diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,7 +1,7 @@
 {
-  "best_global_step": 6100,
-  "best_metric": 98.58235971529302,
-  "best_model_checkpoint": "./output/string-repetition-tiny/checkpoint-3500",
+  "best_global_step": 5000,
+  "best_metric": 98.56639547519084,
+  "best_model_checkpoint": "./output/string-repetition-tiny/checkpoint-5000",
   "epoch": 2.163565556036348,
   "eval_steps": 100,
   "global_step": 10000,
@@ -11,11959 +11,11921 @@
   "log_history": [
     {
       "epoch": 0,
-      "eval_loss": 4.988397121429443,
-      "eval_runtime": 2.0878,
-      "eval_samples_per_second": 15.327,
-      "eval_steps_per_second": 0.479,
+      "eval_loss": 4.9941487312316895,
+      "eval_runtime": 2.0943,
+      "eval_samples_per_second": 15.279,
+      "eval_steps_per_second": 0.477,
       "num_input_tokens_seen": 0,
       "step": 0
     },
     {
       "epoch": 0,
-      "eval_byte_accuracy": 0.29080118694362017,
+      "eval_byte_accuracy": 0.29154302670623145,
       "eval_chrf": 1.9898598915486052,
       "eval_sacrebleu": 0.027097937214200497,
       "eval_word_accuracy": 0.0,
       "num_input_tokens_seen": 0,
-      "perplexity": 146.70109090640656,
+      "perplexity": 147.5472895081926,
       "step": 0
     },
     {
       "epoch": 0.0021635655560363477,
       "grad_norm": 18.625,
       "learning_rate": 1.0799999999999998e-05,
-      "loss": 4.9798,
+      "loss": 4.9829,
       "num_input_tokens_seen": 655360,
       "step": 10,
-      "train_runtime": 7.4881,
-      "train_tokens_per_second": 87520.259
+      "train_runtime": 7.5789,
+      "train_tokens_per_second": 86472.05
     },
     {
       "epoch": 0.004327131112072695,
-      "grad_norm": 16.5,
+      "grad_norm": 16.625,
       "learning_rate": 2.28e-05,
-      "loss": 4.7276,
+      "loss": 4.7329,
       "num_input_tokens_seen": 1310720,
       "step": 20,
-      "train_runtime": 11.2856,
-      "train_tokens_per_second": 116140.595
+      "train_runtime": 11.379,
+      "train_tokens_per_second": 115187.192
     },
     {
       "epoch": 0.006490696668109044,
-      "grad_norm": 16.375,
+      "grad_norm": 16.25,
       "learning_rate": 3.48e-05,
-      "loss": 4.0073,
+      "loss": 4.0165,
       "num_input_tokens_seen": 1966080,
       "step": 30,
-      "train_runtime": 15.0717,
-      "train_tokens_per_second": 130448.408
+      "train_runtime": 15.1734,
+      "train_tokens_per_second": 129573.818
     },
     {
       "epoch": 0.00865426222414539,
-      "grad_norm": 10.8125,
+      "grad_norm": 20.375,
       "learning_rate": 4.68e-05,
-      "loss": 2.5584,
+      "loss": 2.5654,
       "num_input_tokens_seen": 2621440,
       "step": 40,
-      "train_runtime": 18.8656,
-      "train_tokens_per_second": 138953.652
+      "train_runtime": 18.9678,
+      "train_tokens_per_second": 138204.687
     },
     {
       "epoch": 0.010817827780181739,
-      "grad_norm": 13.0625,
+      "grad_norm": 3.15625,
       "learning_rate": 5.88e-05,
-      "loss": 1.5522,
+      "loss": 1.5888,
       "num_input_tokens_seen": 3276800,
       "step": 50,
-      "train_runtime": 22.6701,
-      "train_tokens_per_second": 144542.984
+      "train_runtime": 22.7693,
+      "train_tokens_per_second": 143912.922
     },
     {
       "epoch": 0.012981393336218087,
-      "grad_norm": 10.0,
+      "grad_norm": 3.671875,
       "learning_rate": 7.079999999999999e-05,
-      "loss": 1.2928,
+      "loss": 1.2862,
       "num_input_tokens_seen": 3932160,
       "step": 60,
-      "train_runtime": 26.4611,
-      "train_tokens_per_second": 148601.312
+      "train_runtime": 28.5663,
+      "train_tokens_per_second": 137650.179
     },
     {
       "epoch": 0.015144958892254435,
-      "grad_norm": 11.5625,
+      "grad_norm": 3.6875,
       "learning_rate": 8.28e-05,
-      "loss": 1.1674,
+      "loss": 1.1568,
       "num_input_tokens_seen": 4587520,
       "step": 70,
-      "train_runtime": 30.2488,
-      "train_tokens_per_second": 151659.614
+      "train_runtime": 32.3629,
+      "train_tokens_per_second": 141752.538
     },
     {
       "epoch": 0.01730852444829078,
-      "grad_norm": 9.375,
+      "grad_norm": 2.375,
       "learning_rate": 9.479999999999999e-05,
-      "loss": 1.0741,
+      "loss": 1.0683,
       "num_input_tokens_seen": 5242880,
       "step": 80,
-      "train_runtime": 34.0502,
-      "train_tokens_per_second": 153975.057
+      "train_runtime": 36.1566,
+      "train_tokens_per_second": 145004.585
     },
     {
       "epoch": 0.01947209000432713,
-      "grad_norm": 26.375,
+      "grad_norm": 8.875,
       "learning_rate": 0.00010679999999999998,
-      "loss": 1.0159,
+      "loss": 1.0143,
       "num_input_tokens_seen": 5898240,
       "step": 90,
-      "train_runtime": 37.8468,
-      "train_tokens_per_second": 155845.247
+      "train_runtime": 39.9526,
+      "train_tokens_per_second": 147630.985
     },
     {
       "epoch": 0.021635655560363478,
-      "grad_norm": 3.5625,
+      "grad_norm": 3.484375,
       "learning_rate": 0.0001188,
-      "loss": 0.9462,
+      "loss": 0.9351,
       "num_input_tokens_seen": 6553600,
       "step": 100,
-      "train_runtime": 41.6439,
-      "train_tokens_per_second": 157372.435
+      "train_runtime": 43.7483,
+      "train_tokens_per_second": 149802.332
     },
     {
       "epoch": 0.021635655560363478,
-      "eval_loss": 0.9727591276168823,
-      "eval_runtime": 0.0828,
-      "eval_samples_per_second": 386.585,
-      "eval_steps_per_second": 12.081,
+      "eval_loss": 0.9468593597412109,
+      "eval_runtime": 1.2791,
+      "eval_samples_per_second": 25.018,
+      "eval_steps_per_second": 0.782,
       "num_input_tokens_seen": 6553600,
       "step": 100
     },
     {
       "epoch": 0.021635655560363478,
-      "eval_byte_accuracy": 0.7091988130563798,
-      "eval_chrf": 0.0,
-      "eval_sacrebleu": 0.0,
-      "eval_word_accuracy": 0.5659340659340659,
+      "eval_byte_accuracy": 0.7288575667655787,
+      "eval_chrf": 4.12565998742775,
+      "eval_sacrebleu": 0.0736970000889057,
+      "eval_word_accuracy": 0.6057692307692307,
       "num_input_tokens_seen": 6553600,
-      "perplexity": 2.6452329350572574,
+      "perplexity": 2.5776016143530303,
       "step": 100
     },
     {
       "epoch": 0.023799221116399826,
-      "grad_norm": 2.28125,
+      "grad_norm": 3.046875,
       "learning_rate": 0.00013079999999999998,
-      "loss": 0.9063,
+      "loss": 0.8818,
       "num_input_tokens_seen": 7204864,
       "step": 110,
-      "train_runtime": 45.5286,
-      "train_tokens_per_second": 158249.227
+      "train_runtime": 50.6655,
+      "train_tokens_per_second": 142204.571
     },
     {
       "epoch": 0.025962786672436174,
-      "grad_norm": 1.6640625,
+      "grad_norm": 1.1015625,
       "learning_rate": 0.00014279999999999997,
-      "loss": 0.8488,
+      "loss": 0.8464,
       "num_input_tokens_seen": 7860224,
       "step": 120,
-      "train_runtime": 49.3213,
-      "train_tokens_per_second": 159367.885
+      "train_runtime": 54.4574,
+      "train_tokens_per_second": 144337.182
     },
     {
       "epoch": 0.028126352228472522,
-      "grad_norm": 3.40625,
+      "grad_norm": 3.890625,
       "learning_rate": 0.0001548,
-      "loss": 0.8492,
+      "loss": 0.883,
       "num_input_tokens_seen": 8515584,
       "step": 130,
-      "train_runtime": 53.131,
-      "train_tokens_per_second": 160275.187
+      "train_runtime": 58.2578,
+      "train_tokens_per_second": 146170.797
     },
     {
       "epoch": 0.03028991778450887,
-      "grad_norm": 2.90625,
+      "grad_norm": 2.015625,
       "learning_rate": 0.0001668,
-      "loss": 0.8508,
+      "loss": 0.8342,
       "num_input_tokens_seen": 9170944,
       "step": 140,
-      "train_runtime": 56.9152,
-      "train_tokens_per_second": 161133.514
+      "train_runtime": 62.0415,
+      "train_tokens_per_second": 147819.414
     },
     {
       "epoch": 0.032453483340545215,
-      "grad_norm": 1.03125,
+      "grad_norm": 3.171875,
       "learning_rate": 0.00017879999999999998,
-      "loss": 0.8418,
+      "loss": 0.8202,
       "num_input_tokens_seen": 9826304,
       "step": 150,
-      "train_runtime": 60.7146,
-      "train_tokens_per_second": 161844.175
+      "train_runtime": 65.8305,
+      "train_tokens_per_second": 149266.694
     },
     {
       "epoch": 0.03461704889658156,
-      "grad_norm": 9.8125,
+      "grad_norm": 2.390625,
       "learning_rate": 0.00019079999999999998,
-      "loss": 0.8424,
+      "loss": 0.8595,
       "num_input_tokens_seen": 10481664,
       "step": 160,
-      "train_runtime": 64.5125,
-      "train_tokens_per_second": 162474.904
+      "train_runtime": 71.4958,
+      "train_tokens_per_second": 146605.34
     },
     {
       "epoch": 0.03678061445261791,
-      "grad_norm": 2.125,
+      "grad_norm": 2.25,
       "learning_rate": 0.0002028,
-      "loss": 0.8222,
+      "loss": 0.8288,
       "num_input_tokens_seen": 11137024,
       "step": 170,
-      "train_runtime": 68.3125,
-      "train_tokens_per_second": 163030.526
+      "train_runtime": 75.288,
+      "train_tokens_per_second": 147925.655
     },
     {
       "epoch": 0.03894418000865426,
-      "grad_norm": 2.71875,
+      "grad_norm": 5.78125,
       "learning_rate": 0.00021479999999999996,
-      "loss": 0.7886,
+      "loss": 0.8148,
       "num_input_tokens_seen": 11792384,
       "step": 180,
-      "train_runtime": 72.1123,
-      "train_tokens_per_second": 163528.046
+      "train_runtime": 79.0884,
+      "train_tokens_per_second": 149103.819
     },
     {
       "epoch": 0.04110774556469061,
-      "grad_norm": 0.76171875,
+      "grad_norm": 0.97265625,
       "learning_rate": 0.00022679999999999998,
-      "loss": 0.7946,
+      "loss": 0.7932,
       "num_input_tokens_seen": 12447744,
       "step": 190,
-      "train_runtime": 75.9133,
-      "train_tokens_per_second": 163973.238
+      "train_runtime": 82.8997,
+      "train_tokens_per_second": 150154.262
     },
     {
       "epoch": 0.043271311120726956,
-      "grad_norm": 1.28125,
+      "grad_norm": 1.40625,
       "learning_rate": 0.0002388,
-      "loss": 0.7727,
+      "loss": 0.7827,
       "num_input_tokens_seen": 13103104,
       "step": 200,
-      "train_runtime": 79.7148,
-      "train_tokens_per_second": 164374.865
+      "train_runtime": 86.6986,
+      "train_tokens_per_second": 151133.933
     },
     {
       "epoch": 0.043271311120726956,
-      "eval_loss": 0.8417919874191284,
-      "eval_runtime": 1.5809,
-      "eval_samples_per_second": 20.242,
-      "eval_steps_per_second": 0.633,
+      "eval_loss": 0.8652347326278687,
+      "eval_runtime": 1.5369,
+      "eval_samples_per_second": 20.821,
+      "eval_steps_per_second": 0.651,
       "num_input_tokens_seen": 13103104,
       "step": 200
     },
     {
       "epoch": 0.043271311120726956,
-      "eval_byte_accuracy": 0.7440652818991098,
-      "eval_chrf": 6.6820302805599665,
-      "eval_sacrebleu": 0.4581175341354607,
-      "eval_word_accuracy": 0.6263736263736264,
+      "eval_byte_accuracy": 0.7399851632047477,
+      "eval_chrf": 5.853085649666837,
+      "eval_sacrebleu": 0.21629231639858632,
+      "eval_word_accuracy": 0.6195054945054945,
       "num_input_tokens_seen": 13103104,
-      "perplexity": 2.320521598664858,
+      "perplexity": 2.3755636428327103,
       "step": 200
     },
     {
       "epoch": 0.045434876676763304,
-      "grad_norm": 3.46875,
+      "grad_norm": 1.8046875,
       "learning_rate": 0.00025079999999999997,
-      "loss": 0.7692,
+      "loss": 0.7815,
       "num_input_tokens_seen": 13758464,
       "step": 210,
-      "train_runtime": 85.1122,
-      "train_tokens_per_second": 161650.863
+      "train_runtime": 93.9602,
+      "train_tokens_per_second": 146428.646
     },
     {
       "epoch": 0.04759844223279965,
-      "grad_norm": 0.90234375,
+      "grad_norm": 0.984375,
       "learning_rate": 0.0002628,
-      "loss": 0.7411,
+      "loss": 0.7407,
       "num_input_tokens_seen": 14413824,
       "step": 220,
-      "train_runtime": 88.9079,
-      "train_tokens_per_second": 162120.908
+      "train_runtime": 97.7619,
+      "train_tokens_per_second": 147438.046
     },
     {
       "epoch": 0.049762007788836,
-      "grad_norm": 0.7109375,
+      "grad_norm": 1.703125,
       "learning_rate": 0.0002748,
-      "loss": 0.7572,
+      "loss": 0.7636,
       "num_input_tokens_seen": 15069184,
       "step": 230,
-      "train_runtime": 92.7125,
-      "train_tokens_per_second": 162536.706
+      "train_runtime": 101.5612,
+      "train_tokens_per_second": 148375.388
     },
     {
       "epoch": 0.05192557334487235,
-      "grad_norm": 1.125,
+      "grad_norm": 0.73828125,
       "learning_rate": 0.0002868,
-      "loss": 0.7636,
+      "loss": 0.773,
       "num_input_tokens_seen": 15724544,
       "step": 240,
-      "train_runtime": 96.5159,
-      "train_tokens_per_second": 162921.839
+      "train_runtime": 105.3581,
+      "train_tokens_per_second": 149248.582
     },
     {
       "epoch": 0.0540891389009087,
-      "grad_norm": 4.34375,
+      "grad_norm": 1.4453125,
       "learning_rate": 0.0002988,
-      "loss": 0.7582,
+      "loss": 0.7638,
       "num_input_tokens_seen": 16379904,
       "step": 250,
-      "train_runtime": 100.3175,
-      "train_tokens_per_second": 163280.687
+      "train_runtime": 109.1647,
+      "train_tokens_per_second": 150047.624
     },
     {
       "epoch": 0.056252704456945045,
-      "grad_norm": 1.203125,
+      "grad_norm": 0.89453125,
       "learning_rate": 0.00031079999999999997,
-      "loss": 0.7522,
+      "loss": 0.7616,
       "num_input_tokens_seen": 17035264,
       "step": 260,
-      "train_runtime": 104.1187,
-      "train_tokens_per_second": 163613.953
+      "train_runtime": 114.9034,
+      "train_tokens_per_second": 148257.262
     },
     {
       "epoch": 0.05841627001298139,
-      "grad_norm": 1.953125,
+      "grad_norm": 0.71484375,
       "learning_rate": 0.0003228,
-      "loss": 0.749,
+      "loss": 0.761,
       "num_input_tokens_seen": 17690624,
       "step": 270,
-      "train_runtime": 107.9263,
-      "train_tokens_per_second": 163913.881
+      "train_runtime": 118.6967,
+      "train_tokens_per_second": 149040.514
     },
     {
       "epoch": 0.06057983556901774,
-      "grad_norm": 1.5078125,
+      "grad_norm": 1.3828125,
       "learning_rate": 0.0003348,
-      "loss": 0.7412,
+      "loss": 0.7694,
       "num_input_tokens_seen": 18345984,
       "step": 280,
-      "train_runtime": 111.7314,
-      "train_tokens_per_second": 164197.152
+      "train_runtime": 122.4954,
+      "train_tokens_per_second": 149768.786
     },
     {
       "epoch": 0.06274340112505408,
-      "grad_norm": 2.421875,
+      "grad_norm": 0.8828125,
       "learning_rate": 0.0003467999999999999,
-      "loss": 0.6982,
+      "loss": 0.754,
       "num_input_tokens_seen": 19001344,
       "step": 290,
-      "train_runtime": 115.5403,
-      "train_tokens_per_second": 164456.388
+      "train_runtime": 126.2997,
+      "train_tokens_per_second": 150446.434
     },
     {
       "epoch": 0.06490696668109043,
-      "grad_norm": 0.9296875,
+      "grad_norm": 0.625,
       "learning_rate": 0.00035879999999999994,
-      "loss": 0.6894,
+      "loss": 0.7533,
       "num_input_tokens_seen": 19656704,
       "step": 300,
-      "train_runtime": 119.3511,
-      "train_tokens_per_second": 164696.484
+      "train_runtime": 130.1045,
+      "train_tokens_per_second": 151083.958
     },
     {
       "epoch": 0.06490696668109043,
-      "eval_loss": 0.7334468364715576,
-      "eval_runtime": 1.9947,
-      "eval_samples_per_second": 16.042,
-      "eval_steps_per_second": 0.501,
+      "eval_loss": 0.8242582082748413,
+      "eval_runtime": 1.7792,
+      "eval_samples_per_second": 17.985,
+      "eval_steps_per_second": 0.562,
       "num_input_tokens_seen": 19656704,
       "step": 300
     },
     {
       "epoch": 0.06490696668109043,
-      "eval_byte_accuracy": 0.771513353115727,
-      "eval_chrf": 12.160441031342272,
-      "eval_sacrebleu": 0.710868857582003,
-      "eval_word_accuracy": 0.6510989010989011,
+      "eval_byte_accuracy": 0.7436943620178041,
+      "eval_chrf": 7.359687391784456,
+      "eval_sacrebleu": 0.18818523987470104,
+      "eval_word_accuracy": 0.6195054945054945,
       "num_input_tokens_seen": 19656704,
-      "perplexity": 2.0822454120550398,
+      "perplexity": 2.280188712712687,
       "step": 300
     },
     {
       "epoch": 0.06707053223712678,
-      "grad_norm": 2.640625,
+      "grad_norm": 0.875,
       "learning_rate": 0.00037079999999999996,
-      "loss": 0.6499,
+      "loss": 0.743,
       "num_input_tokens_seen": 20312064,
       "step": 310,
-      "train_runtime": 125.165,
-      "train_tokens_per_second": 162282.302
+      "train_runtime": 137.6374,
+      "train_tokens_per_second": 147576.607
     },
     {
       "epoch": 0.06923409779316313,
-      "grad_norm": 3.328125,
+      "grad_norm": 0.73828125,
       "learning_rate": 0.0003828,
-      "loss": 0.6507,
+      "loss": 0.7551,
       "num_input_tokens_seen": 20967424,
       "step": 320,
-      "train_runtime": 128.9715,
-      "train_tokens_per_second": 162574.135
+      "train_runtime": 141.444,
+      "train_tokens_per_second": 148238.373
     },
     {
       "epoch": 0.07139766334919947,
-      "grad_norm": 3.890625,
+      "grad_norm": 1.203125,
       "learning_rate": 0.0003948,
-      "loss": 0.6283,
+      "loss": 0.7512,
       "num_input_tokens_seen": 21622784,
       "step": 330,
-      "train_runtime": 132.7747,
-      "train_tokens_per_second": 162853.208
+      "train_runtime": 145.2419,
+      "train_tokens_per_second": 148874.327
     },
     {
       "epoch": 0.07356122890523582,
-      "grad_norm": 2.234375,
+      "grad_norm": 1.2109375,
       "learning_rate": 0.00040679999999999997,
-      "loss": 0.5841,
+      "loss": 0.7349,
       "num_input_tokens_seen": 22278144,
       "step": 340,
-      "train_runtime": 136.5774,
-      "train_tokens_per_second": 163117.394
+      "train_runtime": 149.0492,
+      "train_tokens_per_second": 149468.345
     },
     {
       "epoch": 0.07572479446127217,
-      "grad_norm": 4.53125,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.00041879999999999993,
-      "loss": 0.5767,
+      "loss": 0.7356,
       "num_input_tokens_seen": 22933504,
       "step": 350,
-      "train_runtime": 140.3834,
-      "train_tokens_per_second": 163363.348
+      "train_runtime": 152.8623,
+      "train_tokens_per_second": 150027.169
     },
     {
       "epoch": 0.07788836001730852,
-      "grad_norm": 1.6171875,
+      "grad_norm": 0.96484375,
       "learning_rate": 0.00043079999999999995,
-      "loss": 0.5292,
+      "loss": 0.7216,
       "num_input_tokens_seen": 23588864,
       "step": 360,
-      "train_runtime": 144.187,
-      "train_tokens_per_second": 163599.098
+      "train_runtime": 158.5961,
+      "train_tokens_per_second": 148735.445
     },
     {
       "epoch": 0.08005192557334487,
-      "grad_norm": 1.7890625,
+      "grad_norm": 0.9453125,
       "learning_rate": 0.0004428,
-      "loss": 0.5157,
+      "loss": 0.7263,
       "num_input_tokens_seen": 24244224,
       "step": 370,
-      "train_runtime": 147.9913,
-      "train_tokens_per_second": 163821.98
+      "train_runtime": 162.4005,
+      "train_tokens_per_second": 149286.589
     },
     {
       "epoch": 0.08221549112938122,
-      "grad_norm": 2.4375,
+      "grad_norm": 0.78125,
       "learning_rate": 0.00045479999999999994,
-      "loss": 0.4898,
+      "loss": 0.7355,
       "num_input_tokens_seen": 24899584,
       "step": 380,
-      "train_runtime": 151.7906,
-      "train_tokens_per_second": 164039.076
+      "train_runtime": 166.2066,
+      "train_tokens_per_second": 149811.031
     },
     {
       "epoch": 0.08437905668541756,
-      "grad_norm": 2.78125,
+      "grad_norm": 0.6328125,
       "learning_rate": 0.00046679999999999996,
-      "loss": 0.4582,
+      "loss": 0.7117,
       "num_input_tokens_seen": 25554944,
       "step": 390,
-      "train_runtime": 155.5937,
-      "train_tokens_per_second": 164241.491
+      "train_runtime": 170.012,
+      "train_tokens_per_second": 150312.596
     },
     {
       "epoch": 0.08654262224145391,
-      "grad_norm": 3.453125,
+      "grad_norm": 0.625,
       "learning_rate": 0.0004788,
-      "loss": 0.431,
+      "loss": 0.7047,
       "num_input_tokens_seen": 26210304,
       "step": 400,
-      "train_runtime": 159.3994,
-      "train_tokens_per_second": 164431.67
+      "train_runtime": 173.8217,
+      "train_tokens_per_second": 150788.449
     },
     {
       "epoch": 0.08654262224145391,
-      "eval_loss": 0.4691297113895416,
-      "eval_runtime": 1.7697,
-      "eval_samples_per_second": 18.082,
-      "eval_steps_per_second": 0.565,
+      "eval_loss": 0.7835195064544678,
+      "eval_runtime": 1.666,
+      "eval_samples_per_second": 19.208,
+      "eval_steps_per_second": 0.6,
       "num_input_tokens_seen": 26210304,
       "step": 400
     },
     {
       "epoch": 0.08654262224145391,
-      "eval_byte_accuracy": 0.8479228486646885,
-      "eval_chrf": 25.30382858102524,
-      "eval_sacrebleu": 6.954022634991616,
-      "eval_word_accuracy": 0.7348901098901099,
+      "eval_byte_accuracy": 0.766320474777448,
+      "eval_chrf": 7.609548718464071,
+      "eval_sacrebleu": 0.4190094244340075,
+      "eval_word_accuracy": 0.6456043956043956,
       "num_input_tokens_seen": 26210304,
-      "perplexity": 1.5986023422380802,
+      "perplexity": 2.1891634979487655,
       "step": 400
     },
     {
       "epoch": 0.08870618779749026,
-      "grad_norm": 1.328125,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.0004907999999999999,
-      "loss": 0.4126,
+      "loss": 0.6961,
       "num_input_tokens_seen": 26865664,
       "step": 410,
-      "train_runtime": 164.9942,
-      "train_tokens_per_second": 162827.884
+      "train_runtime": 181.276,
+      "train_tokens_per_second": 148203.132
     },
     {
       "epoch": 0.09086975335352661,
-      "grad_norm": 1.7421875,
+      "grad_norm": 0.80859375,
       "learning_rate": 0.0005028,
-      "loss": 0.4102,
+      "loss": 0.6971,
       "num_input_tokens_seen": 27521024,
       "step": 420,
-      "train_runtime": 168.7954,
-      "train_tokens_per_second": 163043.709
+      "train_runtime": 185.0858,
+      "train_tokens_per_second": 148693.323
     },
     {
       "epoch": 0.09303331890956296,
-      "grad_norm": 2.65625,
+      "grad_norm": 1.546875,
       "learning_rate": 0.0005147999999999999,
-      "loss": 0.3765,
+      "loss": 0.668,
       "num_input_tokens_seen": 28176384,
       "step": 430,
-      "train_runtime": 172.5853,
-      "train_tokens_per_second": 163260.64
+      "train_runtime": 188.8745,
+      "train_tokens_per_second": 149180.456
     },
     {
       "epoch": 0.0951968844655993,
-      "grad_norm": 5.875,
+      "grad_norm": 0.859375,
       "learning_rate": 0.0005267999999999999,
-      "loss": 0.3727,
+      "loss": 0.6698,
       "num_input_tokens_seen": 28831744,
       "step": 440,
-      "train_runtime": 176.393,
-      "train_tokens_per_second": 163451.711
+      "train_runtime": 192.6748,
+      "train_tokens_per_second": 149639.425
     },
     {
       "epoch": 0.09736045002163565,
-      "grad_norm": 3.4375,
+      "grad_norm": 1.828125,
       "learning_rate": 0.0005388,
-      "loss": 0.3512,
+      "loss": 0.6434,
       "num_input_tokens_seen": 29487104,
       "step": 450,
-      "train_runtime": 180.1886,
-      "train_tokens_per_second": 163645.825
+      "train_runtime": 196.4814,
+      "train_tokens_per_second": 150075.792
     },
     {
       "epoch": 0.099524015577672,
-      "grad_norm": 2.515625,
+      "grad_norm": 2.3125,
       "learning_rate": 0.0005507999999999999,
-      "loss": 0.3391,
+      "loss": 0.593,
       "num_input_tokens_seen": 30142464,
       "step": 460,
-      "train_runtime": 184.0009,
-      "train_tokens_per_second": 163816.961
+      "train_runtime": 202.2445,
+      "train_tokens_per_second": 149039.706
     },
     {
       "epoch": 0.10168758113370835,
-      "grad_norm": 1.921875,
+      "grad_norm": 2.703125,
       "learning_rate": 0.0005627999999999999,
-      "loss": 0.3332,
+      "loss": 0.5579,
       "num_input_tokens_seen": 30797824,
       "step": 470,
-      "train_runtime": 187.7962,
-      "train_tokens_per_second": 163995.993
+      "train_runtime": 206.0426,
+      "train_tokens_per_second": 149473.092
     },
     {
       "epoch": 0.1038511466897447,
-      "grad_norm": 4.65625,
+      "grad_norm": 1.5390625,
       "learning_rate": 0.0005747999999999999,
-      "loss": 0.3333,
+      "loss": 0.5097,
       "num_input_tokens_seen": 31453184,
       "step": 480,
-      "train_runtime": 191.5972,
-      "train_tokens_per_second": 164163.059
+      "train_runtime": 209.8528,
+      "train_tokens_per_second": 149882.114
     },
     {
       "epoch": 0.10601471224578105,
-      "grad_norm": 2.234375,
+      "grad_norm": 2.25,
       "learning_rate": 0.0005868,
-      "loss": 0.311,
+      "loss": 0.4667,
       "num_input_tokens_seen": 32108544,
       "step": 490,
-      "train_runtime": 195.3941,
-      "train_tokens_per_second": 164327.106
+      "train_runtime": 213.6418,
+      "train_tokens_per_second": 150291.498
     },
     {
       "epoch": 0.1081782778018174,
-      "grad_norm": 3.078125,
+      "grad_norm": 2.25,
       "learning_rate": 0.0005987999999999999,
-      "loss": 0.2887,
+      "loss": 0.4021,
       "num_input_tokens_seen": 32763904,
       "step": 500,
-      "train_runtime": 199.2107,
-      "train_tokens_per_second": 164468.582
-    },
-    {
-      "epoch": 0.1081782778018174,
-      "eval_loss": 0.31848961114883423,
-      "eval_runtime": 1.8473,
-      "eval_samples_per_second": 17.322,
-      "eval_steps_per_second": 0.541,
-      "num_input_tokens_seen": 32763904,
-      "step": 500
-    },
-    {
-      "epoch": 0.1081782778018174,
-      "eval_byte_accuracy": 0.9031899109792285,
-      "eval_chrf": 43.27469201356104,
-      "eval_sacrebleu": 23.601674532788458,
-      "eval_word_accuracy": 0.8255494505494505,
-      "num_input_tokens_seen": 32763904,
-      "perplexity": 1.375049335926744,
-      "step": 500
+      "train_runtime": 217.4483,
+      "train_tokens_per_second": 150674.436
     },
     {
       "epoch": 0.1081782778018174,
-      "eval_loss": 0.41804632544517517,
-      "eval_runtime": 2.9141,
-      "eval_samples_per_second": 10.981,
-      "eval_steps_per_second": 0.343,
+      "eval_loss": 0.3849869966506958,
+      "eval_runtime": 2.1359,
+      "eval_samples_per_second": 14.982,
+      "eval_steps_per_second": 0.468,
       "num_input_tokens_seen": 32763904,
       "step": 500
     },
     {
       "epoch": 0.1081782778018174,
-      "eval_byte_accuracy": 0.8816765578635015,
-      "eval_chrf": 33.679806731632475,
-      "eval_sacrebleu": 17.00545584726125,
-      "eval_word_accuracy": 0.7912087912087912,
+      "eval_byte_accuracy": 0.8775964391691394,
+      "eval_chrf": 33.58088554992905,
+      "eval_sacrebleu": 13.160711204980268,
+      "eval_word_accuracy": 0.7884615384615384,
       "num_input_tokens_seen": 32763904,
-      "perplexity": 1.51899104070853,
+      "perplexity": 1.4695952116570261,
       "step": 500
     },
     {
       "epoch": 0.11034184335785374,
-      "grad_norm": 3.140625,
+      "grad_norm": 1.65625,
       "learning_rate": 0.0005994315789473684,
-      "loss": 0.358,
+      "loss": 0.3626,
       "num_input_tokens_seen": 33419264,
       "step": 510,
-      "train_runtime": 8.3489,
-      "train_tokens_per_second": 4002858.123
+      "train_runtime": 225.4187,
+      "train_tokens_per_second": 148254.162
     },
     {
       "epoch": 0.11250540891389009,
-      "grad_norm": 1.5859375,
+      "grad_norm": 2.640625,
       "learning_rate": 0.0005987999999999999,
-      "loss": 0.3043,
+      "loss": 0.336,
       "num_input_tokens_seen": 34074624,
       "step": 520,
-      "train_runtime": 12.1656,
-      "train_tokens_per_second": 2800901.822
+      "train_runtime": 229.2259,
+      "train_tokens_per_second": 148650.834
     },
     {
       "epoch": 0.11466897446992644,
-      "grad_norm": 3.078125,
+      "grad_norm": 1.7890625,
       "learning_rate": 0.0005981684210526315,
-      "loss": 0.3086,
+      "loss": 0.3142,
       "num_input_tokens_seen": 34729984,
       "step": 530,
-      "train_runtime": 15.9521,
-      "train_tokens_per_second": 2177143.961
+      "train_runtime": 233.0268,
+      "train_tokens_per_second": 149038.583
     },
     {
       "epoch": 0.11683254002596279,
-      "grad_norm": 3.78125,
+      "grad_norm": 1.9296875,
       "learning_rate": 0.0005975368421052631,
-      "loss": 0.2929,
+      "loss": 0.2976,
       "num_input_tokens_seen": 35385344,
       "step": 540,
-      "train_runtime": 19.7607,
-      "train_tokens_per_second": 1790696.752
+      "train_runtime": 236.8458,
+      "train_tokens_per_second": 149402.428
     },
     {
       "epoch": 0.11899610558199913,
-      "grad_norm": 2.609375,
+      "grad_norm": 2.71875,
       "learning_rate": 0.0005969052631578947,
-      "loss": 0.2783,
+      "loss": 0.2723,
       "num_input_tokens_seen": 36040704,
       "step": 550,
-      "train_runtime": 23.5553,
-      "train_tokens_per_second": 1530045.774
+      "train_runtime": 240.644,
+      "train_tokens_per_second": 149767.721
     },
     {
       "epoch": 0.12115967113803548,
-      "grad_norm": 1.171875,
+      "grad_norm": 2.015625,
       "learning_rate": 0.0005962736842105263,
-      "loss": 0.2455,
+      "loss": 0.2423,
       "num_input_tokens_seen": 36696064,
       "step": 560,
-      "train_runtime": 27.3702,
-      "train_tokens_per_second": 1340728.723
+      "train_runtime": 246.4279,
+      "train_tokens_per_second": 148911.956
     },
     {
       "epoch": 0.12332323669407183,
-      "grad_norm": 2.21875,
+      "grad_norm": 2.0,
       "learning_rate": 0.0005956421052631579,
-      "loss": 0.2591,
+      "loss": 0.2458,
       "num_input_tokens_seen": 37351424,
       "step": 570,
-      "train_runtime": 31.1633,
-      "train_tokens_per_second": 1198571.413
+      "train_runtime": 250.2418,
+      "train_tokens_per_second": 149261.331
     },
     {
       "epoch": 0.12548680225010816,
-      "grad_norm": 2.203125,
+      "grad_norm": 1.7734375,
       "learning_rate": 0.0005950105263157894,
-      "loss": 0.245,
+      "loss": 0.2317,
       "num_input_tokens_seen": 38006784,
       "step": 580,
-      "train_runtime": 34.9733,
-      "train_tokens_per_second": 1086737.214
+      "train_runtime": 254.0523,
+      "train_tokens_per_second": 149602.203
     },
     {
       "epoch": 0.12765036780614453,
-      "grad_norm": 1.453125,
+      "grad_norm": 2.671875,
       "learning_rate": 0.000594378947368421,
-      "loss": 0.2374,
+      "loss": 0.217,
       "num_input_tokens_seen": 38662144,
       "step": 590,
-      "train_runtime": 38.778,
-      "train_tokens_per_second": 997012.875
+      "train_runtime": 257.8636,
+      "train_tokens_per_second": 149932.561
     },
     {
       "epoch": 0.12981393336218086,
-      "grad_norm": 1.9140625,
+      "grad_norm": 2.53125,
       "learning_rate": 0.0005937473684210525,
-      "loss": 0.2287,
+      "loss": 0.207,
       "num_input_tokens_seen": 39317504,
       "step": 600,
-      "train_runtime": 42.5907,
-      "train_tokens_per_second": 923146.805
+      "train_runtime": 261.6775,
+      "train_tokens_per_second": 150251.784
     },
     {
       "epoch": 0.12981393336218086,
-      "eval_loss": 0.21300141513347626,
-      "eval_runtime": 1.6985,
-      "eval_samples_per_second": 18.84,
-      "eval_steps_per_second": 0.589,
+      "eval_loss": 0.18635742366313934,
+      "eval_runtime": 2.3524,
+      "eval_samples_per_second": 13.603,
+      "eval_steps_per_second": 0.425,
       "num_input_tokens_seen": 39317504,
       "step": 600
     },
     {
       "epoch": 0.12981393336218086,
-      "eval_byte_accuracy": 0.9362017804154302,
-      "eval_chrf": 58.769769707171605,
-      "eval_sacrebleu": 45.27561934180984,
-      "eval_word_accuracy": 0.8736263736263736,
+      "eval_byte_accuracy": 0.9425074183976261,
+      "eval_chrf": 61.6495042526747,
+      "eval_sacrebleu": 48.627627138112864,
+      "eval_word_accuracy": 0.885989010989011,
       "num_input_tokens_seen": 39317504,
-      "perplexity": 1.2373864023092827,
+      "perplexity": 1.2048528263365477,
       "step": 600
     },
     {
       "epoch": 0.13197749891821722,
-      "grad_norm": 2.46875,
+      "grad_norm": 1.0078125,
       "learning_rate": 0.0005931157894736842,
-      "loss": 0.2159,
+      "loss": 0.1849,
       "num_input_tokens_seen": 39972864,
       "step": 610,
-      "train_runtime": 48.1225,
-      "train_tokens_per_second": 830647.665
+      "train_runtime": 269.8376,
+      "train_tokens_per_second": 148136.726
     },
     {
       "epoch": 0.13414106447425356,
-      "grad_norm": 1.703125,
+      "grad_norm": 2.015625,
       "learning_rate": 0.0005924842105263157,
-      "loss": 0.2167,
+      "loss": 0.1942,
       "num_input_tokens_seen": 40628224,
       "step": 620,
-      "train_runtime": 51.9415,
-      "train_tokens_per_second": 782191.417
+      "train_runtime": 273.6548,
+      "train_tokens_per_second": 148465.26
     },
     {
       "epoch": 0.13630463003028992,
-      "grad_norm": 3.65625,
+      "grad_norm": 2.1875,
       "learning_rate": 0.0005918526315789474,
-      "loss": 0.2096,
+      "loss": 0.1857,
       "num_input_tokens_seen": 41283584,
       "step": 630,
-      "train_runtime": 55.7509,
-      "train_tokens_per_second": 740500.215
+      "train_runtime": 277.4573,
+      "train_tokens_per_second": 148792.569
     },
     {
       "epoch": 0.13846819558632625,
-      "grad_norm": 2.421875,
+      "grad_norm": 1.6875,
       "learning_rate": 0.0005912210526315788,
-      "loss": 0.2046,
+      "loss": 0.1754,
       "num_input_tokens_seen": 41938944,
       "step": 640,
-      "train_runtime": 59.5559,
-      "train_tokens_per_second": 704194.506
+      "train_runtime": 281.2729,
+      "train_tokens_per_second": 149104.106
     },
     {
       "epoch": 0.14063176114236262,
-      "grad_norm": 3.15625,
+      "grad_norm": 1.4375,
       "learning_rate": 0.0005905894736842105,
-      "loss": 0.2014,
+      "loss": 0.1831,
       "num_input_tokens_seen": 42594304,
       "step": 650,
-      "train_runtime": 63.3661,
-      "train_tokens_per_second": 672194.358
+      "train_runtime": 285.0868,
+      "train_tokens_per_second": 149408.204
     },
     {
       "epoch": 0.14279532669839895,
-      "grad_norm": 2.34375,
+      "grad_norm": 2.046875,
       "learning_rate": 0.000589957894736842,
-      "loss": 0.198,
+      "loss": 0.1667,
       "num_input_tokens_seen": 43249664,
       "step": 660,
-      "train_runtime": 67.1728,
-      "train_tokens_per_second": 643856.801
+      "train_runtime": 290.9113,
+      "train_tokens_per_second": 148669.581
     },
     {
       "epoch": 0.1449588922544353,
-      "grad_norm": 2.21875,
+      "grad_norm": 1.765625,
       "learning_rate": 0.0005893263157894736,
-      "loss": 0.1932,
+      "loss": 0.1622,
       "num_input_tokens_seen": 43905024,
       "step": 670,
-      "train_runtime": 70.9897,
-      "train_tokens_per_second": 618470.148
+      "train_runtime": 294.7222,
+      "train_tokens_per_second": 148970.863
     },
     {
       "epoch": 0.14712245781047165,
-      "grad_norm": 2.984375,
+      "grad_norm": 1.5390625,
       "learning_rate": 0.0005886947368421052,
-      "loss": 0.1864,
+      "loss": 0.1577,
       "num_input_tokens_seen": 44560384,
       "step": 680,
-      "train_runtime": 74.8037,
-      "train_tokens_per_second": 595697.625
+      "train_runtime": 298.5377,
+      "train_tokens_per_second": 149262.168
     },
     {
       "epoch": 0.149286023366508,
-      "grad_norm": 3.0625,
+      "grad_norm": 2.3125,
       "learning_rate": 0.0005880631578947368,
-      "loss": 0.1815,
+      "loss": 0.1535,
       "num_input_tokens_seen": 45215744,
       "step": 690,
-      "train_runtime": 78.6205,
-      "train_tokens_per_second": 575113.667
+      "train_runtime": 302.3486,
+      "train_tokens_per_second": 149548.37
     },
     {
       "epoch": 0.15144958892254434,
-      "grad_norm": 2.34375,
+      "grad_norm": 1.4609375,
       "learning_rate": 0.0005874315789473684,
-      "loss": 0.1814,
+      "loss": 0.1465,
       "num_input_tokens_seen": 45871104,
       "step": 700,
-      "train_runtime": 82.4441,
-      "train_tokens_per_second": 556390.233
+      "train_runtime": 306.1617,
+      "train_tokens_per_second": 149826.412
     },
     {
       "epoch": 0.15144958892254434,
-      "eval_loss": 0.1680818498134613,
-      "eval_runtime": 1.3101,
-      "eval_samples_per_second": 24.426,
-      "eval_steps_per_second": 0.763,
+      "eval_loss": 0.14650069177150726,
+      "eval_runtime": 2.2585,
+      "eval_samples_per_second": 14.169,
+      "eval_steps_per_second": 0.443,
       "num_input_tokens_seen": 45871104,
       "step": 700
     },
     {
       "epoch": 0.15144958892254434,
-      "eval_byte_accuracy": 0.9506676557863502,
-      "eval_chrf": 67.20800143820422,
-      "eval_sacrebleu": 55.06406011918448,
-      "eval_word_accuracy": 0.8997252747252747,
+      "eval_byte_accuracy": 0.9584569732937686,
+      "eval_chrf": 73.74830346733725,
+      "eval_sacrebleu": 63.990915184214565,
+      "eval_word_accuracy": 0.9217032967032966,
       "num_input_tokens_seen": 45871104,
-      "perplexity": 1.183033437751315,
+      "perplexity": 1.1577757317344424,
       "step": 700
     },
     {
       "epoch": 0.1536131544785807,
-      "grad_norm": 1.921875,
+      "grad_norm": 2.3125,
       "learning_rate": 0.0005868,
-      "loss": 0.1903,
+      "loss": 0.1534,
       "num_input_tokens_seen": 46526464,
       "step": 710,
-      "train_runtime": 87.5833,
-      "train_tokens_per_second": 531225.494
+      "train_runtime": 314.2434,
+      "train_tokens_per_second": 148058.7
     },
     {
       "epoch": 0.15577672003461704,
-      "grad_norm": 2.265625,
+      "grad_norm": 2.671875,
       "learning_rate": 0.0005861684210526315,
-      "loss": 0.175,
+      "loss": 0.1441,
       "num_input_tokens_seen": 47181824,
       "step": 720,
-      "train_runtime": 91.4013,
-      "train_tokens_per_second": 516205.405
+      "train_runtime": 318.0598,
+      "train_tokens_per_second": 148342.621
     },
     {
       "epoch": 0.1579402855906534,
-      "grad_norm": 2.4375,
+      "grad_norm": 1.8671875,
       "learning_rate": 0.0005855368421052631,
-      "loss": 0.1664,
+      "loss": 0.1346,
       "num_input_tokens_seen": 47837184,
       "step": 730,
-      "train_runtime": 95.225,
-      "train_tokens_per_second": 502359.361
+      "train_runtime": 321.8647,
+      "train_tokens_per_second": 148625.111
     },
     {
       "epoch": 0.16010385114668974,
-      "grad_norm": 3.46875,
+      "grad_norm": 1.8046875,
       "learning_rate": 0.0005849052631578946,
-      "loss": 0.1707,
+      "loss": 0.1403,
       "num_input_tokens_seen": 48492544,
       "step": 740,
-      "train_runtime": 99.0437,
-      "train_tokens_per_second": 489607.502
+      "train_runtime": 325.6813,
+      "train_tokens_per_second": 148895.694
     },
     {
       "epoch": 0.1622674167027261,
-      "grad_norm": 2.296875,
+      "grad_norm": 1.6328125,
       "learning_rate": 0.0005842736842105263,
-      "loss": 0.1741,
+      "loss": 0.1426,
       "num_input_tokens_seen": 49147904,
       "step": 750,
-      "train_runtime": 102.8647,
-      "train_tokens_per_second": 477791.681
+      "train_runtime": 329.4803,
+      "train_tokens_per_second": 149167.94
     },
     {
       "epoch": 0.16443098225876243,
-      "grad_norm": 4.125,
+      "grad_norm": 1.1171875,
       "learning_rate": 0.0005836421052631578,
-      "loss": 0.1601,
+      "loss": 0.1282,
       "num_input_tokens_seen": 49803264,
       "step": 760,
-      "train_runtime": 106.6841,
-      "train_tokens_per_second": 466829.3
+      "train_runtime": 335.3012,
+      "train_tokens_per_second": 148532.903
     },
     {
       "epoch": 0.1665945478147988,
-      "grad_norm": 1.484375,
+      "grad_norm": 2.265625,
       "learning_rate": 0.0005830105263157895,
-      "loss": 0.1658,
+      "loss": 0.132,
       "num_input_tokens_seen": 50458624,
       "step": 770,
-      "train_runtime": 110.5067,
-      "train_tokens_per_second": 456611.574
+      "train_runtime": 339.1103,
+      "train_tokens_per_second": 148797.094
     },
     {
       "epoch": 0.16875811337083513,
-      "grad_norm": 1.40625,
+      "grad_norm": 1.234375,
       "learning_rate": 0.000582378947368421,
-      "loss": 0.1514,
+      "loss": 0.1239,
       "num_input_tokens_seen": 51113984,
       "step": 780,
-      "train_runtime": 114.331,
-      "train_tokens_per_second": 447070.255
+      "train_runtime": 342.9157,
+      "train_tokens_per_second": 149057.003
     },
     {
       "epoch": 0.1709216789268715,
-      "grad_norm": 1.875,
+      "grad_norm": 1.4375,
       "learning_rate": 0.0005817473684210526,
-      "loss": 0.1582,
+      "loss": 0.1318,
       "num_input_tokens_seen": 51769344,
       "step": 790,
-      "train_runtime": 118.1477,
-      "train_tokens_per_second": 438174.843
+      "train_runtime": 346.716,
+      "train_tokens_per_second": 149313.417
     },
     {
       "epoch": 0.17308524448290782,
-      "grad_norm": 1.546875,
+      "grad_norm": 1.7734375,
       "learning_rate": 0.0005811157894736841,
-      "loss": 0.1502,
+      "loss": 0.1268,
       "num_input_tokens_seen": 52424704,
       "step": 800,
-      "train_runtime": 121.9652,
-      "train_tokens_per_second": 429833.159
+      "train_runtime": 350.5228,
+      "train_tokens_per_second": 149561.474
     },
     {
       "epoch": 0.17308524448290782,
-      "eval_loss": 0.1423906683921814,
-      "eval_runtime": 1.704,
-      "eval_samples_per_second": 18.779,
-      "eval_steps_per_second": 0.587,
+      "eval_loss": 0.11049608141183853,
+      "eval_runtime": 1.8345,
+      "eval_samples_per_second": 17.443,
+      "eval_steps_per_second": 0.545,
       "num_input_tokens_seen": 52424704,
       "step": 800
     },
     {
       "epoch": 0.17308524448290782,
-      "eval_byte_accuracy": 0.9577151335311572,
-      "eval_chrf": 68.88294130098063,
-      "eval_sacrebleu": 58.6583351753512,
-      "eval_word_accuracy": 0.9120879120879121,
+      "eval_byte_accuracy": 0.9669881305637982,
+      "eval_chrf": 77.17903139546479,
+      "eval_sacrebleu": 67.74599411512361,
+      "eval_word_accuracy": 0.9313186813186813,
       "num_input_tokens_seen": 52424704,
-      "perplexity": 1.1530270117687667,
+      "perplexity": 1.1168319726389973,
       "step": 800
     },
     {
       "epoch": 0.17524881003894419,
-      "grad_norm": 1.6328125,
+      "grad_norm": 1.109375,
       "learning_rate": 0.0005804842105263157,
-      "loss": 0.1441,
+      "loss": 0.122,
       "num_input_tokens_seen": 53080064,
       "step": 810,
-      "train_runtime": 127.4996,
-      "train_tokens_per_second": 416315.372
+      "train_runtime": 358.1954,
+      "train_tokens_per_second": 148187.448
     },
     {
       "epoch": 0.17741237559498052,
-      "grad_norm": 1.9375,
+      "grad_norm": 1.3515625,
       "learning_rate": 0.0005798526315789473,
-      "loss": 0.1451,
+      "loss": 0.1235,
       "num_input_tokens_seen": 53735424,
       "step": 820,
-      "train_runtime": 131.3126,
-      "train_tokens_per_second": 409217.56
+      "train_runtime": 362.0007,
+      "train_tokens_per_second": 148440.091
     },
     {
       "epoch": 0.17957594115101688,
-      "grad_norm": 2.609375,
+      "grad_norm": 1.5703125,
       "learning_rate": 0.0005792210526315789,
-      "loss": 0.148,
+      "loss": 0.1217,
       "num_input_tokens_seen": 54390784,
       "step": 830,
-      "train_runtime": 135.1268,
-      "train_tokens_per_second": 402516.49
+      "train_runtime": 365.806,
+      "train_tokens_per_second": 148687.519
     },
     {
       "epoch": 0.18173950670705322,
-      "grad_norm": 1.1640625,
+      "grad_norm": 1.1171875,
       "learning_rate": 0.0005785894736842105,
-      "loss": 0.1378,
+      "loss": 0.1169,
       "num_input_tokens_seen": 55046144,
       "step": 840,
-      "train_runtime": 138.9294,
-      "train_tokens_per_second": 396216.745
+      "train_runtime": 369.611,
+      "train_tokens_per_second": 148929.926
     },
     {
       "epoch": 0.18390307226308958,
-      "grad_norm": 1.7109375,
+      "grad_norm": 1.3515625,
       "learning_rate": 0.000577957894736842,
-      "loss": 0.1417,
+      "loss": 0.1193,
       "num_input_tokens_seen": 55701504,
       "step": 850,
-      "train_runtime": 142.7317,
-      "train_tokens_per_second": 390253.256
+      "train_runtime": 373.4083,
+      "train_tokens_per_second": 149170.488
     },
     {
       "epoch": 0.1860666378191259,
-      "grad_norm": 2.25,
+      "grad_norm": 1.15625,
       "learning_rate": 0.0005773263157894736,
-      "loss": 0.1471,
+      "loss": 0.1186,
       "num_input_tokens_seen": 56356864,
       "step": 860,
-      "train_runtime": 146.5477,
-      "train_tokens_per_second": 384563.353
+      "train_runtime": 379.2379,
+      "train_tokens_per_second": 148605.558
     },
     {
       "epoch": 0.18823020337516227,
-      "grad_norm": 1.4140625,
+      "grad_norm": 1.21875,
       "learning_rate": 0.0005766947368421052,
-      "loss": 0.1388,
+      "loss": 0.1143,
       "num_input_tokens_seen": 57012224,
       "step": 870,
-      "train_runtime": 150.3727,
-      "train_tokens_per_second": 379139.434
+      "train_runtime": 383.0534,
+      "train_tokens_per_second": 148836.24
     },
     {
       "epoch": 0.1903937689311986,
-      "grad_norm": 1.5,
+      "grad_norm": 1.6796875,
       "learning_rate": 0.0005760631578947368,
-      "loss": 0.131,
+      "loss": 0.1117,
       "num_input_tokens_seen": 57667584,
       "step": 880,
-      "train_runtime": 154.1907,
-      "train_tokens_per_second": 374001.588
+      "train_runtime": 386.8673,
+      "train_tokens_per_second": 149062.955
     },
     {
       "epoch": 0.19255733448723497,
-      "grad_norm": 1.515625,
+      "grad_norm": 1.78125,
       "learning_rate": 0.0005754315789473684,
-      "loss": 0.1321,
+      "loss": 0.1103,
       "num_input_tokens_seen": 58322944,
       "step": 890,
-      "train_runtime": 158.0167,
-      "train_tokens_per_second": 369093.521
+      "train_runtime": 390.6796,
+      "train_tokens_per_second": 149285.862
     },
     {
       "epoch": 0.1947209000432713,
-      "grad_norm": 1.9609375,
+      "grad_norm": 1.390625,
       "learning_rate": 0.0005747999999999999,
-      "loss": 0.1588,
+      "loss": 0.1326,
       "num_input_tokens_seen": 58978304,
       "step": 900,
-      "train_runtime": 161.8381,
-      "train_tokens_per_second": 364427.771
+      "train_runtime": 394.4919,
+      "train_tokens_per_second": 149504.475
     },
     {
       "epoch": 0.1947209000432713,
-      "eval_loss": 0.12399701774120331,
-      "eval_runtime": 1.6341,
-      "eval_samples_per_second": 19.583,
-      "eval_steps_per_second": 0.612,
+      "eval_loss": 0.10183807462453842,
+      "eval_runtime": 2.1572,
+      "eval_samples_per_second": 14.834,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 58978304,
       "step": 900
     },
     {
       "epoch": 0.1947209000432713,
-      "eval_byte_accuracy": 0.9636498516320475,
-      "eval_chrf": 74.82778143236575,
-      "eval_sacrebleu": 62.86084965698353,
-      "eval_word_accuracy": 0.9189560439560439,
+      "eval_byte_accuracy": 0.9725519287833828,
+      "eval_chrf": 79.43792789971359,
+      "eval_sacrebleu": 70.42945591585108,
+      "eval_word_accuracy": 0.9354395604395604,
       "num_input_tokens_seen": 58978304,
-      "perplexity": 1.13201249503992,
+      "perplexity": 1.107204172760412,
       "step": 900
     },
     {
       "epoch": 0.19688446559930767,
-      "grad_norm": 2.78125,
+      "grad_norm": 1.8984375,
       "learning_rate": 0.0005741684210526316,
-      "loss": 0.1457,
+      "loss": 0.1167,
       "num_input_tokens_seen": 59633664,
       "step": 910,
-      "train_runtime": 167.3034,
-      "train_tokens_per_second": 356440.14
+      "train_runtime": 402.4861,
+      "train_tokens_per_second": 148163.277
     },
     {
       "epoch": 0.199048031155344,
-      "grad_norm": 2.15625,
+      "grad_norm": 1.0234375,
       "learning_rate": 0.000573536842105263,
-      "loss": 0.1314,
+      "loss": 0.1048,
       "num_input_tokens_seen": 60289024,
       "step": 920,
-      "train_runtime": 171.1208,
-      "train_tokens_per_second": 352318.563
+      "train_runtime": 406.3049,
+      "train_tokens_per_second": 148383.715
     },
     {
       "epoch": 0.20121159671138036,
-      "grad_norm": 1.7578125,
+      "grad_norm": 1.6328125,
       "learning_rate": 0.0005729052631578947,
-      "loss": 0.1392,
+      "loss": 0.1103,
       "num_input_tokens_seen": 60944384,
       "step": 930,
-      "train_runtime": 174.9323,
-      "train_tokens_per_second": 348388.345
+      "train_runtime": 410.1125,
+      "train_tokens_per_second": 148604.052
     },
     {
       "epoch": 0.2033751622674167,
-      "grad_norm": 1.6171875,
+      "grad_norm": 1.484375,
       "learning_rate": 0.0005722736842105262,
-      "loss": 0.1306,
+      "loss": 0.1031,
       "num_input_tokens_seen": 61599744,
       "step": 940,
-      "train_runtime": 178.748,
-      "train_tokens_per_second": 344617.844
+      "train_runtime": 413.9199,
+      "train_tokens_per_second": 148820.456
     },
     {
       "epoch": 0.20553872782345306,
-      "grad_norm": 2.203125,
+      "grad_norm": 0.90234375,
       "learning_rate": 0.0005716421052631578,
-      "loss": 0.127,
+      "loss": 0.1013,
       "num_input_tokens_seen": 62255104,
       "step": 950,
-      "train_runtime": 182.5636,
