": 2.4792,
-      "step": 59240
-    },
-    {
-      "epoch": 15.316061506654606,
-      "learning_rate": 1.0428497342149619e-07,
-      "loss": 2.4445,
-      "step": 59260
-    },
-    {
-      "epoch": 15.321230133092131,
-      "learning_rate": 1.0425256061195385e-07,
-      "loss": 2.4796,
-      "step": 59280
-    },
-    {
-      "epoch": 15.326398759529654,
-      "learning_rate": 1.0422014780241151e-07,
-      "loss": 2.4082,
-      "step": 59300
-    },
-    {
-      "epoch": 15.33156738596718,
-      "learning_rate": 1.041877349928692e-07,
-      "loss": 2.4877,
-      "step": 59320
-    },
-    {
-      "epoch": 15.336736012404703,
-      "learning_rate": 1.0415532218332686e-07,
-      "loss": 2.3853,
-      "step": 59340
-    },
-    {
-      "epoch": 15.341904638842228,
-      "learning_rate": 1.0412290937378452e-07,
-      "loss": 2.4456,
-      "step": 59360
-    },
-    {
-      "epoch": 15.347073265279752,
-      "learning_rate": 1.0409049656424218e-07,
-      "loss": 2.3743,
-      "step": 59380
-    },
-    {
-      "epoch": 15.352241891717275,
-      "learning_rate": 1.0405808375469987e-07,
-      "loss": 2.4828,
-      "step": 59400
-    },
-    {
-      "epoch": 15.3574105181548,
-      "learning_rate": 1.0402567094515753e-07,
-      "loss": 2.4157,
-      "step": 59420
-    },
-    {
-      "epoch": 15.362579144592324,
-      "learning_rate": 1.039932581356152e-07,
-      "loss": 2.4593,
-      "step": 59440
-    },
-    {
-      "epoch": 15.36774777102985,
-      "learning_rate": 1.0396084532607286e-07,
-      "loss": 2.3706,
-      "step": 59460
-    },
-    {
-      "epoch": 15.372916397467373,
-      "learning_rate": 1.0392843251653055e-07,
-      "loss": 2.4284,
-      "step": 59480
-    },
-    {
-      "epoch": 15.378085023904898,
-      "learning_rate": 1.0389601970698821e-07,
-      "loss": 2.4174,
-      "step": 59500
-    },
-    {
-      "epoch": 15.383253650342422,
-      "learning_rate": 1.0386360689744587e-07,
-      "loss": 2.3777,
-      "step": 59520
-    },
-    {
-      "epoch": 15.388422276779945,
-      "learning_rate": 1.0383119408790354e-07,
-      "loss": 2.3894,
-      "step": 59540
-    },
-    {
-      "epoch": 15.39359090321747,
-      "learning_rate": 1.0379878127836122e-07,
-      "loss": 2.4183,
-      "step": 59560
-    },
-    {
-      "epoch": 15.398759529654994,
-      "learning_rate": 1.0376636846881888e-07,
-      "loss": 2.3614,
-      "step": 59580
-    },
-    {
-      "epoch": 15.403928156092519,
-      "learning_rate": 1.0373395565927654e-07,
-      "loss": 2.4399,
-      "step": 59600
-    },
-    {
-      "epoch": 15.409096782530042,
-      "learning_rate": 1.0370154284973422e-07,
-      "loss": 2.3998,
-      "step": 59620
-    },
-    {
-      "epoch": 15.414265408967568,
-      "learning_rate": 1.036691300401919e-07,
-      "loss": 2.4962,
-      "step": 59640
-    },
-    {
-      "epoch": 15.419434035405091,
-      "learning_rate": 1.0363671723064956e-07,
-      "loss": 2.3728,
-      "step": 59660
-    },
-    {
-      "epoch": 15.424602661842615,
-      "learning_rate": 1.0360430442110722e-07,
-      "loss": 2.4273,
-      "step": 59680
-    },
-    {
-      "epoch": 15.42977128828014,