-      "train_tokens_per_second": 341005.098
+      "train_runtime": 417.7339,
+      "train_tokens_per_second": 149030.524
     },
     {
       "epoch": 0.2077022933794894,
-      "grad_norm": 2.0625,
+      "grad_norm": 0.69140625,
       "learning_rate": 0.0005710105263157894,
-      "loss": 0.1286,
+      "loss": 0.1018,
       "num_input_tokens_seen": 62910464,
       "step": 960,
-      "train_runtime": 186.3727,
-      "train_tokens_per_second": 337551.932
+      "train_runtime": 423.5646,
+      "train_tokens_per_second": 148526.263
     },
     {
       "epoch": 0.20986585893552576,
-      "grad_norm": 1.359375,
+      "grad_norm": 1.109375,
       "learning_rate": 0.000570378947368421,
-      "loss": 0.1128,
+      "loss": 0.0874,
       "num_input_tokens_seen": 63565824,
       "step": 970,
-      "train_runtime": 190.1949,
-      "train_tokens_per_second": 334214.163
+      "train_runtime": 427.3766,
+      "train_tokens_per_second": 148734.931
     },
     {
       "epoch": 0.2120294244915621,
-      "grad_norm": 2.28125,
+      "grad_norm": 1.9609375,
       "learning_rate": 0.0005697473684210526,
-      "loss": 0.1255,
+      "loss": 0.1025,
       "num_input_tokens_seen": 64221184,
       "step": 980,
-      "train_runtime": 193.9964,
-      "train_tokens_per_second": 331043.206
+      "train_runtime": 431.1598,
+      "train_tokens_per_second": 148949.857
     },
     {
       "epoch": 0.21419299004759845,
-      "grad_norm": 1.8125,
+      "grad_norm": 1.828125,
       "learning_rate": 0.0005691157894736842,
-      "loss": 0.1216,
+      "loss": 0.0988,
       "num_input_tokens_seen": 64876544,
       "step": 990,
-      "train_runtime": 197.8194,
-      "train_tokens_per_second": 327958.403
+      "train_runtime": 434.9749,
+      "train_tokens_per_second": 149150.081
     },
     {
       "epoch": 0.2163565556036348,
-      "grad_norm": 1.3671875,
+      "grad_norm": 1.875,
       "learning_rate": 0.0005684842105263157,
-      "loss": 0.1215,
+      "loss": 0.1019,
       "num_input_tokens_seen": 65531904,
       "step": 1000,
-      "train_runtime": 201.6448,
-      "train_tokens_per_second": 324986.864
+      "train_runtime": 438.7886,
+      "train_tokens_per_second": 149347.327
     },
     {
       "epoch": 0.2163565556036348,
-      "eval_loss": 0.10376948118209839,
-      "eval_runtime": 1.7832,
-      "eval_samples_per_second": 17.945,
-      "eval_steps_per_second": 0.561,
+      "eval_loss": 0.08433977514505386,
+      "eval_runtime": 2.1587,
+      "eval_samples_per_second": 14.824,
+      "eval_steps_per_second": 0.463,
       "num_input_tokens_seen": 65531904,
       "step": 1000
     },
     {
       "epoch": 0.2163565556036348,
-      "eval_byte_accuracy": 0.9721810089020771,
-      "eval_chrf": 77.58364870895588,
-      "eval_sacrebleu": 66.59314384617089,
-      "eval_word_accuracy": 0.9326923076923077,
+      "eval_byte_accuracy": 0.9747774480712166,
+      "eval_chrf": 82.22196329181519,
+      "eval_sacrebleu": 74.53680220343439,
+      "eval_word_accuracy": 0.9464285714285714,
       "num_input_tokens_seen": 65531904,
-      "perplexity": 1.109344700609595,
+      "perplexity": 1.0879985058629236,
       "step": 1000
     },
     {
       "epoch": 0.21852012115967115,
-      "grad_norm": 1.3828125,
+      "grad_norm": 1.90625,
       "learning_rate": 0.0005678526315789473,
-      "loss": 0.117,
+      "loss": 0.0908,
       "num_input_tokens_seen": 66187264,
       "step": 1010,
-      "train_runtime": 207.2888,
-      "train_tokens_per_second": 319299.811
+      "train_runtime": 446.6399,
+      "train_tokens_per_second": 148189.311
     },
     {
       "epoch": 0.22068368671570748,
-      "grad_norm": 2.375,
+      "grad_norm": 1.1796875,
       "learning_rate": 0.0005672210526315789,
-      "loss": 0.1187,
+      "loss": 0.0935,
       "num_input_tokens_seen": 66842624,
       "step": 1020,
-      "train_runtime": 211.106,
-      "train_tokens_per_second": 316630.621
+      "train_runtime": 450.4535,
+      "train_tokens_per_second": 148389.609
     },
     {
       "epoch": 0.22284725227174385,
-      "grad_norm": 1.3671875,
+      "grad_norm": 1.0859375,
       "learning_rate": 0.0005665894736842105,
-      "loss": 0.1187,
+      "loss": 0.0938,
       "num_input_tokens_seen": 67497984,
       "step": 1030,
-      "train_runtime": 214.9217,
-      "train_tokens_per_second": 314058.489
+      "train_runtime": 454.2611,
+      "train_tokens_per_second": 148588.505
     },
     {
       "epoch": 0.22501081782778018,
-      "grad_norm": 1.1328125,
+      "grad_norm": 1.1484375,
       "learning_rate": 0.000565957894736842,
-      "loss": 0.1259,
+      "loss": 0.1033,
       "num_input_tokens_seen": 68153344,
       "step": 1040,
-      "train_runtime": 218.7377,
-      "train_tokens_per_second": 311575.604
+      "train_runtime": 458.0751,
+      "train_tokens_per_second": 148782.018
     },
     {
       "epoch": 0.22717438338381654,
-      "grad_norm": 2.203125,
+      "grad_norm": 1.6328125,
       "learning_rate": 0.0005653263157894737,
-      "loss": 0.1243,
+      "loss": 0.1008,
       "num_input_tokens_seen": 68808704,
       "step": 1050,
-      "train_runtime": 222.5611,
-      "train_tokens_per_second": 309167.656
+      "train_runtime": 461.8811,
+      "train_tokens_per_second": 148974.931
     },
     {
       "epoch": 0.22933794893985288,
-      "grad_norm": 2.0625,
+      "grad_norm": 1.515625,
       "learning_rate": 0.0005646947368421052,
-      "loss": 0.1136,
+      "loss": 0.0898,
       "num_input_tokens_seen": 69464064,
       "step": 1060,
-      "train_runtime": 226.3782,
-      "train_tokens_per_second": 306849.57
+      "train_runtime": 467.8393,
+      "train_tokens_per_second": 148478.48
     },
     {
       "epoch": 0.23150151449588924,
-      "grad_norm": 1.59375,
+      "grad_norm": 1.640625,
       "learning_rate": 0.0005640631578947368,
-      "loss": 0.1111,
+      "loss": 0.0884,
       "num_input_tokens_seen": 70119424,
       "step": 1070,
-      "train_runtime": 230.1891,
-      "train_tokens_per_second": 304616.596
+      "train_runtime": 471.6499,
+      "train_tokens_per_second": 148668.361
     },
     {
       "epoch": 0.23366508005192557,
-      "grad_norm": 2.765625,
+      "grad_norm": 1.453125,
       "learning_rate": 0.0005634315789473683,
-      "loss": 0.1161,
+      "loss": 0.0984,
       "num_input_tokens_seen": 70774784,
       "step": 1080,
-      "train_runtime": 234.0032,
-      "train_tokens_per_second": 302452.263
+      "train_runtime": 475.461,
+      "train_tokens_per_second": 148855.087
     },
     {
       "epoch": 0.23582864560796193,
-      "grad_norm": 1.046875,
+      "grad_norm": 1.3046875,
       "learning_rate": 0.0005627999999999999,
-      "loss": 0.114,
+      "loss": 0.0928,
       "num_input_tokens_seen": 71430144,
       "step": 1090,
-      "train_runtime": 237.8211,
-      "train_tokens_per_second": 300352.379
+      "train_runtime": 479.2668,
+      "train_tokens_per_second": 149040.46
     },
     {
       "epoch": 0.23799221116399827,
-      "grad_norm": 2.09375,
+      "grad_norm": 1.6875,
       "learning_rate": 0.0005621684210526315,
-      "loss": 0.114,
+      "loss": 0.094,
       "num_input_tokens_seen": 72085504,
       "step": 1100,
-      "train_runtime": 241.6412,
-      "train_tokens_per_second": 298316.23
+      "train_runtime": 483.0765,
+      "train_tokens_per_second": 149221.712
     },
     {
       "epoch": 0.23799221116399827,
-      "eval_loss": 0.10388417541980743,
-      "eval_runtime": 1.1455,
-      "eval_samples_per_second": 27.934,
-      "eval_steps_per_second": 0.873,
+      "eval_loss": 0.0829336941242218,
+      "eval_runtime": 1.6255,
+      "eval_samples_per_second": 19.686,
+      "eval_steps_per_second": 0.615,
       "num_input_tokens_seen": 72085504,
       "step": 1100
     },
     {
       "epoch": 0.23799221116399827,
-      "eval_byte_accuracy": 0.9699554896142433,
-      "eval_chrf": 77.89753777813905,
-      "eval_sacrebleu": 69.6119771947039,
-      "eval_word_accuracy": 0.9326923076923077,
+      "eval_byte_accuracy": 0.9762611275964391,
+      "eval_chrf": 82.85300346706357,
+      "eval_sacrebleu": 77.14647221270954,
+      "eval_word_accuracy": 0.9478021978021978,
       "num_input_tokens_seen": 72085504,
-      "perplexity": 1.1094719433512532,
+      "perplexity": 1.0864697668304737,
       "step": 1100
     },
     {
       "epoch": 0.24015577672003463,
-      "grad_norm": 2.03125,
+      "grad_norm": 1.3046875,
       "learning_rate": 0.0005615368421052631,
-      "loss": 0.1178,
+      "loss": 0.0911,
       "num_input_tokens_seen": 72740864,
       "step": 1110,
-      "train_runtime": 246.6141,
-      "train_tokens_per_second": 294958.283
+      "train_runtime": 490.5196,
+      "train_tokens_per_second": 148293.491
     },
     {
       "epoch": 0.24231934227607096,
-      "grad_norm": 2.25,
+      "grad_norm": 1.078125,
       "learning_rate": 0.0005609052631578947,
-      "loss": 0.1074,
+      "loss": 0.087,
       "num_input_tokens_seen": 73396224,
       "step": 1120,
-      "train_runtime": 250.4279,
-      "train_tokens_per_second": 293083.197
+      "train_runtime": 494.333,
+      "train_tokens_per_second": 148475.267
     },
     {
       "epoch": 0.2444829078321073,
-      "grad_norm": 1.8203125,
+      "grad_norm": 0.98046875,
       "learning_rate": 0.0005602736842105263,
-      "loss": 0.1,
+      "loss": 0.0872,
       "num_input_tokens_seen": 74051584,
       "step": 1130,
-      "train_runtime": 254.2431,
-      "train_tokens_per_second": 291262.949
+      "train_runtime": 498.1388,
+      "train_tokens_per_second": 148656.534
     },
     {
       "epoch": 0.24664647338814366,
-      "grad_norm": 1.6171875,
+      "grad_norm": 1.1640625,
       "learning_rate": 0.0005596421052631578,
-      "loss": 0.1025,
+      "loss": 0.0822,
       "num_input_tokens_seen": 74706944,
       "step": 1140,
-      "train_runtime": 258.0624,
-      "train_tokens_per_second": 289491.737
+      "train_runtime": 501.9551,
+      "train_tokens_per_second": 148831.924
     },
     {
       "epoch": 0.24881003894418,
-      "grad_norm": 1.3515625,
+      "grad_norm": 1.1484375,
       "learning_rate": 0.0005590105263157894,
-      "loss": 0.1037,
+      "loss": 0.0837,
       "num_input_tokens_seen": 75362304,
       "step": 1150,
-      "train_runtime": 261.8777,
-      "train_tokens_per_second": 287776.701
+      "train_runtime": 505.7692,
+      "train_tokens_per_second": 149005.319
     },
     {
       "epoch": 0.25097360450021633,
-      "grad_norm": 1.765625,
+      "grad_norm": 1.3359375,
       "learning_rate": 0.000558378947368421,
-      "loss": 0.1021,
+      "loss": 0.0818,
       "num_input_tokens_seen": 76017664,
       "step": 1160,
-      "train_runtime": 265.6992,
-      "train_tokens_per_second": 286104.222
+      "train_runtime": 511.7654,
+      "train_tokens_per_second": 148540.053
     },
     {
       "epoch": 0.2531371700562527,
-      "grad_norm": 1.5234375,
+      "grad_norm": 1.1796875,
       "learning_rate": 0.0005577473684210526,
-      "loss": 0.0998,
+      "loss": 0.0843,
       "num_input_tokens_seen": 76673024,
       "step": 1170,
-      "train_runtime": 269.5112,
-      "train_tokens_per_second": 284489.237
+      "train_runtime": 515.5749,
+      "train_tokens_per_second": 148713.644
     },
     {
       "epoch": 0.25530073561228905,
-      "grad_norm": 1.3515625,
+      "grad_norm": 1.2421875,
       "learning_rate": 0.0005571157894736842,
-      "loss": 0.0996,
+      "loss": 0.084,
       "num_input_tokens_seen": 77328384,
       "step": 1180,
-      "train_runtime": 273.33,
-      "train_tokens_per_second": 282912.16
+      "train_runtime": 519.3838,
+      "train_tokens_per_second": 148884.869
     },
     {
       "epoch": 0.2574643011683254,
-      "grad_norm": 1.125,
+      "grad_norm": 1.4453125,
       "learning_rate": 0.0005564842105263158,
-      "loss": 0.0978,
+      "loss": 0.0813,
       "num_input_tokens_seen": 77979648,
       "step": 1190,
-      "train_runtime": 277.1245,
-      "train_tokens_per_second": 281388.469
+      "train_runtime": 523.1784,
+      "train_tokens_per_second": 149049.825
     },
     {
       "epoch": 0.2596278667243617,
-      "grad_norm": 1.65625,
+      "grad_norm": 0.984375,
       "learning_rate": 0.0005558526315789473,
-      "loss": 0.1089,
+      "loss": 0.0832,
       "num_input_tokens_seen": 78635008,
       "step": 1200,
-      "train_runtime": 280.9373,
-      "train_tokens_per_second": 279902.327
+      "train_runtime": 526.9952,
+      "train_tokens_per_second": 149213.905
     },
     {
       "epoch": 0.2596278667243617,
-      "eval_loss": 0.08458718657493591,
-      "eval_runtime": 1.7632,
-      "eval_samples_per_second": 18.149,
-      "eval_steps_per_second": 0.567,
+      "eval_loss": 0.0685884952545166,
+      "eval_runtime": 1.1892,
+      "eval_samples_per_second": 26.91,
+      "eval_steps_per_second": 0.841,
       "num_input_tokens_seen": 78635008,
       "step": 1200
     },
     {
       "epoch": 0.2596278667243617,
-      "eval_byte_accuracy": 0.9755192878338279,
-      "eval_chrf": 80.88240618865393,
-      "eval_sacrebleu": 71.88855321657775,
-      "eval_word_accuracy": 0.9395604395604396,
+      "eval_byte_accuracy": 0.9803412462908012,
+      "eval_chrf": 86.18467528671245,
+      "eval_sacrebleu": 80.29154426872704,
+      "eval_word_accuracy": 0.9519230769230769,
       "num_input_tokens_seen": 78635008,
-      "perplexity": 1.0882677224312234,
+      "perplexity": 1.0709953987678011,
       "step": 1200
     },
     {
       "epoch": 0.2617914322803981,
-      "grad_norm": 1.2734375,
+      "grad_norm": 1.515625,
       "learning_rate": 0.000555221052631579,
-      "loss": 0.0951,
+      "loss": 0.0736,
       "num_input_tokens_seen": 79290368,
       "step": 1210,
-      "train_runtime": 286.5287,
-      "train_tokens_per_second": 276727.531
+      "train_runtime": 534.0458,
+      "train_tokens_per_second": 148471.108
     },
     {
       "epoch": 0.26395499783643445,
-      "grad_norm": 2.0,
+      "grad_norm": 1.0,
       "learning_rate": 0.0005545894736842104,
-      "loss": 0.0971,
+      "loss": 0.0792,
       "num_input_tokens_seen": 79945728,
       "step": 1220,
-      "train_runtime": 290.3417,
-      "train_tokens_per_second": 275350.474
+      "train_runtime": 537.8548,
+      "train_tokens_per_second": 148638.118
     },
     {
       "epoch": 0.2661185633924708,
-      "grad_norm": 1.1171875,
+      "grad_norm": 0.9453125,
       "learning_rate": 0.0005539578947368421,
-      "loss": 0.0993,
+      "loss": 0.0785,
       "num_input_tokens_seen": 80601088,
       "step": 1230,
-      "train_runtime": 294.1684,
-      "train_tokens_per_second": 273996.464
+      "train_runtime": 541.6708,
+      "train_tokens_per_second": 148800.87
     },
     {
       "epoch": 0.2682821289485071,
-      "grad_norm": 1.2890625,
+      "grad_norm": 1.0859375,
       "learning_rate": 0.0005533263157894736,
-      "loss": 0.0951,
+      "loss": 0.0775,
       "num_input_tokens_seen": 81256448,
       "step": 1240,
-      "train_runtime": 297.9859,
-      "train_tokens_per_second": 272685.574
+      "train_runtime": 545.4848,
+      "train_tokens_per_second": 148961.887
     },
     {
       "epoch": 0.2704456945045435,
-      "grad_norm": 1.359375,
+      "grad_norm": 1.109375,
       "learning_rate": 0.0005526947368421052,
-      "loss": 0.0968,
+      "loss": 0.0781,
       "num_input_tokens_seen": 81911808,
       "step": 1250,
-      "train_runtime": 301.7986,
-      "train_tokens_per_second": 271412.157
+      "train_runtime": 549.2987,
+      "train_tokens_per_second": 149120.689
     },
     {
       "epoch": 0.27260926006057984,
-      "grad_norm": 1.4140625,
+      "grad_norm": 1.1484375,
       "learning_rate": 0.0005520631578947368,
-      "loss": 0.0978,
+      "loss": 0.077,
       "num_input_tokens_seen": 82563072,
       "step": 1260,
-      "train_runtime": 305.5977,
-      "train_tokens_per_second": 270169.191
+      "train_runtime": 554.9402,
+      "train_tokens_per_second": 148778.327
     },
     {
       "epoch": 0.2747728256166162,
-      "grad_norm": 1.640625,
+      "grad_norm": 1.3984375,
       "learning_rate": 0.0005514315789473684,
-      "loss": 0.0992,
+      "loss": 0.0818,
       "num_input_tokens_seen": 83218432,
       "step": 1270,
-      "train_runtime": 309.4192,
-      "train_tokens_per_second": 268950.483
+      "train_runtime": 558.7404,
+      "train_tokens_per_second": 148939.344
     },
     {
       "epoch": 0.2769363911726525,
-      "grad_norm": 2.046875,
+      "grad_norm": 1.4765625,
       "learning_rate": 0.0005507999999999999,
-      "loss": 0.0943,
+      "loss": 0.0764,
       "num_input_tokens_seen": 83873792,
       "step": 1280,
-      "train_runtime": 313.2275,
-      "train_tokens_per_second": 267772.763
+      "train_runtime": 562.5456,
+      "train_tokens_per_second": 149096.867
     },
     {
       "epoch": 0.27909995672868887,
-      "grad_norm": 1.328125,
+      "grad_norm": 1.2421875,
       "learning_rate": 0.0005501684210526315,
-      "loss": 0.093,
+      "loss": 0.0804,
       "num_input_tokens_seen": 84529152,
       "step": 1290,
-      "train_runtime": 317.0294,
-      "train_tokens_per_second": 266628.739
+      "train_runtime": 566.3426,
+      "train_tokens_per_second": 149254.436
     },
     {
       "epoch": 0.28126352228472523,
-      "grad_norm": 1.8828125,
+      "grad_norm": 1.1796875,
       "learning_rate": 0.0005495368421052631,
-      "loss": 0.0966,
+      "loss": 0.0795,
       "num_input_tokens_seen": 85180416,
       "step": 1300,
-      "train_runtime": 320.8229,
-      "train_tokens_per_second": 265506.073
+      "train_runtime": 570.132,
+      "train_tokens_per_second": 149404.72
     },
     {
       "epoch": 0.28126352228472523,
-      "eval_loss": 0.06946705281734467,
-      "eval_runtime": 1.5133,
-      "eval_samples_per_second": 21.145,
-      "eval_steps_per_second": 0.661,
+      "eval_loss": 0.060103051364421844,
+      "eval_runtime": 2.1621,
+      "eval_samples_per_second": 14.8,
+      "eval_steps_per_second": 0.463,
       "num_input_tokens_seen": 85180416,
       "step": 1300
     },
     {
       "epoch": 0.28126352228472523,
-      "eval_byte_accuracy": 0.9799703264094956,
-      "eval_chrf": 84.70162551567762,
-      "eval_sacrebleu": 77.62774508830094,
-      "eval_word_accuracy": 0.9532967032967034,
+      "eval_byte_accuracy": 0.9818249258160238,
+      "eval_chrf": 88.10671185800685,
+      "eval_sacrebleu": 84.31771246247678,
+      "eval_word_accuracy": 0.9629120879120879,
       "num_input_tokens_seen": 85180416,
-      "perplexity": 1.0719367433272844,
+      "perplexity": 1.0619459758885987,
       "step": 1300
     },
     {
       "epoch": 0.2834270878407616,
-      "grad_norm": 2.328125,
+      "grad_norm": 1.09375,
       "learning_rate": 0.0005489052631578947,
-      "loss": 0.0902,
+      "loss": 0.0728,
       "num_input_tokens_seen": 85835776,
       "step": 1310,
-      "train_runtime": 326.1743,
-      "train_tokens_per_second": 263159.235
+      "train_runtime": 578.2913,
+      "train_tokens_per_second": 148430.004
     },
     {
       "epoch": 0.2855906533967979,
-      "grad_norm": 0.9375,
+      "grad_norm": 1.4140625,
       "learning_rate": 0.0005482736842105263,
-      "loss": 0.0901,
+      "loss": 0.0748,
       "num_input_tokens_seen": 86491136,
       "step": 1320,
-      "train_runtime": 329.9891,
-      "train_tokens_per_second": 262103.012
+      "train_runtime": 582.1077,
+      "train_tokens_per_second": 148582.709
     },
     {
       "epoch": 0.28775421895283426,
-      "grad_norm": 1.09375,
+      "grad_norm": 0.8828125,
       "learning_rate": 0.0005476421052631579,
-      "loss": 0.0925,
+      "loss": 0.0738,
       "num_input_tokens_seen": 87146496,
       "step": 1330,
-      "train_runtime": 333.7894,
-      "train_tokens_per_second": 261082.25
+      "train_runtime": 585.9007,
+      "train_tokens_per_second": 148739.354
     },
     {
       "epoch": 0.2899177845088706,
-      "grad_norm": 1.7890625,
+      "grad_norm": 0.75,
       "learning_rate": 0.0005470105263157895,
-      "loss": 0.0848,
+      "loss": 0.0673,
       "num_input_tokens_seen": 87801856,
       "step": 1340,
-      "train_runtime": 337.6083,
-      "train_tokens_per_second": 260070.162
+      "train_runtime": 589.6992,
+      "train_tokens_per_second": 148892.611
     },
     {
       "epoch": 0.292081350064907,
-      "grad_norm": 1.5,
+      "grad_norm": 1.25,
       "learning_rate": 0.000546378947368421,
-      "loss": 0.0891,
+      "loss": 0.0691,
       "num_input_tokens_seen": 88457216,
       "step": 1350,
-      "train_runtime": 341.4242,
-      "train_tokens_per_second": 259083.017
+      "train_runtime": 593.5107,
+      "train_tokens_per_second": 149040.639
     },
     {
       "epoch": 0.2942449156209433,
-      "grad_norm": 1.359375,
+      "grad_norm": 1.5546875,
       "learning_rate": 0.0005457473684210525,
-      "loss": 0.0909,
+      "loss": 0.0731,
       "num_input_tokens_seen": 89112576,
       "step": 1360,
-      "train_runtime": 345.2394,
-      "train_tokens_per_second": 258118.218
+      "train_runtime": 599.3419,
+      "train_tokens_per_second": 148684.049
     },
     {
       "epoch": 0.29640848117697965,
-      "grad_norm": 1.75,
+      "grad_norm": 1.2421875,
       "learning_rate": 0.0005451157894736842,
-      "loss": 0.0877,
+      "loss": 0.0676,
       "num_input_tokens_seen": 89767936,
       "step": 1370,
-      "train_runtime": 349.0489,
-      "train_tokens_per_second": 257178.658
+      "train_runtime": 603.1499,
+      "train_tokens_per_second": 148831.876
     },
     {
       "epoch": 0.298572046733016,
-      "grad_norm": 1.3984375,
+      "grad_norm": 1.2421875,
       "learning_rate": 0.0005444842105263157,
-      "loss": 0.0924,
+      "loss": 0.0776,
       "num_input_tokens_seen": 90423296,
       "step": 1380,
-      "train_runtime": 352.8617,
-      "train_tokens_per_second": 256257.015
+      "train_runtime": 606.9619,
+      "train_tokens_per_second": 148976.882
     },
     {
       "epoch": 0.3007356122890524,
-      "grad_norm": 1.1484375,
+      "grad_norm": 1.25,
       "learning_rate": 0.0005438526315789473,
-      "loss": 0.0905,
+      "loss": 0.0724,
       "num_input_tokens_seen": 91078656,
       "step": 1390,
-      "train_runtime": 356.6816,
-      "train_tokens_per_second": 255350.05
+      "train_runtime": 610.7783,
+      "train_tokens_per_second": 149119.01
     },
     {
       "epoch": 0.3028991778450887,
-      "grad_norm": 1.5234375,
+      "grad_norm": 1.1640625,
       "learning_rate": 0.0005432210526315789,
-      "loss": 0.0863,
+      "loss": 0.0721,
       "num_input_tokens_seen": 91734016,
       "step": 1400,
-      "train_runtime": 360.4899,
-      "train_tokens_per_second": 254470.434
+      "train_runtime": 614.5862,
+      "train_tokens_per_second": 149261.44
     },
     {
       "epoch": 0.3028991778450887,
-      "eval_loss": 0.07211805135011673,
-      "eval_runtime": 1.4771,
-      "eval_samples_per_second": 21.665,
-      "eval_steps_per_second": 0.677,
+      "eval_loss": 0.062408819794654846,
+      "eval_runtime": 2.3491,
+      "eval_samples_per_second": 13.622,
+      "eval_steps_per_second": 0.426,
       "num_input_tokens_seen": 91734016,
       "step": 1400
     },
     {
       "epoch": 0.3028991778450887,
-      "eval_byte_accuracy": 0.9814540059347181,
-      "eval_chrf": 85.4092761981197,
-      "eval_sacrebleu": 78.3597595743492,
-      "eval_word_accuracy": 0.9546703296703297,
+      "eval_byte_accuracy": 0.9803412462908012,
+      "eval_chrf": 86.63365299116445,
+      "eval_sacrebleu": 82.79436667187097,
+      "eval_word_accuracy": 0.9601648351648352,
       "num_input_tokens_seen": 91734016,
-      "perplexity": 1.0747822160666487,
+      "perplexity": 1.064397402519385,
       "step": 1400
     },
     {
       "epoch": 0.30506274340112505,
-      "grad_norm": 1.421875,
+      "grad_norm": 1.453125,
       "learning_rate": 0.0005425894736842105,
-      "loss": 0.0863,
+      "loss": 0.0812,
       "num_input_tokens_seen": 92389376,
       "step": 1410,
-      "train_runtime": 365.7965,
-      "train_tokens_per_second": 252570.395
+      "train_runtime": 622.8076,
+      "train_tokens_per_second": 148343.36
     },
     {
       "epoch": 0.3072263089571614,
-      "grad_norm": 0.96484375,
+      "grad_norm": 1.4140625,
       "learning_rate": 0.000541957894736842,
-      "loss": 0.0933,
+      "loss": 0.0775,
       "num_input_tokens_seen": 93044736,
       "step": 1420,
-      "train_runtime": 369.6021,
-      "train_tokens_per_second": 251742.984
+      "train_runtime": 626.6216,
+      "train_tokens_per_second": 148486.309
     },
     {
       "epoch": 0.30938987451319777,
-      "grad_norm": 1.34375,
+      "grad_norm": 1.015625,
       "learning_rate": 0.0005413263157894736,
-      "loss": 0.0903,
+      "loss": 0.0714,
       "num_input_tokens_seen": 93700096,
       "step": 1430,
-      "train_runtime": 373.4103,
-      "train_tokens_per_second": 250930.673
+      "train_runtime": 630.429,
+      "train_tokens_per_second": 148629.105
     },
     {
       "epoch": 0.3115534400692341,
-      "grad_norm": 1.140625,
+      "grad_norm": 1.3125,
       "learning_rate": 0.0005406947368421052,
-      "loss": 0.0905,
+      "loss": 0.07,
       "num_input_tokens_seen": 94355456,
       "step": 1440,
-      "train_runtime": 377.2166,
-      "train_tokens_per_second": 250136.041
+      "train_runtime": 634.235,
+      "train_tokens_per_second": 148770.494
     },
     {
       "epoch": 0.31371700562527044,
-      "grad_norm": 1.46875,
+      "grad_norm": 0.9921875,
       "learning_rate": 0.0005400631578947368,
-      "loss": 0.0849,
+      "loss": 0.0673,
       "num_input_tokens_seen": 95010816,
       "step": 1450,
-      "train_runtime": 381.0327,
-      "train_tokens_per_second": 249350.817
+      "train_runtime": 638.0388,
+      "train_tokens_per_second": 148910.708
     },
     {
       "epoch": 0.3158805711813068,
-      "grad_norm": 1.46875,
+      "grad_norm": 0.859375,
       "learning_rate": 0.0005394315789473684,
-      "loss": 0.1062,
+      "loss": 0.0797,
       "num_input_tokens_seen": 95666176,
       "step": 1460,
-      "train_runtime": 384.8359,
-      "train_tokens_per_second": 248589.508
+      "train_runtime": 644.0522,
+      "train_tokens_per_second": 148537.927
     },
     {
       "epoch": 0.31804413673734316,
-      "grad_norm": 1.390625,
+      "grad_norm": 1.1171875,
       "learning_rate": 0.0005388,
-      "loss": 0.0876,
+      "loss": 0.0678,
       "num_input_tokens_seen": 96321536,
       "step": 1470,
-      "train_runtime": 388.6531,
-      "train_tokens_per_second": 247834.231
+      "train_runtime": 647.862,
+      "train_tokens_per_second": 148676.001
     },
     {
       "epoch": 0.32020770229337947,
-      "grad_norm": 1.1796875,
+      "grad_norm": 1.15625,
       "learning_rate": 0.0005381684210526316,
-      "loss": 0.0798,
+      "loss": 0.0623,
       "num_input_tokens_seen": 96976896,
       "step": 1480,
-      "train_runtime": 392.4638,
-      "train_tokens_per_second": 247097.674
+      "train_runtime": 651.6702,
+      "train_tokens_per_second": 148812.847
     },
     {
       "epoch": 0.32237126784941583,
-      "grad_norm": 1.6875,
+      "grad_norm": 1.484375,
       "learning_rate": 0.0005375368421052632,
-      "loss": 0.0886,
+      "loss": 0.075,
       "num_input_tokens_seen": 97632256,
       "step": 1490,
-      "train_runtime": 396.2816,
-      "train_tokens_per_second": 246370.891
+      "train_runtime": 655.4755,
+      "train_tokens_per_second": 148948.748
     },
     {
       "epoch": 0.3245348334054522,
-      "grad_norm": 1.140625,
+      "grad_norm": 1.2109375,
       "learning_rate": 0.0005369052631578947,
-      "loss": 0.0813,
+      "loss": 0.069,
       "num_input_tokens_seen": 98287616,
       "step": 1500,
-      "train_runtime": 400.0922,
-      "train_tokens_per_second": 245662.41
+      "train_runtime": 659.2779,
+      "train_tokens_per_second": 149083.749
     },
     {
       "epoch": 0.3245348334054522,
-      "eval_loss": 0.06404151022434235,
-      "eval_runtime": 1.651,
-      "eval_samples_per_second": 19.382,
-      "eval_steps_per_second": 0.606,
+      "eval_loss": 0.05183723196387291,
+      "eval_runtime": 1.9744,
+      "eval_samples_per_second": 16.207,
+      "eval_steps_per_second": 0.506,
       "num_input_tokens_seen": 98287616,
       "step": 1500
     },
     {
       "epoch": 0.3245348334054522,
       "eval_byte_accuracy": 0.9847922848664689,
-      "eval_chrf": 89.0715735988111,
-      "eval_sacrebleu": 85.97753690689314,
-      "eval_word_accuracy": 0.9684065934065934,
+      "eval_chrf": 88.68279915044877,
+      "eval_sacrebleu": 85.65513337936741,
+      "eval_word_accuracy": 0.967032967032967,
       "num_input_tokens_seen": 98287616,
-      "perplexity": 1.0661366534146515,
+      "perplexity": 1.053204300565465,
       "step": 1500
     },
     {
       "epoch": 0.32669839896148856,
-      "grad_norm": 1.5625,
+      "grad_norm": 1.2578125,
       "learning_rate": 0.0005362736842105263,
-      "loss": 0.083,
+      "loss": 0.0688,
       "num_input_tokens_seen": 98942976,
       "step": 1510,
-      "train_runtime": 405.6165,
-      "train_tokens_per_second": 243932.311
+      "train_runtime": 667.1596,
+      "train_tokens_per_second": 148304.8
     },
     {
       "epoch": 0.32886196451752486,
-      "grad_norm": 1.1875,
+      "grad_norm": 1.1328125,
       "learning_rate": 0.0005356421052631578,
-      "loss": 0.0794,
+      "loss": 0.0631,
       "num_input_tokens_seen": 99598336,
       "step": 1520,
-      "train_runtime": 409.4361,
-      "train_tokens_per_second": 243257.336
+      "train_runtime": 670.9705,
+      "train_tokens_per_second": 148439.219
     },
     {
       "epoch": 0.3310255300735612,
-      "grad_norm": 1.046875,
+      "grad_norm": 1.421875,
       "learning_rate": 0.0005350105263157894,
-      "loss": 0.0795,
+      "loss": 0.0648,
       "num_input_tokens_seen": 100253696,
       "step": 1530,
-      "train_runtime": 413.2546,
-      "train_tokens_per_second": 242595.488
+      "train_runtime": 674.7874,
+      "train_tokens_per_second": 148570.779
     },
     {
       "epoch": 0.3331890956295976,
-      "grad_norm": 1.109375,
+      "grad_norm": 1.1015625,
       "learning_rate": 0.000534378947368421,
-      "loss": 0.0851,
+      "loss": 0.0697,
       "num_input_tokens_seen": 100909056,
       "step": 1540,
-      "train_runtime": 417.0767,
-      "train_tokens_per_second": 241943.636
+      "train_runtime": 678.5971,
+      "train_tokens_per_second": 148702.463
     },
     {
       "epoch": 0.33535266118563395,
-      "grad_norm": 1.2734375,
+      "grad_norm": 1.453125,
       "learning_rate": 0.0005337473684210526,
-      "loss": 0.0857,
+      "loss": 0.0649,
       "num_input_tokens_seen": 101564416,
       "step": 1550,
-      "train_runtime": 420.8922,
-      "train_tokens_per_second": 241307.443
+      "train_runtime": 682.4042,
+      "train_tokens_per_second": 148833.213
     },
     {
       "epoch": 0.33751622674167026,
-      "grad_norm": 1.7265625,
+      "grad_norm": 1.1796875,
       "learning_rate": 0.0005331157894736841,
-      "loss": 0.0895,
+      "loss": 0.0753,
       "num_input_tokens_seen": 102219776,
       "step": 1560,
-      "train_runtime": 424.7232,
-      "train_tokens_per_second": 240673.856
+      "train_runtime": 688.2761,
+      "train_tokens_per_second": 148515.665
     },
     {
       "epoch": 0.3396797922977066,
-      "grad_norm": 0.8203125,
+      "grad_norm": 1.6015625,
       "learning_rate": 0.0005324842105263157,
-      "loss": 0.0832,
+      "loss": 0.0689,
       "num_input_tokens_seen": 102875136,
       "step": 1570,
-      "train_runtime": 428.5459,
-      "train_tokens_per_second": 240056.294
+      "train_runtime": 692.0898,
+      "train_tokens_per_second": 148644.199
     },
     {
       "epoch": 0.341843357853743,
-      "grad_norm": 0.99609375,
+      "grad_norm": 1.3125,
       "learning_rate": 0.0005318526315789473,
-      "loss": 0.0818,
+      "loss": 0.0647,
       "num_input_tokens_seen": 103530496,
       "step": 1580,
-      "train_runtime": 432.3726,
-      "train_tokens_per_second": 239447.421
+      "train_runtime": 695.9042,
+      "train_tokens_per_second": 148771.187
     },
     {
       "epoch": 0.34400692340977934,
-      "grad_norm": 1.484375,
+      "grad_norm": 0.98046875,
       "learning_rate": 0.0005312210526315789,
-      "loss": 0.0823,
+      "loss": 0.0651,
       "num_input_tokens_seen": 104185856,
       "step": 1590,
-      "train_runtime": 436.1733,
-      "train_tokens_per_second": 238863.469
+      "train_runtime": 699.6945,
+      "train_tokens_per_second": 148901.914
     },
     {
       "epoch": 0.34617048896581565,
-      "grad_norm": 0.6015625,
+      "grad_norm": 1.421875,
       "learning_rate": 0.0005305894736842105,
-      "loss": 0.081,
+      "loss": 0.0619,
       "num_input_tokens_seen": 104841216,
       "step": 1600,
-      "train_runtime": 439.9906,
-      "train_tokens_per_second": 238280.565
+      "train_runtime": 703.5081,
+      "train_tokens_per_second": 149026.308
     },
     {
       "epoch": 0.34617048896581565,
-      "eval_loss": 0.06596406549215317,
-      "eval_runtime": 1.454,
-      "eval_samples_per_second": 22.009,
-      "eval_steps_per_second": 0.688,
+      "eval_loss": 0.04635748267173767,
+      "eval_runtime": 1.9682,
+      "eval_samples_per_second": 16.259,
+      "eval_steps_per_second": 0.508,
       "num_input_tokens_seen": 104841216,
       "step": 1600
     },
     {
       "epoch": 0.34617048896581565,
-      "eval_byte_accuracy": 0.983679525222552,
-      "eval_chrf": 88.0318341339548,
-      "eval_sacrebleu": 83.92707672444418,
-      "eval_word_accuracy": 0.9656593406593407,
+      "eval_byte_accuracy": 0.9885014836795252,
+      "eval_chrf": 91.37848332537078,
+      "eval_sacrebleu": 88.58898373597441,
+      "eval_word_accuracy": 0.9739010989010989,
       "num_input_tokens_seen": 104841216,
-      "perplexity": 1.0681883316543295,
+      "perplexity": 1.047448788827688,
       "step": 1600
     },
     {
       "epoch": 0.348334054521852,
-      "grad_norm": 0.9609375,
+      "grad_norm": 1.421875,
       "learning_rate": 0.0005299578947368421,
-      "loss": 0.0782,
+      "loss": 0.0631,
       "num_input_tokens_seen": 105492480,
       "step": 1610,
-      "train_runtime": 445.2599,
-      "train_tokens_per_second": 236923.37
+      "train_runtime": 711.5187,
+      "train_tokens_per_second": 148263.814
     },
     {
       "epoch": 0.35049762007788837,
       "grad_norm": 1.03125,
       "learning_rate": 0.0005293263157894737,
-      "loss": 0.0809,
+      "loss": 0.0639,
       "num_input_tokens_seen": 106147840,
       "step": 1620,
-      "train_runtime": 449.0622,
-      "train_tokens_per_second": 236376.719
+      "train_runtime": 715.3139,
+      "train_tokens_per_second": 148393.364
     },
     {
       "epoch": 0.35266118563392473,
-      "grad_norm": 5.84375,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.0005286947368421053,
-      "loss": 0.0895,
+      "loss": 0.0642,
       "num_input_tokens_seen": 106803200,
       "step": 1630,
-      "train_runtime": 452.8883,
-      "train_tokens_per_second": 235826.785
+      "train_runtime": 719.1211,
+      "train_tokens_per_second": 148519.067
     },
     {
       "epoch": 0.35482475118996104,
-      "grad_norm": 1.046875,
+      "grad_norm": 0.93359375,
       "learning_rate": 0.0005280631578947368,
-      "loss": 0.0932,
+      "loss": 0.0652,
       "num_input_tokens_seen": 107458560,
       "step": 1640,
-      "train_runtime": 456.7037,
-      "train_tokens_per_second": 235291.625
+      "train_runtime": 722.9348,
+      "train_tokens_per_second": 148642.112
     },
     {
       "epoch": 0.3569883167459974,
-      "grad_norm": 1.21875,
+      "grad_norm": 0.97265625,
       "learning_rate": 0.0005274315789473684,
-      "loss": 0.0863,
+      "loss": 0.0678,
       "num_input_tokens_seen": 108113920,
       "step": 1650,
-      "train_runtime": 460.5239,
-      "train_tokens_per_second": 234762.901
+      "train_runtime": 726.7458,
+      "train_tokens_per_second": 148764.421
     },
     {
       "epoch": 0.35915188230203376,
-      "grad_norm": 1.453125,
+      "grad_norm": 0.95703125,
       "learning_rate": 0.0005267999999999999,
-      "loss": 0.0716,
+      "loss": 0.0592,
       "num_input_tokens_seen": 108765184,
       "step": 1660,
-      "train_runtime": 464.3225,
-      "train_tokens_per_second": 234244.92
+      "train_runtime": 732.6359,
+      "train_tokens_per_second": 148457.343
     },
     {
       "epoch": 0.3613154478580701,
-      "grad_norm": 1.1796875,
+      "grad_norm": 1.2421875,
       "learning_rate": 0.0005261684210526315,
-      "loss": 0.0744,
+      "loss": 0.0594,
       "num_input_tokens_seen": 109420544,
       "step": 1670,
-      "train_runtime": 468.1425,
-      "train_tokens_per_second": 233733.396
+      "train_runtime": 736.4517,
+      "train_tokens_per_second": 148578.034
     },
     {
       "epoch": 0.36347901341410643,
-      "grad_norm": 1.1171875,
+      "grad_norm": 1.1640625,
       "learning_rate": 0.0005255368421052631,
-      "loss": 0.0738,
+      "loss": 0.061,
       "num_input_tokens_seen": 110071808,
       "step": 1680,
-      "train_runtime": 471.9326,
-      "train_tokens_per_second": 233236.282
+      "train_runtime": 740.2501,
+      "train_tokens_per_second": 148695.435
     },
     {
       "epoch": 0.3656425789701428,
-      "grad_norm": 1.2890625,
+      "grad_norm": 1.2265625,
       "learning_rate": 0.0005249052631578947,
-      "loss": 0.079,
+      "loss": 0.0657,
       "num_input_tokens_seen": 110727168,
       "step": 1690,
-      "train_runtime": 475.748,
-      "train_tokens_per_second": 232743.306
+      "train_runtime": 744.0635,
+      "train_tokens_per_second": 148814.144
     },
     {
       "epoch": 0.36780614452617916,
-      "grad_norm": 1.0078125,
+      "grad_norm": 0.7578125,
       "learning_rate": 0.0005242736842105262,
-      "loss": 0.079,
+      "loss": 0.064,
       "num_input_tokens_seen": 111382528,
       "step": 1700,
-      "train_runtime": 479.5698,
-      "train_tokens_per_second": 232255.089
+      "train_runtime": 747.8708,
+      "train_tokens_per_second": 148932.838
     },
     {
       "epoch": 0.36780614452617916,
-      "eval_loss": 0.05933792144060135,
-      "eval_runtime": 1.8241,
-      "eval_samples_per_second": 17.543,
-      "eval_steps_per_second": 0.548,
+      "eval_loss": 0.048420779407024384,
+      "eval_runtime": 2.1981,
+      "eval_samples_per_second": 14.558,
+      "eval_steps_per_second": 0.455,
       "num_input_tokens_seen": 111382528,
       "step": 1700
     },
     {
       "epoch": 0.36780614452617916,
-      "eval_byte_accuracy": 0.9821958456973294,
-      "eval_chrf": 85.84022704344234,
-      "eval_sacrebleu": 79.27134876780914,
-      "eval_word_accuracy": 0.9560439560439561,
+      "eval_byte_accuracy": 0.9866468842729971,
+      "eval_chrf": 89.9671333112741,
+      "eval_sacrebleu": 87.01567892940223,
+      "eval_word_accuracy": 0.9711538461538461,
       "num_input_tokens_seen": 111382528,
-      "perplexity": 1.0611337600099267,
+      "perplexity": 1.0496122176243317,
       "step": 1700
     },
     {
       "epoch": 0.36996971008221546,
-      "grad_norm": 0.8984375,
+      "grad_norm": 0.859375,
       "learning_rate": 0.0005236421052631578,
-      "loss": 0.0775,
+      "loss": 0.0638,
       "num_input_tokens_seen": 112037888,
       "step": 1710,
-      "train_runtime": 485.2237,
-      "train_tokens_per_second": 230899.473
+      "train_runtime": 755.9586,
+      "train_tokens_per_second": 148206.383
     },
     {
       "epoch": 0.3721332756382518,
-      "grad_norm": 1.796875,
+      "grad_norm": 0.84765625,
       "learning_rate": 0.0005230105263157894,
-      "loss": 0.0844,
+      "loss": 0.0679,
       "num_input_tokens_seen": 112693248,
       "step": 1720,
-      "train_runtime": 489.0379,
-      "train_tokens_per_second": 230438.671
+      "train_runtime": 759.7618,
+      "train_tokens_per_second": 148327.078
     },
     {
       "epoch": 0.3742968411942882,
-      "grad_norm": 1.5,
+      "grad_norm": 1.4765625,
       "learning_rate": 0.000522378947368421,
-      "loss": 0.0771,
+      "loss": 0.0642,
       "num_input_tokens_seen": 113348608,
       "step": 1730,
-      "train_runtime": 492.8493,
-      "train_tokens_per_second": 229986.364
+      "train_runtime": 763.5726,
+      "train_tokens_per_second": 148445.092
     },
     {
       "epoch": 0.37646040675032455,
-      "grad_norm": 1.046875,
+      "grad_norm": 2.640625,
       "learning_rate": 0.0005217473684210526,
-      "loss": 0.0761,
+      "loss": 0.0628,
       "num_input_tokens_seen": 114003968,
       "step": 1740,
-      "train_runtime": 496.6631,
-      "train_tokens_per_second": 229539.85
+      "train_runtime": 767.3829,
+      "train_tokens_per_second": 148562.035
     },
     {
       "epoch": 0.37862397230636086,
-      "grad_norm": 1.4375,
+      "grad_norm": 0.73828125,
       "learning_rate": 0.0005211157894736842,
-      "loss": 0.0772,
+      "loss": 0.0647,
       "num_input_tokens_seen": 114659328,
       "step": 1750,
-      "train_runtime": 500.4771,
-      "train_tokens_per_second": 229100.028
+      "train_runtime": 771.2005,
+      "train_tokens_per_second": 148676.419
     },
     {
       "epoch": 0.3807875378623972,
-      "grad_norm": 0.84765625,
+      "grad_norm": 0.953125,
       "learning_rate": 0.0005204842105263158,
-      "loss": 0.0767,
+      "loss": 0.0594,
       "num_input_tokens_seen": 115314688,
       "step": 1760,
-      "train_runtime": 504.2884,
-      "train_tokens_per_second": 228668.122
+      "train_runtime": 777.1082,
+      "train_tokens_per_second": 148389.491
     },
     {
       "epoch": 0.3829511034184336,
-      "grad_norm": 1.1484375,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.0005198526315789474,
-      "loss": 0.074,
+      "loss": 0.0599,
       "num_input_tokens_seen": 115970048,
       "step": 1770,
-      "train_runtime": 508.1019,
-      "train_tokens_per_second": 228241.718
+      "train_runtime": 780.9251,
+      "train_tokens_per_second": 148503.426
     },
     {
       "epoch": 0.38511466897446994,
-      "grad_norm": 1.2578125,
+      "grad_norm": 0.9140625,
       "learning_rate": 0.0005192210526315789,
-      "loss": 0.0737,
+      "loss": 0.0584,
       "num_input_tokens_seen": 116625408,
       "step": 1780,
-      "train_runtime": 511.9131,
-      "train_tokens_per_second": 227822.652
+      "train_runtime": 784.7363,
+      "train_tokens_per_second": 148617.319
     },
     {
       "epoch": 0.38727823453050625,
-      "grad_norm": 1.2890625,
+      "grad_norm": 0.81640625,
       "learning_rate": 0.0005185894736842105,
-      "loss": 0.0741,
+      "loss": 0.0607,
       "num_input_tokens_seen": 117280768,
       "step": 1790,
-      "train_runtime": 515.7164,
-      "train_tokens_per_second": 227413.314
+      "train_runtime": 788.5519,
+      "train_tokens_per_second": 148729.304
     },
     {
       "epoch": 0.3894418000865426,
-      "grad_norm": 1.203125,
+      "grad_norm": 1.09375,
       "learning_rate": 0.0005179578947368421,
-      "loss": 0.0703,
+      "loss": 0.0586,
       "num_input_tokens_seen": 117936128,
       "step": 1800,
-      "train_runtime": 519.5147,
-      "train_tokens_per_second": 227012.128
+      "train_runtime": 792.3423,
+      "train_tokens_per_second": 148844.921
     },
     {
       "epoch": 0.3894418000865426,
-      "eval_loss": 0.05443021282553673,
-      "eval_runtime": 1.9889,
-      "eval_samples_per_second": 16.089,
-      "eval_steps_per_second": 0.503,
+      "eval_loss": 0.044286634773015976,
+      "eval_runtime": 2.2358,
+      "eval_samples_per_second": 14.312,
+      "eval_steps_per_second": 0.447,
       "num_input_tokens_seen": 117936128,
       "step": 1800
     },
     {
       "epoch": 0.3894418000865426,
-      "eval_byte_accuracy": 0.9847922848664689,
-      "eval_chrf": 87.69637221255437,
-      "eval_sacrebleu": 84.14156310397632,
-      "eval_word_accuracy": 0.9642857142857143,
+      "eval_byte_accuracy": 0.9892433234421365,
+      "eval_chrf": 91.74603528929435,
+      "eval_sacrebleu": 87.57438387373494,
+      "eval_word_accuracy": 0.9711538461538461,
       "num_input_tokens_seen": 117936128,
-      "perplexity": 1.0559387828582605,
+      "perplexity": 1.045281926100255,
       "step": 1800
     },
     {
       "epoch": 0.391605365642579,
-      "grad_norm": 1.40625,
+      "grad_norm": 1.1328125,
       "learning_rate": 0.0005173263157894736,
-      "loss": 0.0786,
+      "loss": 0.0646,
       "num_input_tokens_seen": 118591488,
       "step": 1810,
-      "train_runtime": 525.3311,
-      "train_tokens_per_second": 225746.181
+      "train_runtime": 800.6612,
+      "train_tokens_per_second": 148116.933
     },
     {
       "epoch": 0.39376893119861534,
-      "grad_norm": 0.67578125,
+      "grad_norm": 1.140625,
       "learning_rate": 0.0005166947368421052,
-      "loss": 0.0729,
+      "loss": 0.0613,
       "num_input_tokens_seen": 119246848,
       "step": 1820,
-      "train_runtime": 529.1476,
-      "train_tokens_per_second": 225356.508
+      "train_runtime": 804.478,
+      "train_tokens_per_second": 148228.853
     },
     {
       "epoch": 0.39593249675465164,
-      "grad_norm": 1.1640625,
+      "grad_norm": 1.125,
       "learning_rate": 0.0005160631578947368,
-      "loss": 0.0694,
+      "loss": 0.0591,
       "num_input_tokens_seen": 119902208,
       "step": 1830,
-      "train_runtime": 532.9676,
-      "train_tokens_per_second": 224970.907
+      "train_runtime": 808.2907,
+      "train_tokens_per_second": 148340.459
     },
     {
       "epoch": 0.398096062310688,
-      "grad_norm": 0.8984375,
+      "grad_norm": 1.0703125,
       "learning_rate": 0.0005154315789473684,
-      "loss": 0.0746,
+      "loss": 0.0627,
       "num_input_tokens_seen": 120557568,
       "step": 1840,
-      "train_runtime": 536.7692,
-      "train_tokens_per_second": 224598.51
+      "train_runtime": 812.0918,
+      "train_tokens_per_second": 148453.136
     },
     {
       "epoch": 0.40025962786672437,
-      "grad_norm": 1.2734375,
+      "grad_norm": 1.2890625,
       "learning_rate": 0.0005147999999999999,
-      "loss": 0.0704,
+      "loss": 0.0585,
       "num_input_tokens_seen": 121212928,
       "step": 1850,
-      "train_runtime": 540.578,
-      "train_tokens_per_second": 224228.364
+      "train_runtime": 815.9044,
+      "train_tokens_per_second": 148562.656
     },
     {
       "epoch": 0.4024231934227607,
-      "grad_norm": 1.6640625,
+      "grad_norm": 0.83203125,
       "learning_rate": 0.0005141684210526315,
-      "loss": 0.0699,
+      "loss": 0.0587,
       "num_input_tokens_seen": 121868288,
       "step": 1860,
-      "train_runtime": 544.3824,
-      "train_tokens_per_second": 223865.241
+      "train_runtime": 821.7746,
+      "train_tokens_per_second": 148298.931
     },
     {
       "epoch": 0.40458675897879703,
-      "grad_norm": 1.0625,
+      "grad_norm": 0.84375,
       "learning_rate": 0.0005135368421052631,
-      "loss": 0.072,
+      "loss": 0.0587,
       "num_input_tokens_seen": 122523648,
       "step": 1870,
-      "train_runtime": 548.1984,
-      "train_tokens_per_second": 223502.362
+      "train_runtime": 825.5838,
+      "train_tokens_per_second": 148408.498
     },
     {
       "epoch": 0.4067503245348334,
-      "grad_norm": 1.28125,
+      "grad_norm": 1.484375,
       "learning_rate": 0.0005129052631578947,
-      "loss": 0.0701,
+      "loss": 0.0578,
       "num_input_tokens_seen": 123179008,
       "step": 1880,
-      "train_runtime": 552.006,
-      "train_tokens_per_second": 223147.947
+      "train_runtime": 829.3992,
+      "train_tokens_per_second": 148515.94
     },
     {
       "epoch": 0.40891389009086976,
-      "grad_norm": 1.0546875,
+      "grad_norm": 1.3046875,
       "learning_rate": 0.0005122736842105263,
-      "loss": 0.0673,
+      "loss": 0.0583,
       "num_input_tokens_seen": 123834368,
       "step": 1890,
-      "train_runtime": 555.8224,
-      "train_tokens_per_second": 222794.848
+      "train_runtime": 833.208,
+      "train_tokens_per_second": 148623.591
     },
     {
       "epoch": 0.4110774556469061,
-      "grad_norm": 1.3046875,
+      "grad_norm": 1.0078125,
       "learning_rate": 0.0005116421052631579,
-      "loss": 0.074,
+      "loss": 0.0633,
       "num_input_tokens_seen": 124489728,
       "step": 1900,
-      "train_runtime": 559.6214,
-      "train_tokens_per_second": 222453.48
+      "train_runtime": 837.0084,
+      "train_tokens_per_second": 148731.755
     },
     {
       "epoch": 0.4110774556469061,
-      "eval_loss": 0.05228148028254509,
-      "eval_runtime": 1.8822,
-      "eval_samples_per_second": 17.002,
-      "eval_steps_per_second": 0.531,
+      "eval_loss": 0.043784890323877335,
+      "eval_runtime": 2.1418,
+      "eval_samples_per_second": 14.94,
+      "eval_steps_per_second": 0.467,
       "num_input_tokens_seen": 124489728,
       "step": 1900
     },
     {
       "epoch": 0.4110774556469061,
-      "eval_byte_accuracy": 0.9888724035608308,
-      "eval_chrf": 91.12158224407392,
-      "eval_sacrebleu": 89.04307319092496,
+      "eval_byte_accuracy": 0.987759643916914,
+      "eval_chrf": 92.28593517149247,
+      "eval_sacrebleu": 88.12775358804903,
       "eval_word_accuracy": 0.9739010989010989,
       "num_input_tokens_seen": 124489728,
-      "perplexity": 1.0536722887489947,
+      "perplexity": 1.0447575932475983,
       "step": 1900
     },
     {
       "epoch": 0.4132410212029424,
-      "grad_norm": 0.9453125,
+      "grad_norm": 0.9375,
       "learning_rate": 0.0005110105263157895,
-      "loss": 0.069,
+      "loss": 0.0548,
       "num_input_tokens_seen": 125145088,
       "step": 1910,
-      "train_runtime": 565.3393,
-      "train_tokens_per_second": 221362.787
+      "train_runtime": 845.0802,
+      "train_tokens_per_second": 148086.646
     },
     {
       "epoch": 0.4154045867589788,
-      "grad_norm": 1.65625,
+      "grad_norm": 0.7265625,
       "learning_rate": 0.000510378947368421,
-      "loss": 0.0655,
+      "loss": 0.0515,
       "num_input_tokens_seen": 125800448,
       "step": 1920,
-      "train_runtime": 569.151,
-      "train_tokens_per_second": 221031.744
+      "train_runtime": 848.8975,
+      "train_tokens_per_second": 148192.739
     },
     {
       "epoch": 0.41756815231501515,
-      "grad_norm": 1.265625,
+      "grad_norm": 1.171875,
       "learning_rate": 0.0005097473684210526,
-      "loss": 0.0741,
+      "loss": 0.0573,
       "num_input_tokens_seen": 126455808,
       "step": 1930,
-      "train_runtime": 572.9578,
-      "train_tokens_per_second": 220707.034
+      "train_runtime": 852.6999,
+      "train_tokens_per_second": 148300.488
     },
     {
       "epoch": 0.4197317178710515,
-      "grad_norm": 0.90625,
+      "grad_norm": 0.7734375,
       "learning_rate": 0.0005091157894736842,
-      "loss": 0.0697,
+      "loss": 0.0562,
       "num_input_tokens_seen": 127111168,
       "step": 1940,
-      "train_runtime": 576.7661,
-      "train_tokens_per_second": 220385.973
+      "train_runtime": 856.5073,
+      "train_tokens_per_second": 148406.397
     },
     {
       "epoch": 0.4218952834270878,
-      "grad_norm": 0.78515625,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.0005084842105263157,
-      "loss": 0.0714,
+      "loss": 0.0565,
       "num_input_tokens_seen": 127766528,
       "step": 1950,
-      "train_runtime": 580.5827,
-      "train_tokens_per_second": 220066.025
+      "train_runtime": 860.3189,
+      "train_tokens_per_second": 148510.659
     },
     {
       "epoch": 0.4240588489831242,
-      "grad_norm": 1.1953125,
+      "grad_norm": 0.71875,
       "learning_rate": 0.0005078526315789474,
-      "loss": 0.0673,
+      "loss": 0.0509,
       "num_input_tokens_seen": 128421888,
       "step": 1960,
-      "train_runtime": 584.3962,
-      "train_tokens_per_second": 219751.41
+      "train_runtime": 866.227,
+      "train_tokens_per_second": 148254.317
     },
     {
       "epoch": 0.42622241453916054,
-      "grad_norm": 1.1640625,
+      "grad_norm": 0.99609375,
       "learning_rate": 0.0005072210526315789,
-      "loss": 0.0672,
+      "loss": 0.0567,
       "num_input_tokens_seen": 129077248,
       "step": 1970,
-      "train_runtime": 588.2047,
-      "train_tokens_per_second": 219442.719
+      "train_runtime": 870.0425,
+      "train_tokens_per_second": 148357.399
     },
     {
       "epoch": 0.4283859800951969,
-      "grad_norm": 1.0234375,
+      "grad_norm": 0.80859375,
       "learning_rate": 0.0005065894736842105,
-      "loss": 0.0721,
+      "loss": 0.0578,
       "num_input_tokens_seen": 129732608,
       "step": 1980,
-      "train_runtime": 592.02,
-      "train_tokens_per_second": 219135.502
+      "train_runtime": 873.8588,
+      "train_tokens_per_second": 148459.456
     },
     {
       "epoch": 0.4305495456512332,
-      "grad_norm": 0.8671875,
+      "grad_norm": 0.78515625,
       "learning_rate": 0.000505957894736842,
-      "loss": 0.0687,
+      "loss": 0.0546,
       "num_input_tokens_seen": 130387968,
       "step": 1990,
-      "train_runtime": 595.8415,
-      "train_tokens_per_second": 218829.967
+      "train_runtime": 877.6726,
+      "train_tokens_per_second": 148561.062
     },
     {
       "epoch": 0.4327131112072696,
-      "grad_norm": 1.828125,
+      "grad_norm": 1.0625,
       "learning_rate": 0.0005053263157894736,
-      "loss": 0.067,
+      "loss": 0.0545,
       "num_input_tokens_seen": 131043328,
       "step": 2000,
-      "train_runtime": 599.6411,
-      "train_tokens_per_second": 218536.272
+      "train_runtime": 881.4783,
+      "train_tokens_per_second": 148663.139
     },
     {
       "epoch": 0.4327131112072696,
-      "eval_loss": 0.056328125298023224,
-      "eval_runtime": 2.0705,
-      "eval_samples_per_second": 15.455,
-      "eval_steps_per_second": 0.483,
+      "eval_loss": 0.038986582309007645,
+      "eval_runtime": 2.2975,
+      "eval_samples_per_second": 13.928,
+      "eval_steps_per_second": 0.435,
       "num_input_tokens_seen": 131043328,
       "step": 2000
     },
     {
       "epoch": 0.4327131112072696,
-      "eval_byte_accuracy": 0.9862759643916914,
-      "eval_chrf": 88.37329071319488,
-      "eval_sacrebleu": 86.52596851533677,
-      "eval_word_accuracy": 0.9684065934065934,
+      "eval_byte_accuracy": 0.9873887240356083,
+      "eval_chrf": 89.9749087738313,
+      "eval_sacrebleu": 86.45994331005899,
+      "eval_word_accuracy": 0.9697802197802198,
       "num_input_tokens_seen": 131043328,
-      "perplexity": 1.0579447652317715,
+      "perplexity": 1.039756532424702,
       "step": 2000
     },
     {
       "epoch": 0.43487667676330594,
-      "grad_norm": 1.7421875,
+      "grad_norm": 0.90625,
       "learning_rate": 0.0005046947368421052,
-      "loss": 0.0674,
+      "loss": 0.0533,
       "num_input_tokens_seen": 131698688,
       "step": 2010,
-      "train_runtime": 605.5878,
-      "train_tokens_per_second": 217472.475
+      "train_runtime": 889.92,
+      "train_tokens_per_second": 147989.351
     },
     {
       "epoch": 0.4370402423193423,
-      "grad_norm": 1.03125,
+      "grad_norm": 0.79296875,
       "learning_rate": 0.0005040631578947368,
-      "loss": 0.0645,
+      "loss": 0.052,
       "num_input_tokens_seen": 132354048,
       "step": 2020,
-      "train_runtime": 609.4048,
-      "train_tokens_per_second": 217185.764
+      "train_runtime": 893.7334,
+      "train_tokens_per_second": 148091.201
     },
     {
       "epoch": 0.4392038078753786,
-      "grad_norm": 0.94140625,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.0005034315789473684,
-      "loss": 0.064,
+      "loss": 0.0495,
       "num_input_tokens_seen": 133009408,
       "step": 2030,
-      "train_runtime": 613.218,
-      "train_tokens_per_second": 216903.949
+      "train_runtime": 897.5434,
+      "train_tokens_per_second": 148192.726
     },
     {
       "epoch": 0.44136737343141497,
-      "grad_norm": 1.21875,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.0005028,
-      "loss": 0.0673,
+      "loss": 0.0528,
       "num_input_tokens_seen": 133664768,
       "step": 2040,
-      "train_runtime": 617.0395,
-      "train_tokens_per_second": 216622.719
+      "train_runtime": 901.3573,
+      "train_tokens_per_second": 148292.767
     },
     {
       "epoch": 0.44353093898745133,
-      "grad_norm": 0.9140625,
+      "grad_norm": 1.078125,
       "learning_rate": 0.0005021684210526316,
-      "loss": 0.0627,
+      "loss": 0.0521,
       "num_input_tokens_seen": 134320128,
       "step": 2050,
-      "train_runtime": 620.8539,
-      "train_tokens_per_second": 216347.415
+      "train_runtime": 905.1614,
+      "train_tokens_per_second": 148393.559
     },
     {
       "epoch": 0.4456945045434877,
-      "grad_norm": 1.2421875,
+      "grad_norm": 0.8515625,
       "learning_rate": 0.0005015368421052631,
-      "loss": 0.0622,
+      "loss": 0.0462,
       "num_input_tokens_seen": 134975488,
       "step": 2060,
-      "train_runtime": 624.6588,
-      "train_tokens_per_second": 216078.733
+      "train_runtime": 911.0618,
+      "train_tokens_per_second": 148151.843
     },
     {
       "epoch": 0.447858070099524,
-      "grad_norm": 1.0390625,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.0005009052631578947,
-      "loss": 0.0686,
+      "loss": 0.0527,
       "num_input_tokens_seen": 135630848,
       "step": 2070,
-      "train_runtime": 628.4714,
-      "train_tokens_per_second": 215810.681
+      "train_runtime": 914.8729,
+      "train_tokens_per_second": 148251.033
     },
     {
       "epoch": 0.45002163565556036,
-      "grad_norm": 0.91015625,
+      "grad_norm": 1.0234375,
       "learning_rate": 0.0005002736842105263,
-      "loss": 0.0657,
+      "loss": 0.0515,
       "num_input_tokens_seen": 136286208,
       "step": 2080,
-      "train_runtime": 632.28,
-      "train_tokens_per_second": 215547.249
+      "train_runtime": 918.6826,
+      "train_tokens_per_second": 148349.607
     },
     {
       "epoch": 0.4521852012115967,
-      "grad_norm": 1.7109375,
+      "grad_norm": 0.95703125,
       "learning_rate": 0.0004996421052631578,
-      "loss": 0.0664,
+      "loss": 0.052,
       "num_input_tokens_seen": 136941568,
       "step": 2090,
-      "train_runtime": 636.0935,
-      "train_tokens_per_second": 215285.273
+      "train_runtime": 922.4928,
+      "train_tokens_per_second": 148447.301
     },
     {
       "epoch": 0.4543487667676331,
-      "grad_norm": 1.1328125,
+      "grad_norm": 1.125,
       "learning_rate": 0.0004990105263157895,
-      "loss": 0.0735,
+      "loss": 0.0549,
       "num_input_tokens_seen": 137596928,
       "step": 2100,
-      "train_runtime": 639.9004,
-      "train_tokens_per_second": 215028.649
+      "train_runtime": 926.3048,
+      "train_tokens_per_second": 148543.903
     },
     {
       "epoch": 0.4543487667676331,
-      "eval_loss": 0.05029554292559624,
-      "eval_runtime": 1.8825,
-      "eval_samples_per_second": 16.998,
-      "eval_steps_per_second": 0.531,
+      "eval_loss": 0.039272863417863846,
+      "eval_runtime": 2.2223,
+      "eval_samples_per_second": 14.4,
+      "eval_steps_per_second": 0.45,
       "num_input_tokens_seen": 137596928,
       "step": 2100
     },
     {
       "epoch": 0.4543487667676331,
-      "eval_byte_accuracy": 0.9866468842729971,
-      "eval_chrf": 88.29631077643761,
-      "eval_sacrebleu": 84.17696197936166,
-      "eval_word_accuracy": 0.9642857142857143,
+      "eval_byte_accuracy": 0.9885014836795252,
+      "eval_chrf": 90.0025148718611,
+      "eval_sacrebleu": 87.5087779219263,
+      "eval_word_accuracy": 0.9711538461538461,
       "num_input_tokens_seen": 137596928,
-      "perplexity": 1.0515818380279285,
+      "perplexity": 1.0400542376894084,
       "step": 2100
     },
     {
       "epoch": 0.4565123323236694,
-      "grad_norm": 0.9609375,
+      "grad_norm": 1.15625,
       "learning_rate": 0.000498378947368421,
-      "loss": 0.0646,
+      "loss": 0.0519,
       "num_input_tokens_seen": 138252288,
       "step": 2110,
-      "train_runtime": 645.6128,
-      "train_tokens_per_second": 214141.172
+      "train_runtime": 934.4617,
+      "train_tokens_per_second": 147948.586
     },
     {
       "epoch": 0.45867589787970575,
-      "grad_norm": 1.2109375,
+      "grad_norm": 0.8125,
       "learning_rate": 0.0004977473684210526,
-      "loss": 0.066,
+      "loss": 0.0526,
       "num_input_tokens_seen": 138907648,
       "step": 2120,
-      "train_runtime": 649.4219,
-      "train_tokens_per_second": 213894.317
+      "train_runtime": 938.2686,
+      "train_tokens_per_second": 148046.777
     },
     {
       "epoch": 0.4608394634357421,
-      "grad_norm": 1.1640625,
+      "grad_norm": 1.109375,
       "learning_rate": 0.0004971157894736841,
-      "loss": 0.0638,
+      "loss": 0.0519,
       "num_input_tokens_seen": 139563008,
       "step": 2130,
-      "train_runtime": 653.2367,
-      "train_tokens_per_second": 213648.434
+      "train_runtime": 942.0767,
+      "train_tokens_per_second": 148143.991
     },
     {
       "epoch": 0.4630030289917785,
-      "grad_norm": 0.80078125,
+      "grad_norm": 0.578125,
       "learning_rate": 0.0004964842105263157,
-      "loss": 0.0636,
+      "loss": 0.0491,
       "num_input_tokens_seen": 140218368,
       "step": 2140,
-      "train_runtime": 657.0303,
-      "train_tokens_per_second": 213412.331
+      "train_runtime": 945.8626,
+      "train_tokens_per_second": 148243.903
     },
     {
       "epoch": 0.4651665945478148,
-      "grad_norm": 1.6796875,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.0004958526315789473,
-      "loss": 0.0626,
+      "loss": 0.0491,
       "num_input_tokens_seen": 140873728,
       "step": 2150,
-      "train_runtime": 660.8452,
-      "train_tokens_per_second": 213172.068
+      "train_runtime": 949.6741,
+      "train_tokens_per_second": 148339.015
     },
     {
       "epoch": 0.46733016010385114,
-      "grad_norm": 1.2890625,
+      "grad_norm": 0.640625,
       "learning_rate": 0.0004952210526315789,
-      "loss": 0.0634,
+      "loss": 0.0507,
       "num_input_tokens_seen": 141529088,
       "step": 2160,
-      "train_runtime": 664.6606,
-      "train_tokens_per_second": 212934.373
+      "train_runtime": 955.6014,
+      "train_tokens_per_second": 148104.727
     },
     {
       "epoch": 0.4694937256598875,
-      "grad_norm": 1.3203125,
+      "grad_norm": 1.0703125,
       "learning_rate": 0.0004945894736842105,
-      "loss": 0.0607,
+      "loss": 0.0513,
       "num_input_tokens_seen": 142184448,
       "step": 2170,
-      "train_runtime": 668.4544,
-      "train_tokens_per_second": 212706.285
+      "train_runtime": 959.4008,
+      "train_tokens_per_second": 148201.3
     },
     {
       "epoch": 0.47165729121592387,
-      "grad_norm": 0.9375,
+      "grad_norm": 0.98828125,
       "learning_rate": 0.0004939578947368421,
-      "loss": 0.062,
+      "loss": 0.0546,
       "num_input_tokens_seen": 142839808,
       "step": 2180,
-      "train_runtime": 672.2623,
-      "train_tokens_per_second": 212476.311
+      "train_runtime": 963.2144,
+      "train_tokens_per_second": 148294.929
     },
     {
       "epoch": 0.4738208567719602,
-      "grad_norm": 1.4765625,
+      "grad_norm": 0.7578125,
       "learning_rate": 0.0004933263157894737,
-      "loss": 0.0611,
+      "loss": 0.0487,
       "num_input_tokens_seen": 143495168,
       "step": 2190,
-      "train_runtime": 676.0777,
-      "train_tokens_per_second": 212246.565
+      "train_runtime": 967.0254,
+      "train_tokens_per_second": 148388.209
     },
     {
       "epoch": 0.47598442232799654,
-      "grad_norm": 0.921875,
+      "grad_norm": 0.9140625,
       "learning_rate": 0.0004926947368421052,
-      "loss": 0.06,
+      "loss": 0.0505,
       "num_input_tokens_seen": 144150528,
       "step": 2200,
-      "train_runtime": 679.8922,
-      "train_tokens_per_second": 212019.69
+      "train_runtime": 970.8322,
+      "train_tokens_per_second": 148481.401
     },
     {
       "epoch": 0.47598442232799654,
-      "eval_loss": 0.044721540063619614,
-      "eval_runtime": 1.9205,
-      "eval_samples_per_second": 16.663,
-      "eval_steps_per_second": 0.521,
+      "eval_loss": 0.03245285898447037,
+      "eval_runtime": 2.3707,
+      "eval_samples_per_second": 13.498,
+      "eval_steps_per_second": 0.422,
       "num_input_tokens_seen": 144150528,
       "step": 2200
     },
     {
       "epoch": 0.47598442232799654,
-      "eval_byte_accuracy": 0.9896142433234422,
-      "eval_chrf": 92.35975049219142,
-      "eval_sacrebleu": 89.92764168467501,
-      "eval_word_accuracy": 0.9752747252747253,
+      "eval_byte_accuracy": 0.9903560830860534,
+      "eval_chrf": 92.28294859196265,
+      "eval_sacrebleu": 89.4412322999391,
+      "eval_word_accuracy": 0.9766483516483516,
       "num_input_tokens_seen": 144150528,
-      "perplexity": 1.0457366236080974,
+      "perplexity": 1.032985196024998,
       "step": 2200
     },
     {
       "epoch": 0.4781479878840329,
-      "grad_norm": 1.0390625,
+      "grad_norm": 0.87890625,
       "learning_rate": 0.0004920631578947368,
-      "loss": 0.0589,
+      "loss": 0.0504,
       "num_input_tokens_seen": 144805888,
       "step": 2210,
-      "train_runtime": 685.64,
-      "train_tokens_per_second": 211198.131
+      "train_runtime": 979.3143,
+      "train_tokens_per_second": 147864.564
     },
     {
       "epoch": 0.48031155344006926,
-      "grad_norm": 0.91796875,
+      "grad_norm": 0.74609375,
       "learning_rate": 0.0004914315789473684,
-      "loss": 0.0596,
+      "loss": 0.0489,
       "num_input_tokens_seen": 145461248,
       "step": 2220,
-      "train_runtime": 689.4556,
-      "train_tokens_per_second": 210979.865
+      "train_runtime": 983.1232,
+      "train_tokens_per_second": 147958.314
     },
     {
       "epoch": 0.48247511899610557,
-      "grad_norm": 0.94140625,
+      "grad_norm": 0.546875,
       "learning_rate": 0.0004907999999999999,
-      "loss": 0.0605,
+      "loss": 0.0493,
       "num_input_tokens_seen": 146116608,
       "step": 2230,
-      "train_runtime": 693.2702,
-      "train_tokens_per_second": 210764.3
+      "train_runtime": 986.9267,
+      "train_tokens_per_second": 148052.14
     },
     {
       "epoch": 0.48463868455214193,
-      "grad_norm": 1.2421875,
+      "grad_norm": 0.69140625,
       "learning_rate": 0.0004901684210526316,
-      "loss": 0.0596,
+      "loss": 0.0488,
       "num_input_tokens_seen": 146771968,
       "step": 2240,
-      "train_runtime": 697.0767,
-      "train_tokens_per_second": 210553.54
+      "train_runtime": 990.7291,
+      "train_tokens_per_second": 148145.403
     },
     {
       "epoch": 0.4868022501081783,
-      "grad_norm": 1.359375,
+      "grad_norm": 0.87109375,
       "learning_rate": 0.0004895368421052631,
-      "loss": 0.0624,
+      "loss": 0.0489,
       "num_input_tokens_seen": 147427328,
       "step": 2250,
-      "train_runtime": 700.8957,
-      "train_tokens_per_second": 210341.331
+      "train_runtime": 994.5395,
+      "train_tokens_per_second": 148236.777
     },
     {
       "epoch": 0.4889658156642146,
-      "grad_norm": 0.8671875,
+      "grad_norm": 0.69921875,
       "learning_rate": 0.0004889052631578948,
-      "loss": 0.0576,
+      "loss": 0.0454,
       "num_input_tokens_seen": 148074496,
       "step": 2260,
-      "train_runtime": 704.6655,
-      "train_tokens_per_second": 210134.435
+      "train_runtime": 1000.4269,
+      "train_tokens_per_second": 148011.307
     },
     {
       "epoch": 0.49112938122025096,
-      "grad_norm": 0.8515625,
+      "grad_norm": 0.66796875,
       "learning_rate": 0.00048827368421052624,
-      "loss": 0.0554,
+      "loss": 0.0432,
       "num_input_tokens_seen": 148729856,
       "step": 2270,
-      "train_runtime": 708.4862,
-      "train_tokens_per_second": 209926.25
+      "train_runtime": 1004.2395,
+      "train_tokens_per_second": 148101.981
     },
     {
       "epoch": 0.4932929467762873,
-      "grad_norm": 1.46875,
+      "grad_norm": 0.90234375,
       "learning_rate": 0.0004876421052631579,
-      "loss": 0.0587,
+      "loss": 0.0466,
       "num_input_tokens_seen": 149385216,
       "step": 2280,
-      "train_runtime": 712.2958,
-      "train_tokens_per_second": 209723.561
+      "train_runtime": 1008.0401,
+      "train_tokens_per_second": 148193.725
     },
     {
       "epoch": 0.4954565123323237,
-      "grad_norm": 0.89453125,
+      "grad_norm": 0.8515625,
       "learning_rate": 0.0004870105263157894,
-      "loss": 0.0612,
+      "loss": 0.0506,
       "num_input_tokens_seen": 150040576,
       "step": 2290,
-      "train_runtime": 716.1094,
-      "train_tokens_per_second": 209521.872
+      "train_runtime": 1011.8488,
+      "train_tokens_per_second": 148283.587
     },
     {
       "epoch": 0.49762007788836,
-      "grad_norm": 1.109375,
+      "grad_norm": 0.45703125,
       "learning_rate": 0.000486378947368421,
-      "loss": 0.0593,
+      "loss": 0.0476,
       "num_input_tokens_seen": 150695936,
       "step": 2300,
-      "train_runtime": 719.9311,
-      "train_tokens_per_second": 209319.935
+      "train_runtime": 1015.6644,
+      "train_tokens_per_second": 148371.775
     },
     {
       "epoch": 0.49762007788836,
-      "eval_loss": 0.046136274933815,
-      "eval_runtime": 2.1122,
-      "eval_samples_per_second": 15.15,
-      "eval_steps_per_second": 0.473,
+      "eval_loss": 0.035914402455091476,
+      "eval_runtime": 1.9959,
+      "eval_samples_per_second": 16.032,
+      "eval_steps_per_second": 0.501,
       "num_input_tokens_seen": 150695936,
       "step": 2300
     },
     {
       "epoch": 0.49762007788836,
-      "eval_byte_accuracy": 0.9881305637982196,
-      "eval_chrf": 91.02875313945079,
-      "eval_sacrebleu": 88.18216573136273,
-      "eval_word_accuracy": 0.9725274725274725,
+      "eval_byte_accuracy": 0.9888724035608308,
+      "eval_chrf": 92.1404853165006,
+      "eval_sacrebleu": 90.1751620485485,
+      "eval_word_accuracy": 0.9752747252747253,
       "num_input_tokens_seen": 150695936,
-      "perplexity": 1.0472171106759662,
+      "perplexity": 1.036567115093198,
       "step": 2300
     },
     {
       "epoch": 0.49978364344439635,
-      "grad_norm": 0.84765625,
+      "grad_norm": 0.86328125,
       "learning_rate": 0.0004857473684210526,
-      "loss": 0.0572,
+      "loss": 0.0472,
       "num_input_tokens_seen": 151351296,
       "step": 2310,
-      "train_runtime": 725.8732,
-      "train_tokens_per_second": 208509.263
+      "train_runtime": 1023.6035,
+      "train_tokens_per_second": 147861.259
     },
     {
       "epoch": 0.5019472090004327,
-      "grad_norm": 0.83203125,
+      "grad_norm": 0.71484375,
       "learning_rate": 0.0004851157894736842,
-      "loss": 0.0607,
+      "loss": 0.0497,
       "num_input_tokens_seen": 152006656,
       "step": 2320,
-      "train_runtime": 729.6707,
-      "train_tokens_per_second": 208322.262
+      "train_runtime": 1027.3926,
+      "train_tokens_per_second": 147953.823
     },
     {
       "epoch": 0.504110774556469,
-      "grad_norm": 1.65625,
+      "grad_norm": 1.0859375,
       "learning_rate": 0.00048448421052631576,
-      "loss": 0.0657,
+      "loss": 0.0512,
       "num_input_tokens_seen": 152662016,
       "step": 2330,
-      "train_runtime": 733.4793,
-      "train_tokens_per_second": 208134.061
+      "train_runtime": 1031.1996,
+      "train_tokens_per_second": 148043.133
     },
     {
       "epoch": 0.5062743401125054,
-      "grad_norm": 0.81640625,
+      "grad_norm": 1.296875,
       "learning_rate": 0.0004838526315789473,
-      "loss": 0.057,
+      "loss": 0.0481,
       "num_input_tokens_seen": 153317376,
       "step": 2340,
-      "train_runtime": 737.291,
-      "train_tokens_per_second": 207946.884
+      "train_runtime": 1035.0087,
+      "train_tokens_per_second": 148131.485
     },
     {
       "epoch": 0.5084379056685417,
-      "grad_norm": 0.984375,
+      "grad_norm": 0.8828125,
       "learning_rate": 0.00048322105263157893,
-      "loss": 0.0573,
+      "loss": 0.0474,
       "num_input_tokens_seen": 153968640,
       "step": 2350,
-      "train_runtime": 741.0801,
-      "train_tokens_per_second": 207762.487
+      "train_runtime": 1038.8024,
+      "train_tokens_per_second": 148217.443
     },
     {
       "epoch": 0.5106014712245781,
-      "grad_norm": 0.83984375,
+      "grad_norm": 1.0546875,
       "learning_rate": 0.00048258947368421046,
-      "loss": 0.0568,
+      "loss": 0.0468,
       "num_input_tokens_seen": 154624000,
       "step": 2360,
-      "train_runtime": 744.8993,
-      "train_tokens_per_second": 207577.062
+      "train_runtime": 1044.7364,
+      "train_tokens_per_second": 148002.883
     },
     {
       "epoch": 0.5127650367806145,
-      "grad_norm": 0.875,
+      "grad_norm": 0.98046875,
       "learning_rate": 0.00048195789473684205,
-      "loss": 0.059,
+      "loss": 0.0494,
       "num_input_tokens_seen": 155279360,
       "step": 2370,
-      "train_runtime": 748.7137,
-      "train_tokens_per_second": 207394.832
+      "train_runtime": 1048.5496,
+      "train_tokens_per_second": 148089.669
     },
     {
       "epoch": 0.5149286023366508,
-      "grad_norm": 1.1015625,
+      "grad_norm": 1.0546875,
       "learning_rate": 0.00048132631578947364,
-      "loss": 0.0597,
+      "loss": 0.048,
       "num_input_tokens_seen": 155934720,
       "step": 2380,
-      "train_runtime": 752.5173,
-      "train_tokens_per_second": 207217.454
+      "train_runtime": 1052.3567,
+      "train_tokens_per_second": 148176.674
     },
     {
       "epoch": 0.5170921678926872,
-      "grad_norm": 0.8359375,
+      "grad_norm": 0.7890625,
       "learning_rate": 0.0004806947368421052,
-      "loss": 0.0583,
+      "loss": 0.0471,
       "num_input_tokens_seen": 156590080,
       "step": 2390,
-      "train_runtime": 756.326,
-      "train_tokens_per_second": 207040.443
+      "train_runtime": 1056.1663,
+      "train_tokens_per_second": 148262.709
     },
     {
       "epoch": 0.5192557334487234,
-      "grad_norm": 1.359375,
+      "grad_norm": 0.83984375,
       "learning_rate": 0.0004800631578947368,
-      "loss": 0.0588,
+      "loss": 0.0468,
       "num_input_tokens_seen": 157245440,
       "step": 2400,
-      "train_runtime": 760.1418,
-      "train_tokens_per_second": 206863.304
+      "train_runtime": 1059.976,
+      "train_tokens_per_second": 148348.107
     },
     {
       "epoch": 0.5192557334487234,
-      "eval_loss": 0.04235806316137314,
-      "eval_runtime": 1.7624,
-      "eval_samples_per_second": 18.157,
-      "eval_steps_per_second": 0.567,
+      "eval_loss": 0.0342160128057003,
+      "eval_runtime": 2.3939,
+      "eval_samples_per_second": 13.367,
+      "eval_steps_per_second": 0.418,
       "num_input_tokens_seen": 157245440,
       "step": 2400
     },
     {
       "epoch": 0.5192557334487234,
-      "eval_byte_accuracy": 0.9873887240356083,
-      "eval_chrf": 90.22196022995097,
-      "eval_sacrebleu": 86.9543619692272,
-      "eval_word_accuracy": 0.9697802197802198,
+      "eval_byte_accuracy": 0.9888724035608308,
+      "eval_chrf": 91.05569648119452,
+      "eval_sacrebleu": 88.20971713831192,
+      "eval_word_accuracy": 0.9752747252747253,
       "num_input_tokens_seen": 157245440,
-      "perplexity": 1.0432679677071313,
+      "perplexity": 1.0348081143911412,
       "step": 2400
     },
     {
       "epoch": 0.5214192990047598,
-      "grad_norm": 1.3359375,
+      "grad_norm": 1.125,
       "learning_rate": 0.00047943157894736834,
-      "loss": 0.0579,
+      "loss": 0.0463,
       "num_input_tokens_seen": 157900800,
       "step": 2410,
-      "train_runtime": 765.7372,
-      "train_tokens_per_second": 206207.57
+      "train_runtime": 1068.0988,
+      "train_tokens_per_second": 147833.513
     },
     {
       "epoch": 0.5235828645607962,
-      "grad_norm": 1.1875,
+      "grad_norm": 1.28125,
       "learning_rate": 0.0004788,
-      "loss": 0.0588,
+      "loss": 0.0483,
       "num_input_tokens_seen": 158556160,
       "step": 2420,
-      "train_runtime": 769.5519,
-      "train_tokens_per_second": 206036.994
+      "train_runtime": 1071.9094,
+      "train_tokens_per_second": 147919.371
     },
     {
       "epoch": 0.5257464301168325,
-      "grad_norm": 1.25,
+      "grad_norm": 0.7109375,
       "learning_rate": 0.0004781684210526315,
-      "loss": 0.0547,
+      "loss": 0.0472,
       "num_input_tokens_seen": 159211520,
       "step": 2430,
-      "train_runtime": 773.369,
-      "train_tokens_per_second": 205867.465
+      "train_runtime": 1075.7223,
+      "train_tokens_per_second": 148004.294
     },
     {
       "epoch": 0.5279099956728689,
-      "grad_norm": 1.3984375,
+      "grad_norm": 0.8203125,
       "learning_rate": 0.0004775368421052631,
-      "loss": 0.0592,
+      "loss": 0.0492,
       "num_input_tokens_seen": 159866880,
       "step": 2440,
-      "train_runtime": 777.1861,
-      "train_tokens_per_second": 205699.607
+      "train_runtime": 1079.5268,
+      "train_tokens_per_second": 148089.771
     },
     {
       "epoch": 0.5300735612289053,
-      "grad_norm": 1.3125,
+      "grad_norm": 0.59765625,
       "learning_rate": 0.0004769052631578947,
-      "loss": 0.0546,
+      "loss": 0.0464,
       "num_input_tokens_seen": 160522240,
       "step": 2450,
-      "train_runtime": 781.0098,
-      "train_tokens_per_second": 205531.662
+      "train_runtime": 1083.3371,
+      "train_tokens_per_second": 148173.859
     },
     {
       "epoch": 0.5322371267849416,
-      "grad_norm": 0.9375,
+      "grad_norm": 0.703125,
       "learning_rate": 0.0004762736842105263,
-      "loss": 0.0537,
+      "loss": 0.0459,
       "num_input_tokens_seen": 161177600,
       "step": 2460,
-      "train_runtime": 784.8261,
-      "train_tokens_per_second": 205367.271
+      "train_runtime": 1089.2108,
+      "train_tokens_per_second": 147976.498
     },
     {
       "epoch": 0.534400692340978,
-      "grad_norm": 0.75390625,
+      "grad_norm": 0.94140625,
       "learning_rate": 0.00047564210526315786,
-      "loss": 0.0544,
+      "loss": 0.0445,
       "num_input_tokens_seen": 161832960,
       "step": 2470,
-      "train_runtime": 788.6389,
-      "train_tokens_per_second": 205205.391
+      "train_runtime": 1093.0188,
+      "train_tokens_per_second": 148060.54
     },
     {
       "epoch": 0.5365642578970142,
-      "grad_norm": 1.28125,
+      "grad_norm": 0.9453125,
       "learning_rate": 0.0004750105263157894,
-      "loss": 0.0526,
+      "loss": 0.0444,
       "num_input_tokens_seen": 162488320,
       "step": 2480,
-      "train_runtime": 792.463,
-      "train_tokens_per_second": 205042.153
+      "train_runtime": 1096.8243,
+      "train_tokens_per_second": 148144.343
     },
     {
       "epoch": 0.5387278234530506,
-      "grad_norm": 1.3828125,
+      "grad_norm": 0.8828125,
       "learning_rate": 0.00047437894736842103,
-      "loss": 0.058,
+      "loss": 0.0466,
       "num_input_tokens_seen": 163139584,
       "step": 2490,
-      "train_runtime": 796.2564,
-      "train_tokens_per_second": 204883.23
+      "train_runtime": 1100.616,
+      "train_tokens_per_second": 148225.706
     },
     {
       "epoch": 0.540891389009087,
-      "grad_norm": 0.796875,
+      "grad_norm": 0.71484375,
       "learning_rate": 0.00047374736842105257,
-      "loss": 0.0521,
+      "loss": 0.0436,
       "num_input_tokens_seen": 163794944,
       "step": 2500,
-      "train_runtime": 800.0752,
-      "train_tokens_per_second": 204724.427
+      "train_runtime": 1104.4217,
+      "train_tokens_per_second": 148308.341
     },
     {
       "epoch": 0.540891389009087,
-      "eval_loss": 0.0381394699215889,
-      "eval_runtime": 1.7699,
-      "eval_samples_per_second": 18.08,
-      "eval_steps_per_second": 0.565,
+      "eval_loss": 0.03495877608656883,
+      "eval_runtime": 1.9919,
+      "eval_samples_per_second": 16.065,
+      "eval_steps_per_second": 0.502,
       "num_input_tokens_seen": 163794944,
       "step": 2500
     },
     {
       "epoch": 0.540891389009087,
-      "eval_byte_accuracy": 0.9888724035608308,
-      "eval_chrf": 90.06189872438668,
-      "eval_sacrebleu": 86.93698181397612,
-      "eval_word_accuracy": 0.9711538461538461,
+      "eval_byte_accuracy": 0.990727002967359,
+      "eval_chrf": 94.08434578822838,
+      "eval_sacrebleu": 91.39427252834908,
+      "eval_word_accuracy": 0.9807692307692307,
       "num_input_tokens_seen": 163794944,
-      "perplexity": 1.0388761147451957,
+      "perplexity": 1.0355770173823597,
       "step": 2500
     },
     {
       "epoch": 0.5430549545651233,
-      "grad_norm": 0.91796875,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.00047311578947368415,
-      "loss": 0.056,
+      "loss": 0.0461,
       "num_input_tokens_seen": 164450304,
       "step": 2510,
-      "train_runtime": 805.7206,
-      "train_tokens_per_second": 204103.396
+      "train_runtime": 1112.3764,
+      "train_tokens_per_second": 147836.922
     },
     {
       "epoch": 0.5452185201211597,
-      "grad_norm": 0.9140625,
+      "grad_norm": 0.73046875,
       "learning_rate": 0.00047248421052631574,
-      "loss": 0.0519,
+      "loss": 0.0425,
       "num_input_tokens_seen": 165105664,
       "step": 2520,
-      "train_runtime": 809.5424,
-      "train_tokens_per_second": 203949.384
+      "train_runtime": 1116.5491,
+      "train_tokens_per_second": 147871.39
     },
     {
       "epoch": 0.547382085677196,
-      "grad_norm": 0.953125,
+      "grad_norm": 1.09375,
       "learning_rate": 0.0004718526315789473,
-      "loss": 0.0548,
+      "loss": 0.0446,
       "num_input_tokens_seen": 165761024,
       "step": 2530,
-      "train_runtime": 813.357,
-      "train_tokens_per_second": 203798.62
+      "train_runtime": 1120.3597,
+      "train_tokens_per_second": 147953.403
     },
     {
       "epoch": 0.5495456512332324,
-      "grad_norm": 0.83203125,
+      "grad_norm": 0.8125,
       "learning_rate": 0.0004712210526315789,
-      "loss": 0.0538,
+      "loss": 0.0441,
       "num_input_tokens_seen": 166412288,
       "step": 2540,
-      "train_runtime": 817.142,
-      "train_tokens_per_second": 203651.614
+      "train_runtime": 1124.1458,
+      "train_tokens_per_second": 148034.431
     },
     {
       "epoch": 0.5517092167892688,
-      "grad_norm": 0.6171875,
+      "grad_norm": 0.6796875,
       "learning_rate": 0.0004705894736842105,
-      "loss": 0.0555,
+      "loss": 0.0481,
       "num_input_tokens_seen": 167067648,
       "step": 2550,
-      "train_runtime": 820.9524,
-      "train_tokens_per_second": 203504.673
+      "train_runtime": 1127.9533,
+      "train_tokens_per_second": 148115.745
     },
     {
       "epoch": 0.553872782345305,
-      "grad_norm": 0.86328125,
+      "grad_norm": 0.65625,
       "learning_rate": 0.0004699578947368421,
-      "loss": 0.0521,
+      "loss": 0.0436,
       "num_input_tokens_seen": 167723008,
       "step": 2560,
-      "train_runtime": 824.76,
-      "train_tokens_per_second": 203359.777
+      "train_runtime": 1133.6639,
+      "train_tokens_per_second": 147947.74
     },
     {
       "epoch": 0.5560363479013414,
-      "grad_norm": 1.0,
+      "grad_norm": 0.66796875,
       "learning_rate": 0.0004693263157894736,
-      "loss": 0.0531,
+      "loss": 0.0436,
       "num_input_tokens_seen": 168378368,
       "step": 2570,
-      "train_runtime": 828.5662,
-      "train_tokens_per_second": 203216.547
+      "train_runtime": 1137.4754,
+      "train_tokens_per_second": 148028.141
     },
     {
       "epoch": 0.5581999134573777,
-      "grad_norm": 0.640625,
+      "grad_norm": 0.5703125,
       "learning_rate": 0.0004686947368421052,
-      "loss": 0.056,
+      "loss": 0.0451,
       "num_input_tokens_seen": 169033728,
       "step": 2580,
-      "train_runtime": 832.3752,
-      "train_tokens_per_second": 203073.956
+      "train_runtime": 1141.2806,
+      "train_tokens_per_second": 148108.818
     },
     {
       "epoch": 0.5603634790134141,
-      "grad_norm": 1.015625,
+      "grad_norm": 0.73828125,
       "learning_rate": 0.0004680631578947368,
-      "loss": 0.0533,
+      "loss": 0.0413,
       "num_input_tokens_seen": 169684992,
       "step": 2590,
-      "train_runtime": 836.161,
-      "train_tokens_per_second": 202933.408
+      "train_runtime": 1145.0582,
+      "train_tokens_per_second": 148188.968
     },
     {
       "epoch": 0.5625270445694505,
-      "grad_norm": 0.9921875,
+      "grad_norm": 1.0625,
       "learning_rate": 0.0004674315789473684,
-      "loss": 0.0574,
+      "loss": 0.0466,
       "num_input_tokens_seen": 170340352,
       "step": 2600,
-      "train_runtime": 839.9698,
-      "train_tokens_per_second": 202793.423
+      "train_runtime": 1148.8626,
+      "train_tokens_per_second": 148268.694
     },
     {
       "epoch": 0.5625270445694505,
-      "eval_loss": 0.03629929572343826,
-      "eval_runtime": 1.8211,
-      "eval_samples_per_second": 17.572,
-      "eval_steps_per_second": 0.549,
+      "eval_loss": 0.035643137991428375,
+      "eval_runtime": 2.2313,
+      "eval_samples_per_second": 14.342,
+      "eval_steps_per_second": 0.448,
       "num_input_tokens_seen": 170340352,
       "step": 2600
     },
     {
       "epoch": 0.5625270445694505,
-      "eval_byte_accuracy": 0.990727002967359,
-      "eval_chrf": 94.42432169397047,
-      "eval_sacrebleu": 92.86889065123009,
-      "eval_word_accuracy": 0.9821428571428571,
+      "eval_byte_accuracy": 0.9888724035608308,
+      "eval_chrf": 92.92340614755764,
+      "eval_sacrebleu": 90.15152547055455,
+      "eval_word_accuracy": 0.9766483516483516,
       "num_input_tokens_seen": 170340352,
-      "perplexity": 1.0369661595878308,
+      "perplexity": 1.0362859694048128,
       "step": 2600
     },
     {
       "epoch": 0.5646906101254868,
-      "grad_norm": 1.046875,
+      "grad_norm": 0.859375,
       "learning_rate": 0.00046679999999999996,
-      "loss": 0.0553,
+      "loss": 0.0478,
       "num_input_tokens_seen": 170995712,
       "step": 2610,
-      "train_runtime": 845.6193,
-      "train_tokens_per_second": 202213.575
+      "train_runtime": 1156.9829,
+      "train_tokens_per_second": 147794.504
     },
     {
       "epoch": 0.5668541756815232,
-      "grad_norm": 1.1015625,
+      "grad_norm": 0.9609375,
       "learning_rate": 0.00046616842105263155,
-      "loss": 0.0575,
+      "loss": 0.0508,
       "num_input_tokens_seen": 171651072,
       "step": 2620,
-      "train_runtime": 849.4317,
-      "train_tokens_per_second": 202077.555
+      "train_runtime": 1160.7918,
+      "train_tokens_per_second": 147874.125
     },
     {
       "epoch": 0.5690177412375595,
-      "grad_norm": 0.859375,
+      "grad_norm": 0.83203125,
       "learning_rate": 0.00046553684210526314,
-      "loss": 0.0576,
+      "loss": 0.048,
       "num_input_tokens_seen": 172306432,
       "step": 2630,
-      "train_runtime": 853.258,
-      "train_tokens_per_second": 201939.431
+      "train_runtime": 1164.6014,
+      "train_tokens_per_second": 147953.141
     },
     {
       "epoch": 0.5711813067935958,
-      "grad_norm": 0.98828125,
+      "grad_norm": 0.6328125,
       "learning_rate": 0.00046490526315789467,
-      "loss": 0.0529,
+      "loss": 0.0451,
       "num_input_tokens_seen": 172961792,
       "step": 2640,
-      "train_runtime": 857.0844,
-      "train_tokens_per_second": 201802.519
+      "train_runtime": 1168.4095,
+      "train_tokens_per_second": 148031.825
     },
     {
       "epoch": 0.5733448723496322,
-      "grad_norm": 0.8359375,
+      "grad_norm": 1.171875,
       "learning_rate": 0.00046427368421052625,
-      "loss": 0.0499,
+      "loss": 0.0409,
       "num_input_tokens_seen": 173617152,
       "step": 2650,
-      "train_runtime": 860.9023,
-      "train_tokens_per_second": 201668.821
+      "train_runtime": 1172.221,
+      "train_tokens_per_second": 148109.573
     },
     {
       "epoch": 0.5755084379056685,
-      "grad_norm": 1.2734375,
+      "grad_norm": 1.03125,
       "learning_rate": 0.00046364210526315784,
-      "loss": 0.0496,
+      "loss": 0.0424,
       "num_input_tokens_seen": 174268416,
       "step": 2660,
-      "train_runtime": 864.7001,
-      "train_tokens_per_second": 201536.254
+      "train_runtime": 1178.2999,
+      "train_tokens_per_second": 147898.18
     },
     {
       "epoch": 0.5776720034617049,
-      "grad_norm": 0.7734375,
+      "grad_norm": 0.9765625,
       "learning_rate": 0.0004630105263157894,
-      "loss": 0.0516,
+      "loss": 0.0472,
       "num_input_tokens_seen": 174923776,
       "step": 2670,
-      "train_runtime": 868.5118,
-      "train_tokens_per_second": 201406.34
+      "train_runtime": 1182.1031,
+      "train_tokens_per_second": 147976.753
     },
     {
       "epoch": 0.5798355690177412,
-      "grad_norm": 1.109375,
+      "grad_norm": 0.95703125,
       "learning_rate": 0.000462378947368421,
-      "loss": 0.0584,
+      "loss": 0.0511,
       "num_input_tokens_seen": 175579136,
       "step": 2680,
-      "train_runtime": 872.3297,
-      "train_tokens_per_second": 201276.128
+      "train_runtime": 1185.9109,
+      "train_tokens_per_second": 148054.242
     },
     {
       "epoch": 0.5819991345737776,
-      "grad_norm": 0.96875,
+      "grad_norm": 0.953125,
       "learning_rate": 0.0004617473684210526,
-      "loss": 0.0504,
+      "loss": 0.0438,
       "num_input_tokens_seen": 176234496,
       "step": 2690,
-      "train_runtime": 876.1365,
-      "train_tokens_per_second": 201149.593
+      "train_runtime": 1189.7267,
+      "train_tokens_per_second": 148130.241
     },
     {
       "epoch": 0.584162700129814,
-      "grad_norm": 0.91015625,
+      "grad_norm": 1.265625,
       "learning_rate": 0.0004611157894736842,
-      "loss": 0.0586,
+      "loss": 0.0485,
       "num_input_tokens_seen": 176889856,
       "step": 2700,
-      "train_runtime": 879.9402,
-      "train_tokens_per_second": 201024.85
+      "train_runtime": 1193.5458,
+      "train_tokens_per_second": 148205.339
     },
     {
       "epoch": 0.584162700129814,
-      "eval_loss": 0.03447730839252472,
-      "eval_runtime": 1.7192,
-      "eval_samples_per_second": 18.614,
-      "eval_steps_per_second": 0.582,
+      "eval_loss": 0.031199950724840164,
+      "eval_runtime": 1.9652,
+      "eval_samples_per_second": 16.283,
+      "eval_steps_per_second": 0.509,
       "num_input_tokens_seen": 176889856,
       "step": 2700
     },
     {
       "epoch": 0.584162700129814,
-      "eval_byte_accuracy": 0.990727002967359,
-      "eval_chrf": 93.44542123318921,
-      "eval_sacrebleu": 92.43982388879606,
-      "eval_word_accuracy": 0.9766483516483516,
+      "eval_byte_accuracy": 0.9918397626112759,
+      "eval_chrf": 92.79812823458536,
+      "eval_sacrebleu": 90.75543017558854,
+      "eval_word_accuracy": 0.978021978021978,
       "num_input_tokens_seen": 176889856,
-      "perplexity": 1.0350785405136793,
+      "perplexity": 1.0316917707816073,
       "step": 2700
     },
     {
       "epoch": 0.5863262656858503,
-      "grad_norm": 0.76953125,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.0004604842105263157,
-      "loss": 0.0491,
+      "loss": 0.0387,
       "num_input_tokens_seen": 177545216,
       "step": 2710,
-      "train_runtime": 885.4951,
-      "train_tokens_per_second": 200503.892
+      "train_runtime": 1201.2331,
+      "train_tokens_per_second": 147802.468
     },
     {
       "epoch": 0.5884898312418866,
-      "grad_norm": 1.1875,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.00045985263157894736,
-      "loss": 0.053,
+      "loss": 0.0434,
       "num_input_tokens_seen": 178200576,
       "step": 2720,
-      "train_runtime": 889.3102,
-      "train_tokens_per_second": 200380.684
+      "train_runtime": 1205.3285,
+      "train_tokens_per_second": 147843.994
     },
     {
       "epoch": 0.590653396797923,
-      "grad_norm": 0.95703125,
+      "grad_norm": 0.75390625,
       "learning_rate": 0.0004592210526315789,
-      "loss": 0.0509,
+      "loss": 0.042,
       "num_input_tokens_seen": 178855936,
       "step": 2730,
-      "train_runtime": 893.1217,
-      "train_tokens_per_second": 200259.314
+      "train_runtime": 1209.1463,
+      "train_tokens_per_second": 147919.183
     },
     {
       "epoch": 0.5928169623539593,
-      "grad_norm": 1.2265625,
+      "grad_norm": 0.90234375,
       "learning_rate": 0.0004585894736842105,
-      "loss": 0.0489,
+      "loss": 0.0414,
       "num_input_tokens_seen": 179503104,
       "step": 2740,
-      "train_runtime": 896.8967,
-      "train_tokens_per_second": 200137.981
+      "train_runtime": 1212.915,
+      "train_tokens_per_second": 147993.142
     },
     {
       "epoch": 0.5949805279099957,
-      "grad_norm": 0.9453125,
+      "grad_norm": 0.72265625,
       "learning_rate": 0.00045795789473684206,
-      "loss": 0.0476,
+      "loss": 0.038,
       "num_input_tokens_seen": 180158464,
       "step": 2750,
-      "train_runtime": 900.7088,
-      "train_tokens_per_second": 200018.543
+      "train_runtime": 1216.7277,
+      "train_tokens_per_second": 148068.023
     },
     {
       "epoch": 0.597144093466032,
-      "grad_norm": 0.80078125,
+      "grad_norm": 0.84375,
       "learning_rate": 0.00045732631578947365,
-      "loss": 0.0494,
+      "loss": 0.0431,
       "num_input_tokens_seen": 180813824,
       "step": 2760,
-      "train_runtime": 904.5344,
-      "train_tokens_per_second": 199897.123
+      "train_runtime": 1222.4427,
+      "train_tokens_per_second": 147911.901
     },
     {
       "epoch": 0.5993076590220684,
-      "grad_norm": 0.88671875,
+      "grad_norm": 0.6875,
       "learning_rate": 0.00045669473684210524,
-      "loss": 0.0529,
+      "loss": 0.0422,
       "num_input_tokens_seen": 181469184,
       "step": 2770,
-      "train_runtime": 908.3444,
-      "train_tokens_per_second": 199780.16
+      "train_runtime": 1226.2503,
+      "train_tokens_per_second": 147987.072
     },
     {
       "epoch": 0.6014712245781048,
-      "grad_norm": 0.8984375,
+      "grad_norm": 1.109375,
       "learning_rate": 0.0004560631578947368,
-      "loss": 0.0503,
+      "loss": 0.0398,
       "num_input_tokens_seen": 182124544,
       "step": 2780,
-      "train_runtime": 912.1602,
-      "train_tokens_per_second": 199662.888
+      "train_runtime": 1230.0502,
+      "train_tokens_per_second": 148062.688
     },
     {
       "epoch": 0.6036347901341411,
-      "grad_norm": 1.1875,
+      "grad_norm": 0.79296875,
       "learning_rate": 0.0004554315789473684,
-      "loss": 0.0477,
+      "loss": 0.0384,
       "num_input_tokens_seen": 182779904,
       "step": 2790,
-      "train_runtime": 915.9796,
-      "train_tokens_per_second": 199545.819
+      "train_runtime": 1233.8548,
+      "train_tokens_per_second": 148137.288
     },
     {
       "epoch": 0.6057983556901774,
-      "grad_norm": 1.1484375,
+      "grad_norm": 0.96484375,
       "learning_rate": 0.00045479999999999994,
-      "loss": 0.0546,
+      "loss": 0.0455,
       "num_input_tokens_seen": 183435264,
       "step": 2800,
-      "train_runtime": 919.7965,
-      "train_tokens_per_second": 199430.278
+      "train_runtime": 1237.6628,
+      "train_tokens_per_second": 148211.014
     },
     {
       "epoch": 0.6057983556901774,
-      "eval_loss": 0.03884509950876236,
-      "eval_runtime": 1.7475,
-      "eval_samples_per_second": 18.312,
-      "eval_steps_per_second": 0.572,
+      "eval_loss": 0.03436446934938431,
+      "eval_runtime": 2.0082,
+      "eval_samples_per_second": 15.935,
+      "eval_steps_per_second": 0.498,
       "num_input_tokens_seen": 183435264,
       "step": 2800
     },
     {
       "epoch": 0.6057983556901774,
-      "eval_byte_accuracy": 0.9918397626112759,
-      "eval_chrf": 93.07953774652597,
-      "eval_sacrebleu": 91.3351215187772,
-      "eval_word_accuracy": 0.9807692307692307,
+      "eval_byte_accuracy": 0.9896142433234422,
+      "eval_chrf": 92.24485602910795,
+      "eval_sacrebleu": 89.54669354297549,
+      "eval_word_accuracy": 0.9766483516483516,
       "num_input_tokens_seen": 183435264,
-      "perplexity": 1.0396094351650327,
+      "perplexity": 1.0349617498309909,
       "step": 2800
     },
     {
       "epoch": 0.6079619212462137,
-      "grad_norm": 0.63671875,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.00045416842105263153,
-      "loss": 0.0481,
+      "loss": 0.0402,
       "num_input_tokens_seen": 184090624,
       "step": 2810,
-      "train_runtime": 925.3841,
-      "train_tokens_per_second": 198934.282
+      "train_runtime": 1245.5671,
+      "train_tokens_per_second": 147796.639
     },
     {
       "epoch": 0.6101254868022501,
-      "grad_norm": 0.859375,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.0004535368421052631,
-      "loss": 0.0503,
+      "loss": 0.041,
       "num_input_tokens_seen": 184745984,
       "step": 2820,
-      "train_runtime": 929.1957,
-      "train_tokens_per_second": 198823.555
+      "train_runtime": 1249.3742,
+      "train_tokens_per_second": 147870.819
     },
     {
       "epoch": 0.6122890523582865,
-      "grad_norm": 0.63671875,
+      "grad_norm": 0.67578125,
       "learning_rate": 0.0004529052631578947,
-      "loss": 0.048,
+      "loss": 0.0399,
       "num_input_tokens_seen": 185401344,
       "step": 2830,
-      "train_runtime": 933.0099,
-      "train_tokens_per_second": 198713.165
+      "train_runtime": 1253.1873,
+      "train_tokens_per_second": 147943.844
     },
     {
       "epoch": 0.6144526179143228,
-      "grad_norm": 1.109375,
+      "grad_norm": 0.84765625,
       "learning_rate": 0.0004522736842105263,
-      "loss": 0.0513,
+      "loss": 0.0418,
       "num_input_tokens_seen": 186056704,
       "step": 2840,
-      "train_runtime": 936.8218,
-      "train_tokens_per_second": 198604.161
+      "train_runtime": 1256.9935,
+      "train_tokens_per_second": 148017.236
     },
     {
       "epoch": 0.6166161834703592,
-      "grad_norm": 0.8828125,
+      "grad_norm": 0.5546875,
       "learning_rate": 0.0004516421052631579,
-      "loss": 0.05,
+      "loss": 0.0414,
       "num_input_tokens_seen": 186712064,
       "step": 2850,
-      "train_runtime": 940.6403,
-      "train_tokens_per_second": 198494.647
+      "train_runtime": 1260.8015,
+      "train_tokens_per_second": 148089.971
     },
     {
       "epoch": 0.6187797490263955,
-      "grad_norm": 0.70703125,
+      "grad_norm": 0.55859375,
       "learning_rate": 0.00045101052631578946,
-      "loss": 0.0503,
+      "loss": 0.0415,
       "num_input_tokens_seen": 187367424,
       "step": 2860,
-      "train_runtime": 944.4565,
-      "train_tokens_per_second": 198386.503
+      "train_runtime": 1266.8963,
+      "train_tokens_per_second": 147894.834
     },
     {
       "epoch": 0.6209433145824318,
-      "grad_norm": 1.0625,
+      "grad_norm": 0.7421875,
       "learning_rate": 0.000450378947368421,
-      "loss": 0.0535,
+      "loss": 0.0428,
       "num_input_tokens_seen": 188022784,
       "step": 2870,
-      "train_runtime": 948.2753,
-      "train_tokens_per_second": 198278.689
+      "train_runtime": 1270.7058,
+      "train_tokens_per_second": 147967.208
     },
     {
       "epoch": 0.6231068801384682,
-      "grad_norm": 1.1875,
+      "grad_norm": 0.494140625,
       "learning_rate": 0.0004497473684210526,