-      "learning_rate": 1.0357189161156489e-07,
-      "loss": 2.4366,
-      "step": 59700
-    },
-    {
-      "epoch": 15.434939914717663,
-      "learning_rate": 1.0353947880202257e-07,
-      "loss": 2.4697,
-      "step": 59720
-    },
-    {
-      "epoch": 15.440108541155189,
-      "learning_rate": 1.0350706599248023e-07,
-      "loss": 2.3705,
-      "step": 59740
-    },
-    {
-      "epoch": 15.445277167592712,
-      "learning_rate": 1.0347465318293789e-07,
-      "loss": 2.4004,
-      "step": 59760
-    },
-    {
-      "epoch": 15.450445794030237,
-      "learning_rate": 1.0344224037339557e-07,
-      "loss": 2.4214,
-      "step": 59780
-    },
-    {
-      "epoch": 15.45561442046776,
-      "learning_rate": 1.0340982756385324e-07,
-      "loss": 2.4458,
-      "step": 59800
-    },
-    {
-      "epoch": 15.460783046905284,
-      "learning_rate": 1.033774147543109e-07,
-      "loss": 2.4129,
-      "step": 59820
-    },
-    {
-      "epoch": 15.46595167334281,
-      "learning_rate": 1.0334500194476858e-07,
-      "loss": 2.3648,
-      "step": 59840
-    },
-    {
-      "epoch": 15.471120299780333,
-      "learning_rate": 1.0331258913522624e-07,
-      "loss": 2.3726,
-      "step": 59860
-    },
-    {
-      "epoch": 15.476288926217858,
-      "learning_rate": 1.0328017632568392e-07,
-      "loss": 2.4154,
-      "step": 59880
-    },
-    {
-      "epoch": 15.481457552655382,
-      "learning_rate": 1.0324776351614158e-07,
-      "loss": 2.41,
-      "step": 59900
-    },
-    {
-      "epoch": 15.486626179092905,
-      "learning_rate": 1.0321535070659925e-07,
-      "loss": 2.4175,
-      "step": 59920
-    },
-    {
-      "epoch": 15.49179480553043,
-      "learning_rate": 1.0318293789705691e-07,
-      "loss": 2.4498,
-      "step": 59940
-    },
-    {
-      "epoch": 15.496963431967954,
-      "learning_rate": 1.0315052508751459e-07,
-      "loss": 2.4872,
-      "step": 59960
-    },
-    {
-      "epoch": 15.50213205840548,
-      "learning_rate": 1.0311811227797225e-07,
-      "loss": 2.3658,
-      "step": 59980
-    },
-    {
-      "epoch": 15.507300684843003,
-      "learning_rate": 1.0308569946842993e-07,
-      "loss": 2.4036,
-      "step": 60000
-    },
-    {
-      "epoch": 15.512469311280528,
-      "learning_rate": 1.0305328665888759e-07,
-      "loss": 2.4134,
-      "step": 60020
-    },
-    {
-      "epoch": 15.517637937718051,
-      "learning_rate": 1.0302087384934526e-07,
-      "loss": 2.4383,
-      "step": 60040
-    },
-    {
-      "epoch": 15.522806564155577,
-      "learning_rate": 1.0298846103980293e-07,
-      "loss": 2.3699,
-      "step": 60060
-    },
-    {
-      "epoch": 15.5279751905931,
-      "learning_rate": 1.029560482302606e-07,
-      "loss": 2.3865,
-      "step": 60080
-    },
-    {
-      "epoch": 15.533143817030624,
-      "learning_rate": 1.0292363542071826e-07,
-      "loss": 2.3725,
-      "step": 60100
-    },
-    {
-      "epoch": 15.538312443468149,
-      "learning_rate": 1.0289122261117594e-07,
-      "loss": 2.3363,
-      "step": 60120
-    },
-    {
-      "epoch": 15.543481069905672,
-      "learning_rate": 1.028588098016336e-07,
-      "loss": 2.3921,
-      "step": 60140
-    },
-    {