-      "loss": 0.0528,
+      "loss": 0.0416,
       "num_input_tokens_seen": 188678144,
       "step": 2880,
-      "train_runtime": 952.0987,
-      "train_tokens_per_second": 198170.787
+      "train_runtime": 1274.5129,
+      "train_tokens_per_second": 148039.413
     },
     {
       "epoch": 0.6252704456945045,
-      "grad_norm": 1.1015625,
+      "grad_norm": 0.97265625,
       "learning_rate": 0.00044911578947368417,
-      "loss": 0.051,
+      "loss": 0.0411,
       "num_input_tokens_seen": 189333504,
       "step": 2890,
-      "train_runtime": 955.9177,
-      "train_tokens_per_second": 198064.645
+      "train_runtime": 1278.3223,
+      "train_tokens_per_second": 148110.924
     },
     {
       "epoch": 0.6274340112505409,
-      "grad_norm": 0.75,
+      "grad_norm": 0.640625,
       "learning_rate": 0.00044848421052631575,
-      "loss": 0.0486,
+      "loss": 0.0402,
       "num_input_tokens_seen": 189988864,
       "step": 2900,
-      "train_runtime": 959.7442,
-      "train_tokens_per_second": 197957.814
+      "train_runtime": 1282.1268,
+      "train_tokens_per_second": 148182.58
     },
     {
       "epoch": 0.6274340112505409,
-      "eval_loss": 0.0350818894803524,
-      "eval_runtime": 1.9312,
-      "eval_samples_per_second": 16.57,
-      "eval_steps_per_second": 0.518,
+      "eval_loss": 0.02733401581645012,
+      "eval_runtime": 1.9425,
+      "eval_samples_per_second": 16.474,
+      "eval_steps_per_second": 0.515,
       "num_input_tokens_seen": 189988864,
       "step": 2900
     },
     {
       "epoch": 0.6274340112505409,
-      "eval_byte_accuracy": 0.9918397626112759,
-      "eval_chrf": 93.43306198468976,
-      "eval_sacrebleu": 91.46567212681035,
-      "eval_word_accuracy": 0.9793956043956044,
+      "eval_byte_accuracy": 0.9914688427299704,
+      "eval_chrf": 92.14666454548373,
+      "eval_sacrebleu": 90.64406606563134,
+      "eval_word_accuracy": 0.978021978021978,
       "num_input_tokens_seen": 189988864,
-      "perplexity": 1.0357045186318887,
+      "perplexity": 1.0277110171752541,
       "step": 2900
     },
     {
       "epoch": 0.6295975768065772,
-      "grad_norm": 0.6796875,
+      "grad_norm": 0.79296875,
       "learning_rate": 0.00044785263157894734,
-      "loss": 0.0485,
+      "loss": 0.0409,
       "num_input_tokens_seen": 190644224,
       "step": 2910,
-      "train_runtime": 965.5126,
-      "train_tokens_per_second": 197453.898
+      "train_runtime": 1290.0163,
+      "train_tokens_per_second": 147784.353
     },
     {
       "epoch": 0.6317611423626136,
-      "grad_norm": 1.015625,
+      "grad_norm": 0.8203125,
       "learning_rate": 0.0004472210526315789,
-      "loss": 0.0508,
+      "loss": 0.0421,
       "num_input_tokens_seen": 191299584,
       "step": 2920,
-      "train_runtime": 969.3212,
-      "train_tokens_per_second": 197354.183
+      "train_runtime": 1293.8198,
+      "train_tokens_per_second": 147856.432
     },
     {
       "epoch": 0.63392470791865,
-      "grad_norm": 0.91796875,
+      "grad_norm": 0.80078125,
       "learning_rate": 0.0004465894736842105,
-      "loss": 0.0486,
+      "loss": 0.0406,
       "num_input_tokens_seen": 191954944,
       "step": 2930,
-      "train_runtime": 973.1287,
-      "train_tokens_per_second": 197255.457
+      "train_runtime": 1297.6244,
+      "train_tokens_per_second": 147927.965
     },
     {
       "epoch": 0.6360882734746863,
-      "grad_norm": 0.75,
+      "grad_norm": 1.0703125,
       "learning_rate": 0.00044595789473684204,
-      "loss": 0.0504,
+      "loss": 0.0414,
       "num_input_tokens_seen": 192610304,
       "step": 2940,
-      "train_runtime": 976.9568,
-      "train_tokens_per_second": 197153.346
+      "train_runtime": 1301.4321,
+      "train_tokens_per_second": 147998.74
     },
     {
       "epoch": 0.6382518390307226,
-      "grad_norm": 0.76953125,
+      "grad_norm": 0.6796875,
       "learning_rate": 0.00044532631578947363,
-      "loss": 0.049,
+      "loss": 0.0393,
       "num_input_tokens_seen": 193265664,
       "step": 2950,
-      "train_runtime": 980.7648,
-      "train_tokens_per_second": 197056.071
+      "train_runtime": 1305.2438,
+      "train_tokens_per_second": 148068.632
     },
     {
       "epoch": 0.6404154045867589,
-      "grad_norm": 1.2109375,
+      "grad_norm": 0.796875,
       "learning_rate": 0.0004446947368421052,
-      "loss": 0.053,
+      "loss": 0.0414,
       "num_input_tokens_seen": 193921024,
       "step": 2960,
-      "train_runtime": 984.5799,
-      "train_tokens_per_second": 196958.138
+      "train_runtime": 1311.2552,
+      "train_tokens_per_second": 147889.609
     },
     {
       "epoch": 0.6425789701427953,
-      "grad_norm": 0.71875,
+      "grad_norm": 0.71484375,
       "learning_rate": 0.0004440631578947368,
-      "loss": 0.0513,
+      "loss": 0.0424,
       "num_input_tokens_seen": 194576384,
       "step": 2970,
-      "train_runtime": 988.3971,
-      "train_tokens_per_second": 196860.531
+      "train_runtime": 1315.0601,
+      "train_tokens_per_second": 147960.072
     },
     {
       "epoch": 0.6447425356988317,
-      "grad_norm": 1.265625,
+      "grad_norm": 1.1171875,
       "learning_rate": 0.0004434315789473684,
-      "loss": 0.0486,
+      "loss": 0.04,
       "num_input_tokens_seen": 195231744,
       "step": 2980,
-      "train_runtime": 992.2137,
-      "train_tokens_per_second": 196763.799
+      "train_runtime": 1318.8693,
+      "train_tokens_per_second": 148029.641
     },
     {
       "epoch": 0.646906101254868,
-      "grad_norm": 0.80078125,
+      "grad_norm": 0.51171875,
       "learning_rate": 0.0004428,
-      "loss": 0.0489,
+      "loss": 0.0398,
       "num_input_tokens_seen": 195887104,
       "step": 2990,
-      "train_runtime": 996.0352,
-      "train_tokens_per_second": 196666.855
+      "train_runtime": 1322.6799,
+      "train_tokens_per_second": 148098.645
     },
     {
       "epoch": 0.6490696668109044,
-      "grad_norm": 0.67578125,
+      "grad_norm": 0.5625,
       "learning_rate": 0.00044216842105263156,
-      "loss": 0.0482,
+      "loss": 0.0389,
       "num_input_tokens_seen": 196542464,
       "step": 3000,
-      "train_runtime": 999.8291,
-      "train_tokens_per_second": 196576.059
+      "train_runtime": 1326.4813,
+      "train_tokens_per_second": 148168.289
     },
     {
       "epoch": 0.6490696668109044,
-      "eval_loss": 0.03645855933427811,
-      "eval_runtime": 1.6979,
-      "eval_samples_per_second": 18.847,
-      "eval_steps_per_second": 0.589,
+      "eval_loss": 0.028482545167207718,
+      "eval_runtime": 1.9015,
+      "eval_samples_per_second": 16.829,
+      "eval_steps_per_second": 0.526,
       "num_input_tokens_seen": 196542464,
       "step": 3000
     },
     {
       "epoch": 0.6490696668109044,
-      "eval_byte_accuracy": 0.9903560830860534,
-      "eval_chrf": 92.59396939397627,
-      "eval_sacrebleu": 88.68557212958412,
-      "eval_word_accuracy": 0.978021978021978,
+      "eval_byte_accuracy": 0.9910979228486647,
+      "eval_chrf": 94.2210646218814,
+      "eval_sacrebleu": 91.0806734556956,
+      "eval_word_accuracy": 0.9793956043956044,
       "num_input_tokens_seen": 196542464,
-      "perplexity": 1.037131323714694,
+      "perplexity": 1.0288920515390652,
       "step": 3000
     },
     {
       "epoch": 0.6512332323669408,
-      "grad_norm": 0.85546875,
+      "grad_norm": 0.90234375,
       "learning_rate": 0.00044153684210526315,
-      "loss": 0.0494,
+      "loss": 0.0402,
       "num_input_tokens_seen": 197197824,
       "step": 3010,
-      "train_runtime": 1005.4045,
-      "train_tokens_per_second": 196137.805
+      "train_runtime": 1334.1984,
+      "train_tokens_per_second": 147802.479
     },
     {
       "epoch": 0.6533967979229771,
-      "grad_norm": 1.015625,
+      "grad_norm": 1.03125,
       "learning_rate": 0.0004409052631578947,
-      "loss": 0.0502,
+      "loss": 0.0398,
       "num_input_tokens_seen": 197853184,
       "step": 3020,
-      "train_runtime": 1009.221,
-      "train_tokens_per_second": 196045.449
+      "train_runtime": 1338.0111,
+      "train_tokens_per_second": 147871.114
     },
     {
       "epoch": 0.6555603634790134,
-      "grad_norm": 1.0390625,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.00044027368421052627,
-      "loss": 0.0537,
+      "loss": 0.0432,
       "num_input_tokens_seen": 198508544,
       "step": 3030,
-      "train_runtime": 1013.0394,
-      "train_tokens_per_second": 195953.429
+      "train_runtime": 1341.8307,
+      "train_tokens_per_second": 147938.589
     },
     {
       "epoch": 0.6577239290350497,
-      "grad_norm": 0.81640625,
+      "grad_norm": 0.85546875,
       "learning_rate": 0.00043964210526315785,
-      "loss": 0.0479,
+      "loss": 0.0371,
       "num_input_tokens_seen": 199163904,
       "step": 3040,
-      "train_runtime": 1016.8572,
-      "train_tokens_per_second": 195862.212
+      "train_runtime": 1345.648,
+      "train_tokens_per_second": 148005.948
     },
     {
       "epoch": 0.6598874945910861,
-      "grad_norm": 0.7265625,
+      "grad_norm": 0.8515625,
       "learning_rate": 0.00043901052631578944,
-      "loss": 0.046,
+      "loss": 0.0386,
       "num_input_tokens_seen": 199819264,
       "step": 3050,
-      "train_runtime": 1020.671,
-      "train_tokens_per_second": 195772.461
+      "train_runtime": 1349.4633,
+      "train_tokens_per_second": 148073.135
     },
     {
       "epoch": 0.6620510601471224,
-      "grad_norm": 0.859375,
+      "grad_norm": 0.89453125,
       "learning_rate": 0.000438378947368421,
-      "loss": 0.0456,
+      "loss": 0.0368,
       "num_input_tokens_seen": 200474624,
       "step": 3060,
-      "train_runtime": 1024.488,
-      "train_tokens_per_second": 195682.737
+      "train_runtime": 1355.5535,
+      "train_tokens_per_second": 147891.339
     },
     {
       "epoch": 0.6642146257031588,
-      "grad_norm": 0.984375,
+      "grad_norm": 0.6875,
       "learning_rate": 0.0004377473684210526,
-      "loss": 0.048,
+      "loss": 0.0402,
       "num_input_tokens_seen": 201129984,
       "step": 3070,
-      "train_runtime": 1028.3086,
-      "train_tokens_per_second": 195593.021
+      "train_runtime": 1359.3648,
+      "train_tokens_per_second": 147958.801
     },
     {
       "epoch": 0.6663781912591952,
-      "grad_norm": 1.0390625,
+      "grad_norm": 1.0703125,
       "learning_rate": 0.0004371157894736842,
-      "loss": 0.0495,
+      "loss": 0.0403,
       "num_input_tokens_seen": 201785344,
       "step": 3080,
-      "train_runtime": 1032.1272,
-      "train_tokens_per_second": 195504.331
+      "train_runtime": 1363.1664,
+      "train_tokens_per_second": 148026.936
     },
     {
       "epoch": 0.6685417568152315,
-      "grad_norm": 0.796875,
+      "grad_norm": 0.8203125,
       "learning_rate": 0.00043648421052631573,
-      "loss": 0.0493,
+      "loss": 0.04,
       "num_input_tokens_seen": 202440704,
       "step": 3090,
-      "train_runtime": 1035.9438,
-      "train_tokens_per_second": 195416.689
+      "train_runtime": 1366.9795,
+      "train_tokens_per_second": 148093.446
     },
     {
       "epoch": 0.6707053223712679,
-      "grad_norm": 0.92578125,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.0004358526315789473,
-      "loss": 0.046,
+      "loss": 0.0369,
       "num_input_tokens_seen": 203096064,
       "step": 3100,
-      "train_runtime": 1039.7635,
-      "train_tokens_per_second": 195329.092
+      "train_runtime": 1370.7841,
+      "train_tokens_per_second": 148160.506
     },
     {
       "epoch": 0.6707053223712679,
-      "eval_loss": 0.03350326791405678,
-      "eval_runtime": 1.8809,
-      "eval_samples_per_second": 17.013,
-      "eval_steps_per_second": 0.532,
+      "eval_loss": 0.0291756484657526,
+      "eval_runtime": 2.1304,
+      "eval_samples_per_second": 15.021,
+      "eval_steps_per_second": 0.469,
       "num_input_tokens_seen": 203096064,
       "step": 3100
     },
     {
       "epoch": 0.6707053223712679,
-      "eval_byte_accuracy": 0.9918397626112759,
-      "eval_chrf": 93.82444893287017,
-      "eval_sacrebleu": 91.25242885956851,
+      "eval_byte_accuracy": 0.9914688427299704,
+      "eval_chrf": 93.86777853094796,
+      "eval_sacrebleu": 93.01282984759166,
       "eval_word_accuracy": 0.9807692307692307,
       "num_input_tokens_seen": 203096064,
-      "perplexity": 1.034070822975335,
+      "perplexity": 1.0296054272067894,
       "step": 3100
     },
     {
       "epoch": 0.6728688879273041,
-      "grad_norm": 0.71484375,
+      "grad_norm": 0.890625,
       "learning_rate": 0.0004352210526315789,
-      "loss": 0.0462,
+      "loss": 0.0383,
       "num_input_tokens_seen": 203751424,
       "step": 3110,
-      "train_runtime": 1045.4737,
-      "train_tokens_per_second": 194889.098
+      "train_runtime": 1378.8411,
+      "train_tokens_per_second": 147770.049
     },
     {
       "epoch": 0.6750324534833405,
-      "grad_norm": 0.74609375,
+      "grad_norm": 0.703125,
       "learning_rate": 0.0004345894736842105,
-      "loss": 0.048,
+      "loss": 0.0387,
       "num_input_tokens_seen": 204406784,
       "step": 3120,
-      "train_runtime": 1049.2876,
-      "train_tokens_per_second": 194805.295
+      "train_runtime": 1382.6559,
+      "train_tokens_per_second": 147836.333
     },
     {
       "epoch": 0.6771960190393769,
-      "grad_norm": 1.03125,
+      "grad_norm": 1.0,
       "learning_rate": 0.0004339578947368421,
-      "loss": 0.0468,
+      "loss": 0.0366,
       "num_input_tokens_seen": 205062144,
       "step": 3130,
-      "train_runtime": 1053.1102,
-      "train_tokens_per_second": 194720.49
+      "train_runtime": 1386.4677,
+      "train_tokens_per_second": 147902.572
     },
     {
       "epoch": 0.6793595845954132,
-      "grad_norm": 1.125,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.00043332631578947366,
-      "loss": 0.048,
+      "loss": 0.0396,
       "num_input_tokens_seen": 205717504,
       "step": 3140,
-      "train_runtime": 1056.9229,
-      "train_tokens_per_second": 194638.144
+      "train_runtime": 1390.2775,
+      "train_tokens_per_second": 147968.665
     },
     {
       "epoch": 0.6815231501514496,
-      "grad_norm": 0.91796875,
+      "grad_norm": 0.69140625,
       "learning_rate": 0.00043269473684210525,
-      "loss": 0.0455,
+      "loss": 0.0368,
       "num_input_tokens_seen": 206372864,
       "step": 3150,
-      "train_runtime": 1060.7403,
-      "train_tokens_per_second": 194555.501
+      "train_runtime": 1394.0902,
+      "train_tokens_per_second": 148034.082
     },
     {
       "epoch": 0.683686715707486,
-      "grad_norm": 0.89453125,
+      "grad_norm": 0.392578125,
       "learning_rate": 0.0004320631578947368,
-      "loss": 0.0442,
+      "loss": 0.0364,
       "num_input_tokens_seen": 207028224,
       "step": 3160,
-      "train_runtime": 1064.5576,
-      "train_tokens_per_second": 194473.488
+      "train_runtime": 1400.0318,
+      "train_tokens_per_second": 147873.939
     },
     {
       "epoch": 0.6858502812635223,
-      "grad_norm": 1.4921875,
+      "grad_norm": 0.7734375,
       "learning_rate": 0.00043143157894736837,
-      "loss": 0.0472,
+      "loss": 0.0365,
       "num_input_tokens_seen": 207683584,
       "step": 3170,
-      "train_runtime": 1068.3716,
-      "train_tokens_per_second": 194392.647
+      "train_runtime": 1403.8501,
+      "train_tokens_per_second": 147938.576
     },
     {
       "epoch": 0.6880138468195587,
-      "grad_norm": 0.61328125,
+      "grad_norm": 0.578125,
       "learning_rate": 0.00043079999999999995,
-      "loss": 0.0466,
+      "loss": 0.0381,
       "num_input_tokens_seen": 208338944,
       "step": 3180,
-      "train_runtime": 1072.1806,
-      "train_tokens_per_second": 194313.287
+      "train_runtime": 1407.6691,
+      "train_tokens_per_second": 148002.779
     },
     {
       "epoch": 0.6901774123755949,
-      "grad_norm": 1.1171875,
+      "grad_norm": 0.84375,
       "learning_rate": 0.00043016842105263154,
-      "loss": 0.0433,
+      "loss": 0.0363,
       "num_input_tokens_seen": 208994304,
       "step": 3190,
-      "train_runtime": 1075.9997,
-      "train_tokens_per_second": 194232.684
+      "train_runtime": 1411.485,
+      "train_tokens_per_second": 148066.971
     },
     {
       "epoch": 0.6923409779316313,
-      "grad_norm": 1.0546875,
+      "grad_norm": 0.96875,
       "learning_rate": 0.00042953684210526313,
-      "loss": 0.0475,
+      "loss": 0.0378,
       "num_input_tokens_seen": 209649664,
       "step": 3200,
-      "train_runtime": 1079.7767,
-      "train_tokens_per_second": 194160.206
+      "train_runtime": 1415.2661,
+      "train_tokens_per_second": 148134.448
     },
     {
       "epoch": 0.6923409779316313,
-      "eval_loss": 0.03217955678701401,
-      "eval_runtime": 1.7478,
-      "eval_samples_per_second": 18.309,
-      "eval_steps_per_second": 0.572,
+      "eval_loss": 0.025803592056035995,
+      "eval_runtime": 2.1635,
+      "eval_samples_per_second": 14.791,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 209649664,
       "step": 3200
     },
     {
       "epoch": 0.6923409779316313,
-      "eval_byte_accuracy": 0.9922106824925816,
-      "eval_chrf": 93.56755109473936,
-      "eval_sacrebleu": 92.12472781620804,
-      "eval_word_accuracy": 0.9807692307692307,
+      "eval_byte_accuracy": 0.9936943620178041,
+      "eval_chrf": 95.77357270335774,
+      "eval_sacrebleu": 95.12443110170639,
+      "eval_word_accuracy": 0.9862637362637363,
       "num_input_tokens_seen": 209649664,
-      "perplexity": 1.0327029174764155,
+      "perplexity": 1.026139386752756,
       "step": 3200
     },
     {
       "epoch": 0.6945045434876677,
-      "grad_norm": 0.65234375,
+      "grad_norm": 0.765625,
       "learning_rate": 0.0004289052631578947,
-      "loss": 0.0452,
+      "loss": 0.0383,
       "num_input_tokens_seen": 210305024,
       "step": 3210,
-      "train_runtime": 1085.3612,
-      "train_tokens_per_second": 193765.005
+      "train_runtime": 1423.5396,
+      "train_tokens_per_second": 147733.872
     },
     {
       "epoch": 0.696668109043704,
-      "grad_norm": 0.6796875,
+      "grad_norm": 1.3125,
       "learning_rate": 0.0004282736842105263,
-      "loss": 0.0485,
+      "loss": 0.0417,
       "num_input_tokens_seen": 210960384,
       "step": 3220,
-      "train_runtime": 1089.1789,
-      "train_tokens_per_second": 193687.542
+      "train_runtime": 1427.3546,
+      "train_tokens_per_second": 147798.161
     },
     {
       "epoch": 0.6988316745997404,
-      "grad_norm": 0.498046875,
+      "grad_norm": 1.0078125,
       "learning_rate": 0.00042764210526315783,
-      "loss": 0.0449,
+      "loss": 0.0363,
       "num_input_tokens_seen": 211615744,
       "step": 3230,
-      "train_runtime": 1092.9778,
-      "train_tokens_per_second": 193613.946
+      "train_runtime": 1431.1454,
+      "train_tokens_per_second": 147864.603
     },
     {
       "epoch": 0.7009952401557767,
-      "grad_norm": 0.59375,
+      "grad_norm": 0.36328125,
       "learning_rate": 0.0004270105263157895,
-      "loss": 0.0459,
+      "loss": 0.0377,
       "num_input_tokens_seen": 212271104,
       "step": 3240,
-      "train_runtime": 1096.795,
-      "train_tokens_per_second": 193537.622
+      "train_runtime": 1434.9622,
+      "train_tokens_per_second": 147928.013
     },
     {
       "epoch": 0.7031588057118131,
-      "grad_norm": 0.81640625,
+      "grad_norm": 0.53125,
       "learning_rate": 0.000426378947368421,
-      "loss": 0.0432,
+      "loss": 0.036,
       "num_input_tokens_seen": 212926464,
       "step": 3250,
-      "train_runtime": 1100.6057,
-      "train_tokens_per_second": 193462.988
+      "train_runtime": 1438.7773,
+      "train_tokens_per_second": 147991.258
     },
     {
       "epoch": 0.7053223712678495,
-      "grad_norm": 0.96875,
+      "grad_norm": 0.58203125,
       "learning_rate": 0.0004257473684210526,
-      "loss": 0.0419,
+      "loss": 0.0322,
       "num_input_tokens_seen": 213581824,
       "step": 3260,
-      "train_runtime": 1104.4101,
-      "train_tokens_per_second": 193389.955
+      "train_runtime": 1444.7363,
+      "train_tokens_per_second": 147834.469
     },
     {
       "epoch": 0.7074859368238857,
-      "grad_norm": 0.9609375,
+      "grad_norm": 0.6953125,
       "learning_rate": 0.0004251157894736842,
-      "loss": 0.0461,
+      "loss": 0.0373,
       "num_input_tokens_seen": 214237184,
       "step": 3270,
-      "train_runtime": 1108.2216,
-      "train_tokens_per_second": 193316.195
+      "train_runtime": 1448.5508,
+      "train_tokens_per_second": 147897.598
     },
     {
       "epoch": 0.7096495023799221,
-      "grad_norm": 0.84765625,
+      "grad_norm": 0.6875,
       "learning_rate": 0.00042448421052631576,
-      "loss": 0.0453,
+      "loss": 0.0358,
       "num_input_tokens_seen": 214892544,
       "step": 3280,
-      "train_runtime": 1112.0361,
-      "train_tokens_per_second": 193242.412
+      "train_runtime": 1452.3774,
+      "train_tokens_per_second": 147959.161
     },
     {
       "epoch": 0.7118130679359584,
-      "grad_norm": 0.69140625,
+      "grad_norm": 0.53125,
       "learning_rate": 0.00042385263157894735,
-      "loss": 0.0491,
+      "loss": 0.0407,
       "num_input_tokens_seen": 215547904,
       "step": 3290,
-      "train_runtime": 1115.8415,
-      "train_tokens_per_second": 193170.723
+      "train_runtime": 1456.1897,
+      "train_tokens_per_second": 148021.855
     },
     {
       "epoch": 0.7139766334919948,
-      "grad_norm": 0.71484375,
+      "grad_norm": 0.55859375,
       "learning_rate": 0.0004232210526315789,
-      "loss": 0.0456,
+      "loss": 0.0365,
       "num_input_tokens_seen": 216203264,
       "step": 3300,
-      "train_runtime": 1119.6565,
-      "train_tokens_per_second": 193097.858
+      "train_runtime": 1460.0024,
+      "train_tokens_per_second": 148084.185
     },
     {
       "epoch": 0.7139766334919948,
-      "eval_loss": 0.032814886420965195,
-      "eval_runtime": 1.5657,
-      "eval_samples_per_second": 20.439,
-      "eval_steps_per_second": 0.639,
+      "eval_loss": 0.025559077039361,
+      "eval_runtime": 2.0916,
+      "eval_samples_per_second": 15.3,
+      "eval_steps_per_second": 0.478,
       "num_input_tokens_seen": 216203264,
       "step": 3300
     },
     {
       "epoch": 0.7139766334919948,
-      "eval_byte_accuracy": 0.990727002967359,
-      "eval_chrf": 92.9667765910088,
-      "eval_sacrebleu": 91.00055209302626,
-      "eval_word_accuracy": 0.9793956043956044,
+      "eval_byte_accuracy": 0.9910979228486647,
+      "eval_chrf": 93.62674522126692,
+      "eval_sacrebleu": 92.43274818870434,
+      "eval_word_accuracy": 0.9807692307692307,
       "num_input_tokens_seen": 216203264,
-      "perplexity": 1.0333592327091379,
+      "perplexity": 1.0258885109361953,
       "step": 3300
     },
     {
       "epoch": 0.7161401990480312,
-      "grad_norm": 0.8984375,
+      "grad_norm": 0.7265625,
       "learning_rate": 0.0004225894736842105,
-      "loss": 0.0481,
+      "loss": 0.0393,
       "num_input_tokens_seen": 216858624,
       "step": 3310,
-      "train_runtime": 1125.0552,
-      "train_tokens_per_second": 192753.759
+      "train_runtime": 1468.0444,
+      "train_tokens_per_second": 147719.393
     },
     {
       "epoch": 0.7183037646040675,
-      "grad_norm": 1.1875,
+      "grad_norm": 0.91015625,
       "learning_rate": 0.00042195789473684206,
-      "loss": 0.0485,
+      "loss": 0.0393,
       "num_input_tokens_seen": 217513984,
       "step": 3320,
-      "train_runtime": 1128.8657,
-      "train_tokens_per_second": 192683.671
+      "train_runtime": 1471.8608,
+      "train_tokens_per_second": 147781.623
     },
     {
       "epoch": 0.7204673301601039,
-      "grad_norm": 0.8203125,
+      "grad_norm": 0.8359375,
       "learning_rate": 0.00042132631578947364,
-      "loss": 0.045,
+      "loss": 0.0373,
       "num_input_tokens_seen": 218169344,
       "step": 3330,
-      "train_runtime": 1132.6806,
-      "train_tokens_per_second": 192613.289
+      "train_runtime": 1475.669,
+      "train_tokens_per_second": 147844.362
     },
     {
       "epoch": 0.7226308957161403,
-      "grad_norm": 0.96484375,
+      "grad_norm": 0.76953125,
       "learning_rate": 0.00042069473684210523,
-      "loss": 0.0433,
+      "loss": 0.0358,
       "num_input_tokens_seen": 218824704,
       "step": 3340,
-      "train_runtime": 1136.5047,
-      "train_tokens_per_second": 192541.834
+      "train_runtime": 1479.4858,
+      "train_tokens_per_second": 147905.919
     },
     {
       "epoch": 0.7247944612721765,
-      "grad_norm": 0.66796875,
+      "grad_norm": 0.6875,
       "learning_rate": 0.0004200631578947368,
-      "loss": 0.0425,
+      "loss": 0.0347,
       "num_input_tokens_seen": 219480064,
       "step": 3350,
-      "train_runtime": 1140.3158,
-      "train_tokens_per_second": 192473.056
+      "train_runtime": 1483.2947,
+      "train_tokens_per_second": 147967.941
     },
     {
       "epoch": 0.7269580268282129,
-      "grad_norm": 0.609375,
+      "grad_norm": 0.58203125,
       "learning_rate": 0.0004194315789473684,
-      "loss": 0.0448,
+      "loss": 0.0365,
       "num_input_tokens_seen": 220135424,
       "step": 3360,
-      "train_runtime": 1144.1264,
-      "train_tokens_per_second": 192404.804
+      "train_runtime": 1489.4844,
+      "train_tokens_per_second": 147793.038
     },
     {
       "epoch": 0.7291215923842492,
-      "grad_norm": 0.76171875,
+      "grad_norm": 1.1171875,
       "learning_rate": 0.00041879999999999993,
-      "loss": 0.0478,
+      "loss": 0.0389,
       "num_input_tokens_seen": 220790784,
       "step": 3370,
-      "train_runtime": 1147.9385,
-      "train_tokens_per_second": 192336.766
+      "train_runtime": 1493.2964,
+      "train_tokens_per_second": 147854.625
     },
     {
       "epoch": 0.7312851579402856,
-      "grad_norm": 0.734375,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.0004181684210526316,
-      "loss": 0.0436,
+      "loss": 0.0361,
       "num_input_tokens_seen": 221446144,
       "step": 3380,
-      "train_runtime": 1151.7449,
-      "train_tokens_per_second": 192270.137
+      "train_runtime": 1497.1061,
+      "train_tokens_per_second": 147916.135
     },
     {
       "epoch": 0.733448723496322,
-      "grad_norm": 0.71875,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.0004175368421052631,
-      "loss": 0.0415,
+      "loss": 0.0333,
       "num_input_tokens_seen": 222101504,
       "step": 3390,
-      "train_runtime": 1155.5548,
-      "train_tokens_per_second": 192203.356
+      "train_runtime": 1500.9269,
+      "train_tokens_per_second": 147976.229
     },
     {
       "epoch": 0.7356122890523583,
-      "grad_norm": 1.2890625,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.0004169052631578947,
-      "loss": 0.041,
+      "loss": 0.0331,
       "num_input_tokens_seen": 222756864,
       "step": 3400,
-      "train_runtime": 1159.3705,
-      "train_tokens_per_second": 192136.047
+      "train_runtime": 1504.7401,
+      "train_tokens_per_second": 148036.77
     },
     {
       "epoch": 0.7356122890523583,
-      "eval_loss": 0.03272933140397072,
-      "eval_runtime": 1.4683,
-      "eval_samples_per_second": 21.794,
-      "eval_steps_per_second": 0.681,
+      "eval_loss": 0.02346055395901203,
+      "eval_runtime": 2.1561,
+      "eval_samples_per_second": 14.841,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 222756864,
       "step": 3400
     },
     {
       "epoch": 0.7356122890523583,
-      "eval_byte_accuracy": 0.9918397626112759,
-      "eval_chrf": 94.39060549038757,
-      "eval_sacrebleu": 92.48709733402343,
-      "eval_word_accuracy": 0.9793956043956044,
+      "eval_byte_accuracy": 0.9933234421364985,
+      "eval_chrf": 95.48377466124089,
+      "eval_sacrebleu": 94.0722064625731,
+      "eval_word_accuracy": 0.9848901098901099,
       "num_input_tokens_seen": 222756864,
-      "perplexity": 1.0332708274242337,
+      "perplexity": 1.0237379175422994,
       "step": 3400
     },
     {
       "epoch": 0.7377758546083947,
-      "grad_norm": 0.9609375,
+      "grad_norm": 0.98046875,
       "learning_rate": 0.0004162736842105263,
-      "loss": 0.0415,
+      "loss": 0.0319,
       "num_input_tokens_seen": 223412224,
       "step": 3410,
-      "train_runtime": 1164.6558,
-      "train_tokens_per_second": 191826.826
+      "train_runtime": 1512.839,
+      "train_tokens_per_second": 147677.464
     },
     {
       "epoch": 0.7399394201644309,
-      "grad_norm": 0.75390625,
+      "grad_norm": 0.83984375,
       "learning_rate": 0.00041564210526315787,
-      "loss": 0.0435,
+      "loss": 0.0348,
       "num_input_tokens_seen": 224067584,
       "step": 3420,
-      "train_runtime": 1168.4717,
-      "train_tokens_per_second": 191761.234
+      "train_runtime": 1516.657,
+      "train_tokens_per_second": 147737.809
     },
     {
       "epoch": 0.7421029857204673,
-      "grad_norm": 0.7421875,
+      "grad_norm": 0.6171875,
       "learning_rate": 0.00041501052631578945,
-      "loss": 0.0429,
+      "loss": 0.0341,
       "num_input_tokens_seen": 224722944,
       "step": 3430,
-      "train_runtime": 1172.2982,
-      "train_tokens_per_second": 191694.351
+      "train_runtime": 1520.4755,
+      "train_tokens_per_second": 147797.808
     },
     {
       "epoch": 0.7442665512765037,
-      "grad_norm": 1.1875,
+      "grad_norm": 0.6171875,
       "learning_rate": 0.000414378947368421,
-      "loss": 0.0444,
+      "loss": 0.0337,
       "num_input_tokens_seen": 225378304,
       "step": 3440,
-      "train_runtime": 1176.1143,
-      "train_tokens_per_second": 191629.6
+      "train_runtime": 1524.2832,
+      "train_tokens_per_second": 147858.554
     },
     {
       "epoch": 0.74643011683254,
-      "grad_norm": 0.76171875,
+      "grad_norm": 0.62890625,
       "learning_rate": 0.0004137473684210526,
-      "loss": 0.0467,
+      "loss": 0.038,
       "num_input_tokens_seen": 226033664,
       "step": 3450,
-      "train_runtime": 1179.9278,
-      "train_tokens_per_second": 191565.67
+      "train_runtime": 1528.0999,
+      "train_tokens_per_second": 147918.116
     },
     {
       "epoch": 0.7485936823885764,
-      "grad_norm": 0.765625,
+      "grad_norm": 0.625,
       "learning_rate": 0.00041311578947368416,
-      "loss": 0.042,
+      "loss": 0.0326,
       "num_input_tokens_seen": 226684928,
       "step": 3460,
-      "train_runtime": 1183.7227,
-      "train_tokens_per_second": 191501.723
+      "train_runtime": 1534.0356,
+      "train_tokens_per_second": 147770.314
     },
     {
       "epoch": 0.7507572479446127,
-      "grad_norm": 0.6796875,
+      "grad_norm": 0.92578125,
       "learning_rate": 0.0004124842105263158,
-      "loss": 0.0433,
+      "loss": 0.0351,
       "num_input_tokens_seen": 227340288,
       "step": 3470,
-      "train_runtime": 1187.5357,
-      "train_tokens_per_second": 191438.691
+      "train_runtime": 1537.8508,
+      "train_tokens_per_second": 147829.871
     },
     {
       "epoch": 0.7529208135006491,
-      "grad_norm": 0.7265625,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.00041185263157894733,
-      "loss": 0.0386,
+      "loss": 0.0312,
       "num_input_tokens_seen": 227995648,
       "step": 3480,
-      "train_runtime": 1191.3433,
-      "train_tokens_per_second": 191376.943
+      "train_runtime": 1541.6591,
+      "train_tokens_per_second": 147889.795
     },
     {
       "epoch": 0.7550843790566855,
-      "grad_norm": 0.64453125,
+      "grad_norm": 0.78125,
       "learning_rate": 0.0004112210526315789,
-      "loss": 0.042,
+      "loss": 0.0345,
       "num_input_tokens_seen": 228651008,
       "step": 3490,
-      "train_runtime": 1195.1587,
-      "train_tokens_per_second": 191314.349
+      "train_runtime": 1545.4724,
+      "train_tokens_per_second": 147948.943
     },
     {
       "epoch": 0.7572479446127217,
-      "grad_norm": 0.7890625,
+      "grad_norm": 0.95703125,
       "learning_rate": 0.0004105894736842105,
-      "loss": 0.0452,
+      "loss": 0.0375,
       "num_input_tokens_seen": 229306368,
       "step": 3500,
-      "train_runtime": 1198.972,
-      "train_tokens_per_second": 191252.485
+      "train_runtime": 1549.2818,
+      "train_tokens_per_second": 148008.174
     },
     {
       "epoch": 0.7572479446127217,
-      "eval_loss": 0.02840009331703186,
-      "eval_runtime": 1.49,
-      "eval_samples_per_second": 21.476,
-      "eval_steps_per_second": 0.671,
+      "eval_loss": 0.02329455316066742,
+      "eval_runtime": 2.0292,
+      "eval_samples_per_second": 15.77,
+      "eval_steps_per_second": 0.493,
       "num_input_tokens_seen": 229306368,
       "step": 3500
     },
     {
       "epoch": 0.7572479446127217,
-      "eval_byte_accuracy": 0.9944362017804155,
-      "eval_chrf": 96.0273311209701,
-      "eval_sacrebleu": 94.71325789004388,
-      "eval_word_accuracy": 0.9876373626373627,
+      "eval_byte_accuracy": 0.9940652818991098,
+      "eval_chrf": 95.63532154828998,
+      "eval_sacrebleu": 94.57672912284198,
+      "eval_word_accuracy": 0.9835164835164835,
       "num_input_tokens_seen": 229306368,
-      "perplexity": 1.0288072209830508,
+      "perplexity": 1.023567990335108,
       "step": 3500
     },
     {
       "epoch": 0.7594115101687581,
-      "grad_norm": 0.8203125,
+      "grad_norm": 0.62890625,
       "learning_rate": 0.00040995789473684204,
-      "loss": 0.0414,
+      "loss": 0.0336,
       "num_input_tokens_seen": 229961728,
       "step": 3510,
-      "train_runtime": 1204.34,
-      "train_tokens_per_second": 190944.188
+      "train_runtime": 1557.3127,
+      "train_tokens_per_second": 147665.739
     },
     {
       "epoch": 0.7615750757247944,
-      "grad_norm": 0.9453125,
+      "grad_norm": 0.67578125,
       "learning_rate": 0.0004093263157894737,
-      "loss": 0.0403,
+      "loss": 0.0311,
       "num_input_tokens_seen": 230617088,
       "step": 3520,
-      "train_runtime": 1208.1589,
-      "train_tokens_per_second": 190883.08
+      "train_runtime": 1561.1204,
+      "train_tokens_per_second": 147725.367
     },
     {
       "epoch": 0.7637386412808308,
-      "grad_norm": 0.97265625,
+      "grad_norm": 0.68359375,
       "learning_rate": 0.0004086947368421052,
-      "loss": 0.0409,
+      "loss": 0.0314,
       "num_input_tokens_seen": 231272448,
       "step": 3530,
-      "train_runtime": 1211.9754,
-      "train_tokens_per_second": 190822.731
+      "train_runtime": 1564.9367,
+      "train_tokens_per_second": 147783.895
     },
     {
       "epoch": 0.7659022068368672,
-      "grad_norm": 0.88671875,
+      "grad_norm": 0.8515625,
       "learning_rate": 0.00040806315789473685,
-      "loss": 0.0452,
+      "loss": 0.0365,
       "num_input_tokens_seen": 231927808,
       "step": 3540,
-      "train_runtime": 1215.7804,
-      "train_tokens_per_second": 190764.55
+      "train_runtime": 1568.7575,
+      "train_tokens_per_second": 147841.718
     },
     {
       "epoch": 0.7680657723929035,
-      "grad_norm": 0.85546875,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0004074315789473684,
-      "loss": 0.0409,
+      "loss": 0.032,
       "num_input_tokens_seen": 232583168,
       "step": 3550,
-      "train_runtime": 1219.5993,
-      "train_tokens_per_second": 190704.58
+      "train_runtime": 1572.5709,
+      "train_tokens_per_second": 147899.954
     },
     {
       "epoch": 0.7702293379489399,
-      "grad_norm": 1.015625,
+      "grad_norm": 0.80078125,
       "learning_rate": 0.00040679999999999997,
-      "loss": 0.04,
+      "loss": 0.0318,
       "num_input_tokens_seen": 233238528,
       "step": 3560,
-      "train_runtime": 1223.4146,
-      "train_tokens_per_second": 190645.526
+      "train_runtime": 1578.5239,
+      "train_tokens_per_second": 147757.366
     },
     {
       "epoch": 0.7723929035049762,
-      "grad_norm": 0.9375,
+      "grad_norm": 0.74609375,
       "learning_rate": 0.00040616842105263155,
-      "loss": 0.0435,
+      "loss": 0.0362,
       "num_input_tokens_seen": 233893888,
       "step": 3570,
-      "train_runtime": 1227.2262,
-      "train_tokens_per_second": 190587.425
+      "train_runtime": 1582.3272,
+      "train_tokens_per_second": 147816.385
     },
     {
       "epoch": 0.7745564690610125,
-      "grad_norm": 0.80078125,
+      "grad_norm": 0.451171875,
       "learning_rate": 0.0004055368421052631,
-      "loss": 0.0438,
+      "loss": 0.0354,
       "num_input_tokens_seen": 234549248,
       "step": 3580,
-      "train_runtime": 1231.0532,
-      "train_tokens_per_second": 190527.305
+      "train_runtime": 1586.1479,
+      "train_tokens_per_second": 147873.501
     },
     {
       "epoch": 0.7767200346170489,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.470703125,
       "learning_rate": 0.0004049052631578947,
-      "loss": 0.042,
+      "loss": 0.0333,
       "num_input_tokens_seen": 235204608,
       "step": 3590,
-      "train_runtime": 1234.8666,
-      "train_tokens_per_second": 190469.65
+      "train_runtime": 1589.9649,
+      "train_tokens_per_second": 147930.69
     },
     {
       "epoch": 0.7788836001730852,
-      "grad_norm": 0.70703125,
+      "grad_norm": 1.0703125,
       "learning_rate": 0.00040427368421052626,
-      "loss": 0.0447,
+      "loss": 0.0375,
       "num_input_tokens_seen": 235859968,
       "step": 3600,
-      "train_runtime": 1238.6912,
-      "train_tokens_per_second": 190410.628
+      "train_runtime": 1593.7746,
+      "train_tokens_per_second": 147988.287
     },
     {
       "epoch": 0.7788836001730852,
-      "eval_loss": 0.03048066981136799,
-      "eval_runtime": 1.5257,
-      "eval_samples_per_second": 20.974,
-      "eval_steps_per_second": 0.655,
+      "eval_loss": 0.020246392115950584,
+      "eval_runtime": 2.1515,
+      "eval_samples_per_second": 14.873,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 235859968,
       "step": 3600
     },
     {
       "epoch": 0.7788836001730852,
-      "eval_byte_accuracy": 0.9933234421364985,
-      "eval_chrf": 94.5949510570737,
-      "eval_sacrebleu": 92.65942721026246,
-      "eval_word_accuracy": 0.9821428571428571,
+      "eval_byte_accuracy": 0.9944362017804155,
+      "eval_chrf": 96.81607520480924,
+      "eval_sacrebleu": 96.20892605110268,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 235859968,
-      "perplexity": 1.0309499613989452,
+      "perplexity": 1.020452740563904,
       "step": 3600
     },
     {
       "epoch": 0.7810471657291216,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.64453125,
       "learning_rate": 0.0004036421052631579,
-      "loss": 0.0394,
+      "loss": 0.0331,
       "num_input_tokens_seen": 236515328,
       "step": 3610,
-      "train_runtime": 1244.0566,
-      "train_tokens_per_second": 190116.21
+      "train_runtime": 1602.0598,
+      "train_tokens_per_second": 147632.018
     },
     {
       "epoch": 0.783210731285158,
-      "grad_norm": 0.7890625,
+      "grad_norm": 0.75,
       "learning_rate": 0.00040301052631578943,
-      "loss": 0.0396,
+      "loss": 0.0335,
       "num_input_tokens_seen": 237170688,
       "step": 3620,
-      "train_runtime": 1247.873,
-      "train_tokens_per_second": 190059.955
+      "train_runtime": 1605.8778,
+      "train_tokens_per_second": 147689.128
     },
     {
       "epoch": 0.7853742968411943,
-      "grad_norm": 0.61328125,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.000402378947368421,
-      "loss": 0.0422,
+      "loss": 0.0363,
       "num_input_tokens_seen": 237821952,
       "step": 3630,
-      "train_runtime": 1251.6719,
-      "train_tokens_per_second": 190003.424
+      "train_runtime": 1609.6669,
+      "train_tokens_per_second": 147746.067
     },
     {
       "epoch": 0.7875378623972307,
-      "grad_norm": 0.74609375,
+      "grad_norm": 0.734375,
       "learning_rate": 0.0004017473684210526,
-      "loss": 0.0433,
+      "loss": 0.0356,
       "num_input_tokens_seen": 238477312,
       "step": 3640,
-      "train_runtime": 1255.4702,
-      "train_tokens_per_second": 189950.593
+      "train_runtime": 1613.462,
+      "train_tokens_per_second": 147804.73
     },
     {
       "epoch": 0.789701427953267,
-      "grad_norm": 0.90625,
+      "grad_norm": 0.60546875,
       "learning_rate": 0.00040111578947368414,
-      "loss": 0.0405,
+      "loss": 0.0343,
       "num_input_tokens_seen": 239132672,
       "step": 3650,
-      "train_runtime": 1259.2841,
-      "train_tokens_per_second": 189895.733
+      "train_runtime": 1617.2673,
+      "train_tokens_per_second": 147862.178
     },
     {
       "epoch": 0.7918649935093033,
-      "grad_norm": 0.81640625,
+      "grad_norm": 0.984375,
       "learning_rate": 0.0004004842105263158,
-      "loss": 0.0427,
+      "loss": 0.0359,
       "num_input_tokens_seen": 239788032,
       "step": 3660,
-      "train_runtime": 1263.0992,
-      "train_tokens_per_second": 189841.019
+      "train_runtime": 1623.2242,
+      "train_tokens_per_second": 147723.298
     },
     {
       "epoch": 0.7940285590653396,
-      "grad_norm": 0.92578125,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.0003998526315789473,
-      "loss": 0.0447,
+      "loss": 0.0372,
       "num_input_tokens_seen": 240443392,
       "step": 3670,
-      "train_runtime": 1266.9137,
-      "train_tokens_per_second": 189786.715
+      "train_runtime": 1627.027,
+      "train_tokens_per_second": 147780.827
     },
     {
       "epoch": 0.796192124621376,
-      "grad_norm": 0.77734375,
+      "grad_norm": 0.9921875,
       "learning_rate": 0.00039922105263157895,
-      "loss": 0.04,
+      "loss": 0.0322,
       "num_input_tokens_seen": 241098752,
       "step": 3680,
-      "train_runtime": 1270.7206,
-      "train_tokens_per_second": 189733.882
+      "train_runtime": 1630.8244,
+      "train_tokens_per_second": 147838.573
     },
     {
       "epoch": 0.7983556901774124,
-      "grad_norm": 0.83984375,
+      "grad_norm": 0.859375,
       "learning_rate": 0.0003985894736842105,
-      "loss": 0.0432,
+      "loss": 0.0368,
       "num_input_tokens_seen": 241754112,
       "step": 3690,
-      "train_runtime": 1274.5299,
-      "train_tokens_per_second": 189680.998
+      "train_runtime": 1634.6421,
+      "train_tokens_per_second": 147894.218
     },
     {
       "epoch": 0.8005192557334487,
-      "grad_norm": 0.70703125,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.0003979578947368421,
-      "loss": 0.0751,
+      "loss": 0.0657,
       "num_input_tokens_seen": 242409472,
       "step": 3700,
-      "train_runtime": 1278.3442,
-      "train_tokens_per_second": 189627.701
+      "train_runtime": 1638.4625,
+      "train_tokens_per_second": 147949.357
     },
     {
       "epoch": 0.8005192557334487,
-      "eval_loss": 0.02744060382246971,
-      "eval_runtime": 1.8855,
-      "eval_samples_per_second": 16.971,
-      "eval_steps_per_second": 0.53,
+      "eval_loss": 0.027874412015080452,
+      "eval_runtime": 2.3954,
+      "eval_samples_per_second": 13.359,
+      "eval_steps_per_second": 0.417,
       "num_input_tokens_seen": 242409472,
       "step": 3700
     },
     {
       "epoch": 0.8005192557334487,
-      "eval_byte_accuracy": 0.994807121661721,
-      "eval_chrf": 95.85159781134789,
-      "eval_sacrebleu": 93.59317163487975,
-      "eval_word_accuracy": 0.9848901098901099,
+      "eval_byte_accuracy": 0.9925816023738873,
+      "eval_chrf": 93.66102977336031,
+      "eval_sacrebleu": 92.05374753299421,
+      "eval_word_accuracy": 0.9821428571428571,
       "num_input_tokens_seen": 242409472,
-      "perplexity": 1.0278205646814604,
+      "perplexity": 1.0282665383894833,
       "step": 3700
     },
     {
       "epoch": 0.8026828212894851,
-      "grad_norm": 0.8515625,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.00039732631578947366,
-      "loss": 0.0393,
+      "loss": 0.0331,
       "num_input_tokens_seen": 243064832,
       "step": 3710,
-      "train_runtime": 1284.0643,
-      "train_tokens_per_second": 189293.349
+      "train_runtime": 1646.8333,
+      "train_tokens_per_second": 147595.283
     },
     {
       "epoch": 0.8048463868455215,
-      "grad_norm": 0.9140625,
+      "grad_norm": 0.6484375,
       "learning_rate": 0.0003966947368421052,
-      "loss": 0.0396,
+      "loss": 0.0335,
       "num_input_tokens_seen": 243720192,
       "step": 3720,
-      "train_runtime": 1287.8863,
-      "train_tokens_per_second": 189240.46
+      "train_runtime": 1650.6505,
+      "train_tokens_per_second": 147650.997
     },
     {
       "epoch": 0.8070099524015578,
-      "grad_norm": 0.59375,
+      "grad_norm": 0.35546875,
       "learning_rate": 0.00039606315789473683,
-      "loss": 0.0359,
+      "loss": 0.0308,
       "num_input_tokens_seen": 244375552,
       "step": 3730,
-      "train_runtime": 1291.7103,
-      "train_tokens_per_second": 189187.584
+      "train_runtime": 1654.463,
+      "train_tokens_per_second": 147706.872
     },
     {
       "epoch": 0.8091735179575941,
-      "grad_norm": 0.875,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.00039543157894736836,
-      "loss": 0.0384,
+      "loss": 0.0319,
       "num_input_tokens_seen": 245030912,
       "step": 3740,
-      "train_runtime": 1295.5185,
-      "train_tokens_per_second": 189137.327
+      "train_runtime": 1658.2795,
+      "train_tokens_per_second": 147762.13
     },
     {
       "epoch": 0.8113370835136304,
-      "grad_norm": 0.86328125,
+      "grad_norm": 0.51953125,
       "learning_rate": 0.0003948,
-      "loss": 0.0383,
+      "loss": 0.0321,
       "num_input_tokens_seen": 245686272,
       "step": 3750,
-      "train_runtime": 1299.337,
-      "train_tokens_per_second": 189085.878
+      "train_runtime": 1662.0901,
+      "train_tokens_per_second": 147817.663
     },
     {
       "epoch": 0.8135006490696668,
-      "grad_norm": 0.66015625,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.00039416842105263153,
-      "loss": 0.0401,
+      "loss": 0.0323,
       "num_input_tokens_seen": 246337536,
       "step": 3760,
-      "train_runtime": 1303.1231,
-      "train_tokens_per_second": 189036.279
+      "train_runtime": 1668.0409,
+      "train_tokens_per_second": 147680.756
     },
     {
       "epoch": 0.8156642146257032,
-      "grad_norm": 0.75,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.0003935368421052632,
-      "loss": 0.0424,
+      "loss": 0.0361,
       "num_input_tokens_seen": 246988800,
       "step": 3770,
-      "train_runtime": 1306.9192,
-      "train_tokens_per_second": 188985.517
+      "train_runtime": 1671.8351,
+      "train_tokens_per_second": 147735.145
     },
     {
       "epoch": 0.8178277801817395,
-      "grad_norm": 0.89453125,
+      "grad_norm": 0.94140625,
       "learning_rate": 0.0003929052631578947,
-      "loss": 0.0392,
+      "loss": 0.033,
       "num_input_tokens_seen": 247644160,
       "step": 3780,
-      "train_runtime": 1310.7374,
-      "train_tokens_per_second": 188934.986
+      "train_runtime": 1675.649,
+      "train_tokens_per_second": 147789.994
     },
     {
       "epoch": 0.8199913457377759,
-      "grad_norm": 0.80078125,
+      "grad_norm": 0.6875,
       "learning_rate": 0.00039227368421052624,
-      "loss": 0.0377,
+      "loss": 0.0325,
       "num_input_tokens_seen": 248299520,
       "step": 3790,
-      "train_runtime": 1314.5417,
-      "train_tokens_per_second": 188886.762
+      "train_runtime": 1679.4499,
+      "train_tokens_per_second": 147845.743
     },
     {
       "epoch": 0.8221549112938122,
-      "grad_norm": 0.875,
+      "grad_norm": 0.6328125,
       "learning_rate": 0.0003916421052631579,
-      "loss": 0.0386,
+      "loss": 0.0329,
       "num_input_tokens_seen": 248954880,
       "step": 3800,
-      "train_runtime": 1318.3632,
-      "train_tokens_per_second": 188836.345
+      "train_runtime": 1683.2644,
+      "train_tokens_per_second": 147900.044
     },
     {
       "epoch": 0.8221549112938122,
-      "eval_loss": 0.02655559405684471,
-      "eval_runtime": 1.8736,
-      "eval_samples_per_second": 17.079,
-      "eval_steps_per_second": 0.534,
+      "eval_loss": 0.023587677627801895,
+      "eval_runtime": 2.0373,
+      "eval_samples_per_second": 15.707,
+      "eval_steps_per_second": 0.491,
       "num_input_tokens_seen": 248954880,
       "step": 3800
     },
     {
       "epoch": 0.8221549112938122,
       "eval_byte_accuracy": 0.9940652818991098,
-      "eval_chrf": 94.92109122001662,
-      "eval_sacrebleu": 94.00295930011256,
-      "eval_word_accuracy": 0.9835164835164835,
+      "eval_chrf": 95.56456555243946,
+      "eval_sacrebleu": 94.62541130797075,
+      "eval_word_accuracy": 0.9848901098901099,
       "num_input_tokens_seen": 248954880,
-      "perplexity": 1.026911335841954,
+      "perplexity": 1.0238680671346283,
       "step": 3800
     },
     {
       "epoch": 0.8243184768498486,
-      "grad_norm": 0.8984375,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.0003910105263157894,
-      "loss": 0.038,
+      "loss": 0.0315,
       "num_input_tokens_seen": 249610240,
       "step": 3810,
-      "train_runtime": 1324.0597,
-      "train_tokens_per_second": 188518.873
+      "train_runtime": 1691.2895,
+      "train_tokens_per_second": 147585.76
     },
     {
       "epoch": 0.8264820424058849,
-      "grad_norm": 0.7421875,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.00039037894736842105,
-      "loss": 0.0379,
+      "loss": 0.032,
       "num_input_tokens_seen": 250265600,
       "step": 3820,
-      "train_runtime": 1327.8807,
-      "train_tokens_per_second": 188469.945
+      "train_runtime": 1695.1146,
+      "train_tokens_per_second": 147639.343
     },
     {
       "epoch": 0.8286456079619212,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.453125,
       "learning_rate": 0.0003897473684210526,