-      "epoch": 15.548649696343197,
-      "learning_rate": 1.0282639699209127e-07,
-      "loss": 2.4028,
-      "step": 60160
-    },
-    {
-      "epoch": 15.553818322780721,
-      "learning_rate": 1.0279398418254894e-07,
-      "loss": 2.4546,
-      "step": 60180
-    },
-    {
-      "epoch": 15.558986949218244,
-      "learning_rate": 1.0276157137300661e-07,
-      "loss": 2.3757,
-      "step": 60200
-    },
-    {
-      "epoch": 15.56415557565577,
-      "learning_rate": 1.0272915856346427e-07,
-      "loss": 2.3857,
-      "step": 60220
-    },
-    {
-      "epoch": 15.569324202093293,
-      "learning_rate": 1.0269674575392195e-07,
-      "loss": 2.4224,
-      "step": 60240
-    },
-    {
-      "epoch": 15.574492828530818,
-      "learning_rate": 1.0266433294437962e-07,
-      "loss": 2.4753,
-      "step": 60260
-    },
-    {
-      "epoch": 15.579661454968342,
-      "learning_rate": 1.0263192013483729e-07,
-      "loss": 2.3691,
-      "step": 60280
-    },
-    {
-      "epoch": 15.584830081405867,
-      "learning_rate": 1.0259950732529495e-07,
-      "loss": 2.363,
-      "step": 60300
-    },
-    {
-      "epoch": 15.58999870784339,
-      "learning_rate": 1.0256709451575262e-07,
-      "loss": 2.4099,
-      "step": 60320
-    },
-    {
-      "epoch": 15.595167334280914,
-      "learning_rate": 1.025346817062103e-07,
-      "loss": 2.4276,
-      "step": 60340
-    },
-    {
-      "epoch": 15.60033596071844,
-      "learning_rate": 1.0250226889666796e-07,
-      "loss": 2.4374,
-      "step": 60360
-    },
-    {
-      "epoch": 15.605504587155963,
-      "learning_rate": 1.0246985608712562e-07,
-      "loss": 2.426,
-      "step": 60380
-    },
-    {
-      "epoch": 15.610673213593488,
-      "learning_rate": 1.024374432775833e-07,
-      "loss": 2.4147,
-      "step": 60400
-    },
-    {
-      "epoch": 15.615841840031012,
-      "learning_rate": 1.0240503046804097e-07,
-      "loss": 2.4362,
-      "step": 60420
-    },
-    {
-      "epoch": 15.621010466468537,
-      "learning_rate": 1.0237261765849863e-07,
-      "loss": 2.4182,
-      "step": 60440
-    },
-    {
-      "epoch": 15.62617909290606,
-      "learning_rate": 1.023402048489563e-07,
-      "loss": 2.389,
-      "step": 60460
-    },
-    {
-      "epoch": 15.631347719343584,
-      "learning_rate": 1.0230779203941397e-07,
-      "loss": 2.4295,
-      "step": 60480
-    },
-    {
-      "epoch": 15.636516345781109,
-      "learning_rate": 1.0227537922987165e-07,
-      "loss": 2.4227,
-      "step": 60500
-    },
-    {
-      "epoch": 15.641684972218632,
-      "learning_rate": 1.0224296642032931e-07,
-      "loss": 2.4355,
-      "step": 60520
-    },
-    {
-      "epoch": 15.646853598656158,
-      "learning_rate": 1.0221055361078698e-07,
-      "loss": 2.3482,
-      "step": 60540
-    },
-    {
-      "epoch": 15.652022225093681,
-      "learning_rate": 1.0217814080124464e-07,
-      "loss": 2.4117,
-      "step": 60560
-    },
-    {
-      "epoch": 15.657190851531206,
-      "learning_rate": 1.0214572799170232e-07,
-      "loss": 2.3671,
-      "step": 60580
-    },
-    {
-      "epoch": 15.66235947796873,
-      "learning_rate": 1.0211331518215998e-07,
-      "loss": 2.44,