-      "loss": 0.0359,
+      "loss": 0.0306,
       "num_input_tokens_seen": 250920960,
       "step": 3830,
-      "train_runtime": 1331.6986,
-      "train_tokens_per_second": 188421.738
+      "train_runtime": 1698.9403,
+      "train_tokens_per_second": 147692.632
     },
     {
       "epoch": 0.8308091735179576,
-      "grad_norm": 0.97265625,
+      "grad_norm": 0.66796875,
       "learning_rate": 0.0003891157894736842,
-      "loss": 0.041,
+      "loss": 0.0326,
       "num_input_tokens_seen": 251576320,
       "step": 3840,
-      "train_runtime": 1335.5172,
-      "train_tokens_per_second": 188373.698
+      "train_runtime": 1702.7553,
+      "train_tokens_per_second": 147746.612
     },
     {
       "epoch": 0.8329727390739939,
-      "grad_norm": 0.9140625,
+      "grad_norm": 0.5859375,
       "learning_rate": 0.00038848421052631576,
-      "loss": 0.0386,
+      "loss": 0.0306,
       "num_input_tokens_seen": 252231680,
       "step": 3850,
-      "train_runtime": 1339.3407,
-      "train_tokens_per_second": 188325.258
+      "train_runtime": 1706.5717,
+      "train_tokens_per_second": 147800.22
     },
     {
       "epoch": 0.8351363046300303,
       "grad_norm": 0.859375,
       "learning_rate": 0.0003878526315789473,
-      "loss": 0.0411,
+      "loss": 0.0349,
       "num_input_tokens_seen": 252887040,
       "step": 3860,
-      "train_runtime": 1343.1562,
-      "train_tokens_per_second": 188278.21
+      "train_runtime": 1712.5469,
+      "train_tokens_per_second": 147667.224
     },
     {
       "epoch": 0.8372998701860667,
-      "grad_norm": 0.9609375,
+      "grad_norm": 0.7109375,
       "learning_rate": 0.00038722105263157893,
-      "loss": 0.0393,
+      "loss": 0.0327,
       "num_input_tokens_seen": 253542400,
       "step": 3870,
-      "train_runtime": 1346.9779,
-      "train_tokens_per_second": 188230.558
+      "train_runtime": 1716.3594,
+      "train_tokens_per_second": 147721.046
     },
     {
       "epoch": 0.839463435742103,
-      "grad_norm": 0.71484375,
+      "grad_norm": 0.8515625,
       "learning_rate": 0.00038658947368421046,
-      "loss": 0.0748,
+      "loss": 0.0721,
       "num_input_tokens_seen": 254197760,
       "step": 3880,
-      "train_runtime": 1350.7997,
-      "train_tokens_per_second": 188183.161
+      "train_runtime": 1720.1745,
+      "train_tokens_per_second": 147774.405
     },
     {
       "epoch": 0.8416270012981394,
-      "grad_norm": 0.5703125,
+      "grad_norm": 0.6953125,
       "learning_rate": 0.0003859578947368421,
-      "loss": 0.0418,
+      "loss": 0.0352,
       "num_input_tokens_seen": 254853120,
       "step": 3890,
-      "train_runtime": 1354.6168,
-      "train_tokens_per_second": 188136.694
+      "train_runtime": 1723.9877,
+      "train_tokens_per_second": 147827.693
     },
     {
       "epoch": 0.8437905668541756,
-      "grad_norm": 0.77734375,
+      "grad_norm": 0.68359375,
       "learning_rate": 0.00038532631578947363,
-      "loss": 0.0395,
+      "loss": 0.032,
       "num_input_tokens_seen": 255508480,
       "step": 3900,
-      "train_runtime": 1358.4315,
-      "train_tokens_per_second": 188090.809
+      "train_runtime": 1727.7959,
+      "train_tokens_per_second": 147881.17
     },
     {
       "epoch": 0.8437905668541756,
-      "eval_loss": 0.026101363822817802,
-      "eval_runtime": 1.759,
-      "eval_samples_per_second": 18.192,
-      "eval_steps_per_second": 0.568,
+      "eval_loss": 0.021019412204623222,
+      "eval_runtime": 2.1668,
+      "eval_samples_per_second": 14.769,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 255508480,
       "step": 3900
     },
     {
       "epoch": 0.8437905668541756,
-      "eval_byte_accuracy": 0.9929525222551929,
-      "eval_chrf": 95.31269349287258,
-      "eval_sacrebleu": 94.24185400796716,
-      "eval_word_accuracy": 0.9821428571428571,
+      "eval_byte_accuracy": 0.9940652818991098,
+      "eval_chrf": 95.6553581195663,
+      "eval_sacrebleu": 94.6463989521252,
+      "eval_word_accuracy": 0.9835164835164835,
       "num_input_tokens_seen": 255508480,
-      "perplexity": 1.026444987588306,
+      "perplexity": 1.0212418760013775,
       "step": 3900
     },
     {
       "epoch": 0.845954132410212,
-      "grad_norm": 0.625,
+      "grad_norm": 0.74609375,
       "learning_rate": 0.0003846947368421053,
-      "loss": 0.0377,
+      "loss": 0.032,
       "num_input_tokens_seen": 256163840,
       "step": 3910,
-      "train_runtime": 1364.0191,
-      "train_tokens_per_second": 187800.768
+      "train_runtime": 1735.9497,
+      "train_tokens_per_second": 147564.095
     },
     {
       "epoch": 0.8481176979662484,
-      "grad_norm": 0.765625,
+      "grad_norm": 0.6796875,
       "learning_rate": 0.0003840631578947368,
-      "loss": 0.0383,
+      "loss": 0.033,
       "num_input_tokens_seen": 256819200,
       "step": 3920,
-      "train_runtime": 1367.8422,
-      "train_tokens_per_second": 187754.995
+      "train_runtime": 1739.7714,
+      "train_tokens_per_second": 147616.637
     },
     {
       "epoch": 0.8502812635222847,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.75,
       "learning_rate": 0.00038343157894736834,
-      "loss": 0.0391,
+      "loss": 0.0322,
       "num_input_tokens_seen": 257474560,
       "step": 3930,
-      "train_runtime": 1371.6627,
-      "train_tokens_per_second": 187709.814
+      "train_runtime": 1743.5864,
+      "train_tokens_per_second": 147669.512
     },
     {
       "epoch": 0.8524448290783211,
-      "grad_norm": 0.52734375,
+      "grad_norm": 0.6484375,
       "learning_rate": 0.0003828,
-      "loss": 0.0406,
+      "loss": 0.0341,
       "num_input_tokens_seen": 258129920,
       "step": 3940,
-      "train_runtime": 1375.4817,
-      "train_tokens_per_second": 187665.108
+      "train_runtime": 1747.3972,
+      "train_tokens_per_second": 147722.52
     },
     {
       "epoch": 0.8546083946343574,
-      "grad_norm": 0.61328125,
+      "grad_norm": 0.64453125,
       "learning_rate": 0.0003821684210526315,
-      "loss": 0.0362,
+      "loss": 0.0287,
       "num_input_tokens_seen": 258785280,
       "step": 3950,
-      "train_runtime": 1379.2745,
-      "train_tokens_per_second": 187624.208
+      "train_runtime": 1751.1973,
+      "train_tokens_per_second": 147776.198
     },
     {
       "epoch": 0.8567719601903938,
-      "grad_norm": 0.94140625,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.00038153684210526315,
-      "loss": 0.0413,
+      "loss": 0.0342,
       "num_input_tokens_seen": 259440640,
       "step": 3960,
-      "train_runtime": 1383.0876,
-      "train_tokens_per_second": 187580.778
+      "train_runtime": 1756.985,
+      "train_tokens_per_second": 147662.414
     },
     {
       "epoch": 0.8589355257464301,
-      "grad_norm": 0.6640625,
+      "grad_norm": 0.58203125,
       "learning_rate": 0.0003809052631578947,
-      "loss": 0.0332,
+      "loss": 0.0253,
       "num_input_tokens_seen": 260096000,
       "step": 3970,
-      "train_runtime": 1386.8873,
-      "train_tokens_per_second": 187539.385
+      "train_runtime": 1760.7782,
+      "train_tokens_per_second": 147716.507
     },
     {
       "epoch": 0.8610990913024664,
-      "grad_norm": 0.71484375,
+      "grad_norm": 0.71875,
       "learning_rate": 0.0003802736842105263,
-      "loss": 0.0369,
+      "loss": 0.0289,
       "num_input_tokens_seen": 260751360,
       "step": 3980,
-      "train_runtime": 1390.6976,
-      "train_tokens_per_second": 187496.803
+      "train_runtime": 1764.5928,
+      "train_tokens_per_second": 147768.576
     },
     {
       "epoch": 0.8632626568585028,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.78125,
       "learning_rate": 0.00037964210526315786,
-      "loss": 0.0393,
+      "loss": 0.0324,
       "num_input_tokens_seen": 261406720,
       "step": 3990,
-      "train_runtime": 1394.514,
-      "train_tokens_per_second": 187453.636
+      "train_runtime": 1768.3999,
+      "train_tokens_per_second": 147821.046
     },
     {
       "epoch": 0.8654262224145391,
-      "grad_norm": 0.8359375,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.0003790105263157894,
-      "loss": 0.0374,
+      "loss": 0.0306,
       "num_input_tokens_seen": 262062080,
       "step": 4000,
-      "train_runtime": 1398.333,
-      "train_tokens_per_second": 187410.351
+      "train_runtime": 1772.2207,
+      "train_tokens_per_second": 147872.143
     },
     {
       "epoch": 0.8654262224145391,
-      "eval_loss": 0.024643762037158012,
-      "eval_runtime": 1.7566,
-      "eval_samples_per_second": 18.217,
-      "eval_steps_per_second": 0.569,
+      "eval_loss": 0.018921982496976852,
+      "eval_runtime": 2.1514,
+      "eval_samples_per_second": 14.874,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 262062080,
       "step": 4000
     },
     {
       "epoch": 0.8654262224145391,
-      "eval_byte_accuracy": 0.9933234421364985,
-      "eval_chrf": 94.19624263012378,
-      "eval_sacrebleu": 93.71978495314545,
-      "eval_word_accuracy": 0.9821428571428571,
+      "eval_byte_accuracy": 0.9936943620178041,
+      "eval_chrf": 95.4335711604272,
+      "eval_sacrebleu": 93.94661550016679,
+      "eval_word_accuracy": 0.9835164835164835,
       "num_input_tokens_seen": 262062080,
-      "perplexity": 1.0249499294059543,
+      "perplexity": 1.0191021377117173,
       "step": 4000
     },
     {
       "epoch": 0.8675897879705755,
-      "grad_norm": 0.703125,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.00037837894736842103,
-      "loss": 0.0389,
+      "loss": 0.0315,
       "num_input_tokens_seen": 262717440,
       "step": 4010,
-      "train_runtime": 1403.9596,
-      "train_tokens_per_second": 187126.066
+      "train_runtime": 1780.524,
+      "train_tokens_per_second": 147550.629
     },
     {
       "epoch": 0.8697533535266119,
-      "grad_norm": 0.94921875,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.00037774736842105256,
-      "loss": 0.0375,
+      "loss": 0.0305,
       "num_input_tokens_seen": 263372800,
       "step": 4020,
-      "train_runtime": 1407.7746,
-      "train_tokens_per_second": 187084.494
+      "train_runtime": 1784.3244,
+      "train_tokens_per_second": 147603.655
     },
     {
       "epoch": 0.8719169190826482,
-      "grad_norm": 0.76171875,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.0003771157894736842,
-      "loss": 0.0361,
+      "loss": 0.0286,
       "num_input_tokens_seen": 264028160,
       "step": 4030,
-      "train_runtime": 1411.5751,
-      "train_tokens_per_second": 187045.062
+      "train_runtime": 1788.12,
+      "train_tokens_per_second": 147656.847
     },
     {
       "epoch": 0.8740804846386846,
-      "grad_norm": 0.70703125,
+      "grad_norm": 0.494140625,
       "learning_rate": 0.00037648421052631574,
-      "loss": 0.0344,
+      "loss": 0.0289,
       "num_input_tokens_seen": 264683520,
       "step": 4040,
-      "train_runtime": 1415.3795,
-      "train_tokens_per_second": 187005.335
+      "train_runtime": 1791.923,
+      "train_tokens_per_second": 147709.207
     },
     {
       "epoch": 0.8762440501947208,
-      "grad_norm": 0.6953125,
+      "grad_norm": 0.609375,
       "learning_rate": 0.0003758526315789474,
-      "loss": 0.0375,
+      "loss": 0.031,
       "num_input_tokens_seen": 265338880,
       "step": 4050,
-      "train_runtime": 1419.1956,
-      "train_tokens_per_second": 186964.276
+      "train_runtime": 1795.7271,
+      "train_tokens_per_second": 147761.251
     },
     {
       "epoch": 0.8784076157507572,
-      "grad_norm": 0.6796875,
+      "grad_norm": 0.59375,
       "learning_rate": 0.0003752210526315789,
-      "loss": 0.037,
+      "loss": 0.0295,
       "num_input_tokens_seen": 265994240,
       "step": 4060,
-      "train_runtime": 1423.0053,
-      "train_tokens_per_second": 186924.278
+      "train_runtime": 1801.6832,
+      "train_tokens_per_second": 147636.518
     },
     {
       "epoch": 0.8805711813067936,
-      "grad_norm": 0.8203125,
+      "grad_norm": 0.703125,
       "learning_rate": 0.00037458947368421044,
-      "loss": 0.0369,
+      "loss": 0.0298,
       "num_input_tokens_seen": 266649600,
       "step": 4070,
-      "train_runtime": 1426.8192,
-      "train_tokens_per_second": 186883.94
+      "train_runtime": 1805.4894,
+      "train_tokens_per_second": 147688.264
     },
     {
       "epoch": 0.8827347468628299,
-      "grad_norm": 0.73046875,
+      "grad_norm": 0.81640625,
       "learning_rate": 0.0003739578947368421,
-      "loss": 0.037,
+      "loss": 0.0309,
       "num_input_tokens_seen": 267304960,
       "step": 4080,
-      "train_runtime": 1430.6353,
-      "train_tokens_per_second": 186843.537
+      "train_runtime": 1809.3053,
+      "train_tokens_per_second": 147739.004
     },
     {
       "epoch": 0.8848983124188663,
-      "grad_norm": 0.82421875,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.0003733263157894736,
-      "loss": 0.0365,
+      "loss": 0.0295,
       "num_input_tokens_seen": 267960320,
       "step": 4090,
-      "train_runtime": 1434.4492,
-      "train_tokens_per_second": 186803.634
+      "train_runtime": 1813.1124,
+      "train_tokens_per_second": 147790.242
     },
     {
       "epoch": 0.8870618779749027,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.4765625,
       "learning_rate": 0.00037269473684210525,
-      "loss": 0.037,
+      "loss": 0.0302,
       "num_input_tokens_seen": 268615680,
       "step": 4100,
-      "train_runtime": 1438.2647,
-      "train_tokens_per_second": 186763.736
+      "train_runtime": 1816.9268,
+      "train_tokens_per_second": 147840.673
     },
     {
       "epoch": 0.8870618779749027,
-      "eval_loss": 0.0261303149163723,
-      "eval_runtime": 1.6603,
-      "eval_samples_per_second": 19.273,
-      "eval_steps_per_second": 0.602,
+      "eval_loss": 0.018717404454946518,
+      "eval_runtime": 2.135,
+      "eval_samples_per_second": 14.988,
+      "eval_steps_per_second": 0.468,
       "num_input_tokens_seen": 268615680,
       "step": 4100
     },
     {
       "epoch": 0.8870618779749027,
-      "eval_byte_accuracy": 0.9933234421364985,
-      "eval_chrf": 94.41058320134658,
-      "eval_sacrebleu": 92.20217616355528,
-      "eval_word_accuracy": 0.9807692307692307,
+      "eval_byte_accuracy": 0.9944362017804155,
+      "eval_chrf": 95.60721087616592,
+      "eval_sacrebleu": 95.5313400244968,
+      "eval_word_accuracy": 0.9862637362637363,
       "num_input_tokens_seen": 268615680,
-      "perplexity": 1.0264747047233398,
+      "perplexity": 1.0188936731161218,
       "step": 4100
     },
     {
       "epoch": 0.889225443530939,
-      "grad_norm": 0.63671875,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.0003720631578947368,
-      "loss": 0.0336,
+      "loss": 0.028,
       "num_input_tokens_seen": 269271040,
       "step": 4110,
-      "train_runtime": 1443.7537,
-      "train_tokens_per_second": 186507.598
+      "train_runtime": 1825.0501,
+      "train_tokens_per_second": 147541.722
     },
     {
       "epoch": 0.8913890090869754,
-      "grad_norm": 0.83203125,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.00037143157894736843,
-      "loss": 0.038,
+      "loss": 0.0317,
       "num_input_tokens_seen": 269926400,
       "step": 4120,
-      "train_runtime": 1447.5712,
-      "train_tokens_per_second": 186468.482
+      "train_runtime": 1828.8662,
+      "train_tokens_per_second": 147592.21
     },
     {
       "epoch": 0.8935525746430116,
-      "grad_norm": 1.359375,
+      "grad_norm": 0.90625,
       "learning_rate": 0.00037079999999999996,
-      "loss": 0.0386,
+      "loss": 0.031,
       "num_input_tokens_seen": 270581760,
       "step": 4130,
-      "train_runtime": 1451.3821,
-      "train_tokens_per_second": 186430.404
+      "train_runtime": 1832.6868,
+      "train_tokens_per_second": 147642.115
     },
     {
       "epoch": 0.895716140199048,
-      "grad_norm": 0.74609375,
+      "grad_norm": 0.78125,
       "learning_rate": 0.0003701684210526315,
-      "loss": 0.0405,
+      "loss": 0.0334,
       "num_input_tokens_seen": 271237120,
       "step": 4140,
-      "train_runtime": 1455.1991,
-      "train_tokens_per_second": 186391.754
+      "train_runtime": 1836.4971,
+      "train_tokens_per_second": 147692.644
     },
     {
       "epoch": 0.8978797057550844,
-      "grad_norm": 0.890625,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.00036953684210526313,
-      "loss": 0.0353,
+      "loss": 0.0298,
       "num_input_tokens_seen": 271892480,
       "step": 4150,
-      "train_runtime": 1459.023,
-      "train_tokens_per_second": 186352.426
+      "train_runtime": 1840.3089,
+      "train_tokens_per_second": 147742.851
     },
     {
       "epoch": 0.9000432713111207,
-      "grad_norm": 0.87109375,
+      "grad_norm": 0.515625,
       "learning_rate": 0.00036890526315789466,
-      "loss": 0.04,
+      "loss": 0.0329,
       "num_input_tokens_seen": 272547840,
       "step": 4160,
-      "train_runtime": 1462.8409,
-      "train_tokens_per_second": 186314.065
+      "train_runtime": 1846.4577,
+      "train_tokens_per_second": 147605.789
     },
     {
       "epoch": 0.9022068368671571,
-      "grad_norm": 0.734375,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.0003682736842105263,
-      "loss": 0.0354,
+      "loss": 0.0292,
       "num_input_tokens_seen": 273203200,
       "step": 4170,
-      "train_runtime": 1466.6408,
-      "train_tokens_per_second": 186278.193
+      "train_runtime": 1850.249,
+      "train_tokens_per_second": 147657.532
     },
     {
       "epoch": 0.9043704024231934,
-      "grad_norm": 0.6328125,
+      "grad_norm": 0.48828125,
       "learning_rate": 0.00036764210526315784,
-      "loss": 0.0353,
+      "loss": 0.0287,
       "num_input_tokens_seen": 273858560,
       "step": 4180,
-      "train_runtime": 1470.4587,
-      "train_tokens_per_second": 186240.221
+      "train_runtime": 1854.0593,
+      "train_tokens_per_second": 147707.553
     },
     {
       "epoch": 0.9065339679792298,
-      "grad_norm": 0.73828125,
+      "grad_norm": 0.578125,
       "learning_rate": 0.0003670105263157895,
-      "loss": 0.0401,
+      "loss": 0.0327,
       "num_input_tokens_seen": 274513920,
       "step": 4190,
-      "train_runtime": 1474.2745,
-      "train_tokens_per_second": 186202.717
+      "train_runtime": 1857.8704,
+      "train_tokens_per_second": 147757.306
     },
     {
       "epoch": 0.9086975335352662,
-      "grad_norm": 0.9921875,
+      "grad_norm": 0.6015625,
       "learning_rate": 0.000366378947368421,
-      "loss": 0.0387,
+      "loss": 0.031,
       "num_input_tokens_seen": 275165184,
       "step": 4200,
-      "train_runtime": 1478.0643,
-      "train_tokens_per_second": 186165.908
+      "train_runtime": 1861.6603,
+      "train_tokens_per_second": 147806.335
     },
     {
       "epoch": 0.9086975335352662,
-      "eval_loss": 0.025060011073946953,
-      "eval_runtime": 1.8349,
-      "eval_samples_per_second": 17.439,
-      "eval_steps_per_second": 0.545,
+      "eval_loss": 0.017941121011972427,
+      "eval_runtime": 1.9902,
+      "eval_samples_per_second": 16.079,
+      "eval_steps_per_second": 0.502,
       "num_input_tokens_seen": 275165184,
       "step": 4200
     },
     {
       "epoch": 0.9086975335352662,
-      "eval_byte_accuracy": 0.9936943620178041,
-      "eval_chrf": 95.7358495183548,
-      "eval_sacrebleu": 94.40739967319129,
-      "eval_word_accuracy": 0.9862637362637363,
+      "eval_byte_accuracy": 0.9955489614243324,
+      "eval_chrf": 97.07386912321206,
+      "eval_sacrebleu": 95.99172930598161,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 275165184,
-      "perplexity": 1.025376652632231,
+      "perplexity": 1.0181030297489129,
       "step": 4200
     },
     {
       "epoch": 0.9108610990913024,
-      "grad_norm": 0.84765625,
+      "grad_norm": 0.70703125,
       "learning_rate": 0.00036574736842105254,
-      "loss": 0.0399,
+      "loss": 0.0324,
       "num_input_tokens_seen": 275820544,
       "step": 4210,
-      "train_runtime": 1483.7398,
-      "train_tokens_per_second": 185895.497
+      "train_runtime": 1869.6358,
+      "train_tokens_per_second": 147526.35
     },
     {
       "epoch": 0.9130246646473388,
-      "grad_norm": 1.2265625,
+      "grad_norm": 0.46484375,
       "learning_rate": 0.0003651157894736842,
-      "loss": 0.0403,
+      "loss": 0.0293,
       "num_input_tokens_seen": 276475904,
       "step": 4220,
-      "train_runtime": 1487.5562,
-      "train_tokens_per_second": 185859.129
+      "train_runtime": 1873.4536,
+      "train_tokens_per_second": 147575.532
     },
     {
       "epoch": 0.9151882302033751,
-      "grad_norm": 0.6171875,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.0003644842105263157,
-      "loss": 0.0697,
+      "loss": 0.0647,
       "num_input_tokens_seen": 277131264,
       "step": 4230,
-      "train_runtime": 1491.3729,
-      "train_tokens_per_second": 185822.919
+      "train_runtime": 1877.2653,
+      "train_tokens_per_second": 147624.983
     },
     {
       "epoch": 0.9173517957594115,
-      "grad_norm": 0.76953125,
+      "grad_norm": 0.466796875,
       "learning_rate": 0.00036385263157894736,
-      "loss": 0.0414,
+      "loss": 0.0344,
       "num_input_tokens_seen": 277786624,
       "step": 4240,
-      "train_runtime": 1495.1934,
-      "train_tokens_per_second": 185786.418
+      "train_runtime": 1881.0625,
+      "train_tokens_per_second": 147675.379
     },
     {
       "epoch": 0.9195153613154479,
-      "grad_norm": 0.484375,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0003632210526315789,
-      "loss": 0.0373,
+      "loss": 0.0298,
       "num_input_tokens_seen": 278441984,
       "step": 4250,
-      "train_runtime": 1499.009,
-      "train_tokens_per_second": 185750.71
+      "train_runtime": 1884.8718,
+      "train_tokens_per_second": 147724.625
     },
     {
       "epoch": 0.9216789268714842,
-      "grad_norm": 0.7421875,
+      "grad_norm": 0.703125,
       "learning_rate": 0.00036258947368421053,
-      "loss": 0.0331,
+      "loss": 0.0272,
       "num_input_tokens_seen": 279097344,
       "step": 4260,
-      "train_runtime": 1502.8257,
-      "train_tokens_per_second": 185715.05
+      "train_runtime": 1890.8449,
+      "train_tokens_per_second": 147604.568
     },
     {
       "epoch": 0.9238424924275206,
-      "grad_norm": 0.64453125,
+      "grad_norm": 0.5,
       "learning_rate": 0.00036195789473684206,
-      "loss": 0.0345,
+      "loss": 0.0279,
       "num_input_tokens_seen": 279748608,
       "step": 4270,
-      "train_runtime": 1506.619,
-      "train_tokens_per_second": 185679.735
+      "train_runtime": 1894.6384,
+      "train_tokens_per_second": 147652.772
     },
     {
       "epoch": 0.926006057983557,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.67578125,
       "learning_rate": 0.0003613263157894736,
-      "loss": 0.0363,
+      "loss": 0.0305,
       "num_input_tokens_seen": 280403968,
       "step": 4280,
-      "train_runtime": 1510.4341,
-      "train_tokens_per_second": 185644.628
+      "train_runtime": 1898.4464,
+      "train_tokens_per_second": 147701.806
     },
     {
       "epoch": 0.9281696235395932,
-      "grad_norm": 0.58984375,
+      "grad_norm": 0.69140625,
       "learning_rate": 0.00036069473684210523,
-      "loss": 0.0367,
+      "loss": 0.0312,
       "num_input_tokens_seen": 281059328,
       "step": 4290,
-      "train_runtime": 1514.2516,
-      "train_tokens_per_second": 185609.403
+      "train_runtime": 1902.2624,
+      "train_tokens_per_second": 147750.028
     },
     {
       "epoch": 0.9303331890956296,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.59765625,
       "learning_rate": 0.00036006315789473677,
-      "loss": 0.0371,
+      "loss": 0.0315,
       "num_input_tokens_seen": 281714688,
       "step": 4300,
-      "train_runtime": 1518.064,
-      "train_tokens_per_second": 185574.972
+      "train_runtime": 1906.0694,
+      "train_tokens_per_second": 147798.754
     },
     {
       "epoch": 0.9303331890956296,
-      "eval_loss": 0.026061272248625755,
-      "eval_runtime": 1.6122,
-      "eval_samples_per_second": 19.849,
-      "eval_steps_per_second": 0.62,
+      "eval_loss": 0.01897530071437359,
+      "eval_runtime": 2.0156,
+      "eval_samples_per_second": 15.876,
+      "eval_steps_per_second": 0.496,
       "num_input_tokens_seen": 281714688,
       "step": 4300
     },
     {
       "epoch": 0.9303331890956296,
       "eval_byte_accuracy": 0.9936943620178041,
-      "eval_chrf": 94.55529374351134,
-      "eval_sacrebleu": 92.55941688947077,
-      "eval_word_accuracy": 0.9821428571428571,
+      "eval_chrf": 95.0236625071108,
+      "eval_sacrebleu": 94.9115842118578,
+      "eval_word_accuracy": 0.9848901098901099,
       "num_input_tokens_seen": 281714688,
-      "perplexity": 1.0264038366178412,
+      "perplexity": 1.0191564758696394,
       "step": 4300
     },
     {
       "epoch": 0.9324967546516659,
-      "grad_norm": 0.68359375,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.0003594315789473684,
-      "loss": 0.0369,
+      "loss": 0.031,
       "num_input_tokens_seen": 282370048,
       "step": 4310,
-      "train_runtime": 1523.4951,
-      "train_tokens_per_second": 185343.586
+      "train_runtime": 1914.2808,
+      "train_tokens_per_second": 147507.122
     },
     {
       "epoch": 0.9346603202077023,
-      "grad_norm": 0.875,
+      "grad_norm": 0.51171875,
       "learning_rate": 0.00035879999999999994,
-      "loss": 0.0359,
+      "loss": 0.0299,
       "num_input_tokens_seen": 283025408,
       "step": 4320,
-      "train_runtime": 1527.3009,
-      "train_tokens_per_second": 185310.831
+      "train_runtime": 1918.0892,
+      "train_tokens_per_second": 147555.918
     },
     {
       "epoch": 0.9368238857637387,
-      "grad_norm": 0.83203125,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0003581684210526316,
-      "loss": 0.043,
+      "loss": 0.0344,
       "num_input_tokens_seen": 283680768,
       "step": 4330,
-      "train_runtime": 1531.1173,
-      "train_tokens_per_second": 185276.971
+      "train_runtime": 1921.9064,
+      "train_tokens_per_second": 147603.843
     },
     {
       "epoch": 0.938987451319775,
-      "grad_norm": 0.66796875,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0003575368421052631,
-      "loss": 0.0338,
+      "loss": 0.026,
       "num_input_tokens_seen": 284336128,
       "step": 4340,
-      "train_runtime": 1534.9394,
-      "train_tokens_per_second": 185242.579
+      "train_runtime": 1925.715,
+      "train_tokens_per_second": 147652.234
     },
     {
       "epoch": 0.9411510168758114,
-      "grad_norm": 0.56640625,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00035690526315789475,
-      "loss": 0.0357,
+      "loss": 0.0288,
       "num_input_tokens_seen": 284991488,
       "step": 4350,
-      "train_runtime": 1538.7526,
-      "train_tokens_per_second": 185209.432
+      "train_runtime": 1929.5298,
+      "train_tokens_per_second": 147699.968
     },
     {
       "epoch": 0.9433145824318477,
-      "grad_norm": 0.93359375,
+      "grad_norm": 0.59765625,
       "learning_rate": 0.0003562736842105263,
-      "loss": 0.0425,
+      "loss": 0.0342,
       "num_input_tokens_seen": 285646848,
       "step": 4360,
-      "train_runtime": 1542.5657,
-      "train_tokens_per_second": 185176.458
+      "train_runtime": 1935.5282,
+      "train_tokens_per_second": 147580.828
     },
     {
       "epoch": 0.945478147987884,
-      "grad_norm": 0.78515625,
+      "grad_norm": 0.6953125,
       "learning_rate": 0.0003556421052631578,
-      "loss": 0.0356,
+      "loss": 0.0295,
       "num_input_tokens_seen": 286302208,
       "step": 4370,
-      "train_runtime": 1546.3779,
-      "train_tokens_per_second": 185143.756
+      "train_runtime": 1939.3378,
+      "train_tokens_per_second": 147628.849
     },
     {
       "epoch": 0.9476417135439203,
       "grad_norm": 0.76953125,
       "learning_rate": 0.00035501052631578946,
-      "loss": 0.0384,
+      "loss": 0.0318,
       "num_input_tokens_seen": 286957568,
       "step": 4380,
-      "train_runtime": 1550.1751,
-      "train_tokens_per_second": 185112.997
+      "train_runtime": 1943.1309,
+      "train_tokens_per_second": 147677.942
     },
     {
       "epoch": 0.9498052790999567,
-      "grad_norm": 0.478515625,
+      "grad_norm": 0.6015625,
       "learning_rate": 0.000354378947368421,
-      "loss": 0.0358,
+      "loss": 0.0308,
       "num_input_tokens_seen": 287612928,
       "step": 4390,
-      "train_runtime": 1553.9759,
-      "train_tokens_per_second": 185081.972
+      "train_runtime": 1946.9273,
+      "train_tokens_per_second": 147726.585
     },
     {
       "epoch": 0.9519688446559931,
-      "grad_norm": 0.68359375,
+      "grad_norm": 0.51953125,
       "learning_rate": 0.00035374736842105263,
-      "loss": 0.0389,
+      "loss": 0.0322,
       "num_input_tokens_seen": 288268288,
       "step": 4400,
-      "train_runtime": 1557.8016,
-      "train_tokens_per_second": 185048.144
+      "train_runtime": 1950.7432,
+      "train_tokens_per_second": 147773.567
     },
     {
       "epoch": 0.9519688446559931,
-      "eval_loss": 0.025347542017698288,
-      "eval_runtime": 1.5674,
-      "eval_samples_per_second": 20.416,
-      "eval_steps_per_second": 0.638,
+      "eval_loss": 0.019506528973579407,
+      "eval_runtime": 2.1658,
+      "eval_samples_per_second": 14.775,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 288268288,
       "step": 4400
     },
     {
       "epoch": 0.9519688446559931,
       "eval_byte_accuracy": 0.9933234421364985,
-      "eval_chrf": 94.8674310164017,
-      "eval_sacrebleu": 94.82084502261215,
-      "eval_word_accuracy": 0.9835164835164835,
+      "eval_chrf": 94.24507398874036,
+      "eval_sacrebleu": 92.8561495846159,
+      "eval_word_accuracy": 0.9821428571428571,
       "num_input_tokens_seen": 288268288,
-      "perplexity": 1.0256715225389428,
+      "perplexity": 1.0196980244203853,
       "step": 4400
     },
     {
       "epoch": 0.9541324102120294,
-      "grad_norm": 0.9296875,
+      "grad_norm": 0.75,
       "learning_rate": 0.00035311578947368416,
-      "loss": 0.0384,
+      "loss": 0.0307,
       "num_input_tokens_seen": 288923648,
       "step": 4410,
-      "train_runtime": 1563.1975,
-      "train_tokens_per_second": 184828.623
+      "train_runtime": 1958.9228,
+      "train_tokens_per_second": 147491.081
     },
     {
       "epoch": 0.9562959757680658,
-      "grad_norm": 0.421875,
+      "grad_norm": 0.408203125,
       "learning_rate": 0.0003524842105263158,
-      "loss": 0.0373,
+      "loss": 0.0283,
       "num_input_tokens_seen": 289579008,
       "step": 4420,
-      "train_runtime": 1567.0228,
-      "train_tokens_per_second": 184795.663
+      "train_runtime": 1962.7389,
+      "train_tokens_per_second": 147538.225
     },
     {
       "epoch": 0.9584595413241022,
-      "grad_norm": 0.98828125,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.00035185263157894734,
-      "loss": 0.0397,
+      "loss": 0.0303,
       "num_input_tokens_seen": 290234368,
       "step": 4430,
-      "train_runtime": 1570.8514,
-      "train_tokens_per_second": 184762.465
+      "train_runtime": 1966.5423,
+      "train_tokens_per_second": 147586.133
     },
     {
       "epoch": 0.9606231068801385,
-      "grad_norm": 0.8828125,
+      "grad_norm": 0.6953125,
       "learning_rate": 0.00035122105263157887,
-      "loss": 0.0374,
+      "loss": 0.031,
       "num_input_tokens_seen": 290889728,
       "step": 4440,
-      "train_runtime": 1574.6696,
-      "train_tokens_per_second": 184730.646
+      "train_runtime": 1970.3612,
+      "train_tokens_per_second": 147632.695
     },
     {
       "epoch": 0.9627866724361748,
-      "grad_norm": 1.0703125,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.0003505894736842105,
-      "loss": 0.039,
+      "loss": 0.0316,
       "num_input_tokens_seen": 291545088,
       "step": 4450,
-      "train_runtime": 1578.4859,
-      "train_tokens_per_second": 184699.202
+      "train_runtime": 1974.1835,
+      "train_tokens_per_second": 147678.822
     },
     {
       "epoch": 0.9649502379922111,
-      "grad_norm": 0.8359375,
+      "grad_norm": 0.59375,
       "learning_rate": 0.00034995789473684204,
-      "loss": 0.0353,
+      "loss": 0.0284,
       "num_input_tokens_seen": 292200448,
       "step": 4460,
-      "train_runtime": 1582.3064,
-      "train_tokens_per_second": 184667.428
+      "train_runtime": 1980.166,
+      "train_tokens_per_second": 147563.609
     },
     {
       "epoch": 0.9671138035482475,
-      "grad_norm": 0.96484375,
+      "grad_norm": 0.60546875,
       "learning_rate": 0.0003493263157894737,
-      "loss": 0.0359,
+      "loss": 0.0303,
       "num_input_tokens_seen": 292855808,
       "step": 4470,
-      "train_runtime": 1586.1211,
-      "train_tokens_per_second": 184636.469
+      "train_runtime": 1983.9865,
+      "train_tokens_per_second": 147609.78
     },
     {
       "epoch": 0.9692773691042839,
-      "grad_norm": 0.7578125,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.0003486947368421052,
-      "loss": 0.0332,
+      "loss": 0.0268,
       "num_input_tokens_seen": 293511168,
       "step": 4480,
-      "train_runtime": 1589.9408,
-      "train_tokens_per_second": 184605.094
+      "train_runtime": 1987.7995,
+      "train_tokens_per_second": 147656.327
     },
     {
       "epoch": 0.9714409346603202,
-      "grad_norm": 0.7421875,
+      "grad_norm": 0.6875,
       "learning_rate": 0.00034806315789473685,
-      "loss": 0.0365,
+      "loss": 0.0304,
       "num_input_tokens_seen": 294166528,
       "step": 4490,
-      "train_runtime": 1593.7578,
-      "train_tokens_per_second": 184574.168
+      "train_runtime": 1991.6159,
+      "train_tokens_per_second": 147702.439
     },
     {
       "epoch": 0.9736045002163566,
-      "grad_norm": 0.640625,
+      "grad_norm": 0.55859375,
       "learning_rate": 0.0003474315789473684,
-      "loss": 0.0392,
+      "loss": 0.0309,
       "num_input_tokens_seen": 294821888,
       "step": 4500,
-      "train_runtime": 1597.5794,
-      "train_tokens_per_second": 184542.868
-    },
-    {
-      "epoch": 0.9736045002163566,
-      "eval_loss": 0.024211181327700615,
-      "eval_runtime": 1.7479,
-      "eval_samples_per_second": 18.308,
-      "eval_steps_per_second": 0.572,
-      "num_input_tokens_seen": 294821888,
-      "step": 4500
-    },
-    {
-      "epoch": 0.9736045002163566,
-      "eval_byte_accuracy": 0.9936943620178041,
-      "eval_chrf": 94.58267760158972,
-      "eval_sacrebleu": 93.44801868564782,
-      "eval_word_accuracy": 0.9835164835164835,
-      "num_input_tokens_seen": 294821888,
-      "perplexity": 1.0245066517219334,
-      "step": 4500
+      "train_runtime": 1995.424,
+      "train_tokens_per_second": 147748.994
     },
     {
       "epoch": 0.9736045002163566,
-      "eval_loss": 0.024180177599191666,
-      "eval_runtime": 2.8794,
-      "eval_samples_per_second": 11.113,
-      "eval_steps_per_second": 0.347,
+      "eval_loss": 0.017310751602053642,
+      "eval_runtime": 2.0408,
+      "eval_samples_per_second": 15.68,
+      "eval_steps_per_second": 0.49,
       "num_input_tokens_seen": 294821888,
       "step": 4500
     },
     {
       "epoch": 0.9736045002163566,
       "eval_byte_accuracy": 0.9940652818991098,
-      "eval_chrf": 95.26030523847231,
-      "eval_sacrebleu": 93.68260424037459,
-      "eval_word_accuracy": 0.9848901098901099,
+      "eval_chrf": 95.3397660833719,
+      "eval_sacrebleu": 95.29534882340593,
+      "eval_word_accuracy": 0.9876373626373627,
       "num_input_tokens_seen": 294821888,
-      "perplexity": 1.0244748886882367,
+      "perplexity": 1.0174614509798736,
       "step": 4500
     },
     {
       "epoch": 0.975768065772393,
-      "grad_norm": 0.439453125,
+      "grad_norm": 0.45703125,
       "learning_rate": 0.0003467999999999999,
-      "loss": 0.0303,
+      "loss": 0.0277,
       "num_input_tokens_seen": 295477248,
       "step": 4510,
-      "train_runtime": 8.1902,
-      "train_tokens_per_second": 36076993.874
+      "train_runtime": 2003.6989,
+      "train_tokens_per_second": 147465.895
     },
     {
       "epoch": 0.9779316313284292,
-      "grad_norm": 0.4375,
+      "grad_norm": 0.61328125,
       "learning_rate": 0.00034616842105263156,
-      "loss": 0.0342,
+      "loss": 0.0323,
       "num_input_tokens_seen": 296132608,
       "step": 4520,
-      "train_runtime": 12.0015,
-      "train_tokens_per_second": 24674670.481
+      "train_runtime": 2007.5052,
+      "train_tokens_per_second": 147512.751
     },
     {
       "epoch": 0.9800951968844656,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.0003455368421052631,
-      "loss": 0.0322,
+      "loss": 0.0314,
       "num_input_tokens_seen": 296787968,
       "step": 4530,
-      "train_runtime": 15.8385,
-      "train_tokens_per_second": 18738336.118
+      "train_runtime": 2011.3248,
+      "train_tokens_per_second": 147558.447
     },
     {
       "epoch": 0.9822587624405019,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.6875,
       "learning_rate": 0.00034490526315789473,
-      "loss": 0.0279,
+      "loss": 0.0275,
       "num_input_tokens_seen": 297443328,
       "step": 4540,
-      "train_runtime": 19.6683,
-      "train_tokens_per_second": 15123011.172
+      "train_runtime": 2015.1382,
+      "train_tokens_per_second": 147604.429
     },
     {
       "epoch": 0.9844223279965383,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.5546875,
       "learning_rate": 0.00034427368421052626,
-      "loss": 0.0302,
+      "loss": 0.0296,
       "num_input_tokens_seen": 298098688,
       "step": 4550,
-      "train_runtime": 23.5035,
-      "train_tokens_per_second": 12683170.2
+      "train_runtime": 2018.9595,
+      "train_tokens_per_second": 147649.663
     },
     {
       "epoch": 0.9865858935525746,
-      "grad_norm": 0.466796875,
+      "grad_norm": 0.69921875,
       "learning_rate": 0.0003436421052631579,
-      "loss": 0.0295,
+      "loss": 0.0294,
       "num_input_tokens_seen": 298754048,
       "step": 4560,
-      "train_runtime": 27.3289,
-      "train_tokens_per_second": 10931815.69
+      "train_runtime": 2024.9701,
+      "train_tokens_per_second": 147535.042
     },
     {
       "epoch": 0.988749459108611,
       "grad_norm": 0.640625,
       "learning_rate": 0.00034301052631578944,
-      "loss": 0.0296,
+      "loss": 0.0292,
       "num_input_tokens_seen": 299405312,
       "step": 4570,
-      "train_runtime": 31.1358,
-      "train_tokens_per_second": 9616105.982
+      "train_runtime": 2028.7655,
+      "train_tokens_per_second": 147580.048
     },
     {
       "epoch": 0.9909130246646474,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.000342378947368421,
       "loss": 0.0277,
       "num_input_tokens_seen": 300056576,
       "step": 4580,
-      "train_runtime": 34.9281,
-      "train_tokens_per_second": 8590693.677
+      "train_runtime": 2032.5336,
+      "train_tokens_per_second": 147626.871
     },
     {
       "epoch": 0.9930765902206837,
-      "grad_norm": 0.451171875,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.0003417473684210526,
-      "loss": 0.0323,
+      "loss": 0.0321,
       "num_input_tokens_seen": 300711936,
       "step": 4590,
-      "train_runtime": 38.7343,
-      "train_tokens_per_second": 7763459.853
+      "train_runtime": 2036.3318,
+      "train_tokens_per_second": 147673.349
     },
     {
       "epoch": 0.99524015577672,
-      "grad_norm": 0.498046875,
+      "grad_norm": 0.640625,
       "learning_rate": 0.00034111578947368414,
-      "loss": 0.0284,
+      "loss": 0.0275,
       "num_input_tokens_seen": 301367296,
       "step": 4600,
-      "train_runtime": 42.5467,
-      "train_tokens_per_second": 7083205.808
+      "train_runtime": 2040.1253,
+      "train_tokens_per_second": 147719.992
     },
     {
       "epoch": 0.99524015577672,
-      "eval_loss": 0.020042533054947853,
-      "eval_runtime": 1.6332,
-      "eval_samples_per_second": 19.593,
-      "eval_steps_per_second": 0.612,
+      "eval_loss": 0.017312878742814064,
+      "eval_runtime": 2.0657,
+      "eval_samples_per_second": 15.491,
+      "eval_steps_per_second": 0.484,
       "num_input_tokens_seen": 301367296,
       "step": 4600
     },
     {
       "epoch": 0.99524015577672,
       "eval_byte_accuracy": 0.9951780415430267,
-      "eval_chrf": 96.63362661333757,
-      "eval_sacrebleu": 95.8953860263406,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_chrf": 96.36001642042964,
+      "eval_sacrebleu": 95.65640954170594,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 301367296,
-      "perplexity": 1.0202447332292253,
+      "perplexity": 1.0174636152659,
       "step": 4600
     },
     {
       "epoch": 0.9974037213327563,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.796875,
       "learning_rate": 0.0003404842105263158,
-      "loss": 0.0309,
+      "loss": 0.0316,
       "num_input_tokens_seen": 302022656,
       "step": 4610,
-      "train_runtime": 48.0092,
-      "train_tokens_per_second": 6290930.799
+      "train_runtime": 2048.1812,
+      "train_tokens_per_second": 147458.95
     },
     {
       "epoch": 0.9995672868887927,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.6796875,
       "learning_rate": 0.0003398526315789473,
-      "loss": 0.0289,
+      "loss": 0.0275,
       "num_input_tokens_seen": 302678016,
       "step": 4620,
-      "train_runtime": 51.8316,
-      "train_tokens_per_second": 5839641.358
+      "train_runtime": 2051.9727,
+      "train_tokens_per_second": 147505.87
     },
     {
       "epoch": 1.0017308524448292,
-      "grad_norm": 0.48828125,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.00033922105263157896,
-      "loss": 0.0265,
+      "loss": 0.0268,
       "num_input_tokens_seen": 303321088,
       "step": 4630,
-      "train_runtime": 55.6955,
-      "train_tokens_per_second": 5446064.044
+      "train_runtime": 2055.7464,
+      "train_tokens_per_second": 147547.911
     },
     {
       "epoch": 1.0038944180008653,
-      "grad_norm": 0.455078125,
+      "grad_norm": 0.69921875,
       "learning_rate": 0.0003385894736842105,
-      "loss": 0.0274,
+      "loss": 0.0278,
       "num_input_tokens_seen": 303976448,
       "step": 4640,
-      "train_runtime": 59.5237,
-      "train_tokens_per_second": 5106813.968
+      "train_runtime": 2059.555,
+      "train_tokens_per_second": 147593.262
     },
     {
       "epoch": 1.0060579835569017,
-      "grad_norm": 0.52734375,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.0003379578947368421,
-      "loss": 0.0263,
+      "loss": 0.0257,
       "num_input_tokens_seen": 304631808,
       "step": 4650,
-      "train_runtime": 63.3483,
-      "train_tokens_per_second": 4808840.309
+      "train_runtime": 2063.3612,
+      "train_tokens_per_second": 147638.623
     },
     {
       "epoch": 1.008221549112938,
-      "grad_norm": 0.73046875,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.00033732631578947366,
-      "loss": 0.0269,
+      "loss": 0.0265,
       "num_input_tokens_seen": 305287168,
       "step": 4660,
-      "train_runtime": 67.1647,
-      "train_tokens_per_second": 4545353.422
+      "train_runtime": 2069.364,
+      "train_tokens_per_second": 147527.051
     },
     {
       "epoch": 1.0103851146689744,
-      "grad_norm": 0.59765625,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0003366947368421052,
-      "loss": 0.0265,
+      "loss": 0.0247,
       "num_input_tokens_seen": 305942528,
       "step": 4670,
-      "train_runtime": 70.9861,
-      "train_tokens_per_second": 4309893.533
+      "train_runtime": 2073.1806,
+      "train_tokens_per_second": 147571.575
     },
     {
       "epoch": 1.0125486802250108,
-      "grad_norm": 0.6796875,
+      "grad_norm": 0.51171875,
       "learning_rate": 0.00033606315789473683,
-      "loss": 0.0269,
+      "loss": 0.026,
       "num_input_tokens_seen": 306597888,
       "step": 4680,
-      "train_runtime": 74.8102,
-      "train_tokens_per_second": 4098342.089
+      "train_runtime": 2076.9852,
+      "train_tokens_per_second": 147616.792
     },
     {
       "epoch": 1.0147122457810471,
-      "grad_norm": 0.474609375,
+      "grad_norm": 0.70703125,
       "learning_rate": 0.00033543157894736837,
-      "loss": 0.0282,
+      "loss": 0.027,
       "num_input_tokens_seen": 307253248,
       "step": 4690,
-      "train_runtime": 78.6348,
-      "train_tokens_per_second": 3907343.298
+      "train_runtime": 2080.8031,
+      "train_tokens_per_second": 147660.893
     },
     {
       "epoch": 1.0168758113370835,
-      "grad_norm": 0.404296875,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.0003348,
-      "loss": 0.0272,
+      "loss": 0.0268,
       "num_input_tokens_seen": 307908608,
       "step": 4700,
-      "train_runtime": 82.46,
-      "train_tokens_per_second": 3734036.121
+      "train_runtime": 2084.6037,
+      "train_tokens_per_second": 147706.064
     },
     {
       "epoch": 1.0168758113370835,
-      "eval_loss": 0.020144794136285782,
-      "eval_runtime": 1.6689,
-      "eval_samples_per_second": 19.175,
-      "eval_steps_per_second": 0.599,
+      "eval_loss": 0.015669453889131546,
+      "eval_runtime": 2.0897,
+      "eval_samples_per_second": 15.313,
+      "eval_steps_per_second": 0.479,
       "num_input_tokens_seen": 307908608,
       "step": 4700
     },
     {
       "epoch": 1.0168758113370835,
-      "eval_byte_accuracy": 0.9944362017804155,
-      "eval_chrf": 96.01941443962484,
-      "eval_sacrebleu": 94.43788582746568,
-      "eval_word_accuracy": 0.9862637362637363,
+      "eval_byte_accuracy": 0.9955489614243324,
+      "eval_chrf": 96.59396178137179,
+      "eval_sacrebleu": 95.77338932371731,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 307908608,
-      "perplexity": 1.0203490698935738,
+      "perplexity": 1.0157928635263578,
       "step": 4700
     },
     {
       "epoch": 1.0190393768931199,
-      "grad_norm": 0.44921875,
+      "grad_norm": 0.67578125,
       "learning_rate": 0.00033416842105263154,
-      "loss": 0.0281,
+      "loss": 0.0275,
       "num_input_tokens_seen": 308563968,
       "step": 4710,
-      "train_runtime": 87.973,
-      "train_tokens_per_second": 3507485.0
+      "train_runtime": 2092.891,
+      "train_tokens_per_second": 147434.321
     },
     {
       "epoch": 1.0212029424491562,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.765625,
       "learning_rate": 0.0003335368421052631,
-      "loss": 0.0266,
+      "loss": 0.0276,
       "num_input_tokens_seen": 309219328,
       "step": 4720,
-      "train_runtime": 91.7926,
-      "train_tokens_per_second": 3368675.086
+      "train_runtime": 2096.697,
+      "train_tokens_per_second": 147479.265
     },
     {
       "epoch": 1.0233665080051926,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.494140625,
       "learning_rate": 0.0003329052631578947,
-      "loss": 0.0279,
+      "loss": 0.0276,
       "num_input_tokens_seen": 309874688,
       "step": 4730,
-      "train_runtime": 95.6175,
-      "train_tokens_per_second": 3240774.088
+      "train_runtime": 2100.5115,
+      "train_tokens_per_second": 147523.446
     },
     {
       "epoch": 1.025530073561229,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.373046875,
       "learning_rate": 0.00033227368421052624,
-      "loss": 0.0232,
+      "loss": 0.0241,
       "num_input_tokens_seen": 310530048,
       "step": 4740,
-      "train_runtime": 99.4401,
-      "train_tokens_per_second": 3122786.399
+      "train_runtime": 2104.32,
+      "train_tokens_per_second": 147567.887
     },
     {
       "epoch": 1.0276936391172653,
-      "grad_norm": 0.921875,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0003316421052631579,
-      "loss": 0.0279,
+      "loss": 0.0285,
       "num_input_tokens_seen": 311185408,
       "step": 4750,
-      "train_runtime": 103.265,
-      "train_tokens_per_second": 3013463.926
+      "train_runtime": 2108.1335,
+      "train_tokens_per_second": 147611.81
     },
     {
       "epoch": 1.0298572046733017,
-      "grad_norm": 0.8125,
+      "grad_norm": 0.55859375,
       "learning_rate": 0.0003310105263157894,
-      "loss": 0.0267,
+      "loss": 0.0266,
       "num_input_tokens_seen": 311840768,
       "step": 4760,
-      "train_runtime": 107.0815,
-      "train_tokens_per_second": 2912180.92
+      "train_runtime": 2114.1392,
+      "train_tokens_per_second": 147502.479
     },
     {
       "epoch": 1.032020770229338,
-      "grad_norm": 0.90234375,
+      "grad_norm": 0.5703125,
       "learning_rate": 0.00033037894736842106,
-      "loss": 0.0289,
+      "loss": 0.027,
       "num_input_tokens_seen": 312496128,
       "step": 4770,
-      "train_runtime": 110.8903,
-      "train_tokens_per_second": 2818065.378
+      "train_runtime": 2117.9312,
+      "train_tokens_per_second": 147547.821
     },
     {
       "epoch": 1.0341843357853744,
-      "grad_norm": 0.53515625,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.0003297473684210526,
-      "loss": 0.0247,
+      "loss": 0.023,
       "num_input_tokens_seen": 313151488,
       "step": 4780,
-      "train_runtime": 114.7073,
-      "train_tokens_per_second": 2730004.96
+      "train_runtime": 2121.7355,
+      "train_tokens_per_second": 147592.141
     },
     {
       "epoch": 1.0363479013414107,
-      "grad_norm": 0.703125,
+      "grad_norm": 0.421875,
       "learning_rate": 0.0003291157894736842,
-      "loss": 0.0267,
+      "loss": 0.0247,
       "num_input_tokens_seen": 313806848,
       "step": 4790,
-      "train_runtime": 118.5372,
-      "train_tokens_per_second": 2647328.086
+      "train_runtime": 2125.5459,
+      "train_tokens_per_second": 147635.883
     },
     {