-      "step": 60600
-    },
-    {
-      "epoch": 15.667528104406253,
-      "learning_rate": 1.0208090237261766e-07,
-      "loss": 2.4273,
-      "step": 60620
-    },
-    {
-      "epoch": 15.672696730843779,
-      "learning_rate": 1.0204848956307532e-07,
-      "loss": 2.4038,
-      "step": 60640
-    },
-    {
-      "epoch": 15.677865357281302,
-      "learning_rate": 1.0201607675353299e-07,
-      "loss": 2.4434,
-      "step": 60660
-    },
-    {
-      "epoch": 15.683033983718827,
-      "learning_rate": 1.0198366394399066e-07,
-      "loss": 2.3874,
-      "step": 60680
-    },
-    {
-      "epoch": 15.68820261015635,
-      "learning_rate": 1.0195125113444833e-07,
-      "loss": 2.4237,
-      "step": 60700
-    },
-    {
-      "epoch": 15.693371236593876,
-      "learning_rate": 1.0191883832490599e-07,
-      "loss": 2.3947,
-      "step": 60720
-    },
-    {
-      "epoch": 15.6985398630314,
-      "learning_rate": 1.0188642551536367e-07,
-      "loss": 2.4482,
-      "step": 60740
-    },
-    {
-      "epoch": 15.703708489468923,
-      "learning_rate": 1.0185401270582133e-07,
-      "loss": 2.4592,
-      "step": 60760
-    },
-    {
-      "epoch": 15.708877115906448,
-      "learning_rate": 1.01821599896279e-07,
-      "loss": 2.4356,
-      "step": 60780
-    },
-    {
-      "epoch": 15.714045742343972,
-      "learning_rate": 1.0178918708673667e-07,
-      "loss": 2.3921,
-      "step": 60800
-    },
-    {
-      "epoch": 15.719214368781497,
-      "learning_rate": 1.0175677427719434e-07,
-      "loss": 2.4722,
-      "step": 60820
-    },
-    {
-      "epoch": 15.72438299521902,
-      "learning_rate": 1.01724361467652e-07,
-      "loss": 2.3778,
-      "step": 60840
-    },
-    {
-      "epoch": 15.729551621656544,
-      "learning_rate": 1.0169194865810968e-07,
-      "loss": 2.4242,
-      "step": 60860
-    },
-    {
-      "epoch": 15.734720248094069,
-      "learning_rate": 1.0165953584856734e-07,
-      "loss": 2.4437,
-      "step": 60880
-    },
-    {
-      "epoch": 15.739888874531593,
-      "learning_rate": 1.0162712303902502e-07,
-      "loss": 2.4182,
-      "step": 60900
-    },
-    {
-      "epoch": 15.745057500969118,
-      "learning_rate": 1.0159471022948268e-07,
-      "loss": 2.3767,
-      "step": 60920
-    },
-    {
-      "epoch": 15.750226127406641,
-      "learning_rate": 1.0156229741994035e-07,
-      "loss": 2.4421,
-      "step": 60940
-    },
-    {
-      "epoch": 15.755394753844167,
-      "learning_rate": 1.0152988461039801e-07,
-      "loss": 2.4444,
-      "step": 60960
-    },
-    {
-      "epoch": 15.76056338028169,
-      "learning_rate": 1.0149747180085569e-07,
-      "loss": 2.4428,
-      "step": 60980
-    },
-    {
-      "epoch": 15.765732006719215,
-      "learning_rate": 1.0146505899131335e-07,
-      "loss": 2.4303,
-      "step": 61000
-    },
-    {
-      "epoch": 15.770900633156739,
-      "learning_rate": 1.0143264618177103e-07,
-      "loss": 2.385,
-      "step": 61020
-    },
-    {
-      "epoch": 15.776069259594262,
-      "learning_rate": 1.0140023337222869e-07,
-      "loss": 2.4281,
-      "step": 61040
-    },
-    {
-      "epoch": 15.781237886031787,
-      "learning_rate": 1.0136782056268636e-07,
-      "loss": 2.4551,
-      "step": 61060
-    },
-    {
-      "epoch": 15.786406512469311,
-      "learning_rate": 1.0133540775314404e-07,
-      "loss": 2.4592,
-      "step": 61080
-    },
-    {
-      "epoch": 15.791575138906836,
-      "learning_rate": 1.013029949436017e-07,
-      "loss": 2.4028,
-      "step": 61100
-    },
-    {
-      "epoch": 15.79674376534436,
-      "learning_rate": 1.0127058213405936e-07,
-      "loss": 2.466,
-      "step": 61120
-    },
-    {
-      "epoch": 15.801912391781883,
-      "learning_rate": 1.0123816932451704e-07,
-      "loss": 2.4135,
-      "step": 61140
-    },
-    {
-      "epoch": 15.807081018219408,
-      "learning_rate": 1.0120575651497471e-07,
-      "loss": 2.3495,
-      "step": 61160
-    },
-    {
-      "epoch": 15.812249644656932,
-      "learning_rate": 1.0117334370543237e-07,
-      "loss": 2.435,
-      "step": 61180
-    },
-    {
-      "epoch": 15.817418271094457,
-      "learning_rate": 1.0114093089589004e-07,
-      "loss": 2.4339,
-      "step": 61200
-    },
-    {
-      "epoch": 15.82258689753198,
-      "learning_rate": 1.0110851808634771e-07,
-      "loss": 2.4557,
-      "step": 61220
-    },
-    {
-      "epoch": 15.827755523969506,
-      "learning_rate": 1.0107610527680539e-07,
-      "loss": 2.4107,
-      "step": 61240
-    },
-    {
-      "epoch": 15.83292415040703,
-      "learning_rate": 1.0104369246726305e-07,
-      "loss": 2.424,
-      "step": 61260
-    },
-    {
-      "epoch": 15.838092776844553,
-      "learning_rate": 1.0101127965772071e-07,
-      "loss": 2.3895,
-      "step": 61280
-    },
-    {
-      "epoch": 15.843261403282078,
-      "learning_rate": 1.0097886684817838e-07,
-      "loss": 2.3533,
-      "step": 61300
-    },
-    {
-      "epoch": 15.848430029719601,
-      "learning_rate": 1.0094645403863606e-07,
-      "loss": 2.4579,
-      "step": 61320
-    },
-    {
-      "epoch": 15.853598656157127,
-      "learning_rate": 1.0091404122909372e-07,
-      "loss": 2.4648,
-      "step": 61340
-    },
-    {
-      "epoch": 15.85876728259465,
-      "learning_rate": 1.0088162841955138e-07,
-      "loss": 2.4004,
-      "step": 61360
-    },
-    {
-      "epoch": 15.863935909032175,
-      "learning_rate": 1.0084921561000907e-07,
-      "loss": 2.4231,
-      "step": 61380
-    },
-    {
-      "epoch": 15.869104535469699,
-      "learning_rate": 1.0081680280046673e-07,
-      "loss": 2.4427,
-      "step": 61400
-    },
-    {
-      "epoch": 15.874273161907222,
-      "learning_rate": 1.007843899909244e-07,
-      "loss": 2.4253,
-      "step": 61420
-    },
-    {
-      "epoch": 15.879441788344748,
-      "learning_rate": 1.0075197718138206e-07,
-      "loss": 2.3745,
-      "step": 61440
-    },
-    {
-      "epoch": 15.884610414782271,
-      "learning_rate": 1.0071956437183975e-07,
-      "loss": 2.4168,
-      "step": 61460
-    },
-    {
-      "epoch": 15.889779041219796,
-      "learning_rate": 1.0068715156229741e-07,
-      "loss": 2.4033,
-      "step": 61480
-    },
-    {
-      "epoch": 15.89494766765732,
-      "learning_rate": 1.0065473875275507e-07,
-      "loss": 2.3622,
-      "step": 61500
-    },
-    {
-      "epoch": 15.900116294094845,