       "epoch": 1.0385114668974469,
-      "grad_norm": 0.4375,
+      "grad_norm": 0.66015625,
       "learning_rate": 0.00032848421052631576,
-      "loss": 0.025,
+      "loss": 0.0256,
       "num_input_tokens_seen": 314462208,
       "step": 4800,
-      "train_runtime": 122.361,
-      "train_tokens_per_second": 2569953.792
+      "train_runtime": 2129.3486,
+      "train_tokens_per_second": 147680.003
     },
     {
       "epoch": 1.0385114668974469,
-      "eval_loss": 0.018589144572615623,
-      "eval_runtime": 1.8825,
-      "eval_samples_per_second": 16.998,
-      "eval_steps_per_second": 0.531,
+      "eval_loss": 0.015679502859711647,
+      "eval_runtime": 2.1061,
+      "eval_samples_per_second": 15.194,
+      "eval_steps_per_second": 0.475,
       "num_input_tokens_seen": 314462208,
       "step": 4800
     },
     {
       "epoch": 1.0385114668974469,
-      "eval_byte_accuracy": 0.9951780415430267,
-      "eval_chrf": 96.23348374614099,
-      "eval_sacrebleu": 94.52630209470658,
+      "eval_byte_accuracy": 0.9955489614243324,
+      "eval_chrf": 96.5138850641643,
+      "eval_sacrebleu": 95.39446008456702,
       "eval_word_accuracy": 0.9876373626373627,
       "num_input_tokens_seen": 314462208,
-      "perplexity": 1.0187629983138469,
+      "perplexity": 1.0158030712502473,
       "step": 4800
     },
     {
       "epoch": 1.0406750324534833,
-      "grad_norm": 0.55859375,
+      "grad_norm": 0.78515625,
       "learning_rate": 0.00032785263157894735,
-      "loss": 0.0261,
+      "loss": 0.0256,
       "num_input_tokens_seen": 315117568,
       "step": 4810,
-      "train_runtime": 128.0843,
-      "train_tokens_per_second": 2460235.566
+      "train_runtime": 2137.468,
+      "train_tokens_per_second": 147425.63
     },
     {
       "epoch": 1.0428385980095196,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.6484375,
       "learning_rate": 0.00032722105263157894,
-      "loss": 0.0267,
+      "loss": 0.026,
       "num_input_tokens_seen": 315772928,
       "step": 4820,
-      "train_runtime": 131.9081,
-      "train_tokens_per_second": 2393886.109
+      "train_runtime": 2141.2729,
+      "train_tokens_per_second": 147469.726
     },
     {
       "epoch": 1.045002163565556,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.5625,
       "learning_rate": 0.00032658947368421047,
-      "loss": 0.0268,
+      "loss": 0.0275,
       "num_input_tokens_seen": 316428288,
       "step": 4830,
-      "train_runtime": 135.7375,
-      "train_tokens_per_second": 2331178.327
+      "train_runtime": 2145.0816,
+      "train_tokens_per_second": 147513.402
     },
     {
       "epoch": 1.0471657291215923,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.578125,
       "learning_rate": 0.0003259578947368421,
-      "loss": 0.0285,
+      "loss": 0.0273,
       "num_input_tokens_seen": 317083648,
       "step": 4840,
-      "train_runtime": 139.5602,
-      "train_tokens_per_second": 2272020.269
+      "train_runtime": 2148.8867,
+      "train_tokens_per_second": 147557.175
     },
     {
       "epoch": 1.0493292946776287,
-      "grad_norm": 0.640625,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00032532631578947364,
-      "loss": 0.0257,
+      "loss": 0.0252,
       "num_input_tokens_seen": 317739008,
       "step": 4850,
-      "train_runtime": 143.3904,
-      "train_tokens_per_second": 2215901.171
+      "train_runtime": 2152.6994,
+      "train_tokens_per_second": 147600.267
     },
     {
       "epoch": 1.051492860233665,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.4375,
       "learning_rate": 0.0003246947368421052,
-      "loss": 0.0272,
+      "loss": 0.0256,
       "num_input_tokens_seen": 318394368,
       "step": 4860,
-      "train_runtime": 147.2228,
-      "train_tokens_per_second": 2162670.788
+      "train_runtime": 2158.693,
+      "train_tokens_per_second": 147494.048
     },
     {
       "epoch": 1.0536564257897014,
-      "grad_norm": 0.69921875,
+      "grad_norm": 0.57421875,
       "learning_rate": 0.0003240631578947368,
-      "loss": 0.0267,
+      "loss": 0.026,
       "num_input_tokens_seen": 319049728,
       "step": 4870,
-      "train_runtime": 151.0482,
-      "train_tokens_per_second": 2112237.293
+      "train_runtime": 2162.5068,
+      "train_tokens_per_second": 147536.981
     },
     {
       "epoch": 1.0558199913457378,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.66796875,
       "learning_rate": 0.0003234315789473684,
-      "loss": 0.0261,
+      "loss": 0.0249,
       "num_input_tokens_seen": 319705088,
       "step": 4880,
-      "train_runtime": 154.8725,
-      "train_tokens_per_second": 2064311.503
+      "train_runtime": 2166.3052,
+      "train_tokens_per_second": 147580.812
     },
     {
       "epoch": 1.0579835569017741,
-      "grad_norm": 0.66015625,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.0003228,
-      "loss": 0.0242,
+      "loss": 0.025,
       "num_input_tokens_seen": 320360448,
       "step": 4890,
-      "train_runtime": 158.6951,
-      "train_tokens_per_second": 2018717.011
+      "train_runtime": 2170.1088,
+      "train_tokens_per_second": 147624.142
     },
     {
       "epoch": 1.0601471224578105,
-      "grad_norm": 0.77734375,
+      "grad_norm": 0.63671875,
       "learning_rate": 0.0003221684210526315,
-      "loss": 0.0295,
+      "loss": 0.0277,
       "num_input_tokens_seen": 321015808,
       "step": 4900,
-      "train_runtime": 162.5123,
-      "train_tokens_per_second": 1975332.311
+      "train_runtime": 2173.8996,
+      "train_tokens_per_second": 147668.186
     },
     {
       "epoch": 1.0601471224578105,
-      "eval_loss": 0.020329073071479797,
-      "eval_runtime": 1.7307,
-      "eval_samples_per_second": 18.49,
-      "eval_steps_per_second": 0.578,
+      "eval_loss": 0.01509543601423502,
+      "eval_runtime": 2.2745,
+      "eval_samples_per_second": 14.069,
+      "eval_steps_per_second": 0.44,
       "num_input_tokens_seen": 321015808,
       "step": 4900
     },
     {
       "epoch": 1.0601471224578105,
-      "eval_byte_accuracy": 0.9944362017804155,
-      "eval_chrf": 95.49067447945914,
-      "eval_sacrebleu": 93.90610177072331,
-      "eval_word_accuracy": 0.9848901098901099,
+      "eval_byte_accuracy": 0.995919881305638,
+      "eval_chrf": 96.76141864085723,
+      "eval_sacrebleu": 96.18078243947666,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 321015808,
-      "perplexity": 1.0205371160596415,
+      "perplexity": 1.0152099475835925,
       "step": 4900
     },
     {
       "epoch": 1.0623106880138469,
-      "grad_norm": 0.46484375,
+      "grad_norm": 0.75,
       "learning_rate": 0.00032153684210526316,
-      "loss": 0.0281,
+      "loss": 0.0278,
       "num_input_tokens_seen": 321671168,
       "step": 4910,
-      "train_runtime": 168.09,
-      "train_tokens_per_second": 1913684.312
+      "train_runtime": 2182.1878,
+      "train_tokens_per_second": 147407.643
     },
     {
       "epoch": 1.0644742535698832,
-      "grad_norm": 0.6328125,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0003209052631578947,
-      "loss": 0.0271,
+      "loss": 0.0273,
       "num_input_tokens_seen": 322326528,
       "step": 4920,
-      "train_runtime": 171.9119,
-      "train_tokens_per_second": 1874951.703
+      "train_runtime": 2186.0007,
+      "train_tokens_per_second": 147450.33
     },
     {
       "epoch": 1.0666378191259196,
-      "grad_norm": 0.6484375,
+      "grad_norm": 0.498046875,
       "learning_rate": 0.0003202736842105263,
-      "loss": 0.0262,
+      "loss": 0.0259,
       "num_input_tokens_seen": 322981888,
       "step": 4930,
-      "train_runtime": 175.7226,
-      "train_tokens_per_second": 1838021.179
+      "train_runtime": 2189.8029,
+      "train_tokens_per_second": 147493.591
     },
     {
       "epoch": 1.068801384681956,
-      "grad_norm": 0.486328125,
+      "grad_norm": 0.7578125,
       "learning_rate": 0.00031964210526315786,
-      "loss": 0.0256,
+      "loss": 0.0242,
       "num_input_tokens_seen": 323637248,
       "step": 4940,
-      "train_runtime": 179.5451,
-      "train_tokens_per_second": 1802539.801
+      "train_runtime": 2193.6076,
+      "train_tokens_per_second": 147536.527
     },
     {
       "epoch": 1.0709649502379923,
-      "grad_norm": 0.65234375,
+      "grad_norm": 0.45703125,
       "learning_rate": 0.00031901052631578945,
-      "loss": 0.025,
+      "loss": 0.0245,
       "num_input_tokens_seen": 324292608,
       "step": 4950,
-      "train_runtime": 183.36,
-      "train_tokens_per_second": 1768611.34
+      "train_runtime": 2197.4066,
+      "train_tokens_per_second": 147579.7
     },
     {
       "epoch": 1.0731285157940285,
-      "grad_norm": 0.5546875,
+      "grad_norm": 0.423828125,
       "learning_rate": 0.00031837894736842104,
-      "loss": 0.0265,
+      "loss": 0.0252,
       "num_input_tokens_seen": 324947968,
       "step": 4960,
-      "train_runtime": 187.1847,
-      "train_tokens_per_second": 1735974.791
+      "train_runtime": 2203.6181,
+      "train_tokens_per_second": 147461.111
     },
     {
       "epoch": 1.0752920813500648,
-      "grad_norm": 0.76171875,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.00031774736842105257,
-      "loss": 0.0275,
+      "loss": 0.0254,
       "num_input_tokens_seen": 325603328,
       "step": 4970,
-      "train_runtime": 190.9969,
-      "train_tokens_per_second": 1704757.413
+      "train_runtime": 2207.4386,
+      "train_tokens_per_second": 147502.782
     },
     {
       "epoch": 1.0774556469061012,
-      "grad_norm": 0.53515625,
+      "grad_norm": 0.703125,
       "learning_rate": 0.0003171157894736842,
-      "loss": 0.0274,
+      "loss": 0.0258,
       "num_input_tokens_seen": 326258688,
       "step": 4980,
-      "train_runtime": 194.8145,
-      "train_tokens_per_second": 1674714.319
+      "train_runtime": 2211.2502,
+      "train_tokens_per_second": 147544.899
     },
     {
       "epoch": 1.0796192124621375,
-      "grad_norm": 0.51953125,
+      "grad_norm": 0.455078125,
       "learning_rate": 0.00031648421052631574,
-      "loss": 0.0252,
+      "loss": 0.0241,
       "num_input_tokens_seen": 326914048,
       "step": 4990,
-      "train_runtime": 198.6353,
-      "train_tokens_per_second": 1645800.089
+      "train_runtime": 2215.0686,
+      "train_tokens_per_second": 147586.421
     },
     {
       "epoch": 1.081782778018174,
-      "grad_norm": 0.59765625,
+      "grad_norm": 0.73046875,
       "learning_rate": 0.00031585263157894733,
-      "loss": 0.0239,
+      "loss": 0.0241,
       "num_input_tokens_seen": 327569408,
       "step": 5000,
-      "train_runtime": 202.4467,
-      "train_tokens_per_second": 1618052.623
+      "train_runtime": 2218.8662,
+      "train_tokens_per_second": 147629.183
     },
     {
       "epoch": 1.081782778018174,
-      "eval_loss": 0.01753578893840313,
-      "eval_runtime": 1.8451,
-      "eval_samples_per_second": 17.344,
-      "eval_steps_per_second": 0.542,
+      "eval_loss": 0.01370089128613472,
+      "eval_runtime": 2.0517,
+      "eval_samples_per_second": 15.597,
+      "eval_steps_per_second": 0.487,
       "num_input_tokens_seen": 327569408,
       "step": 5000
     },
     {
       "epoch": 1.081782778018174,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 96.50768150331436,
-      "eval_sacrebleu": 95.31161095839255,
-      "eval_word_accuracy": 0.9903846153846154,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 98.56639547519084,
+      "eval_sacrebleu": 97.89877731533007,
+      "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 327569408,
-      "perplexity": 1.0176904435596141,
+      "perplexity": 1.0137951786118617,
       "step": 5000
     },
     {
       "epoch": 1.0839463435742103,
-      "grad_norm": 0.5,
+      "grad_norm": 0.3984375,
       "learning_rate": 0.0003152210526315789,
-      "loss": 0.0233,
+      "loss": 0.0227,
       "num_input_tokens_seen": 328224768,
       "step": 5010,
-      "train_runtime": 208.189,
-      "train_tokens_per_second": 1576571.175
+      "train_runtime": 2226.993,
+      "train_tokens_per_second": 147384.73
     },
     {
       "epoch": 1.0861099091302466,
-      "grad_norm": 0.46875,
+      "grad_norm": 0.431640625,
       "learning_rate": 0.0003145894736842105,
       "loss": 0.0251,
       "num_input_tokens_seen": 328880128,
       "step": 5020,
-      "train_runtime": 212.0222,
-      "train_tokens_per_second": 1551158.906
+      "train_runtime": 2230.8082,
+      "train_tokens_per_second": 147426.446
     },
     {
       "epoch": 1.088273474686283,
-      "grad_norm": 0.640625,
+      "grad_norm": 0.515625,
       "learning_rate": 0.0003139578947368421,
-      "loss": 0.026,
+      "loss": 0.0244,
       "num_input_tokens_seen": 329535488,
       "step": 5030,
-      "train_runtime": 215.8507,
-      "train_tokens_per_second": 1526682.433
+      "train_runtime": 2234.6284,
+      "train_tokens_per_second": 147467.688
     },
     {
       "epoch": 1.0904370402423194,
-      "grad_norm": 0.490234375,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0003133263157894737,
-      "loss": 0.0272,
+      "loss": 0.026,
       "num_input_tokens_seen": 330186752,
       "step": 5040,
-      "train_runtime": 219.6399,
-      "train_tokens_per_second": 1503309.199
+      "train_runtime": 2238.4126,
+      "train_tokens_per_second": 147509.335
     },
     {
       "epoch": 1.0926006057983557,
-      "grad_norm": 0.478515625,
+      "grad_norm": 0.373046875,
       "learning_rate": 0.00031269473684210526,
-      "loss": 0.0256,
+      "loss": 0.0257,
       "num_input_tokens_seen": 330842112,
       "step": 5050,
-      "train_runtime": 223.4569,
-      "train_tokens_per_second": 1480563.289
+      "train_runtime": 2242.2218,
+      "train_tokens_per_second": 147551.017
     },
     {
       "epoch": 1.094764171354392,
-      "grad_norm": 0.63671875,
+      "grad_norm": 0.37109375,
       "learning_rate": 0.0003120631578947368,
-      "loss": 0.0246,
+      "loss": 0.0228,
       "num_input_tokens_seen": 331497472,
       "step": 5060,
-      "train_runtime": 227.2758,
-      "train_tokens_per_second": 1458569.335
+      "train_runtime": 2248.2359,
+      "train_tokens_per_second": 147447.817
     },
     {
       "epoch": 1.0969277369104284,
-      "grad_norm": 0.70703125,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0003114315789473684,
-      "loss": 0.024,
+      "loss": 0.0237,
       "num_input_tokens_seen": 332152832,
       "step": 5070,
-      "train_runtime": 231.1046,
-      "train_tokens_per_second": 1437240.386
+      "train_runtime": 2252.0457,
+      "train_tokens_per_second": 147489.382
     },
     {
       "epoch": 1.0990913024664648,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.54296875,
       "learning_rate": 0.00031079999999999997,
-      "loss": 0.0261,
+      "loss": 0.0259,
       "num_input_tokens_seen": 332808192,
       "step": 5080,
-      "train_runtime": 234.9359,
-      "train_tokens_per_second": 1416591.304
+      "train_runtime": 2255.8557,
+      "train_tokens_per_second": 147530.798
     },
     {
       "epoch": 1.1012548680225012,
-      "grad_norm": 0.6015625,
+      "grad_norm": 0.65234375,
       "learning_rate": 0.00031016842105263155,
-      "loss": 0.0253,
+      "loss": 0.0255,
       "num_input_tokens_seen": 333463552,
       "step": 5090,
-      "train_runtime": 238.7688,
-      "train_tokens_per_second": 1396596.039
+      "train_runtime": 2259.6673,
+      "train_tokens_per_second": 147571.971
     },
     {
       "epoch": 1.1034184335785375,
-      "grad_norm": 0.5078125,
+      "grad_norm": 0.76171875,
       "learning_rate": 0.00030953684210526314,
-      "loss": 0.0249,
+      "loss": 0.0251,
       "num_input_tokens_seen": 334118912,
       "step": 5100,
-      "train_runtime": 242.5974,
-      "train_tokens_per_second": 1377256.985
+      "train_runtime": 2263.4742,
+      "train_tokens_per_second": 147613.306
     },
     {
       "epoch": 1.1034184335785375,
-      "eval_loss": 0.018276168033480644,
-      "eval_runtime": 1.8788,
-      "eval_samples_per_second": 17.032,
-      "eval_steps_per_second": 0.532,
+      "eval_loss": 0.013377854600548744,
+      "eval_runtime": 2.0698,
+      "eval_samples_per_second": 15.46,
+      "eval_steps_per_second": 0.483,
       "num_input_tokens_seen": 334118912,
       "step": 5100
     },
     {
       "epoch": 1.1034184335785375,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 96.9547828716358,
-      "eval_sacrebleu": 96.30918177844974,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 96.85382382959251,
+      "eval_sacrebleu": 96.41686458971391,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 334118912,
-      "perplexity": 1.018444199287346,
+      "perplexity": 1.013467738467937,
       "step": 5100
     },
     {
       "epoch": 1.1055819991345737,
-      "grad_norm": 0.72265625,
+      "grad_norm": 0.462890625,
       "learning_rate": 0.0003089052631578947,
       "loss": 0.0233,
       "num_input_tokens_seen": 334774272,
       "step": 5110,
-      "train_runtime": 248.3149,
-      "train_tokens_per_second": 1348184.603
+      "train_runtime": 2271.5876,
+      "train_tokens_per_second": 147374.581
     },
     {
       "epoch": 1.10774556469061,
-      "grad_norm": 0.470703125,
+      "grad_norm": 0.65625,
       "learning_rate": 0.0003082736842105263,
-      "loss": 0.023,
+      "loss": 0.0231,
       "num_input_tokens_seen": 335429632,
       "step": 5120,
-      "train_runtime": 252.1428,
-      "train_tokens_per_second": 1330316.02
+      "train_runtime": 2275.3995,
+      "train_tokens_per_second": 147415.708
     },
     {
       "epoch": 1.1099091302466464,
-      "grad_norm": 0.423828125,
+      "grad_norm": 0.408203125,
       "learning_rate": 0.00030764210526315784,
-      "loss": 0.0269,
+      "loss": 0.0255,
       "num_input_tokens_seen": 336084992,
       "step": 5130,
-      "train_runtime": 255.9686,
-      "train_tokens_per_second": 1312992.994
+      "train_runtime": 2279.2058,
+      "train_tokens_per_second": 147457.06
     },
     {
       "epoch": 1.1120726958026828,
-      "grad_norm": 0.490234375,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.00030701052631578943,
-      "loss": 0.0216,
+      "loss": 0.0224,
       "num_input_tokens_seen": 336740352,
       "step": 5140,
-      "train_runtime": 259.7909,
-      "train_tokens_per_second": 1296197.746
+      "train_runtime": 2283.014,
+      "train_tokens_per_second": 147498.152
     },
     {
       "epoch": 1.1142362613587191,
-      "grad_norm": 0.54296875,
+      "grad_norm": 0.474609375,
       "learning_rate": 0.000306378947368421,
-      "loss": 0.0246,
+      "loss": 0.0243,
       "num_input_tokens_seen": 337395712,
       "step": 5150,
-      "train_runtime": 263.6165,
-      "train_tokens_per_second": 1279873.366
+      "train_runtime": 2286.8196,
+      "train_tokens_per_second": 147539.279
     },
     {
       "epoch": 1.1163998269147555,
-      "grad_norm": 0.470703125,
+      "grad_norm": 0.51171875,
       "learning_rate": 0.0003057473684210526,
-      "loss": 0.0238,
+      "loss": 0.0232,
       "num_input_tokens_seen": 338051072,
       "step": 5160,
-      "train_runtime": 267.4291,
-      "train_tokens_per_second": 1264077.394
+      "train_runtime": 2292.84,
+      "train_tokens_per_second": 147437.708
     },
     {
       "epoch": 1.1185633924707918,
-      "grad_norm": 0.51171875,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.0003051157894736842,
-      "loss": 0.0237,
+      "loss": 0.0235,
       "num_input_tokens_seen": 338706432,
       "step": 5170,
-      "train_runtime": 271.2587,
-      "train_tokens_per_second": 1248647.458
+      "train_runtime": 2296.6595,
+      "train_tokens_per_second": 147477.859
     },
     {
       "epoch": 1.1207269580268282,
-      "grad_norm": 0.71875,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.0003044842105263158,
-      "loss": 0.024,
+      "loss": 0.0243,
       "num_input_tokens_seen": 339361792,
       "step": 5180,
-      "train_runtime": 275.0804,
-      "train_tokens_per_second": 1233682.374
+      "train_runtime": 2300.471,
+      "train_tokens_per_second": 147518.394
     },
     {
       "epoch": 1.1228905235828646,
-      "grad_norm": 0.5625,
+      "grad_norm": 0.388671875,
       "learning_rate": 0.00030385263157894736,
-      "loss": 0.0238,
+      "loss": 0.0236,
       "num_input_tokens_seen": 340017152,
       "step": 5190,
-      "train_runtime": 278.9047,
-      "train_tokens_per_second": 1219115.936
+      "train_runtime": 2304.2794,
+      "train_tokens_per_second": 147558.997
     },
     {
       "epoch": 1.125054089138901,
-      "grad_norm": 0.67578125,
+      "grad_norm": 0.490234375,
       "learning_rate": 0.0003032210526315789,
-      "loss": 0.0262,
+      "loss": 0.0249,
       "num_input_tokens_seen": 340672512,
       "step": 5200,
-      "train_runtime": 282.7305,
-      "train_tokens_per_second": 1204937.424
+      "train_runtime": 2308.0903,
+      "train_tokens_per_second": 147599.302
     },
     {
       "epoch": 1.125054089138901,
-      "eval_loss": 0.016979500651359558,
-      "eval_runtime": 1.6264,
-      "eval_samples_per_second": 19.675,
-      "eval_steps_per_second": 0.615,
+      "eval_loss": 0.012978856451809406,
+      "eval_runtime": 2.1092,
+      "eval_samples_per_second": 15.172,
+      "eval_steps_per_second": 0.474,
       "num_input_tokens_seen": 340672512,
       "step": 5200
     },
     {
       "epoch": 1.125054089138901,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.02684713927562,
-      "eval_sacrebleu": 96.68143920500805,
-      "eval_word_accuracy": 0.9903846153846154,
+      "eval_byte_accuracy": 0.9970326409495549,
+      "eval_chrf": 97.87302942819403,
+      "eval_sacrebleu": 97.56427984542309,
+      "eval_word_accuracy": 0.9931318681318682,
       "num_input_tokens_seen": 340672512,
-      "perplexity": 1.017124471722373,
+      "perplexity": 1.013063447377543,
       "step": 5200
     },
     {
       "epoch": 1.1272176546949373,
-      "grad_norm": 0.376953125,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.0003025894736842105,
-      "loss": 0.0229,
+      "loss": 0.0216,
       "num_input_tokens_seen": 341327872,
       "step": 5210,
-      "train_runtime": 288.1928,
-      "train_tokens_per_second": 1184373.426
+      "train_runtime": 2316.2667,
+      "train_tokens_per_second": 147361.212
     },
     {
       "epoch": 1.1293812202509736,
-      "grad_norm": 0.828125,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.00030195789473684207,
-      "loss": 0.0241,
+      "loss": 0.0239,
       "num_input_tokens_seen": 341983232,
       "step": 5220,
-      "train_runtime": 292.0166,
-      "train_tokens_per_second": 1171108.686
+      "train_runtime": 2320.082,
+      "train_tokens_per_second": 147401.356
     },
     {
       "epoch": 1.13154478580701,
-      "grad_norm": 0.640625,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.00030132631578947365,
-      "loss": 0.0273,
+      "loss": 0.0276,
       "num_input_tokens_seen": 342638592,
       "step": 5230,
-      "train_runtime": 295.8511,
-      "train_tokens_per_second": 1158145.403
+      "train_runtime": 2323.8936,
+      "train_tokens_per_second": 147441.603
     },
     {
       "epoch": 1.1337083513630464,
-      "grad_norm": 0.6875,
+      "grad_norm": 0.640625,
       "learning_rate": 0.00030069473684210524,
-      "loss": 0.0242,
+      "loss": 0.0234,
       "num_input_tokens_seen": 343293952,
       "step": 5240,
-      "train_runtime": 299.6779,
-      "train_tokens_per_second": 1145543.28
+      "train_runtime": 2327.7124,
+      "train_tokens_per_second": 147481.259
     },
     {
       "epoch": 1.1358719169190827,
-      "grad_norm": 0.478515625,
+      "grad_norm": 0.6484375,
       "learning_rate": 0.0003000631578947368,
-      "loss": 0.0261,
+      "loss": 0.0263,
       "num_input_tokens_seen": 343949312,
       "step": 5250,
-      "train_runtime": 303.4995,
-      "train_tokens_per_second": 1133278.165
+      "train_runtime": 2331.5303,
+      "train_tokens_per_second": 147520.839
     },
     {
       "epoch": 1.1380354824751189,
-      "grad_norm": 0.546875,
+      "grad_norm": 0.640625,
       "learning_rate": 0.0002994315789473684,
-      "loss": 0.0232,
+      "loss": 0.0239,
       "num_input_tokens_seen": 344604672,
       "step": 5260,
-      "train_runtime": 307.3262,
-      "train_tokens_per_second": 1121299.327
+      "train_runtime": 2337.5477,
+      "train_tokens_per_second": 147421.452
     },
     {
       "epoch": 1.1401990480311555,
-      "grad_norm": 0.6015625,
+      "grad_norm": 0.53125,
       "learning_rate": 0.0002988,
-      "loss": 0.0254,
+      "loss": 0.0253,
       "num_input_tokens_seen": 345260032,
       "step": 5270,
-      "train_runtime": 311.1516,
-      "train_tokens_per_second": 1109619.945
+      "train_runtime": 2341.3694,
+      "train_tokens_per_second": 147460.726
     },
     {
       "epoch": 1.1423626135871916,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.53125,
       "learning_rate": 0.00029816842105263153,
-      "loss": 0.0228,
+      "loss": 0.0227,
       "num_input_tokens_seen": 345915392,
       "step": 5280,
-      "train_runtime": 314.9762,
-      "train_tokens_per_second": 1098227.147
+      "train_runtime": 2345.1866,
+      "train_tokens_per_second": 147500.159
     },
     {
       "epoch": 1.144526179143228,
-      "grad_norm": 0.51953125,
+      "grad_norm": 0.5859375,
       "learning_rate": 0.0002975368421052631,
-      "loss": 0.0241,
+      "loss": 0.024,
       "num_input_tokens_seen": 346570752,
       "step": 5290,
-      "train_runtime": 318.8009,
-      "train_tokens_per_second": 1087107.321
+      "train_runtime": 2348.9992,
+      "train_tokens_per_second": 147539.746
     },
     {
       "epoch": 1.1466897446992643,
-      "grad_norm": 0.400390625,
+      "grad_norm": 0.46875,
       "learning_rate": 0.0002969052631578947,
-      "loss": 0.0254,
+      "loss": 0.0259,
       "num_input_tokens_seen": 347226112,
       "step": 5300,
-      "train_runtime": 322.6315,
-      "train_tokens_per_second": 1076231.314
+      "train_runtime": 2352.8085,
+      "train_tokens_per_second": 147579.42
     },
     {
       "epoch": 1.1466897446992643,
-      "eval_loss": 0.017398551106452942,
-      "eval_runtime": 1.5555,
-      "eval_samples_per_second": 20.573,
-      "eval_steps_per_second": 0.643,
+      "eval_loss": 0.015913764014840126,
+      "eval_runtime": 2.0214,
+      "eval_samples_per_second": 15.831,
+      "eval_steps_per_second": 0.495,
       "num_input_tokens_seen": 347226112,
       "step": 5300
     },
     {
       "epoch": 1.1466897446992643,
-      "eval_byte_accuracy": 0.994807121661721,
-      "eval_chrf": 96.3910110996503,
-      "eval_sacrebleu": 95.04714140916596,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.9944362017804155,
+      "eval_chrf": 96.27201255408502,
+      "eval_sacrebleu": 96.39469981452005,
+      "eval_word_accuracy": 0.9876373626373627,
       "num_input_tokens_seen": 347226112,
-      "perplexity": 1.0175507875128094,
+      "perplexity": 1.0160410623260452,
       "step": 5300
     },
     {
       "epoch": 1.1488533102553007,
-      "grad_norm": 0.39453125,
+      "grad_norm": 0.416015625,
       "learning_rate": 0.0002962736842105263,
-      "loss": 0.0242,
+      "loss": 0.025,
       "num_input_tokens_seen": 347881472,
       "step": 5310,
-      "train_runtime": 328.0278,
-      "train_tokens_per_second": 1060524.345
+      "train_runtime": 2360.8879,
+      "train_tokens_per_second": 147351.964
     },
     {
       "epoch": 1.151016875811337,
-      "grad_norm": 0.66796875,
+      "grad_norm": 0.376953125,
       "learning_rate": 0.0002956421052631579,
-      "loss": 0.0252,
+      "loss": 0.0249,
       "num_input_tokens_seen": 348536832,
       "step": 5320,
-      "train_runtime": 331.855,
-      "train_tokens_per_second": 1050268.479
+      "train_runtime": 2364.6967,
+      "train_tokens_per_second": 147391.77
     },
     {
       "epoch": 1.1531804413673734,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.390625,
       "learning_rate": 0.00029501052631578946,
-      "loss": 0.025,
+      "loss": 0.0252,
       "num_input_tokens_seen": 349192192,
       "step": 5330,
-      "train_runtime": 335.6748,
-      "train_tokens_per_second": 1040269.264
+      "train_runtime": 2368.5236,
+      "train_tokens_per_second": 147430.318
     },
     {
       "epoch": 1.1553440069234098,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.62890625,
       "learning_rate": 0.00029437894736842105,
-      "loss": 0.0245,
+      "loss": 0.0246,
       "num_input_tokens_seen": 349847552,
       "step": 5340,
-      "train_runtime": 339.4914,
-      "train_tokens_per_second": 1030504.966
+      "train_runtime": 2372.3427,
+      "train_tokens_per_second": 147469.233
     },
     {
       "epoch": 1.1575075724794461,
-      "grad_norm": 0.66015625,
+      "grad_norm": 0.46875,
       "learning_rate": 0.00029374736842105264,
-      "loss": 0.0258,
+      "loss": 0.0256,
       "num_input_tokens_seen": 350502912,
       "step": 5350,
-      "train_runtime": 343.3195,
-      "train_tokens_per_second": 1020923.503
+      "train_runtime": 2376.1584,
+      "train_tokens_per_second": 147508.228
     },
     {
       "epoch": 1.1596711380354825,
-      "grad_norm": 0.63671875,
+      "grad_norm": 0.7421875,
       "learning_rate": 0.00029311578947368417,
-      "loss": 0.0242,
+      "loss": 0.0247,
       "num_input_tokens_seen": 351158272,
       "step": 5360,
-      "train_runtime": 347.1488,
-      "train_tokens_per_second": 1011549.652
+      "train_runtime": 2382.1833,
+      "train_tokens_per_second": 147410.268
     },
     {
       "epoch": 1.1618347035915189,
-      "grad_norm": 0.59765625,
+      "grad_norm": 0.6328125,
       "learning_rate": 0.00029248421052631575,
-      "loss": 0.0243,
+      "loss": 0.024,
       "num_input_tokens_seen": 351813632,
       "step": 5370,
-      "train_runtime": 350.956,
-      "train_tokens_per_second": 1002443.602
+      "train_runtime": 2385.9801,
+      "train_tokens_per_second": 147450.361
     },
     {
       "epoch": 1.1639982691475552,
-      "grad_norm": 0.4921875,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00029185263157894734,
-      "loss": 0.0245,
+      "loss": 0.0247,
       "num_input_tokens_seen": 352468992,
       "step": 5380,
-      "train_runtime": 354.7755,
-      "train_tokens_per_second": 993498.788
+      "train_runtime": 2389.7895,
+      "train_tokens_per_second": 147489.558
     },
     {
       "epoch": 1.1661618347035916,
-      "grad_norm": 0.70703125,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.00029122105263157893,
-      "loss": 0.0228,
+      "loss": 0.0234,
       "num_input_tokens_seen": 353124352,
       "step": 5390,
-      "train_runtime": 358.6124,
-      "train_tokens_per_second": 984696.408
+      "train_runtime": 2393.6014,
+      "train_tokens_per_second": 147528.467
     },
     {
       "epoch": 1.168325400259628,
-      "grad_norm": 0.52734375,
+      "grad_norm": 0.53125,
       "learning_rate": 0.0002905894736842105,
-      "loss": 0.0239,
+      "loss": 0.0236,
       "num_input_tokens_seen": 353779712,
       "step": 5400,
-      "train_runtime": 362.4359,
-      "train_tokens_per_second": 976116.534
+      "train_runtime": 2397.4054,
+      "train_tokens_per_second": 147567.749
     },
     {
       "epoch": 1.168325400259628,
-      "eval_loss": 0.016983406618237495,
-      "eval_runtime": 1.6338,
-      "eval_samples_per_second": 19.586,
-      "eval_steps_per_second": 0.612,
+      "eval_loss": 0.015481002628803253,
+      "eval_runtime": 2.2867,
+      "eval_samples_per_second": 13.994,
+      "eval_steps_per_second": 0.437,
       "num_input_tokens_seen": 353779712,
       "step": 5400
     },
     {
       "epoch": 1.168325400259628,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.53769868351773,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9903846153846154,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.64325335494277,
+      "eval_sacrebleu": 95.8953860263406,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 353779712,
-      "perplexity": 1.0171284445846291,
+      "perplexity": 1.0156014541172325,
       "step": 5400
     },
     {
       "epoch": 1.1704889658156643,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.640625,
       "learning_rate": 0.0002899578947368421,
-      "loss": 0.0215,
+      "loss": 0.0211,
       "num_input_tokens_seen": 354435072,
       "step": 5410,
-      "train_runtime": 367.8932,
-      "train_tokens_per_second": 963418.478
+      "train_runtime": 2405.5282,
+      "train_tokens_per_second": 147341.887
     },
     {
       "epoch": 1.1726525313717007,
-      "grad_norm": 0.470703125,
+      "grad_norm": 0.6328125,
       "learning_rate": 0.0002893263157894737,
-      "loss": 0.0222,
+      "loss": 0.0224,
       "num_input_tokens_seen": 355090432,
       "step": 5420,
-      "train_runtime": 371.7262,
-      "train_tokens_per_second": 955247.186
+      "train_runtime": 2409.3446,
+      "train_tokens_per_second": 147380.51
     },
     {
       "epoch": 1.1748160969277368,
-      "grad_norm": 0.5,
+      "grad_norm": 0.671875,
       "learning_rate": 0.0002886947368421052,
-      "loss": 0.0237,
+      "loss": 0.025,
       "num_input_tokens_seen": 355745792,
       "step": 5430,
-      "train_runtime": 375.5435,
-      "train_tokens_per_second": 947282.496
+      "train_runtime": 2413.1608,
+      "train_tokens_per_second": 147419.019
     },
     {
       "epoch": 1.1769796624837732,
-      "grad_norm": 0.375,
+      "grad_norm": 0.69140625,
       "learning_rate": 0.0002880631578947368,
-      "loss": 0.0232,
+      "loss": 0.0235,
       "num_input_tokens_seen": 356401152,
       "step": 5440,
-      "train_runtime": 379.3575,
-      "train_tokens_per_second": 939486.215
+      "train_runtime": 2416.9521,
+      "train_tokens_per_second": 147458.924
     },
     {
       "epoch": 1.1791432280398095,
-      "grad_norm": 0.71484375,
+      "grad_norm": 0.373046875,
       "learning_rate": 0.0002874315789473684,
-      "loss": 0.0245,
+      "loss": 0.0233,
       "num_input_tokens_seen": 357056512,
       "step": 5450,
-      "train_runtime": 383.1768,
-      "train_tokens_per_second": 931832.247
+      "train_runtime": 2420.7669,
+      "train_tokens_per_second": 147497.268
     },
     {
       "epoch": 1.181306793595846,
-      "grad_norm": 0.4765625,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.0002868,
-      "loss": 0.0223,
+      "loss": 0.0217,
       "num_input_tokens_seen": 357711872,
       "step": 5460,
-      "train_runtime": 386.9794,
-      "train_tokens_per_second": 924369.237
+      "train_runtime": 2426.9173,
+      "train_tokens_per_second": 147393.515
     },
     {
       "epoch": 1.1834703591518823,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00028616842105263156,
-      "loss": 0.0206,
+      "loss": 0.0211,
       "num_input_tokens_seen": 358367232,
       "step": 5470,
-      "train_runtime": 390.809,
-      "train_tokens_per_second": 916988.113
+      "train_runtime": 2430.738,
+      "train_tokens_per_second": 147431.449
     },
     {
       "epoch": 1.1856339247079186,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.609375,
       "learning_rate": 0.00028553684210526315,
-      "loss": 0.0233,
+      "loss": 0.0228,
       "num_input_tokens_seen": 359022592,
       "step": 5480,
-      "train_runtime": 394.6294,
-      "train_tokens_per_second": 909771.547
+      "train_runtime": 2434.5618,
+      "train_tokens_per_second": 147469.083
     },
     {
       "epoch": 1.187797490263955,
-      "grad_norm": 0.48828125,
+      "grad_norm": 0.296875,
       "learning_rate": 0.00028490526315789474,
-      "loss": 0.0232,
+      "loss": 0.0231,
       "num_input_tokens_seen": 359673856,
       "step": 5490,
-      "train_runtime": 398.4372,
-      "train_tokens_per_second": 902711.587
+      "train_runtime": 2438.3485,
+      "train_tokens_per_second": 147507.158
     },
     {
       "epoch": 1.1899610558199913,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.392578125,
       "learning_rate": 0.0002842736842105263,
-      "loss": 0.0268,
+      "loss": 0.0267,
       "num_input_tokens_seen": 360329216,
       "step": 5500,
-      "train_runtime": 402.2563,
-      "train_tokens_per_second": 895770.167
+      "train_runtime": 2442.168,
+      "train_tokens_per_second": 147544.813
     },
     {
       "epoch": 1.1899610558199913,
-      "eval_loss": 0.015771791338920593,
-      "eval_runtime": 1.6199,
-      "eval_samples_per_second": 19.754,
-      "eval_steps_per_second": 0.617,
+      "eval_loss": 0.01402247790247202,
+      "eval_runtime": 2.1152,
+      "eval_samples_per_second": 15.129,
+      "eval_steps_per_second": 0.473,
       "num_input_tokens_seen": 360329216,
       "step": 5500
     },
     {
       "epoch": 1.1899610558199913,
-      "eval_byte_accuracy": 0.9951780415430267,
-      "eval_chrf": 95.97882814418759,
-      "eval_sacrebleu": 95.13971801972568,
+      "eval_byte_accuracy": 0.9955489614243324,
+      "eval_chrf": 96.96256197348622,
+      "eval_sacrebleu": 95.41032845057128,
       "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 360329216,
-      "perplexity": 1.0158968224968823,
+      "perplexity": 1.014121254000941,
       "step": 5500
     },
     {
       "epoch": 1.1921246213760277,
-      "grad_norm": 0.53515625,
+      "grad_norm": 0.4765625,
       "learning_rate": 0.00028364210526315786,
-      "loss": 0.0223,
+      "loss": 0.0232,
       "num_input_tokens_seen": 360984576,
       "step": 5510,
-      "train_runtime": 407.7707,
-      "train_tokens_per_second": 885263.641
+      "train_runtime": 2450.3602,
+      "train_tokens_per_second": 147318.987
     },
     {
       "epoch": 1.194288186932064,
-      "grad_norm": 0.462890625,
+      "grad_norm": 0.474609375,
       "learning_rate": 0.00028301052631578944,
-      "loss": 0.0215,
+      "loss": 0.0224,
       "num_input_tokens_seen": 361639936,
       "step": 5520,
-      "train_runtime": 411.5863,
-      "train_tokens_per_second": 878649.158
+      "train_runtime": 2454.1587,
+      "train_tokens_per_second": 147358.01
     },
     {
       "epoch": 1.1964517524881004,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.00028237894736842103,
-      "loss": 0.0217,
+      "loss": 0.0225,
       "num_input_tokens_seen": 362295296,
       "step": 5530,
-      "train_runtime": 415.4216,
-      "train_tokens_per_second": 872114.779
+      "train_runtime": 2457.9704,
+      "train_tokens_per_second": 147396.119
     },
     {
       "epoch": 1.1986153180441368,
-      "grad_norm": 0.5078125,
+      "grad_norm": 0.423828125,
       "learning_rate": 0.0002817473684210526,
-      "loss": 0.0201,
+      "loss": 0.0207,
       "num_input_tokens_seen": 362950656,
       "step": 5540,
-      "train_runtime": 419.2418,
-      "train_tokens_per_second": 865731.155
+      "train_runtime": 2461.7834,
+      "train_tokens_per_second": 147434.036
     },
     {
       "epoch": 1.2007788836001732,
-      "grad_norm": 0.5703125,
+      "grad_norm": 0.451171875,
       "learning_rate": 0.0002811157894736842,
-      "loss": 0.0218,
+      "loss": 0.0224,
       "num_input_tokens_seen": 363606016,
       "step": 5550,
-      "train_runtime": 423.0643,
-      "train_tokens_per_second": 859458.108
+      "train_runtime": 2465.5888,
+      "train_tokens_per_second": 147472.285
     },
     {
       "epoch": 1.2029424491562095,
-      "grad_norm": 0.60546875,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.0002804842105263158,
-      "loss": 0.0238,
+      "loss": 0.023,
       "num_input_tokens_seen": 364261376,
       "step": 5560,
-      "train_runtime": 426.8852,
-      "train_tokens_per_second": 853300.583
+      "train_runtime": 2471.6087,
+      "train_tokens_per_second": 147378.254
     },
     {
       "epoch": 1.2051060147122459,
-      "grad_norm": 0.5625,
+      "grad_norm": 0.318359375,
       "learning_rate": 0.0002798526315789474,
-      "loss": 0.0242,
+      "loss": 0.0243,
       "num_input_tokens_seen": 364916736,
       "step": 5570,
-      "train_runtime": 430.7066,
-      "train_tokens_per_second": 847251.408
+      "train_runtime": 2475.4093,
+      "train_tokens_per_second": 147416.726
     },
     {
       "epoch": 1.207269580268282,
-      "grad_norm": 0.353515625,
+      "grad_norm": 0.474609375,
       "learning_rate": 0.00027922105263157896,
-      "loss": 0.022,
+      "loss": 0.0217,
       "num_input_tokens_seen": 365572096,
       "step": 5580,
-      "train_runtime": 434.5293,
-      "train_tokens_per_second": 841306.025
+      "train_runtime": 2479.2175,
+      "train_tokens_per_second": 147454.628
     },
     {
       "epoch": 1.2094331458243186,
-      "grad_norm": 0.54296875,
+      "grad_norm": 0.365234375,
       "learning_rate": 0.0002785894736842105,
-      "loss": 0.0224,
+      "loss": 0.0226,
       "num_input_tokens_seen": 366227456,
       "step": 5590,
-      "train_runtime": 438.3537,
-      "train_tokens_per_second": 835461.001
+      "train_runtime": 2483.0274,
+      "train_tokens_per_second": 147492.311
     },
     {
       "epoch": 1.2115967113803547,
-      "grad_norm": 0.58203125,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.0002779578947368421,
-      "loss": 0.0225,
+      "loss": 0.0228,
       "num_input_tokens_seen": 366878720,
       "step": 5600,
-      "train_runtime": 442.1531,
-      "train_tokens_per_second": 829754.863
+      "train_runtime": 2486.8135,
+      "train_tokens_per_second": 147529.65
     },
     {
       "epoch": 1.2115967113803547,
-      "eval_loss": 0.01541847176849842,
-      "eval_runtime": 1.7839,
-      "eval_samples_per_second": 17.938,
-      "eval_steps_per_second": 0.561,
+      "eval_loss": 0.01578563265502453,
+      "eval_runtime": 2.1536,
+      "eval_samples_per_second": 14.859,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 366878720,
       "step": 5600
     },
     {
       "epoch": 1.2115967113803547,
-      "eval_byte_accuracy": 0.9955489614243324,
-      "eval_chrf": 96.54446339411348,
-      "eval_sacrebleu": 95.8953860263406,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.9951780415430267,
+      "eval_chrf": 97.24990341576824,
+      "eval_sacrebleu": 96.77994309150014,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 366878720,
-      "perplexity": 1.0155379496700943,
+      "perplexity": 1.0159108839432456,
       "step": 5600
     },
     {
       "epoch": 1.213760276936391,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.3984375,
       "learning_rate": 0.00027732631578947367,
-      "loss": 0.0221,
+      "loss": 0.0214,
       "num_input_tokens_seen": 367534080,
       "step": 5610,
-      "train_runtime": 447.7745,
-      "train_tokens_per_second": 820801.684
+      "train_runtime": 2495.1977,
+      "train_tokens_per_second": 147296.576
     },
     {
       "epoch": 1.2159238424924275,
-      "grad_norm": 0.43359375,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.00027669473684210525,
-      "loss": 0.0205,
+      "loss": 0.0218,
       "num_input_tokens_seen": 368189440,
       "step": 5620,
-      "train_runtime": 451.6045,
-      "train_tokens_per_second": 815291.722
+      "train_runtime": 2499.0073,
+      "train_tokens_per_second": 147334.277
     },
     {
       "epoch": 1.2180874080484638,
-      "grad_norm": 0.53125,
+      "grad_norm": 0.400390625,
       "learning_rate": 0.00027606315789473684,
-      "loss": 0.0242,
+      "loss": 0.024,
       "num_input_tokens_seen": 368844800,
       "step": 5630,
-      "train_runtime": 455.4277,
-      "train_tokens_per_second": 809886.68
+      "train_runtime": 2502.8184,
+      "train_tokens_per_second": 147371.776
     },
     {
       "epoch": 1.2202509736045002,
-      "grad_norm": 0.46484375,
+      "grad_norm": 0.51953125,
       "learning_rate": 0.0002754315789473684,
-      "loss": 0.0244,
+      "loss": 0.0247,
       "num_input_tokens_seen": 369500160,
       "step": 5640,
-      "train_runtime": 459.243,
-      "train_tokens_per_second": 804585.215
+      "train_runtime": 2506.6357,
+      "train_tokens_per_second": 147408.8
     },
     {
       "epoch": 1.2224145391605366,
-      "grad_norm": 0.6328125,
+      "grad_norm": 0.5,
       "learning_rate": 0.0002748,
-      "loss": 0.0215,
+      "loss": 0.0228,
       "num_input_tokens_seen": 370151424,
       "step": 5650,
-      "train_runtime": 463.0536,
-      "train_tokens_per_second": 799370.528
+      "train_runtime": 2510.4299,
+      "train_tokens_per_second": 147445.433
     },
     {
       "epoch": 1.224578104716573,
-      "grad_norm": 0.484375,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.00027416842105263154,
-      "loss": 0.0247,
+      "loss": 0.0258,
       "num_input_tokens_seen": 370806784,
       "step": 5660,
-      "train_runtime": 466.8798,
-      "train_tokens_per_second": 794223.279
+      "train_runtime": 2516.4496,
+      "train_tokens_per_second": 147353.151
     },
     {
       "epoch": 1.2267416702726093,
-      "grad_norm": 0.43359375,
+      "grad_norm": 0.68359375,
       "learning_rate": 0.00027353684210526313,
       "loss": 0.0263,
       "num_input_tokens_seen": 371462144,
       "step": 5670,
-      "train_runtime": 470.6811,
-      "train_tokens_per_second": 789201.358
+      "train_runtime": 2520.2461,
+      "train_tokens_per_second": 147391.221
     },
     {
       "epoch": 1.2289052358286456,
-      "grad_norm": 0.546875,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.0002729052631578947,
-      "loss": 0.0248,
+      "loss": 0.0254,
       "num_input_tokens_seen": 372117504,
       "step": 5680,
-      "train_runtime": 474.4964,
-      "train_tokens_per_second": 784236.796
+      "train_runtime": 2524.0507,
+      "train_tokens_per_second": 147428.695
     },
     {
       "epoch": 1.231068801384682,
-      "grad_norm": 0.73828125,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0002722736842105263,
-      "loss": 0.0236,
+      "loss": 0.024,
       "num_input_tokens_seen": 372768768,
       "step": 5690,
-      "train_runtime": 478.295,
-      "train_tokens_per_second": 779369.954
+      "train_runtime": 2527.8427,
+      "train_tokens_per_second": 147465.177
     },
     {
       "epoch": 1.2332323669407184,
-      "grad_norm": 0.4375,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.0002716421052631579,
-      "loss": 0.0213,
+      "loss": 0.0227,
       "num_input_tokens_seen": 373424128,
       "step": 5700,
-      "train_runtime": 482.1137,
-      "train_tokens_per_second": 774556.193
+      "train_runtime": 2531.6566,
+      "train_tokens_per_second": 147501.888
     },
     {
       "epoch": 1.2332323669407184,
-      "eval_loss": 0.01559929084032774,
-      "eval_runtime": 1.636,
-      "eval_samples_per_second": 19.56,
-      "eval_steps_per_second": 0.611,
+      "eval_loss": 0.01439273077994585,
+      "eval_runtime": 2.1944,
+      "eval_samples_per_second": 14.583,
+      "eval_steps_per_second": 0.456,
       "num_input_tokens_seen": 373424128,
       "step": 5700
     },
     {
       "epoch": 1.2332323669407184,
-      "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 96.9001440282776,
-      "eval_sacrebleu": 95.64541250374113,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9951780415430267,
+      "eval_chrf": 96.76431117455587,
+      "eval_sacrebleu": 96.44528222520958,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 373424128,
-      "perplexity": 1.0157215949023408,
+      "perplexity": 1.0144968048334402,
       "step": 5700
     },
     {
       "epoch": 1.2353959324967547,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.56640625,
       "learning_rate": 0.0002710105263157895,
-      "loss": 0.0231,
+      "loss": 0.0232,
       "num_input_tokens_seen": 374079488,
       "step": 5710,
-      "train_runtime": 487.5803,
-      "train_tokens_per_second": 767216.19
+      "train_runtime": 2539.9047,
+      "train_tokens_per_second": 147280.912
     },
     {
       "epoch": 1.237559498052791,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00027037894736842106,
-      "loss": 0.0235,
+      "loss": 0.0242,
       "num_input_tokens_seen": 374734848,
       "step": 5720,
-      "train_runtime": 491.3926,
-      "train_tokens_per_second": 762597.695
+      "train_runtime": 2543.7103,
+      "train_tokens_per_second": 147318.211
     },
     {
       "epoch": 1.2397230636088272,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00026974736842105265,
-      "loss": 0.0225,
+      "loss": 0.0223,
       "num_input_tokens_seen": 375386112,
       "step": 5730,
-      "train_runtime": 495.1891,
-      "train_tokens_per_second": 758066.187
+      "train_runtime": 2547.5088,
+      "train_tokens_per_second": 147354.199
     },
     {
       "epoch": 1.2418866291648638,
-      "grad_norm": 0.48828125,
+      "grad_norm": 0.390625,
       "learning_rate": 0.0002691157894736842,
-      "loss": 0.0213,
+      "loss": 0.022,
       "num_input_tokens_seen": 376041472,
       "step": 5740,
-      "train_runtime": 499.0112,
-      "train_tokens_per_second": 753573.282
+      "train_runtime": 2551.311,
+      "train_tokens_per_second": 147391.465
     },
     {
       "epoch": 1.2440501947209,
-      "grad_norm": 0.44921875,
+      "grad_norm": 0.30859375,
       "learning_rate": 0.00026848421052631577,
       "loss": 0.0248,
       "num_input_tokens_seen": 376696832,
       "step": 5750,
-      "train_runtime": 502.8267,
-      "train_tokens_per_second": 749158.34
+      "train_runtime": 2555.1174,
+      "train_tokens_per_second": 147428.384
     },
     {
       "epoch": 1.2462137602769363,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00026785263157894735,
-      "loss": 0.0213,
+      "loss": 0.0215,
       "num_input_tokens_seen": 377352192,
       "step": 5760,