-      "learning_rate": 1.0062232594321273e-07,
-      "loss": 2.4244,
-      "step": 61520
-    },
-    {
-      "epoch": 15.905284920532369,
-      "learning_rate": 1.0058991313367042e-07,
-      "loss": 2.3934,
-      "step": 61540
-    },
-    {
-      "epoch": 15.910453546969892,
-      "learning_rate": 1.0055750032412808e-07,
-      "loss": 2.4082,
-      "step": 61560
-    },
-    {
-      "epoch": 15.915622173407417,
-      "learning_rate": 1.0052508751458574e-07,
-      "loss": 2.4445,
-      "step": 61580
-    },
-    {
-      "epoch": 15.92079079984494,
-      "learning_rate": 1.0049267470504342e-07,
-      "loss": 2.4324,
-      "step": 61600
-    },
-    {
-      "epoch": 15.925959426282466,
-      "learning_rate": 1.004602618955011e-07,
-      "loss": 2.3891,
-      "step": 61620
-    },
-    {
-      "epoch": 15.93112805271999,
-      "learning_rate": 1.0042784908595876e-07,
-      "loss": 2.4304,
-      "step": 61640
-    },
-    {
-      "epoch": 15.936296679157515,
-      "learning_rate": 1.0039543627641642e-07,
-      "loss": 2.4452,
-      "step": 61660
-    },
-    {
-      "epoch": 15.941465305595038,
-      "learning_rate": 1.003630234668741e-07,
-      "loss": 2.4355,
-      "step": 61680
-    },
-    {
-      "epoch": 15.946633932032562,
-      "learning_rate": 1.0033061065733177e-07,
-      "loss": 2.3643,
-      "step": 61700
-    },
-    {
-      "epoch": 15.951802558470087,
-      "learning_rate": 1.0029819784778943e-07,
-      "loss": 2.3778,
-      "step": 61720
-    },
-    {
-      "epoch": 15.95697118490761,
-      "learning_rate": 1.0026578503824709e-07,
-      "loss": 2.4225,
-      "step": 61740
-    },
-    {
-      "epoch": 15.962139811345136,
-      "learning_rate": 1.0023337222870478e-07,
-      "loss": 2.374,
-      "step": 61760
-    },
-    {
-      "epoch": 15.967308437782659,
-      "learning_rate": 1.0020095941916244e-07,
-      "loss": 2.4335,
-      "step": 61780
-    },
-    {
-      "epoch": 15.972477064220184,
-      "learning_rate": 1.001685466096201e-07,
-      "loss": 2.4207,
-      "step": 61800
-    },
-    {
-      "epoch": 15.977645690657708,
-      "learning_rate": 1.0013613380007777e-07,
-      "loss": 2.3516,
-      "step": 61820
-    },
-    {
-      "epoch": 15.982814317095231,
-      "learning_rate": 1.0010372099053545e-07,
-      "loss": 2.4139,
-      "step": 61840
-    },
-    {
-      "epoch": 15.987982943532757,
-      "learning_rate": 1.0007130818099312e-07,
-      "loss": 2.4381,
-      "step": 61860
-    },
-    {
-      "epoch": 15.99315156997028,
-      "learning_rate": 1.0003889537145078e-07,
-      "loss": 2.4053,
-      "step": 61880
-    },
-    {
-      "epoch": 15.998320196407805,
-      "learning_rate": 1.0000648256190844e-07,
-      "loss": 2.439,
-      "step": 61900
     }
   ],
   "logging_steps": 20,
@@ -18747,12 +12905,12 @@
         "should_evaluate": false,
         "should_log": false,
         "should_save": true,
-        "should_training_stop": true
+        "should_training_stop": false
       },
       "attributes": {}
     }
   },
-  "total_flos": 4.0635244159460966e+17,
+  "total_flos": 2.793935715793797e+17,
   "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null