-      "train_runtime": 506.6473,
-      "train_tokens_per_second": 744802.561
+      "train_runtime": 2561.1355,
+      "train_tokens_per_second": 147337.848
     },
     {
       "epoch": 1.2483773258329727,
-      "grad_norm": 0.48828125,
+      "grad_norm": 0.408203125,
       "learning_rate": 0.00026722105263157894,
-      "loss": 0.0253,
+      "loss": 0.0259,
       "num_input_tokens_seen": 378007552,
       "step": 5770,
-      "train_runtime": 510.4719,
-      "train_tokens_per_second": 740506.029
+      "train_runtime": 2564.9477,
+      "train_tokens_per_second": 147374.368
     },
     {
       "epoch": 1.250540891389009,
-      "grad_norm": 0.484375,
+      "grad_norm": 0.5,
       "learning_rate": 0.0002665894736842105,
-      "loss": 0.0212,
+      "loss": 0.0228,
       "num_input_tokens_seen": 378662912,
       "step": 5780,
-      "train_runtime": 514.2971,
-      "train_tokens_per_second": 736272.655
+      "train_runtime": 2568.7519,
+      "train_tokens_per_second": 147411.241
     },
     {
       "epoch": 1.2527044569450454,
-      "grad_norm": 0.46875,
+      "grad_norm": 0.3359375,
       "learning_rate": 0.0002659578947368421,
-      "loss": 0.0225,
+      "loss": 0.0221,
       "num_input_tokens_seen": 379318272,
       "step": 5790,
-      "train_runtime": 518.1139,
-      "train_tokens_per_second": 732113.734
+      "train_runtime": 2572.5685,
+      "train_tokens_per_second": 147447.299
     },
     {
       "epoch": 1.2548680225010818,
-      "grad_norm": 0.4296875,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.0002653263157894737,
-      "loss": 0.0227,
+      "loss": 0.0235,
       "num_input_tokens_seen": 379973632,
       "step": 5800,
-      "train_runtime": 521.9145,
-      "train_tokens_per_second": 728038.134
+      "train_runtime": 2576.363,
+      "train_tokens_per_second": 147484.511
     },
     {
       "epoch": 1.2548680225010818,
-      "eval_loss": 0.015982676297426224,
-      "eval_runtime": 1.702,
-      "eval_samples_per_second": 18.802,
-      "eval_steps_per_second": 0.588,
+      "eval_loss": 0.013111169449985027,
+      "eval_runtime": 2.2119,
+      "eval_samples_per_second": 14.467,
+      "eval_steps_per_second": 0.452,
       "num_input_tokens_seen": 379973632,
       "step": 5800
     },
     {
       "epoch": 1.2548680225010818,
       "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.25120016329974,
-      "eval_sacrebleu": 95.26949349674726,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 96.67853700001116,
+      "eval_sacrebleu": 96.53538171173645,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 379973632,
-      "perplexity": 1.0161110824474469,
+      "perplexity": 1.0131974977077127,
       "step": 5800
     },
     {
       "epoch": 1.2570315880571181,
-      "grad_norm": 0.6796875,
+      "grad_norm": 0.515625,
       "learning_rate": 0.00026469473684210523,
-      "loss": 0.0226,
+      "loss": 0.0222,
       "num_input_tokens_seen": 380628992,
       "step": 5810,
-      "train_runtime": 527.4555,
-      "train_tokens_per_second": 721632.439
+      "train_runtime": 2584.6272,
+      "train_tokens_per_second": 147266.499
     },
     {
       "epoch": 1.2591951536131545,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.390625,
       "learning_rate": 0.0002640631578947368,
-      "loss": 0.0235,
+      "loss": 0.0233,
       "num_input_tokens_seen": 381284352,
       "step": 5820,
-      "train_runtime": 531.2775,
-      "train_tokens_per_second": 717674.632
+      "train_runtime": 2588.4283,
+      "train_tokens_per_second": 147303.427
     },
     {
       "epoch": 1.2613587191691908,
-      "grad_norm": 0.66015625,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.0002634315789473684,
-      "loss": 0.0213,
+      "loss": 0.0214,
       "num_input_tokens_seen": 381939712,
       "step": 5830,
-      "train_runtime": 535.1066,
-      "train_tokens_per_second": 713763.812
+      "train_runtime": 2592.2337,
+      "train_tokens_per_second": 147340.002
     },
     {
       "epoch": 1.2635222847252272,
-      "grad_norm": 0.337890625,
+      "grad_norm": 0.486328125,
       "learning_rate": 0.0002628,
-      "loss": 0.0217,
+      "loss": 0.0215,
       "num_input_tokens_seen": 382595072,
       "step": 5840,
-      "train_runtime": 538.9122,
-      "train_tokens_per_second": 709939.579
+      "train_runtime": 2596.0288,
+      "train_tokens_per_second": 147377.052
     },
     {
       "epoch": 1.2656858502812636,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.57421875,
       "learning_rate": 0.0002621684210526316,
-      "loss": 0.0209,
+      "loss": 0.0221,
       "num_input_tokens_seen": 383250432,
       "step": 5850,
-      "train_runtime": 542.7301,
-      "train_tokens_per_second": 706152.935
+      "train_runtime": 2599.8368,
+      "train_tokens_per_second": 147413.267
     },
     {
       "epoch": 1.2678494158373,
-      "grad_norm": 0.671875,
+      "grad_norm": 0.66796875,
       "learning_rate": 0.00026153684210526316,
-      "loss": 0.0237,
+      "loss": 0.0248,
       "num_input_tokens_seen": 383905792,
       "step": 5860,
-      "train_runtime": 546.548,
-      "train_tokens_per_second": 702419.116
+      "train_runtime": 2606.0753,
+      "train_tokens_per_second": 147311.857
     },
     {
       "epoch": 1.2700129813933363,
-      "grad_norm": 0.54296875,
+      "grad_norm": 0.640625,
       "learning_rate": 0.00026090526315789475,
-      "loss": 0.0226,
+      "loss": 0.0237,
       "num_input_tokens_seen": 384561152,
       "step": 5870,
-      "train_runtime": 550.3699,
-      "train_tokens_per_second": 698732.126
+      "train_runtime": 2609.8901,
+      "train_tokens_per_second": 147347.642
     },
     {
       "epoch": 1.2721765469493724,
-      "grad_norm": 0.3515625,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.0002602736842105263,
-      "loss": 0.0238,
+      "loss": 0.0235,
       "num_input_tokens_seen": 385216512,
       "step": 5880,
-      "train_runtime": 554.185,
-      "train_tokens_per_second": 695104.493
+      "train_runtime": 2613.6997,
+      "train_tokens_per_second": 147383.617
     },
     {
       "epoch": 1.274340112505409,
-      "grad_norm": 0.416015625,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.00025964210526315787,
-      "loss": 0.0209,
+      "loss": 0.0211,
       "num_input_tokens_seen": 385871872,
       "step": 5890,
-      "train_runtime": 558.0069,
-      "train_tokens_per_second": 691518.079
+      "train_runtime": 2617.5106,
+      "train_tokens_per_second": 147419.412
     },
     {
       "epoch": 1.2765036780614452,
-      "grad_norm": 0.49609375,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00025901052631578946,
-      "loss": 0.0261,
+      "loss": 0.0263,
       "num_input_tokens_seen": 386527232,
       "step": 5900,
-      "train_runtime": 561.8248,
-      "train_tokens_per_second": 687985.319
+      "train_runtime": 2621.3256,
+      "train_tokens_per_second": 147454.874
     },
     {
       "epoch": 1.2765036780614452,
-      "eval_loss": 0.016280530020594597,
-      "eval_runtime": 1.7769,
-      "eval_samples_per_second": 18.009,
-      "eval_steps_per_second": 0.563,
+      "eval_loss": 0.014893613755702972,
+      "eval_runtime": 2.2011,
+      "eval_samples_per_second": 14.538,
+      "eval_steps_per_second": 0.454,
       "num_input_tokens_seen": 386527232,
       "step": 5900
     },
     {
       "epoch": 1.2765036780614452,
-      "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 97.0096297173289,
-      "eval_sacrebleu": 96.23024322319195,
+      "eval_byte_accuracy": 0.9951780415430267,
+      "eval_chrf": 97.5606524584088,
+      "eval_sacrebleu": 96.5155008972932,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 386527232,
-      "perplexity": 1.0164137799940642,
+      "perplexity": 1.0150050762935756,
       "step": 5900
     },
     {
       "epoch": 1.2786672436174817,
-      "grad_norm": 0.60546875,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.00025837894736842104,
-      "loss": 0.0243,
+      "loss": 0.0248,
       "num_input_tokens_seen": 387182592,
       "step": 5910,
-      "train_runtime": 567.4347,
-      "train_tokens_per_second": 682338.577
+      "train_runtime": 2629.6051,
+      "train_tokens_per_second": 147239.825
     },
     {
       "epoch": 1.2808308091735179,
-      "grad_norm": 0.439453125,
+      "grad_norm": 0.54296875,
       "learning_rate": 0.00025774736842105263,
-      "loss": 0.0219,
+      "loss": 0.0227,
       "num_input_tokens_seen": 387837952,
       "step": 5920,
-      "train_runtime": 571.2533,
-      "train_tokens_per_second": 678924.606
+      "train_runtime": 2633.4187,
+      "train_tokens_per_second": 147275.462
     },
     {
       "epoch": 1.2829943747295542,
-      "grad_norm": 0.310546875,
+      "grad_norm": 0.58203125,
       "learning_rate": 0.0002571157894736842,
-      "loss": 0.0237,
+      "loss": 0.0251,
       "num_input_tokens_seen": 388493312,
       "step": 5930,
-      "train_runtime": 575.0688,
-      "train_tokens_per_second": 675559.711
+      "train_runtime": 2637.2287,
+      "train_tokens_per_second": 147311.197
     },
     {
       "epoch": 1.2851579402855906,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.59765625,
       "learning_rate": 0.0002564842105263158,
-      "loss": 0.0215,
+      "loss": 0.0233,
       "num_input_tokens_seen": 389148672,
       "step": 5940,
-      "train_runtime": 578.8832,
-      "train_tokens_per_second": 672240.439
+      "train_runtime": 2641.041,
+      "train_tokens_per_second": 147346.701
     },
     {
       "epoch": 1.287321505841627,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.4140625,
       "learning_rate": 0.00025585263157894733,
-      "loss": 0.0223,
+      "loss": 0.0225,
       "num_input_tokens_seen": 389804032,
       "step": 5950,
-      "train_runtime": 582.6975,
-      "train_tokens_per_second": 668964.678
+      "train_runtime": 2644.8514,
+      "train_tokens_per_second": 147382.205
     },
     {
       "epoch": 1.2894850713976633,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.431640625,
       "learning_rate": 0.0002552210526315789,
-      "loss": 0.0233,
+      "loss": 0.0232,
       "num_input_tokens_seen": 390459392,
       "step": 5960,
-      "train_runtime": 586.5197,
-      "train_tokens_per_second": 665722.568
+      "train_runtime": 2650.9057,
+      "train_tokens_per_second": 147292.823
     },
     {
       "epoch": 1.2916486369536997,
-      "grad_norm": 0.412109375,
+      "grad_norm": 0.326171875,
       "learning_rate": 0.0002545894736842105,
-      "loss": 0.0213,
+      "loss": 0.0216,
       "num_input_tokens_seen": 391114752,
       "step": 5970,
-      "train_runtime": 590.3469,
-      "train_tokens_per_second": 662516.839
+      "train_runtime": 2654.7136,
+      "train_tokens_per_second": 147328.415
     },
     {
       "epoch": 1.293812202509736,
-      "grad_norm": 0.419921875,
+      "grad_norm": 0.5,
       "learning_rate": 0.0002539578947368421,
-      "loss": 0.023,
+      "loss": 0.0233,
       "num_input_tokens_seen": 391770112,
       "step": 5980,
-      "train_runtime": 594.1612,
-      "train_tokens_per_second": 659366.757
+      "train_runtime": 2658.5272,
+      "train_tokens_per_second": 147363.588
     },
     {
       "epoch": 1.2959757680657724,
-      "grad_norm": 0.5625,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0002533263157894737,
-      "loss": 0.023,
+      "loss": 0.024,
       "num_input_tokens_seen": 392425472,
       "step": 5990,
-      "train_runtime": 597.9746,
-      "train_tokens_per_second": 656257.765
+      "train_runtime": 2662.3361,
+      "train_tokens_per_second": 147398.924
     },
     {
       "epoch": 1.2981393336218088,
-      "grad_norm": 0.59375,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.00025269473684210527,
-      "loss": 0.0231,
+      "loss": 0.0242,
       "num_input_tokens_seen": 393080832,
       "step": 6000,
-      "train_runtime": 601.8047,
-      "train_tokens_per_second": 653170.044
+      "train_runtime": 2666.1611,
+      "train_tokens_per_second": 147433.261
     },
     {
       "epoch": 1.2981393336218088,
-      "eval_loss": 0.015439599752426147,
-      "eval_runtime": 1.6914,
-      "eval_samples_per_second": 18.919,
-      "eval_steps_per_second": 0.591,
+      "eval_loss": 0.014728213660418987,
+      "eval_runtime": 2.1539,
+      "eval_samples_per_second": 14.857,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 393080832,
       "step": 6000
     },
     {
       "epoch": 1.2981393336218088,
-      "eval_byte_accuracy": 0.9955489614243324,
-      "eval_chrf": 96.31324033462046,
-      "eval_sacrebleu": 95.31161095839255,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.995919881305638,
+      "eval_chrf": 97.03697158984805,
+      "eval_sacrebleu": 97.0441363353587,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 393080832,
-      "perplexity": 1.0155594061662383,
+      "perplexity": 1.0148372082403216,
       "step": 6000
     },
     {
       "epoch": 1.3003028991778451,
-      "grad_norm": 0.498046875,
+      "grad_norm": 0.65625,
       "learning_rate": 0.00025206315789473685,
-      "loss": 0.0227,
+      "loss": 0.0226,
       "num_input_tokens_seen": 393736192,
       "step": 6010,
-      "train_runtime": 607.3786,
-      "train_tokens_per_second": 648255.009
+      "train_runtime": 2674.4254,
+      "train_tokens_per_second": 147222.724
     },
     {
       "epoch": 1.3024664647338815,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.49609375,
       "learning_rate": 0.0002514315789473684,
-      "loss": 0.0212,
+      "loss": 0.0222,
       "num_input_tokens_seen": 394391552,
       "step": 6020,
-      "train_runtime": 611.1994,
-      "train_tokens_per_second": 645274.738
+      "train_runtime": 2678.2398,
+      "train_tokens_per_second": 147257.744
     },
     {
       "epoch": 1.3046300302899179,
-      "grad_norm": 0.326171875,
+      "grad_norm": 0.39453125,
       "learning_rate": 0.00025079999999999997,
-      "loss": 0.0225,
+      "loss": 0.0236,
       "num_input_tokens_seen": 395046912,
       "step": 6030,
-      "train_runtime": 615.0088,
-      "train_tokens_per_second": 642343.486
+      "train_runtime": 2682.0502,
+      "train_tokens_per_second": 147292.888
     },
     {
       "epoch": 1.3067935958459542,
-      "grad_norm": 0.478515625,
+      "grad_norm": 0.470703125,
       "learning_rate": 0.00025016842105263156,
-      "loss": 0.0255,
+      "loss": 0.0262,
       "num_input_tokens_seen": 395702272,
       "step": 6040,
-      "train_runtime": 618.8251,
-      "train_tokens_per_second": 639441.233
+      "train_runtime": 2685.8612,
+      "train_tokens_per_second": 147327.891
     },
     {
       "epoch": 1.3089571614019904,
-      "grad_norm": 0.42578125,
+      "grad_norm": 0.43359375,
       "learning_rate": 0.00024953684210526314,
-      "loss": 0.0221,
+      "loss": 0.0237,
       "num_input_tokens_seen": 396357632,
       "step": 6050,
-      "train_runtime": 622.6396,
-      "train_tokens_per_second": 636576.348
+      "train_runtime": 2689.6805,
+      "train_tokens_per_second": 147362.345
     },
     {
       "epoch": 1.311120726958027,
-      "grad_norm": 0.416015625,
+      "grad_norm": 0.486328125,
       "learning_rate": 0.00024890526315789473,
-      "loss": 0.0212,
+      "loss": 0.0217,
       "num_input_tokens_seen": 397012992,
       "step": 6060,
-      "train_runtime": 626.4648,
-      "train_tokens_per_second": 633735.535
+      "train_runtime": 2695.7306,
+      "train_tokens_per_second": 147274.727
     },
     {
       "epoch": 1.313284292514063,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.54296875,
       "learning_rate": 0.0002482736842105263,
-      "loss": 0.0236,
+      "loss": 0.024,
       "num_input_tokens_seen": 397668352,
       "step": 6070,
-      "train_runtime": 630.2994,
-      "train_tokens_per_second": 630919.772
+      "train_runtime": 2699.5518,
+      "train_tokens_per_second": 147309.027
     },
     {
       "epoch": 1.3154478580700995,
-      "grad_norm": 0.6875,
+      "grad_norm": 0.41796875,
       "learning_rate": 0.0002476421052631579,
-      "loss": 0.0202,
+      "loss": 0.0213,
       "num_input_tokens_seen": 398323712,
       "step": 6080,
-      "train_runtime": 634.1117,
-      "train_tokens_per_second": 628160.174
+      "train_runtime": 2703.3717,
+      "train_tokens_per_second": 147343.303
     },
     {
       "epoch": 1.3176114236261358,
-      "grad_norm": 0.49609375,
+      "grad_norm": 0.3828125,
       "learning_rate": 0.00024701052631578943,
-      "loss": 0.0232,
+      "loss": 0.0238,
       "num_input_tokens_seen": 398974976,
       "step": 6090,
-      "train_runtime": 637.9159,
-      "train_tokens_per_second": 625435.07
+      "train_runtime": 2707.1731,
+      "train_tokens_per_second": 147376.97
     },
     {
       "epoch": 1.3197749891821722,
-      "grad_norm": 0.6953125,
+      "grad_norm": 0.451171875,
       "learning_rate": 0.000246378947368421,
-      "loss": 0.0223,
+      "loss": 0.0234,
       "num_input_tokens_seen": 399626240,
       "step": 6100,
-      "train_runtime": 641.7163,
-      "train_tokens_per_second": 622745.948
+      "train_runtime": 2710.974,
+      "train_tokens_per_second": 147410.577
     },
     {
       "epoch": 1.3197749891821722,
-      "eval_loss": 0.014141172170639038,
-      "eval_runtime": 1.7688,
-      "eval_samples_per_second": 18.092,
-      "eval_steps_per_second": 0.565,
+      "eval_loss": 0.014207079075276852,
+      "eval_runtime": 2.3254,
+      "eval_samples_per_second": 13.761,
+      "eval_steps_per_second": 0.43,
       "num_input_tokens_seen": 399626240,
       "step": 6100
     },
     {
       "epoch": 1.3197749891821722,
-      "eval_byte_accuracy": 0.9977744807121661,
-      "eval_chrf": 98.58235971529302,
-      "eval_sacrebleu": 97.87900330273517,
-      "eval_word_accuracy": 0.9945054945054945,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 96.607979287624,
+      "eval_sacrebleu": 96.44528222520958,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 399626240,
-      "perplexity": 1.014241631524937,
+      "perplexity": 1.014308479254265,
       "step": 6100
     },
     {
       "epoch": 1.3219385547382085,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.0002457473684210526,
-      "loss": 0.021,
+      "loss": 0.0221,
       "num_input_tokens_seen": 400281600,
       "step": 6110,
-      "train_runtime": 647.3233,
-      "train_tokens_per_second": 618364.309
+      "train_runtime": 2719.3656,
+      "train_tokens_per_second": 147196.683
     },
     {
       "epoch": 1.324102120294245,
-      "grad_norm": 0.546875,
+      "grad_norm": 0.443359375,
       "learning_rate": 0.0002451157894736842,
-      "loss": 0.0203,
+      "loss": 0.0207,
       "num_input_tokens_seen": 400936960,
       "step": 6120,
-      "train_runtime": 651.1178,
-      "train_tokens_per_second": 615767.119
+      "train_runtime": 2723.1458,
+      "train_tokens_per_second": 147233.014
     },
     {
       "epoch": 1.3262656858502813,
-      "grad_norm": 0.42578125,
+      "grad_norm": 0.58984375,
       "learning_rate": 0.0002444842105263158,
-      "loss": 0.0224,
+      "loss": 0.024,
       "num_input_tokens_seen": 401592320,
       "step": 6130,
-      "train_runtime": 654.9362,
-      "train_tokens_per_second": 613177.74
+      "train_runtime": 2726.9626,
+      "train_tokens_per_second": 147267.263
     },
     {
       "epoch": 1.3284292514063176,
-      "grad_norm": 0.51171875,
+      "grad_norm": 0.474609375,
       "learning_rate": 0.00024385263157894737,
-      "loss": 0.0204,
+      "loss": 0.0214,
       "num_input_tokens_seen": 402247680,
       "step": 6140,
-      "train_runtime": 658.7591,
-      "train_tokens_per_second": 610614.248
+      "train_runtime": 2730.774,
+      "train_tokens_per_second": 147301.709
     },
     {
       "epoch": 1.330592816962354,
-      "grad_norm": 0.6171875,
+      "grad_norm": 0.419921875,
       "learning_rate": 0.00024322105263157893,
-      "loss": 0.0209,
+      "loss": 0.0216,
       "num_input_tokens_seen": 402903040,
       "step": 6150,
-      "train_runtime": 662.5781,
-      "train_tokens_per_second": 608083.892
+      "train_runtime": 2734.5891,
+      "train_tokens_per_second": 147335.863
     },
     {
       "epoch": 1.3327563825183903,
-      "grad_norm": 0.58984375,
+      "grad_norm": 0.3828125,
       "learning_rate": 0.00024258947368421049,
-      "loss": 0.0214,
+      "loss": 0.0216,
       "num_input_tokens_seen": 403558400,
       "step": 6160,
-      "train_runtime": 666.4075,
-      "train_tokens_per_second": 605573.016
+      "train_runtime": 2740.6289,
+      "train_tokens_per_second": 147250.292
     },
     {
       "epoch": 1.3349199480744267,
-      "grad_norm": 0.6015625,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.00024195789473684207,
-      "loss": 0.0226,
+      "loss": 0.0229,
       "num_input_tokens_seen": 404213760,
       "step": 6170,
-      "train_runtime": 670.2283,
-      "train_tokens_per_second": 603098.629
+      "train_runtime": 2744.4401,
+      "train_tokens_per_second": 147284.598
     },
     {
       "epoch": 1.337083513630463,
-      "grad_norm": 0.65234375,
+      "grad_norm": 0.48828125,
       "learning_rate": 0.00024132631578947366,
-      "loss": 0.0214,
+      "loss": 0.0224,
       "num_input_tokens_seen": 404869120,
       "step": 6180,
-      "train_runtime": 674.0552,
-      "train_tokens_per_second": 600646.812
+      "train_runtime": 2748.2532,
+      "train_tokens_per_second": 147318.71
     },
     {
       "epoch": 1.3392470791864994,
-      "grad_norm": 0.44140625,
+      "grad_norm": 0.46875,
       "learning_rate": 0.00024069473684210524,
-      "loss": 0.0223,
+      "loss": 0.0238,
       "num_input_tokens_seen": 405524480,
       "step": 6190,
-      "train_runtime": 677.8865,
-      "train_tokens_per_second": 598218.858
+      "train_runtime": 2752.0697,
+      "train_tokens_per_second": 147352.548
     },
     {
       "epoch": 1.3414106447425356,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.4140625,
       "learning_rate": 0.00024006315789473683,
-      "loss": 0.0199,
+      "loss": 0.0207,
       "num_input_tokens_seen": 406179840,
       "step": 6200,
-      "train_runtime": 681.6967,
-      "train_tokens_per_second": 595836.593
+      "train_runtime": 2755.863,
+      "train_tokens_per_second": 147387.528
     },
     {
       "epoch": 1.3414106447425356,
-      "eval_loss": 0.015185019001364708,
-      "eval_runtime": 1.6993,
-      "eval_samples_per_second": 18.831,
-      "eval_steps_per_second": 0.588,
+      "eval_loss": 0.013366620987653732,
+      "eval_runtime": 2.3739,
+      "eval_samples_per_second": 13.48,
+      "eval_steps_per_second": 0.421,
       "num_input_tokens_seen": 406179840,
       "step": 6200
     },
     {
       "epoch": 1.3414106447425356,
       "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.8403707869436,
-      "eval_sacrebleu": 96.97610863887425,
+      "eval_chrf": 97.6580883112683,
+      "eval_sacrebleu": 96.60546188979983,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 406179840,
-      "perplexity": 1.015300897196969,
+      "perplexity": 1.013456353627628,
       "step": 6200
     },
     {
       "epoch": 1.3435742102985722,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.6171875,
       "learning_rate": 0.00023943157894736842,
-      "loss": 0.0232,
+      "loss": 0.0253,
       "num_input_tokens_seen": 406835200,
       "step": 6210,
-      "train_runtime": 687.2335,
-      "train_tokens_per_second": 591989.788
+      "train_runtime": 2764.2874,
+      "train_tokens_per_second": 147175.436
     },
     {
       "epoch": 1.3457377758546083,
-      "grad_norm": 0.609375,
+      "grad_norm": 0.5703125,
       "learning_rate": 0.0002388,
-      "loss": 0.0208,
+      "loss": 0.0212,
       "num_input_tokens_seen": 407490560,
       "step": 6220,
-      "train_runtime": 691.056,
-      "train_tokens_per_second": 589663.599
+      "train_runtime": 2768.0945,
+      "train_tokens_per_second": 147209.774
     },
     {
       "epoch": 1.3479013414106447,
-      "grad_norm": 0.484375,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.00023816842105263154,
-      "loss": 0.0204,
+      "loss": 0.0212,
       "num_input_tokens_seen": 408145920,
       "step": 6230,
-      "train_runtime": 694.8687,
-      "train_tokens_per_second": 587371.244
+      "train_runtime": 2771.8982,
+      "train_tokens_per_second": 147244.197
     },
     {
       "epoch": 1.350064906966681,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.296875,
       "learning_rate": 0.00023753684210526312,
       "loss": 0.0195,
       "num_input_tokens_seen": 408801280,
       "step": 6240,
-      "train_runtime": 698.6775,
-      "train_tokens_per_second": 585107.283
+      "train_runtime": 2775.6897,
+      "train_tokens_per_second": 147279.173
     },
     {
       "epoch": 1.3522284725227174,
-      "grad_norm": 0.60546875,
+      "grad_norm": 0.416015625,
       "learning_rate": 0.0002369052631578947,
-      "loss": 0.0223,
+      "loss": 0.0226,
       "num_input_tokens_seen": 409456640,
       "step": 6250,
-      "train_runtime": 702.4988,
-      "train_tokens_per_second": 582857.396
+      "train_runtime": 2779.506,
+      "train_tokens_per_second": 147312.741
     },
     {
       "epoch": 1.3543920380787537,
-      "grad_norm": 0.59375,
+      "grad_norm": 0.490234375,
       "learning_rate": 0.0002362736842105263,
-      "loss": 0.0215,
+      "loss": 0.022,
       "num_input_tokens_seen": 410112000,
       "step": 6260,
-      "train_runtime": 706.3213,
-      "train_tokens_per_second": 580630.921
+      "train_runtime": 2785.5514,
+      "train_tokens_per_second": 147228.3
     },
     {
       "epoch": 1.35655560363479,
-      "grad_norm": 0.51171875,
+      "grad_norm": 0.453125,
       "learning_rate": 0.00023564210526315788,
-      "loss": 0.0216,
+      "loss": 0.0229,
       "num_input_tokens_seen": 410767360,
       "step": 6270,
-      "train_runtime": 710.1478,
-      "train_tokens_per_second": 578425.136
+      "train_runtime": 2789.3661,
+      "train_tokens_per_second": 147261.903
     },
     {
       "epoch": 1.3587191691908265,
-      "grad_norm": 0.359375,
+      "grad_norm": 0.30859375,
       "learning_rate": 0.00023501052631578947,
-      "loss": 0.02,
+      "loss": 0.0214,
       "num_input_tokens_seen": 411422720,
       "step": 6280,
-      "train_runtime": 713.9708,
-      "train_tokens_per_second": 576245.865
+      "train_runtime": 2793.186,
+      "train_tokens_per_second": 147295.136
     },
     {
       "epoch": 1.3608827347468628,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00023437894736842105,
       "loss": 0.0226,
       "num_input_tokens_seen": 412078080,
       "step": 6290,
-      "train_runtime": 717.7901,
-      "train_tokens_per_second": 574092.768
+      "train_runtime": 2797.0053,
+      "train_tokens_per_second": 147328.315
     },
     {
       "epoch": 1.3630463003028992,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.43359375,
       "learning_rate": 0.00023374736842105264,
-      "loss": 0.0209,
+      "loss": 0.0213,
       "num_input_tokens_seen": 412733440,
       "step": 6300,
-      "train_runtime": 721.6105,
-      "train_tokens_per_second": 571961.491
+      "train_runtime": 2800.8201,
+      "train_tokens_per_second": 147361.637
     },
     {
       "epoch": 1.3630463003028992,
-      "eval_loss": 0.01683359406888485,
-      "eval_runtime": 1.7461,
-      "eval_samples_per_second": 18.326,
-      "eval_steps_per_second": 0.573,
+      "eval_loss": 0.011968844570219517,
+      "eval_runtime": 2.2624,
+      "eval_samples_per_second": 14.144,
+      "eval_steps_per_second": 0.442,
       "num_input_tokens_seen": 412733440,
       "step": 6300
     },
     {
       "epoch": 1.3630463003028992,
       "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 97.50495693707096,
-      "eval_sacrebleu": 96.96578008549908,
-      "eval_word_accuracy": 0.9903846153846154,
+      "eval_chrf": 97.74613558028801,
+      "eval_sacrebleu": 97.11458584551553,
+      "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 412733440,
-      "perplexity": 1.0169760773928707,
+      "perplexity": 1.0120407578101336,
       "step": 6300
     },
     {
       "epoch": 1.3652098658589356,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.4375,
       "learning_rate": 0.00023311578947368417,
-      "loss": 0.0212,
+      "loss": 0.0218,
       "num_input_tokens_seen": 413388800,
       "step": 6310,
-      "train_runtime": 727.2029,
-      "train_tokens_per_second": 568464.163
+      "train_runtime": 2808.9578,
+      "train_tokens_per_second": 147168.037
     },
     {
       "epoch": 1.367373431414972,
-      "grad_norm": 0.380859375,
+      "grad_norm": 0.384765625,
       "learning_rate": 0.00023248421052631576,
-      "loss": 0.0194,
+      "loss": 0.02,
       "num_input_tokens_seen": 414044160,
       "step": 6320,
-      "train_runtime": 731.0202,
-      "train_tokens_per_second": 566392.203
+      "train_runtime": 2812.7793,
+      "train_tokens_per_second": 147201.084
     },
     {
       "epoch": 1.3695369969710083,
-      "grad_norm": 0.5546875,
+      "grad_norm": 0.486328125,
       "learning_rate": 0.00023185263157894735,
       "loss": 0.0223,
       "num_input_tokens_seen": 414699520,
       "step": 6330,
-      "train_runtime": 734.8416,
-      "train_tokens_per_second": 564338.689
+      "train_runtime": 2816.5902,
+      "train_tokens_per_second": 147234.595
     },
     {
       "epoch": 1.3717005625270446,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.54296875,
       "learning_rate": 0.00023122105263157893,
-      "loss": 0.0198,
+      "loss": 0.0209,
       "num_input_tokens_seen": 415354880,
       "step": 6340,
-      "train_runtime": 738.6717,
-      "train_tokens_per_second": 562299.732
+      "train_runtime": 2820.4056,
+      "train_tokens_per_second": 147267.781
     },
     {
       "epoch": 1.3738641280830808,
-      "grad_norm": 0.447265625,
+      "grad_norm": 0.48828125,
       "learning_rate": 0.00023058947368421052,
-      "loss": 0.0219,
+      "loss": 0.0232,
       "num_input_tokens_seen": 416010240,
       "step": 6350,
-      "train_runtime": 742.4996,
-      "train_tokens_per_second": 560283.436
+      "train_runtime": 2824.215,
+      "train_tokens_per_second": 147301.193
     },
     {
       "epoch": 1.3760276936391174,
-      "grad_norm": 0.5,
+      "grad_norm": 0.5859375,
       "learning_rate": 0.0002299578947368421,
-      "loss": 0.0206,
+      "loss": 0.0213,
       "num_input_tokens_seen": 416665600,
       "step": 6360,
-      "train_runtime": 746.3237,
-      "train_tokens_per_second": 558290.723
+      "train_runtime": 2830.3894,
+      "train_tokens_per_second": 147211.405
     },
     {
       "epoch": 1.3781912591951535,
-      "grad_norm": 0.53515625,
+      "grad_norm": 0.52734375,
       "learning_rate": 0.0002293263157894737,
-      "loss": 0.0223,
+      "loss": 0.0228,
       "num_input_tokens_seen": 417320960,
       "step": 6370,
-      "train_runtime": 750.1461,
-      "train_tokens_per_second": 556319.563
+      "train_runtime": 2834.2046,
+      "train_tokens_per_second": 147244.473
     },
     {
       "epoch": 1.38035482475119,
-      "grad_norm": 0.4453125,
+      "grad_norm": 0.57421875,
       "learning_rate": 0.00022869473684210522,
-      "loss": 0.0214,
+      "loss": 0.0231,
       "num_input_tokens_seen": 417976320,
       "step": 6380,
-      "train_runtime": 753.968,
-      "train_tokens_per_second": 554368.783
+      "train_runtime": 2838.014,
+      "train_tokens_per_second": 147277.752
     },
     {
       "epoch": 1.3825183903072262,
-      "grad_norm": 0.4296875,
+      "grad_norm": 0.470703125,
       "learning_rate": 0.0002280631578947368,
-      "loss": 0.0188,
+      "loss": 0.0196,
       "num_input_tokens_seen": 418631680,
       "step": 6390,
-      "train_runtime": 757.7935,
-      "train_tokens_per_second": 552435.046
+      "train_runtime": 2841.8354,
+      "train_tokens_per_second": 147310.319
     },
     {
       "epoch": 1.3846819558632626,
-      "grad_norm": 0.34765625,
+      "grad_norm": 0.390625,
       "learning_rate": 0.0002274315789473684,
-      "loss": 0.0197,
+      "loss": 0.0215,
       "num_input_tokens_seen": 419287040,
       "step": 6400,
-      "train_runtime": 761.6173,
-      "train_tokens_per_second": 550521.964
+      "train_runtime": 2845.6548,
+      "train_tokens_per_second": 147342.904
     },
     {
       "epoch": 1.3846819558632626,
-      "eval_loss": 0.015446390956640244,
-      "eval_runtime": 1.7867,
-      "eval_samples_per_second": 17.91,
-      "eval_steps_per_second": 0.56,
+      "eval_loss": 0.012704423628747463,
+      "eval_runtime": 2.187,
+      "eval_samples_per_second": 14.632,
+      "eval_steps_per_second": 0.457,
       "num_input_tokens_seen": 419287040,
       "step": 6400
     },
     {
       "epoch": 1.3846819558632626,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.04447388777369,
-      "eval_sacrebleu": 94.67835385995274,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.9970326409495549,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
+      "eval_word_accuracy": 0.9931318681318682,
       "num_input_tokens_seen": 419287040,
-      "perplexity": 1.0155663030609763,
+      "perplexity": 1.012785467660862,
       "step": 6400
     },
     {
       "epoch": 1.386845521419299,
-      "grad_norm": 0.66015625,
+      "grad_norm": 0.490234375,
       "learning_rate": 0.00022679999999999998,
-      "loss": 0.022,
+      "loss": 0.0223,
       "num_input_tokens_seen": 419942400,
       "step": 6410,
-      "train_runtime": 767.2444,
-      "train_tokens_per_second": 547338.49
+      "train_runtime": 2853.8879,
+      "train_tokens_per_second": 147147.474
     },
     {
       "epoch": 1.3890090869753353,
-      "grad_norm": 0.470703125,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00022616842105263157,
-      "loss": 0.0199,
+      "loss": 0.0205,
       "num_input_tokens_seen": 420597760,
       "step": 6420,
-      "train_runtime": 771.0717,
-      "train_tokens_per_second": 545471.684
+      "train_runtime": 2857.696,
+      "train_tokens_per_second": 147180.72
     },
     {
       "epoch": 1.3911726525313717,
-      "grad_norm": 0.455078125,
+      "grad_norm": 0.4296875,
       "learning_rate": 0.00022553684210526316,
-      "loss": 0.0213,
+      "loss": 0.0225,
       "num_input_tokens_seen": 421253120,
       "step": 6430,
-      "train_runtime": 774.9024,
-      "train_tokens_per_second": 543620.847
+      "train_runtime": 2861.5115,
+      "train_tokens_per_second": 147213.5
     },
     {
       "epoch": 1.393336218087408,
-      "grad_norm": 0.51171875,
+      "grad_norm": 0.357421875,
       "learning_rate": 0.00022490526315789474,
       "loss": 0.0205,
       "num_input_tokens_seen": 421908480,
       "step": 6440,
-      "train_runtime": 778.7188,
-      "train_tokens_per_second": 541798.28
+      "train_runtime": 2865.3128,
+      "train_tokens_per_second": 147246.916
     },
     {
       "epoch": 1.3954997836434444,
-      "grad_norm": 0.5,
+      "grad_norm": 0.62109375,
       "learning_rate": 0.00022427368421052627,
-      "loss": 0.0223,
+      "loss": 0.0231,
       "num_input_tokens_seen": 422563840,
       "step": 6450,
-      "train_runtime": 782.5419,
-      "train_tokens_per_second": 539988.798
+      "train_runtime": 2869.1293,
+      "train_tokens_per_second": 147279.467
     },
     {
       "epoch": 1.3976633491994808,
-      "grad_norm": 0.62109375,
+      "grad_norm": 0.44921875,
       "learning_rate": 0.00022364210526315786,
-      "loss": 0.0205,
+      "loss": 0.0212,
       "num_input_tokens_seen": 423219200,
       "step": 6460,
-      "train_runtime": 786.3586,
-      "train_tokens_per_second": 538201.273
+      "train_runtime": 2875.1773,
+      "train_tokens_per_second": 147197.602
     },
     {
       "epoch": 1.3998269147555171,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00022301052631578945,
-      "loss": 0.02,
+      "loss": 0.0221,
       "num_input_tokens_seen": 423874560,
       "step": 6470,
-      "train_runtime": 790.1794,
-      "train_tokens_per_second": 536428.278
+      "train_runtime": 2878.9924,
+      "train_tokens_per_second": 147230.179
     },
     {
       "epoch": 1.4019904803115535,
-      "grad_norm": 0.359375,
+      "grad_norm": 0.60546875,
       "learning_rate": 0.00022237894736842103,
-      "loss": 0.0193,
+      "loss": 0.0195,
       "num_input_tokens_seen": 424529920,
       "step": 6480,
-      "train_runtime": 794.0007,
-      "train_tokens_per_second": 534671.99
+      "train_runtime": 2882.8075,
+      "train_tokens_per_second": 147262.669
     },
     {
       "epoch": 1.4041540458675899,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.546875,
       "learning_rate": 0.00022174736842105262,
-      "loss": 0.0225,
+      "loss": 0.0224,
       "num_input_tokens_seen": 425185280,
       "step": 6490,
-      "train_runtime": 797.8185,
-      "train_tokens_per_second": 532934.816
+      "train_runtime": 2886.6208,
+      "train_tokens_per_second": 147295.164
     },
     {
       "epoch": 1.4063176114236262,
-      "grad_norm": 0.53125,
+      "grad_norm": 0.35546875,
       "learning_rate": 0.0002211157894736842,
-      "loss": 0.0226,
+      "loss": 0.0237,
       "num_input_tokens_seen": 425840640,
       "step": 6500,
-      "train_runtime": 801.6401,
-      "train_tokens_per_second": 531211.776
+      "train_runtime": 2890.4362,
+      "train_tokens_per_second": 147327.464
     },
     {
       "epoch": 1.4063176114236262,
-      "eval_loss": 0.014038060791790485,
-      "eval_runtime": 1.82,
-      "eval_samples_per_second": 17.582,
-      "eval_steps_per_second": 0.549,
+      "eval_loss": 0.013664661906659603,
+      "eval_runtime": 2.0687,
+      "eval_samples_per_second": 15.468,
+      "eval_steps_per_second": 0.483,
       "num_input_tokens_seen": 425840640,
       "step": 6500
     },
     {
       "epoch": 1.4063176114236262,
       "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 98.33382876029357,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_chrf": 98.07622351272572,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 425840640,
-      "perplexity": 1.014137057063326,
+      "perplexity": 1.0137584501069554,
       "step": 6500
     },
     {
       "epoch": 1.4084811769796626,
-      "grad_norm": 0.5625,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.0002204842105263158,
-      "loss": 0.0206,
+      "loss": 0.0211,
       "num_input_tokens_seen": 426496000,
       "step": 6510,
-      "train_runtime": 807.3224,
-      "train_tokens_per_second": 528284.6
+      "train_runtime": 2898.7837,
+      "train_tokens_per_second": 147129.293
     },
     {
       "epoch": 1.4106447425356987,
-      "grad_norm": 0.52734375,
+      "grad_norm": 0.46875,
       "learning_rate": 0.00021985263157894733,
-      "loss": 0.0209,
+      "loss": 0.0213,
       "num_input_tokens_seen": 427151360,
       "step": 6520,
-      "train_runtime": 811.1474,
-      "train_tokens_per_second": 526601.42
+      "train_runtime": 2902.6053,
+      "train_tokens_per_second": 147161.368
     },
     {
       "epoch": 1.4128083080917353,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.423828125,
       "learning_rate": 0.0002192210526315789,
-      "loss": 0.0204,
+      "loss": 0.0216,
       "num_input_tokens_seen": 427806720,
       "step": 6530,
-      "train_runtime": 814.968,
-      "train_tokens_per_second": 524936.804
+      "train_runtime": 2906.4268,
+      "train_tokens_per_second": 147193.358
     },
     {
       "epoch": 1.4149718736477714,
-      "grad_norm": 0.4921875,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.0002185894736842105,
-      "loss": 0.0182,
+      "loss": 0.0191,
       "num_input_tokens_seen": 428462080,
       "step": 6540,
-      "train_runtime": 818.7958,
-      "train_tokens_per_second": 523283.211
+      "train_runtime": 2910.2357,
+      "train_tokens_per_second": 147225.906
     },
     {
       "epoch": 1.4171354392038078,
-      "grad_norm": 0.380859375,
+      "grad_norm": 0.482421875,
       "learning_rate": 0.00021795789473684208,
-      "loss": 0.0181,
+      "loss": 0.0191,
       "num_input_tokens_seen": 429117440,
       "step": 6550,
-      "train_runtime": 822.6201,
-      "train_tokens_per_second": 521647.178
+      "train_runtime": 2914.0494,
+      "train_tokens_per_second": 147258.121
     },
     {
       "epoch": 1.4192990047598442,
-      "grad_norm": 0.55859375,
+      "grad_norm": 0.44140625,
       "learning_rate": 0.00021732631578947367,
-      "loss": 0.019,
+      "loss": 0.0198,
       "num_input_tokens_seen": 429772800,
       "step": 6560,
-      "train_runtime": 826.4515,
-      "train_tokens_per_second": 520021.829
+      "train_runtime": 2920.0844,
+      "train_tokens_per_second": 147178.211
     },
     {
       "epoch": 1.4214625703158805,
-      "grad_norm": 0.490234375,
+      "grad_norm": 0.408203125,
       "learning_rate": 0.00021669473684210526,
-      "loss": 0.021,
+      "loss": 0.0212,
       "num_input_tokens_seen": 430428160,
       "step": 6570,
-      "train_runtime": 830.2566,
-      "train_tokens_per_second": 518427.851
+      "train_runtime": 2923.8842,
+      "train_tokens_per_second": 147211.081
     },
     {
       "epoch": 1.4236261358719169,
-      "grad_norm": 0.56640625,
+      "grad_norm": 0.306640625,
       "learning_rate": 0.00021606315789473684,
-      "loss": 0.0201,
+      "loss": 0.0198,
       "num_input_tokens_seen": 431083520,
       "step": 6580,
-      "train_runtime": 834.0781,
-      "train_tokens_per_second": 516838.294
+      "train_runtime": 2927.6985,
+      "train_tokens_per_second": 147243.14
     },
     {
       "epoch": 1.4257897014279532,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.34765625,
       "learning_rate": 0.00021543157894736838,
-      "loss": 0.0207,
+      "loss": 0.0209,
       "num_input_tokens_seen": 431738880,
       "step": 6590,
-      "train_runtime": 837.9037,
-      "train_tokens_per_second": 515260.748
+      "train_runtime": 2931.5141,
+      "train_tokens_per_second": 147275.046
     },
     {
       "epoch": 1.4279532669839896,
-      "grad_norm": 0.5234375,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00021479999999999996,
-      "loss": 0.02,
+      "loss": 0.0207,
       "num_input_tokens_seen": 432394240,
       "step": 6600,
-      "train_runtime": 841.7307,
-      "train_tokens_per_second": 513696.636
+      "train_runtime": 2935.3272,
+      "train_tokens_per_second": 147307.0
     },
     {
       "epoch": 1.4279532669839896,
-      "eval_loss": 0.014414757490158081,
-      "eval_runtime": 2.1559,
-      "eval_samples_per_second": 14.843,
-      "eval_steps_per_second": 0.464,
+      "eval_loss": 0.013251814059913158,
+      "eval_runtime": 2.1468,
+      "eval_samples_per_second": 14.906,
+      "eval_steps_per_second": 0.466,
       "num_input_tokens_seen": 432394240,
       "step": 6600
     },
     {
       "epoch": 1.4279532669839896,
-      "eval_byte_accuracy": 0.9955489614243324,
-      "eval_chrf": 96.67764573411229,
-      "eval_sacrebleu": 95.86027221972736,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.9970326409495549,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 432394240,
-      "perplexity": 1.014519151106678,
+      "perplexity": 1.01334000849601,
       "step": 6600
     },
     {
       "epoch": 1.430116832540026,
-      "grad_norm": 0.3984375,
+      "grad_norm": 0.4921875,
       "learning_rate": 0.00021416842105263155,
-      "loss": 0.0199,
+      "loss": 0.0197,
       "num_input_tokens_seen": 433049600,
       "step": 6610,
-      "train_runtime": 847.7278,
-      "train_tokens_per_second": 510835.646
+      "train_runtime": 2943.524,
+      "train_tokens_per_second": 147119.439
     },
     {
       "epoch": 1.4322803980960623,
-      "grad_norm": 0.404296875,
+      "grad_norm": 0.400390625,
       "learning_rate": 0.00021353684210526314,
-      "loss": 0.0194,
+      "loss": 0.0206,
       "num_input_tokens_seen": 433704960,
       "step": 6620,
-      "train_runtime": 851.5621,
-      "train_tokens_per_second": 509305.165
+      "train_runtime": 2947.3456,
+      "train_tokens_per_second": 147151.037
     },
     {
       "epoch": 1.4344439636520987,
-      "grad_norm": 0.60546875,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.00021290526315789472,
-      "loss": 0.0258,
+      "loss": 0.0256,
       "num_input_tokens_seen": 434360320,
       "step": 6630,
-      "train_runtime": 855.3921,
-      "train_tokens_per_second": 507790.911
+      "train_runtime": 2951.1633,
+      "train_tokens_per_second": 147182.748
     },
     {
       "epoch": 1.436607529208135,
-      "grad_norm": 0.41796875,
+      "grad_norm": 0.419921875,
       "learning_rate": 0.0002122736842105263,
-      "loss": 0.0196,
+      "loss": 0.0204,
       "num_input_tokens_seen": 435015680,
       "step": 6640,
-      "train_runtime": 859.2162,
-      "train_tokens_per_second": 506293.62
+      "train_runtime": 2954.9846,
+      "train_tokens_per_second": 147214.194
     },
     {
       "epoch": 1.4387710947641714,
-      "grad_norm": 0.32421875,
+      "grad_norm": 0.4375,
       "learning_rate": 0.0002116421052631579,
-      "loss": 0.0196,
+      "loss": 0.0198,
       "num_input_tokens_seen": 435671040,
       "step": 6650,
-      "train_runtime": 863.0321,
-      "train_tokens_per_second": 504814.436
+      "train_runtime": 2958.7979,
+      "train_tokens_per_second": 147245.961
     },
     {
       "epoch": 1.4409346603202078,
-      "grad_norm": 0.427734375,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.00021101052631578945,
-      "loss": 0.0186,
+      "loss": 0.0188,
       "num_input_tokens_seen": 436326400,
       "step": 6660,
-      "train_runtime": 866.8539,
-      "train_tokens_per_second": 503344.785
+      "train_runtime": 2964.8365,
+      "train_tokens_per_second": 147167.102
     },
     {
       "epoch": 1.443098225876244,
-      "grad_norm": 0.4453125,
+      "grad_norm": 0.349609375,
       "learning_rate": 0.000210378947368421,
       "loss": 0.0205,
       "num_input_tokens_seen": 436981760,
       "step": 6670,
-      "train_runtime": 870.6753,
-      "train_tokens_per_second": 501888.287
+      "train_runtime": 2968.6512,
+      "train_tokens_per_second": 147198.754
     },
     {
       "epoch": 1.4452617914322805,
-      "grad_norm": 0.515625,
+      "grad_norm": 0.59375,
       "learning_rate": 0.0002097473684210526,
-      "loss": 0.0208,
+      "loss": 0.0214,
       "num_input_tokens_seen": 437637120,
       "step": 6680,
-      "train_runtime": 874.4982,
-      "train_tokens_per_second": 500443.7
+      "train_runtime": 2972.4583,
+      "train_tokens_per_second": 147230.699
     },
     {
       "epoch": 1.4474253569883166,
-      "grad_norm": 0.62109375,
+      "grad_norm": 0.4921875,
       "learning_rate": 0.00020911578947368419,
-      "loss": 0.0201,
+      "loss": 0.0209,
       "num_input_tokens_seen": 438292480,
       "step": 6690,
-      "train_runtime": 878.3125,
-      "train_tokens_per_second": 499016.574
+      "train_runtime": 2976.2746,
+      "train_tokens_per_second": 147262.113
     },
     {
       "epoch": 1.4495889225443532,
-      "grad_norm": 0.73828125,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.00020848421052631577,
-      "loss": 0.0211,
+      "loss": 0.0219,
       "num_input_tokens_seen": 438947840,
       "step": 6700,
-      "train_runtime": 882.1283,
-      "train_tokens_per_second": 497600.901
+      "train_runtime": 2980.0894,
+      "train_tokens_per_second": 147293.516
     },
     {
       "epoch": 1.4495889225443532,
-      "eval_loss": 0.015520436689257622,
-      "eval_runtime": 1.7455,
-      "eval_samples_per_second": 18.333,
-      "eval_steps_per_second": 0.573,
+      "eval_loss": 0.012278531678020954,
+      "eval_runtime": 2.2,
+      "eval_samples_per_second": 14.546,
+      "eval_steps_per_second": 0.455,
       "num_input_tokens_seen": 438947840,
       "step": 6700
     },
     {
       "epoch": 1.4495889225443532,
-      "eval_byte_accuracy": 0.9955489614243324,
-      "eval_chrf": 96.30359773954326,
-      "eval_sacrebleu": 94.84810346446147,
-      "eval_word_accuracy": 0.9876373626373627,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 97.32826860603707,
+      "eval_sacrebleu": 95.31161095839255,
+      "eval_word_accuracy": 0.989010989010989,
       "num_input_tokens_seen": 438947840,
-      "perplexity": 1.0156415041960352,
+      "perplexity": 1.0123542223208506,
       "step": 6700
     },
     {
       "epoch": 1.4517524881003894,
-      "grad_norm": 0.41796875,
+      "grad_norm": 0.3515625,
       "learning_rate": 0.00020785263157894736,
-      "loss": 0.0203,
+      "loss": 0.0211,
       "num_input_tokens_seen": 439603200,
       "step": 6710,
-      "train_runtime": 887.7094,
-      "train_tokens_per_second": 495210.697
+      "train_runtime": 2988.5337,
+      "train_tokens_per_second": 147096.618
     },
     {
       "epoch": 1.4539160536564257,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.404296875,
       "learning_rate": 0.00020722105263157895,
-      "loss": 0.019,
+      "loss": 0.0186,
       "num_input_tokens_seen": 440258560,
       "step": 6720,
-      "train_runtime": 891.5276,
-      "train_tokens_per_second": 493824.937
+      "train_runtime": 2992.3565,
+      "train_tokens_per_second": 147127.71
     },
     {
       "epoch": 1.456079619212462,
-      "grad_norm": 0.33203125,
+      "grad_norm": 0.36328125,
       "learning_rate": 0.0002065894736842105,
-      "loss": 0.0206,
+      "loss": 0.0216,
       "num_input_tokens_seen": 440913920,
       "step": 6730,
-      "train_runtime": 895.3511,
-      "train_tokens_per_second": 492448.083
+      "train_runtime": 2996.1679,
+      "train_tokens_per_second": 147159.282
     },
     {
       "epoch": 1.4582431847684985,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.462890625,
       "learning_rate": 0.00020595789473684206,
-      "loss": 0.02,
+      "loss": 0.0203,
       "num_input_tokens_seen": 441569280,
       "step": 6740,
-      "train_runtime": 899.1684,
-      "train_tokens_per_second": 491086.316
+      "train_runtime": 2999.9752,
+      "train_tokens_per_second": 147190.979
     },
     {
       "epoch": 1.4604067503245348,
-      "grad_norm": 0.5546875,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00020532631578947365,
-      "loss": 0.022,
+      "loss": 0.0225,
       "num_input_tokens_seen": 442224640,
       "step": 6750,
-      "train_runtime": 902.9825,
-      "train_tokens_per_second": 489737.8
+      "train_runtime": 3003.779,
+      "train_tokens_per_second": 147222.762
     },
     {
       "epoch": 1.4625703158805712,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.57421875,
       "learning_rate": 0.00020469473684210524,
-      "loss": 0.019,
+      "loss": 0.0198,
       "num_input_tokens_seen": 442880000,
       "step": 6760,
-      "train_runtime": 906.7891,
-      "train_tokens_per_second": 488404.645
+      "train_runtime": 3009.7963,
+      "train_tokens_per_second": 147146.172
     },
     {
       "epoch": 1.4647338814366075,
-      "grad_norm": 0.306640625,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.00020406315789473682,
-      "loss": 0.0206,
+      "loss": 0.0219,
       "num_input_tokens_seen": 443535360,
       "step": 6770,
-      "train_runtime": 910.6215,
-      "train_tokens_per_second": 487068.867
+      "train_runtime": 3013.6084,
+      "train_tokens_per_second": 147177.502
     },
     {
       "epoch": 1.466897446992644,
-      "grad_norm": 0.349609375,
+      "grad_norm": 0.330078125,
       "learning_rate": 0.0002034315789473684,
-      "loss": 0.0196,
+      "loss": 0.0202,
       "num_input_tokens_seen": 444190720,
       "step": 6780,
-      "train_runtime": 914.4324,
-      "train_tokens_per_second": 485755.666
+      "train_runtime": 3017.3963,
+      "train_tokens_per_second": 147209.935
     },
     {
       "epoch": 1.4690610125486803,
-      "grad_norm": 0.310546875,
+      "grad_norm": 0.359375,
       "learning_rate": 0.0002028,
-      "loss": 0.0202,
+      "loss": 0.0213,
       "num_input_tokens_seen": 444846080,
       "step": 6790,
-      "train_runtime": 918.2489,
-      "train_tokens_per_second": 484450.42
+      "train_runtime": 3021.2137,
+      "train_tokens_per_second": 147240.854
     },
     {
       "epoch": 1.4712245781047166,
-      "grad_norm": 0.39453125,
+      "grad_norm": 0.380859375,
       "learning_rate": 0.00020216842105263156,
-      "loss": 0.0184,
+      "loss": 0.0192,
       "num_input_tokens_seen": 445497344,
       "step": 6800,
-      "train_runtime": 922.0496,
-      "train_tokens_per_second": 483159.857
+      "train_runtime": 3025.0076,
+      "train_tokens_per_second": 147271.478
     },
     {
       "epoch": 1.4712245781047166,
-      "eval_loss": 0.015351799316704273,
-      "eval_runtime": 1.7437,
-      "eval_samples_per_second": 18.352,
-      "eval_steps_per_second": 0.573,
+      "eval_loss": 0.011801213026046753,
+      "eval_runtime": 2.0912,
+      "eval_samples_per_second": 15.302,
+      "eval_steps_per_second": 0.478,
       "num_input_tokens_seen": 445497344,
       "step": 6800
     },
     {
       "epoch": 1.4712245781047166,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.3917095936316,
-      "eval_sacrebleu": 96.37930136022618,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.91173752160901,
+      "eval_sacrebleu": 96.29188240283864,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 445497344,
-      "perplexity": 1.0154702435221923,
+      "perplexity": 1.0118711220736836,
       "step": 6800
     },
     {
       "epoch": 1.473388143660753,
-      "grad_norm": 0.4375,
+      "grad_norm": 0.40625,
       "learning_rate": 0.00020153684210526314,
-      "loss": 0.0192,
+      "loss": 0.0203,
       "num_input_tokens_seen": 446152704,
       "step": 6810,
-      "train_runtime": 927.6261,
-      "train_tokens_per_second": 480961.793
+      "train_runtime": 3033.2934,
+      "train_tokens_per_second": 147085.246
     },
     {
       "epoch": 1.4755517092167891,
-      "grad_norm": 0.56640625,
+      "grad_norm": 0.44140625,
       "learning_rate": 0.0002009052631578947,
-      "loss": 0.0194,
+      "loss": 0.0204,
       "num_input_tokens_seen": 446808064,
       "step": 6820,
-      "train_runtime": 931.455,
-      "train_tokens_per_second": 479688.322
+      "train_runtime": 3037.1092,
+      "train_tokens_per_second": 147116.232
     },
     {
       "epoch": 1.4777152747728257,
-      "grad_norm": 0.400390625,
+      "grad_norm": 0.484375,
       "learning_rate": 0.0002002736842105263,
-      "loss": 0.0205,
+      "loss": 0.0207,
       "num_input_tokens_seen": 447463424,
       "step": 6830,
-      "train_runtime": 935.2803,
-      "train_tokens_per_second": 478427.09
+      "train_runtime": 3040.9219,
+      "train_tokens_per_second": 147147.292
     },
     {
       "epoch": 1.4798788403288619,
-      "grad_norm": 0.4609375,
+      "grad_norm": 0.384765625,
       "learning_rate": 0.00019964210526315787,
-      "loss": 0.0195,
+      "loss": 0.0219,
       "num_input_tokens_seen": 448118784,
       "step": 6840,
-      "train_runtime": 939.0955,
-      "train_tokens_per_second": 477181.257
+      "train_runtime": 3044.7287,
+      "train_tokens_per_second": 147178.561
     },
     {
       "epoch": 1.4820424058848984,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.5234375,
       "learning_rate": 0.00019901052631578946,
-      "loss": 0.0206,
+      "loss": 0.0223,
       "num_input_tokens_seen": 448774144,
       "step": 6850,
-      "train_runtime": 942.9144,
-      "train_tokens_per_second": 475943.686
+      "train_runtime": 3048.5411,
+      "train_tokens_per_second": 147209.476
     },
     {
       "epoch": 1.4842059714409346,
-      "grad_norm": 0.46484375,
+      "grad_norm": 0.30859375,
       "learning_rate": 0.00019837894736842105,
-      "loss": 0.0212,
+      "loss": 0.0214,
       "num_input_tokens_seen": 449429504,
       "step": 6860,
-      "train_runtime": 946.7363,
-      "train_tokens_per_second": 474714.567
+      "train_runtime": 3054.5882,
+      "train_tokens_per_second": 147132.6
     },
     {
       "epoch": 1.486369536996971,
-      "grad_norm": 0.408203125,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.0001977473684210526,
-      "loss": 0.0506,
+      "loss": 0.0488,
       "num_input_tokens_seen": 450084864,
       "step": 6870,
-      "train_runtime": 950.5571,
-      "train_tokens_per_second": 473495.858
+      "train_runtime": 3058.3992,
+      "train_tokens_per_second": 147163.546
     },
     {
       "epoch": 1.4885331025530073,
-      "grad_norm": 0.365234375,
+      "grad_norm": 0.322265625,
       "learning_rate": 0.0001971157894736842,
-      "loss": 0.0199,
+      "loss": 0.02,
       "num_input_tokens_seen": 450740224,
       "step": 6880,
-      "train_runtime": 954.368,
-      "train_tokens_per_second": 472291.835
+      "train_runtime": 3062.2077,
+      "train_tokens_per_second": 147194.528
     },
     {
       "epoch": 1.4906966681090437,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.55078125,
       "learning_rate": 0.00019648421052631578,
-      "loss": 0.0207,
+      "loss": 0.0216,
       "num_input_tokens_seen": 451395584,
       "step": 6890,
-      "train_runtime": 958.1838,
-      "train_tokens_per_second": 471094.968
+      "train_runtime": 3066.0215,
+      "train_tokens_per_second": 147225.186
     },
     {
       "epoch": 1.49286023366508,
-      "grad_norm": 0.6015625,
+      "grad_norm": 0.400390625,
       "learning_rate": 0.00019585263157894734,
-      "loss": 0.0202,
+      "loss": 0.0206,
       "num_input_tokens_seen": 452050944,
       "step": 6900,
-      "train_runtime": 962.0035,
-      "train_tokens_per_second": 469905.704
+      "train_runtime": 3069.8357,
+      "train_tokens_per_second": 147255.746
     },
     {
       "epoch": 1.49286023366508,
-      "eval_loss": 0.014207720756530762,
-      "eval_runtime": 1.7403,
-      "eval_samples_per_second": 18.388,
-      "eval_steps_per_second": 0.575,
+      "eval_loss": 0.010991868562996387,
+      "eval_runtime": 2.029,
+      "eval_samples_per_second": 15.771,
+      "eval_steps_per_second": 0.493,
       "num_input_tokens_seen": 452050944,
       "step": 6900
     },
     {
       "epoch": 1.49286023366508,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.24944108474526,
-      "eval_sacrebleu": 96.68143920500805,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 97.10549441712193,
+      "eval_sacrebleu": 96.36994641700896,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 452050944,
-      "perplexity": 1.0143091301172105,
+      "perplexity": 1.0110525011015739,
       "step": 6900
     },
     {
       "epoch": 1.4950237992211164,
-      "grad_norm": 0.421875,
+      "grad_norm": 0.38671875,
       "learning_rate": 0.00019522105263157892,
-      "loss": 0.0201,
+      "loss": 0.0202,
       "num_input_tokens_seen": 452706304,
       "step": 6910,
-      "train_runtime": 967.588,
-      "train_tokens_per_second": 467870.924
+      "train_runtime": 3078.111,
+      "train_tokens_per_second": 147072.767
     },
     {
       "epoch": 1.4971873647771528,
-      "grad_norm": 0.43359375,
+      "grad_norm": 0.4921875,
       "learning_rate": 0.0001945894736842105,
-      "loss": 0.0226,
+      "loss": 0.0227,
       "num_input_tokens_seen": 453361664,
       "step": 6920,
-      "train_runtime": 971.3999,
-      "train_tokens_per_second": 466709.598
+      "train_runtime": 3081.8971,
+      "train_tokens_per_second": 147104.737
     },
     {
       "epoch": 1.4993509303331891,
-      "grad_norm": 0.5546875,
+      "grad_norm": 0.51171875,
       "learning_rate": 0.0001939578947368421,
-      "loss": 0.0208,
+      "loss": 0.0204,
       "num_input_tokens_seen": 454017024,
       "step": 6930,
-      "train_runtime": 975.2197,
-      "train_tokens_per_second": 465553.591
+      "train_runtime": 3085.7057,
+      "train_tokens_per_second": 147135.556
     },
     {
       "epoch": 1.5015144958892255,
-      "grad_norm": 0.29296875,
+      "grad_norm": 0.392578125,
       "learning_rate": 0.00019332631578947366,
-      "loss": 0.0182,
+      "loss": 0.0195,
       "num_input_tokens_seen": 454672384,
       "step": 6940,
-      "train_runtime": 979.0437,
-      "train_tokens_per_second": 464404.602
+      "train_runtime": 3089.5175,
+      "train_tokens_per_second": 147166.147
     },
     {
       "epoch": 1.5036780614452618,
-      "grad_norm": 0.455078125,
+      "grad_norm": 0.3671875,
       "learning_rate": 0.00019269473684210524,
-      "loss": 0.0198,
+      "loss": 0.0202,
       "num_input_tokens_seen": 455327744,
       "step": 6950,
-      "train_runtime": 982.8674,
-      "train_tokens_per_second": 463264.695
+      "train_runtime": 3093.3221,
+      "train_tokens_per_second": 147197.004
     },
     {
       "epoch": 1.5058416270012982,
-      "grad_norm": 0.392578125,
+      "grad_norm": 0.4375,
       "learning_rate": 0.00019206315789473683,
-      "loss": 0.0181,
+      "loss": 0.0201,
       "num_input_tokens_seen": 455983104,
       "step": 6960,
-      "train_runtime": 986.6982,
-      "train_tokens_per_second": 462130.276
+      "train_runtime": 3099.3571,
+      "train_tokens_per_second": 147121.833
     },
     {
       "epoch": 1.5080051925573343,
-      "grad_norm": 0.431640625,
+      "grad_norm": 0.4453125,
       "learning_rate": 0.0001914315789473684,
-      "loss": 0.0183,
+      "loss": 0.0191,
       "num_input_tokens_seen": 456638464,
       "step": 6970,
-      "train_runtime": 990.5199,
-      "train_tokens_per_second": 461008.87
+      "train_runtime": 3103.1587,
+      "train_tokens_per_second": 147152.793
     },
     {
       "epoch": 1.510168758113371,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00019079999999999998,
-      "loss": 0.0184,
+      "loss": 0.019,
       "num_input_tokens_seen": 457293824,
       "step": 6980,
-      "train_runtime": 994.3509,
-      "train_tokens_per_second": 459891.818
+      "train_runtime": 3106.9748,
+      "train_tokens_per_second": 147182.983
     },
     {
       "epoch": 1.512332323669407,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.341796875,
       "learning_rate": 0.00019016842105263156,
-      "loss": 0.0196,
+      "loss": 0.0198,
       "num_input_tokens_seen": 457949184,
       "step": 6990,
-      "train_runtime": 998.1779,
-      "train_tokens_per_second": 458785.114
+      "train_runtime": 3110.7791,
+      "train_tokens_per_second": 147213.661
     },
     {
       "epoch": 1.5144958892254436,
-      "grad_norm": 0.51953125,
+      "grad_norm": 0.376953125,
       "learning_rate": 0.00018953684210526315,
-      "loss": 0.0213,
+      "loss": 0.0216,
       "num_input_tokens_seen": 458604544,
       "step": 7000,
-      "train_runtime": 1002.0001,
-      "train_tokens_per_second": 457689.118
+      "train_runtime": 3114.5883,
+      "train_tokens_per_second": 147244.034
     },
     {
       "epoch": 1.5144958892254436,
-      "eval_loss": 0.014768999069929123,
-      "eval_runtime": 1.7379,
-      "eval_samples_per_second": 18.413,
-      "eval_steps_per_second": 0.575,
+      "eval_loss": 0.010852305218577385,
+      "eval_runtime": 2.1629,
+      "eval_samples_per_second": 14.795,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 458604544,
       "step": 7000
     },
     {
       "epoch": 1.5144958892254436,
-      "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 96.77834078968087,
-      "eval_sacrebleu": 95.79032669908624,
-      "eval_word_accuracy": 0.989010989010989,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 98.28090039473334,
+      "eval_sacrebleu": 97.54443059083923,
+      "eval_word_accuracy": 0.9931318681318682,
       "num_input_tokens_seen": 458604544,
-      "perplexity": 1.0148785996355225,
+      "perplexity": 1.0109114050792825,
       "step": 7000
     },
     {
       "epoch": 1.5166594547814798,
-      "grad_norm": 0.53125,
+      "grad_norm": 0.4296875,
       "learning_rate": 0.0001889052631578947,
       "loss": 0.02,
       "num_input_tokens_seen": 459259904,
       "step": 7010,
-      "train_runtime": 1007.6295,
-      "train_tokens_per_second": 455782.526
+      "train_runtime": 3122.8668,
+      "train_tokens_per_second": 147063.557
     },
     {
       "epoch": 1.5188230203375164,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.0001882736842105263,
-      "loss": 0.0199,
+      "loss": 0.0196,
       "num_input_tokens_seen": 459915264,
       "step": 7020,
-      "train_runtime": 1011.4394,
-      "train_tokens_per_second": 454713.63
+      "train_runtime": 3126.6547,
+      "train_tokens_per_second": 147094.997
     },
     {
       "epoch": 1.5209865858935525,
-      "grad_norm": 0.6015625,
+      "grad_norm": 0.33984375,
       "learning_rate": 0.00018764210526315788,
-      "loss": 0.0187,
+      "loss": 0.0201,
       "num_input_tokens_seen": 460570624,
       "step": 7030,
-      "train_runtime": 1015.2675,
-      "train_tokens_per_second": 453644.627
+      "train_runtime": 3130.458,
+      "train_tokens_per_second": 147125.634
     },
     {
       "epoch": 1.5231501514495889,
-      "grad_norm": 0.3515625,
+      "grad_norm": 0.435546875,
       "learning_rate": 0.00018701052631578947,
-      "loss": 0.0192,
+      "loss": 0.0195,
       "num_input_tokens_seen": 461225984,
       "step": 7040,
-      "train_runtime": 1019.094,
-      "train_tokens_per_second": 452584.352
+      "train_runtime": 3134.2733,
+      "train_tokens_per_second": 147155.635
     },
     {
       "epoch": 1.5253137170056252,
-      "grad_norm": 0.5,
+      "grad_norm": 0.400390625,
       "learning_rate": 0.00018637894736842103,
-      "loss": 0.0198,
+      "loss": 0.0209,
       "num_input_tokens_seen": 461881344,
       "step": 7050,
-      "train_runtime": 1022.903,
-      "train_tokens_per_second": 451539.732
+      "train_runtime": 3138.0781,
+      "train_tokens_per_second": 147186.058
     },
     {
       "epoch": 1.5274772825616616,
-      "grad_norm": 0.5,
+      "grad_norm": 0.451171875,
       "learning_rate": 0.0001857473684210526,
-      "loss": 0.0192,
+      "loss": 0.0205,
       "num_input_tokens_seen": 462536704,
       "step": 7060,
-      "train_runtime": 1026.7344,
-      "train_tokens_per_second": 450493.05
+      "train_runtime": 3144.1331,
+      "train_tokens_per_second": 147111.046
     },
     {
       "epoch": 1.529640848117698,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.486328125,
       "learning_rate": 0.0001851157894736842,
-      "loss": 0.0196,
+      "loss": 0.0209,
       "num_input_tokens_seen": 463192064,
       "step": 7070,
-      "train_runtime": 1030.5622,
-      "train_tokens_per_second": 449455.697
+      "train_runtime": 3147.9443,
+      "train_tokens_per_second": 147141.124
     },
     {
       "epoch": 1.5318044136737343,
-      "grad_norm": 0.333984375,
+      "grad_norm": 0.47265625,
       "learning_rate": 0.00018448421052631579,
-      "loss": 0.0197,
+      "loss": 0.0198,
       "num_input_tokens_seen": 463847424,
       "step": 7080,
-      "train_runtime": 1034.3948,
-      "train_tokens_per_second": 448423.959
+      "train_runtime": 3151.7527,
+      "train_tokens_per_second": 147171.262
     },
     {
       "epoch": 1.5339679792297707,
-      "grad_norm": 0.404296875,
+      "grad_norm": 0.37890625,
       "learning_rate": 0.00018385263157894735,
-      "loss": 0.0191,
+      "loss": 0.0201,
       "num_input_tokens_seen": 464502784,
       "step": 7090,
-      "train_runtime": 1038.2274,
-      "train_tokens_per_second": 447399.872
+      "train_runtime": 3155.5665,
+      "train_tokens_per_second": 147201.077
     },
     {
       "epoch": 1.536131544785807,
-      "grad_norm": 0.53515625,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.00018322105263157893,
-      "loss": 0.0177,
+      "loss": 0.019,
       "num_input_tokens_seen": 465158144,
       "step": 7100,
-      "train_runtime": 1042.0582,
-      "train_tokens_per_second": 446384.022
+      "train_runtime": 3159.3692,
+      "train_tokens_per_second": 147231.333
     },
     {
       "epoch": 1.536131544785807,
-      "eval_loss": 0.01327989250421524,
-      "eval_runtime": 1.7389,
-      "eval_samples_per_second": 18.403,
-      "eval_steps_per_second": 0.575,
+      "eval_loss": 0.011366115882992744,
+      "eval_runtime": 2.0634,
+      "eval_samples_per_second": 15.508,
+      "eval_steps_per_second": 0.485,
       "num_input_tokens_seen": 465158144,
       "step": 7100
     },
     {
       "epoch": 1.536131544785807,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.50210841922704,
-      "eval_sacrebleu": 96.68143920500805,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 465158144,
-      "perplexity": 1.0133684619064596,
+      "perplexity": 1.0114309556038612,
       "step": 7100
     },
     {
       "epoch": 1.5382951103418434,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.4609375,
       "learning_rate": 0.00018258947368421052,
-      "loss": 0.0189,
+      "loss": 0.0198,
       "num_input_tokens_seen": 465813504,
       "step": 7110,
-      "train_runtime": 1047.6475,
-      "train_tokens_per_second": 444628.075
+      "train_runtime": 3167.5112,
+      "train_tokens_per_second": 147059.78
     },
     {
       "epoch": 1.5404586758978795,
-      "grad_norm": 0.318359375,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.0001819578947368421,
-      "loss": 0.0203,
+      "loss": 0.0212,
       "num_input_tokens_seen": 466468864,
       "step": 7120,
-      "train_runtime": 1051.4769,
-      "train_tokens_per_second": 443632.077
+      "train_runtime": 3171.3238,
+      "train_tokens_per_second": 147089.636
     },
     {
       "epoch": 1.5426222414539161,
-      "grad_norm": 0.3125,
+      "grad_norm": 0.380859375,
       "learning_rate": 0.00018132631578947366,
-      "loss": 0.0204,
+      "loss": 0.0207,
       "num_input_tokens_seen": 467124224,
       "step": 7130,
-      "train_runtime": 1055.3019,
-      "train_tokens_per_second": 442645.107
+      "train_runtime": 3175.134,
+      "train_tokens_per_second": 147119.53
     },
     {
       "epoch": 1.5447858070099523,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.349609375,
       "learning_rate": 0.00018069473684210525,
-      "loss": 0.0195,
+      "loss": 0.0202,
       "num_input_tokens_seen": 467779584,
       "step": 7140,
-      "train_runtime": 1059.1102,
-      "train_tokens_per_second": 441672.241
+      "train_runtime": 3178.9275,
+      "train_tokens_per_second": 147150.128
     },
     {
       "epoch": 1.5469493725659889,
-      "grad_norm": 0.294921875,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00018006315789473684,
-      "loss": 0.0185,
+      "loss": 0.0197,
       "num_input_tokens_seen": 468434944,
       "step": 7150,
-      "train_runtime": 1062.9377,
-      "train_tokens_per_second": 440698.395
+      "train_runtime": 3182.7383,
+      "train_tokens_per_second": 147179.848
     },
     {
       "epoch": 1.549112938122025,
-      "grad_norm": 0.5078125,
+      "grad_norm": 0.384765625,
       "learning_rate": 0.0001794315789473684,
-      "loss": 0.0186,
+      "loss": 0.0192,
       "num_input_tokens_seen": 469090304,
       "step": 7160,
-      "train_runtime": 1066.764,
-      "train_tokens_per_second": 439732.024
+      "train_runtime": 3188.9696,
+      "train_tokens_per_second": 147097.768
     },
     {
       "epoch": 1.5512765036780616,
-      "grad_norm": 0.3671875,
+      "grad_norm": 0.4140625,
       "learning_rate": 0.00017879999999999998,
-      "loss": 0.0218,
+      "loss": 0.0219,
       "num_input_tokens_seen": 469745664,
       "step": 7170,
-      "train_runtime": 1070.5891,
-      "train_tokens_per_second": 438773.072
+      "train_runtime": 3192.7902,
+      "train_tokens_per_second": 147127.004
     },
     {
       "epoch": 1.5534400692340977,
-      "grad_norm": 0.4296875,
+      "grad_norm": 0.482421875,
       "learning_rate": 0.00017816842105263157,
       "loss": 0.0185,
       "num_input_tokens_seen": 470401024,
       "step": 7180,
-      "train_runtime": 1074.4161,
-      "train_tokens_per_second": 437820.156
+      "train_runtime": 3196.6084,
+      "train_tokens_per_second": 147156.289
     },
     {
       "epoch": 1.5556036347901343,
-      "grad_norm": 0.5234375,
+      "grad_norm": 0.50390625,
       "learning_rate": 0.00017753684210526316,
-      "loss": 0.019,
+      "loss": 0.0193,
       "num_input_tokens_seen": 471056384,
       "step": 7190,
-      "train_runtime": 1078.245,
-      "train_tokens_per_second": 436873.226
+      "train_runtime": 3200.4207,
+      "train_tokens_per_second": 147185.77
     },
     {
       "epoch": 1.5577672003461704,
-      "grad_norm": 0.423828125,
+      "grad_norm": 0.443359375,
       "learning_rate": 0.00017690526315789471,
-      "loss": 0.0193,
+      "loss": 0.0213,
       "num_input_tokens_seen": 471711744,
       "step": 7200,
-      "train_runtime": 1082.0752,
-      "train_tokens_per_second": 435932.508
+      "train_runtime": 3204.2436,
+      "train_tokens_per_second": 147214.696
     },
     {
       "epoch": 1.5577672003461704,
-      "eval_loss": 0.01322339940816164,
-      "eval_runtime": 1.724,
-      "eval_samples_per_second": 18.562,
-      "eval_steps_per_second": 0.58,
+      "eval_loss": 0.012076064012944698,
+      "eval_runtime": 2.0546,
+      "eval_samples_per_second": 15.575,
+      "eval_steps_per_second": 0.487,
       "num_input_tokens_seen": 471711744,
       "step": 7200
     },
     {
       "epoch": 1.5577672003461704,
       "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.51595116562792,
-      "eval_sacrebleu": 96.68143920500805,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.89186504468623,
+      "eval_sacrebleu": 96.96578008549908,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 471711744,
-      "perplexity": 1.0133112152016404,
+      "perplexity": 1.0121492740736238,
       "step": 7200
     },
     {
       "epoch": 1.5599307659022068,
-      "grad_norm": 0.78125,
+      "grad_norm": 0.4375,
       "learning_rate": 0.0001762736842105263,
-      "loss": 0.0231,
+      "loss": 0.0238,
       "num_input_tokens_seen": 472367104,
       "step": 7210,
-      "train_runtime": 1087.647,
-      "train_tokens_per_second": 434301.852
+      "train_runtime": 3212.3634,
+      "train_tokens_per_second": 147046.598
     },
     {
       "epoch": 1.5620943314582432,
-      "grad_norm": 0.46875,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.0001756421052631579,
-      "loss": 0.0183,
+      "loss": 0.019,
       "num_input_tokens_seen": 473022464,
       "step": 7220,
-      "train_runtime": 1091.4784,
-      "train_tokens_per_second": 433377.759
+      "train_runtime": 3216.1804,
+      "train_tokens_per_second": 147075.85
     },
     {
       "epoch": 1.5642578970142795,
-      "grad_norm": 0.298828125,
+      "grad_norm": 0.390625,
       "learning_rate": 0.00017501052631578945,
-      "loss": 0.0162,
+      "loss": 0.0176,
       "num_input_tokens_seen": 473673728,
       "step": 7230,
-      "train_runtime": 1095.2826,
-      "train_tokens_per_second": 432467.15
+      "train_runtime": 3219.9613,
+      "train_tokens_per_second": 147105.412
     },
     {
       "epoch": 1.566421462570316,
-      "grad_norm": 0.455078125,
+      "grad_norm": 0.33203125,
       "learning_rate": 0.00017437894736842103,
-      "loss": 0.0193,
+      "loss": 0.0204,
       "num_input_tokens_seen": 474329088,
       "step": 7240,
-      "train_runtime": 1099.1053,
-      "train_tokens_per_second": 431559.287
+      "train_runtime": 3223.7734,
+      "train_tokens_per_second": 147134.751
     },
     {
       "epoch": 1.5685850281263523,
-      "grad_norm": 0.400390625,
+      "grad_norm": 0.28515625,
       "learning_rate": 0.00017374736842105262,
-      "loss": 0.0168,
+      "loss": 0.0174,
       "num_input_tokens_seen": 474984448,
       "step": 7250,
-      "train_runtime": 1102.9285,
-      "train_tokens_per_second": 430657.516
+      "train_runtime": 3227.5767,
+      "train_tokens_per_second": 147164.42
     },
     {
       "epoch": 1.5707485936823886,
-      "grad_norm": 0.34765625,
+      "grad_norm": 0.3828125,
       "learning_rate": 0.0001731157894736842,
-      "loss": 0.0191,
+      "loss": 0.0202,
       "num_input_tokens_seen": 475639808,
       "step": 7260,
-      "train_runtime": 1106.7545,
-      "train_tokens_per_second": 429760.903
+      "train_runtime": 3233.6382,
+      "train_tokens_per_second": 147091.224
     },
     {
       "epoch": 1.572912159238425,
-      "grad_norm": 0.5390625,
+      "grad_norm": 0.43359375,
       "learning_rate": 0.0001724842105263158,
-      "loss": 0.0215,
+      "loss": 0.022,
       "num_input_tokens_seen": 476295168,
       "step": 7270,
-      "train_runtime": 1110.579,
-      "train_tokens_per_second": 428871.048
+      "train_runtime": 3237.4441,
+      "train_tokens_per_second": 147120.739
     },
     {
       "epoch": 1.5750757247944613,
-      "grad_norm": 0.50390625,
+      "grad_norm": 0.4296875,
       "learning_rate": 0.00017185263157894735,
-      "loss": 0.021,
+      "loss": 0.0214,
       "num_input_tokens_seen": 476950528,
       "step": 7280,
-      "train_runtime": 1114.4053,
-      "train_tokens_per_second": 427986.6
+      "train_runtime": 3241.2576,
+      "train_tokens_per_second": 147149.836
     },
     {
       "epoch": 1.5772392903504975,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.380859375,
       "learning_rate": 0.00017122105263157894,
-      "loss": 0.0199,
+      "loss": 0.0207,
       "num_input_tokens_seen": 477605888,
       "step": 7290,
-      "train_runtime": 1118.2285,
-      "train_tokens_per_second": 427109.371
+      "train_runtime": 3245.0654,
+      "train_tokens_per_second": 147179.126
     },
     {
       "epoch": 1.579402855906534,
-      "grad_norm": 0.4765625,
+      "grad_norm": 0.40625,
       "learning_rate": 0.0001705894736842105,
-      "loss": 0.019,
+      "loss": 0.0201,
       "num_input_tokens_seen": 478261248,
       "step": 7300,
-      "train_runtime": 1122.0519,
-      "train_tokens_per_second": 426238.09
+      "train_runtime": 3248.8769,
+      "train_tokens_per_second": 147208.177
     },
     {
       "epoch": 1.579402855906534,
-      "eval_loss": 0.013833579607307911,
-      "eval_runtime": 1.7302,
-      "eval_samples_per_second": 18.495,
-      "eval_steps_per_second": 0.578,
+      "eval_loss": 0.011056940071284771,
+      "eval_runtime": 2.0875,
+      "eval_samples_per_second": 15.33,
+      "eval_steps_per_second": 0.479,
       "num_input_tokens_seen": 478261248,
       "step": 7300
     },
     {
       "epoch": 1.579402855906534,
-      "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.06431064307826,
-      "eval_sacrebleu": 95.76190341678922,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 478261248,
-      "perplexity": 1.0139297063170556,
+      "perplexity": 1.011118293953376,
       "step": 7300
     },
     {
       "epoch": 1.5815664214625702,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00016995789473684208,
-      "loss": 0.0191,
+      "loss": 0.0197,
       "num_input_tokens_seen": 478916608,
       "step": 7310,
-      "train_runtime": 1127.6289,
-      "train_tokens_per_second": 424711.19
+      "train_runtime": 3257.0094,
+      "train_tokens_per_second": 147041.825
     },
     {
       "epoch": 1.5837299870186068,
       "grad_norm": 0.455078125,
       "learning_rate": 0.00016932631578947367,
-      "loss": 0.0191,
+      "loss": 0.0207,
       "num_input_tokens_seen": 479567872,
       "step": 7320,
-      "train_runtime": 1131.433,
-      "train_tokens_per_second": 423858.826
+      "train_runtime": 3260.806,
+      "train_tokens_per_second": 147070.347
     },
     {
       "epoch": 1.585893552574643,
-      "grad_norm": 0.41796875,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00016869473684210526,
-      "loss": 0.0187,
+      "loss": 0.0188,
       "num_input_tokens_seen": 480223232,
       "step": 7330,
-      "train_runtime": 1135.2422,
-      "train_tokens_per_second": 423013.894
+      "train_runtime": 3264.6033,
+      "train_tokens_per_second": 147100.026
     },
     {
       "epoch": 1.5880571181306795,
-      "grad_norm": 0.462890625,
+      "grad_norm": 0.53125,
       "learning_rate": 0.00016806315789473684,
-      "loss": 0.0225,
+      "loss": 0.0233,
       "num_input_tokens_seen": 480878592,
       "step": 7340,
-      "train_runtime": 1139.066,
-      "train_tokens_per_second": 422169.227
+      "train_runtime": 3268.4136,
+      "train_tokens_per_second": 147129.053
     },
     {
       "epoch": 1.5902206836867157,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.42578125,
       "learning_rate": 0.00016743157894736843,
-      "loss": 0.0172,
+      "loss": 0.0179,
       "num_input_tokens_seen": 481533952,
       "step": 7350,
-      "train_runtime": 1142.8924,
-      "train_tokens_per_second": 421329.2
+      "train_runtime": 3272.2159,
+      "train_tokens_per_second": 147158.367
     },
     {
       "epoch": 1.592384249242752,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.0001668,
-      "loss": 0.019,
+      "loss": 0.0203,
       "num_input_tokens_seen": 482189312,
       "step": 7360,
-      "train_runtime": 1146.724,
-      "train_tokens_per_second": 420492.911
+      "train_runtime": 3278.2597,
+      "train_tokens_per_second": 147086.979
     },
     {
       "epoch": 1.5945478147987884,
-      "grad_norm": 0.54296875,
+      "grad_norm": 0.37109375,
       "learning_rate": 0.00016616842105263155,
-      "loss": 0.0197,
+      "loss": 0.0207,
       "num_input_tokens_seen": 482844672,
       "step": 7370,
-      "train_runtime": 1150.5458,
-      "train_tokens_per_second": 419665.757
+      "train_runtime": 3282.076,
+      "train_tokens_per_second": 147115.627
     },
     {
       "epoch": 1.5967113803548247,
-      "grad_norm": 0.314453125,
+      "grad_norm": 0.392578125,
       "learning_rate": 0.00016553684210526313,
-      "loss": 0.0194,
+      "loss": 0.0202,
       "num_input_tokens_seen": 483500032,
       "step": 7380,
-      "train_runtime": 1154.374,
-      "train_tokens_per_second": 418841.767
+      "train_runtime": 3285.8809,
+      "train_tokens_per_second": 147144.724
     },
     {
       "epoch": 1.598874945910861,
-      "grad_norm": 0.609375,
+      "grad_norm": 0.54296875,
       "learning_rate": 0.00016490526315789472,
       "loss": 0.0206,
       "num_input_tokens_seen": 484155392,
       "step": 7390,
-      "train_runtime": 1158.1709,
-      "train_tokens_per_second": 418034.488
+      "train_runtime": 3289.6598,
+      "train_tokens_per_second": 147174.913
     },
     {
       "epoch": 1.6010385114668975,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.46484375,
       "learning_rate": 0.0001642736842105263,
-      "loss": 0.0193,
+      "loss": 0.0205,
       "num_input_tokens_seen": 484810752,
       "step": 7400,
-      "train_runtime": 1161.9921,
-      "train_tokens_per_second": 417223.789
+      "train_runtime": 3293.4779,
+      "train_tokens_per_second": 147203.282
     },
     {
       "epoch": 1.6010385114668975,
-      "eval_loss": 0.013829471543431282,
-      "eval_runtime": 1.7413,
-      "eval_samples_per_second": 18.377,
-      "eval_steps_per_second": 0.574,
+      "eval_loss": 0.012541618198156357,
+      "eval_runtime": 2.0947,
+      "eval_samples_per_second": 15.277,
+      "eval_steps_per_second": 0.477,
       "num_input_tokens_seen": 484810752,
       "step": 7400
     },
     {
       "epoch": 1.6010385114668975,
-      "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 97.63397091399936,
-      "eval_sacrebleu": 97.01614214490985,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.57226152603681,
+      "eval_sacrebleu": 96.77994309150014,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 484810752,
-      "perplexity": 1.0139255410376113,
+      "perplexity": 1.0126205941082222,
       "step": 7400
     },
     {
       "epoch": 1.6032020770229338,
-      "grad_norm": 0.380859375,
+      "grad_norm": 0.359375,
       "learning_rate": 0.0001636421052631579,
-      "loss": 0.0199,
+      "loss": 0.0205,
       "num_input_tokens_seen": 485466112,
       "step": 7410,
-      "train_runtime": 1167.5747,
-      "train_tokens_per_second": 415790.198
+      "train_runtime": 3301.6581,
+      "train_tokens_per_second": 147037.062
     },
     {
       "epoch": 1.6053656425789702,
-      "grad_norm": 0.390625,
+      "grad_norm": 0.361328125,
       "learning_rate": 0.00016301052631578948,
-      "loss": 0.0205,
+      "loss": 0.02,
       "num_input_tokens_seen": 486121472,
       "step": 7420,
-      "train_runtime": 1171.3949,
-      "train_tokens_per_second": 414993.662
+      "train_runtime": 3305.4728,
+      "train_tokens_per_second": 147065.64
     },
     {
       "epoch": 1.6075292081350065,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.5,
       "learning_rate": 0.00016237894736842104,
-      "loss": 0.0197,
+      "loss": 0.0204,
       "num_input_tokens_seen": 486776832,
       "step": 7430,
-      "train_runtime": 1175.211,
-      "train_tokens_per_second": 414203.761
+      "train_runtime": 3309.3,
+      "train_tokens_per_second": 147093.596
     },
     {
       "epoch": 1.6096927736910427,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.466796875,
       "learning_rate": 0.0001617473684210526,
-      "loss": 0.0186,
+      "loss": 0.02,
       "num_input_tokens_seen": 487432192,
       "step": 7440,
-      "train_runtime": 1179.0344,
-      "train_tokens_per_second": 413416.438
+      "train_runtime": 3313.1196,
+      "train_tokens_per_second": 147121.821
     },
     {
       "epoch": 1.6118563392470793,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.3515625,
       "learning_rate": 0.00016111578947368419,
-      "loss": 0.0183,
+      "loss": 0.0191,
       "num_input_tokens_seen": 488087552,
       "step": 7450,
-      "train_runtime": 1182.8525,
-      "train_tokens_per_second": 412636.036
+      "train_runtime": 3316.9304,
+      "train_tokens_per_second": 147150.375
     },
     {
       "epoch": 1.6140199048031154,
-      "grad_norm": 0.373046875,
+      "grad_norm": 0.34765625,
       "learning_rate": 0.00016048421052631577,
-      "loss": 0.0173,
+      "loss": 0.0175,
       "num_input_tokens_seen": 488742912,
       "step": 7460,
-      "train_runtime": 1186.6729,
-      "train_tokens_per_second": 411859.836
+      "train_runtime": 3323.1698,
+      "train_tokens_per_second": 147071.304
     },
     {
       "epoch": 1.616183470359152,
-      "grad_norm": 0.703125,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00015985263157894736,
-      "loss": 0.0197,
+      "loss": 0.0214,
       "num_input_tokens_seen": 489398272,
       "step": 7470,
-      "train_runtime": 1190.4916,
-      "train_tokens_per_second": 411089.213
+      "train_runtime": 3326.9742,
+      "train_tokens_per_second": 147100.111
     },
     {
       "epoch": 1.6183470359151881,
-      "grad_norm": 0.458984375,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00015922105263157894,
-      "loss": 0.0169,
+      "loss": 0.0184,
       "num_input_tokens_seen": 490053632,
       "step": 7480,
-      "train_runtime": 1194.3154,
-      "train_tokens_per_second": 410321.798
+      "train_runtime": 3330.7823,
+      "train_tokens_per_second": 147128.691
     },
     {
       "epoch": 1.6205106014712247,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.00015858947368421053,
-      "loss": 0.0181,
+      "loss": 0.0191,
       "num_input_tokens_seen": 490708992,
       "step": 7490,
-      "train_runtime": 1198.1421,
-      "train_tokens_per_second": 409558.257
+      "train_runtime": 3334.5974,
+      "train_tokens_per_second": 147156.895
     },
     {
       "epoch": 1.6226741670272609,
-      "grad_norm": 0.5078125,
+      "grad_norm": 0.3046875,
       "learning_rate": 0.00015795789473684212,
-      "loss": 0.0188,
+      "loss": 0.0193,
       "num_input_tokens_seen": 491364352,
       "step": 7500,
-      "train_runtime": 1201.9634,
-      "train_tokens_per_second": 408801.423
+      "train_runtime": 3338.4043,
+      "train_tokens_per_second": 147185.395
     },
     {
       "epoch": 1.6226741670272609,
-      "eval_loss": 0.013616210781037807,
-      "eval_runtime": 1.7416,
-      "eval_samples_per_second": 18.374,
-      "eval_steps_per_second": 0.574,
+      "eval_loss": 0.01195054966956377,
+      "eval_runtime": 2.1304,
+      "eval_samples_per_second": 15.021,
+      "eval_steps_per_second": 0.469,
       "num_input_tokens_seen": 491364352,
       "step": 7500
     },
     {
       "epoch": 1.6226741670272609,
       "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.09237409482614,
-      "eval_sacrebleu": 96.08208473900969,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 491364352,
-      "perplexity": 1.0137093335588248,
+      "perplexity": 1.0120222427943755,
       "step": 7500
     },
     {
       "epoch": 1.6248377325832972,
-      "grad_norm": 0.375,
+      "grad_norm": 0.349609375,
       "learning_rate": 0.00015732631578947365,
-      "loss": 0.0175,
+      "loss": 0.0177,
       "num_input_tokens_seen": 492019712,
       "step": 7510,
-      "train_runtime": 1207.5937,
-      "train_tokens_per_second": 407438.132
+      "train_runtime": 3346.6726,
+      "train_tokens_per_second": 147017.581
     },
     {
       "epoch": 1.6270012981393336,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.306640625,
       "learning_rate": 0.00015669473684210524,
-      "loss": 0.0194,
+      "loss": 0.0198,
       "num_input_tokens_seen": 492675072,
       "step": 7520,
-      "train_runtime": 1211.4091,
-      "train_tokens_per_second": 406695.854
+      "train_runtime": 3350.4863,
+      "train_tokens_per_second": 147045.84
     },
     {
       "epoch": 1.62916486369537,
-      "grad_norm": 0.26171875,
+      "grad_norm": 0.404296875,
       "learning_rate": 0.00015606315789473682,
-      "loss": 0.0185,
+      "loss": 0.0191,
       "num_input_tokens_seen": 493330432,
       "step": 7530,
-      "train_runtime": 1215.2321,
-      "train_tokens_per_second": 405955.72
+      "train_runtime": 3354.2991,
+      "train_tokens_per_second": 147074.074
     },
     {
       "epoch": 1.6313284292514063,
-      "grad_norm": 0.296875,
+      "grad_norm": 0.333984375,
       "learning_rate": 0.0001554315789473684,
-      "loss": 0.0182,
+      "loss": 0.0187,
       "num_input_tokens_seen": 493985792,
       "step": 7540,
-      "train_runtime": 1219.0542,
-      "train_tokens_per_second": 405220.534
+      "train_runtime": 3358.1088,
+      "train_tokens_per_second": 147102.38
     },
     {
       "epoch": 1.6334919948074427,
-      "grad_norm": 0.4609375,
+      "grad_norm": 0.546875,
       "learning_rate": 0.0001548,
-      "loss": 0.0194,
+      "loss": 0.0203,
       "num_input_tokens_seen": 494641152,
       "step": 7550,
-      "train_runtime": 1222.8649,
-      "train_tokens_per_second": 404493.711
+      "train_runtime": 3361.9119,
+      "train_tokens_per_second": 147130.909
     },
     {
       "epoch": 1.635655560363479,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.458984375,
       "learning_rate": 0.00015416842105263158,
-      "loss": 0.0208,
+      "loss": 0.0216,
       "num_input_tokens_seen": 495296512,
       "step": 7560,
-      "train_runtime": 1226.6937,
-      "train_tokens_per_second": 403765.416
+      "train_runtime": 3367.9656,
+      "train_tokens_per_second": 147061.037
     },
     {
       "epoch": 1.6378191259195154,
-      "grad_norm": 0.5546875,
+      "grad_norm": 0.470703125,
       "learning_rate": 0.00015353684210526317,
-      "loss": 0.0281,
+      "loss": 0.0283,
       "num_input_tokens_seen": 495951872,
       "step": 7570,
-      "train_runtime": 1230.5146,
-      "train_tokens_per_second": 403044.283
+      "train_runtime": 3371.7831,
+      "train_tokens_per_second": 147088.902
     },
     {
       "epoch": 1.6399826914755518,
-      "grad_norm": 0.41015625,
+      "grad_norm": 0.484375,
       "learning_rate": 0.0001529052631578947,
-      "loss": 0.018,
+      "loss": 0.0193,
       "num_input_tokens_seen": 496603136,
       "step": 7580,
-      "train_runtime": 1234.3122,
-      "train_tokens_per_second": 402331.882
+      "train_runtime": 3375.574,
+      "train_tokens_per_second": 147116.648
     },
     {
       "epoch": 1.642146257031588,
-      "grad_norm": 0.376953125,
+      "grad_norm": 0.431640625,
       "learning_rate": 0.0001522736842105263,
-      "loss": 0.0421,
+      "loss": 0.0434,
       "num_input_tokens_seen": 497258496,
       "step": 7590,
-      "train_runtime": 1238.1343,
-      "train_tokens_per_second": 401619.179
+      "train_runtime": 3379.386,
+      "train_tokens_per_second": 147144.627
     },
     {
       "epoch": 1.6443098225876245,
-      "grad_norm": 0.68359375,
+      "grad_norm": 0.6640625,
       "learning_rate": 0.00015164210526315787,
-      "loss": 0.0217,
+      "loss": 0.0233,
       "num_input_tokens_seen": 497913856,
       "step": 7600,
-      "train_runtime": 1241.953,
-      "train_tokens_per_second": 400912.003
+      "train_runtime": 3383.1944,
+      "train_tokens_per_second": 147172.7
     },
     {
       "epoch": 1.6443098225876245,
-      "eval_loss": 0.014020687900483608,
-      "eval_runtime": 1.7048,
-      "eval_samples_per_second": 18.771,
-      "eval_steps_per_second": 0.587,
+      "eval_loss": 0.012651835568249226,
+      "eval_runtime": 2.1539,
+      "eval_samples_per_second": 14.857,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 497913856,
       "step": 7600
     },
     {
       "epoch": 1.6443098225876245,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.66165640680113,
-      "eval_sacrebleu": 97.01614214490985,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 497913856,
-      "perplexity": 1.0141194387235046,
+      "perplexity": 1.0127322086378236,
       "step": 7600
     },
     {
       "epoch": 1.6464733881436606,
-      "grad_norm": 0.546875,
+      "grad_norm": 0.482421875,
       "learning_rate": 0.00015101052631578946,
-      "loss": 0.0174,
+      "loss": 0.0188,
       "num_input_tokens_seen": 498569216,
       "step": 7610,
-      "train_runtime": 1247.4884,
-      "train_tokens_per_second": 399658.409
+      "train_runtime": 3391.4129,
+      "train_tokens_per_second": 147009.295
     },
     {
       "epoch": 1.6486369536996972,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.314453125,
       "learning_rate": 0.00015037894736842105,
-      "loss": 0.0165,
+      "loss": 0.0175,
       "num_input_tokens_seen": 499224576,
       "step": 7620,
-      "train_runtime": 1251.297,
-      "train_tokens_per_second": 398965.687
+      "train_runtime": 3395.2172,
+      "train_tokens_per_second": 147037.597
     },
     {
       "epoch": 1.6508005192557333,
-      "grad_norm": 0.333984375,
+      "grad_norm": 0.345703125,
       "learning_rate": 0.0001497473684210526,
-      "loss": 0.0201,
+      "loss": 0.0206,
       "num_input_tokens_seen": 499879936,
       "step": 7630,
-      "train_runtime": 1255.1196,
-      "train_tokens_per_second": 398272.738
+      "train_runtime": 3399.0365,
+      "train_tokens_per_second": 147065.187
     },
     {
       "epoch": 1.65296408481177,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.45703125,
       "learning_rate": 0.0001491157894736842,
-      "loss": 0.021,
+      "loss": 0.0221,
       "num_input_tokens_seen": 500535296,
       "step": 7640,
-      "train_runtime": 1258.9407,
-      "train_tokens_per_second": 397584.502
+      "train_runtime": 3402.8462,
+      "train_tokens_per_second": 147093.129
     },
     {
       "epoch": 1.655127650367806,
-      "grad_norm": 0.4453125,
+      "grad_norm": 0.53125,
       "learning_rate": 0.00014848421052631578,
-      "loss": 0.0217,
+      "loss": 0.0223,
       "num_input_tokens_seen": 501190656,
       "step": 7650,
-      "train_runtime": 1262.7516,
-      "train_tokens_per_second": 396903.614
+      "train_runtime": 3406.6556,
+      "train_tokens_per_second": 147121.024
     },
     {
       "epoch": 1.6572912159238427,
-      "grad_norm": 0.431640625,
+      "grad_norm": 0.388671875,
       "learning_rate": 0.00014785263157894736,
-      "loss": 0.018,
+      "loss": 0.0185,
       "num_input_tokens_seen": 501846016,
       "step": 7660,
-      "train_runtime": 1266.5554,
-      "train_tokens_per_second": 396229.029
+      "train_runtime": 3412.9006,
+      "train_tokens_per_second": 147043.841
     },
     {
       "epoch": 1.6594547814798788,
-      "grad_norm": 0.388671875,
+      "grad_norm": 0.3046875,
       "learning_rate": 0.00014722105263157892,
-      "loss": 0.0187,
+      "loss": 0.019,
       "num_input_tokens_seen": 502501376,
       "step": 7670,
-      "train_runtime": 1270.3615,
-      "train_tokens_per_second": 395557.781
+      "train_runtime": 3416.71,
+      "train_tokens_per_second": 147071.709
     },
     {
       "epoch": 1.6616183470359152,
-      "grad_norm": 0.33203125,
+      "grad_norm": 0.298828125,
       "learning_rate": 0.0001465894736842105,
-      "loss": 0.0194,
+      "loss": 0.0196,
       "num_input_tokens_seen": 503156736,
       "step": 7680,
-      "train_runtime": 1274.1782,
-      "train_tokens_per_second": 394887.254
+      "train_runtime": 3420.5206,
+      "train_tokens_per_second": 147099.46
     },
     {
       "epoch": 1.6637819125919515,
-      "grad_norm": 0.427734375,
+      "grad_norm": 0.416015625,
       "learning_rate": 0.0001459578947368421,
-      "loss": 0.0251,
+      "loss": 0.0261,
       "num_input_tokens_seen": 503812096,
       "step": 7690,
-      "train_runtime": 1277.9972,
-      "train_tokens_per_second": 394220.025
+      "train_runtime": 3424.3328,
+      "train_tokens_per_second": 147127.084
     },
     {
       "epoch": 1.6659454781479879,
-      "grad_norm": 0.57421875,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.00014532631578947368,
-      "loss": 0.0198,
+      "loss": 0.0201,
       "num_input_tokens_seen": 504467456,
       "step": 7700,
-      "train_runtime": 1281.8119,
-      "train_tokens_per_second": 393558.1
+      "train_runtime": 3428.1363,
+      "train_tokens_per_second": 147155.016
     },
     {
       "epoch": 1.6659454781479879,
-      "eval_loss": 0.013091851025819778,
-      "eval_runtime": 1.6783,
-      "eval_samples_per_second": 19.067,
-      "eval_steps_per_second": 0.596,
+      "eval_loss": 0.011731253936886787,
+      "eval_runtime": 2.1236,
+      "eval_samples_per_second": 15.069,
+      "eval_steps_per_second": 0.471,
       "num_input_tokens_seen": 504467456,
       "step": 7700
     },
     {
       "epoch": 1.6659454781479879,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 98.33791810109832,
-      "eval_sacrebleu": 97.89877731533007,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 504467456,
-      "perplexity": 1.013177924517751,
+      "perplexity": 1.0118003349677656,
       "step": 7700
     },
     {
       "epoch": 1.6681090437040242,
-      "grad_norm": 0.498046875,
+      "grad_norm": 0.453125,
       "learning_rate": 0.00014469473684210524,
-      "loss": 0.0208,
+      "loss": 0.0218,
       "num_input_tokens_seen": 505122816,
       "step": 7710,
-      "train_runtime": 1287.3267,
-      "train_tokens_per_second": 392381.207
+      "train_runtime": 3436.3206,
+      "train_tokens_per_second": 146995.251
     },
     {
       "epoch": 1.6702726092600606,
-      "grad_norm": 0.26171875,
+      "grad_norm": 0.330078125,
       "learning_rate": 0.00014406315789473683,
-      "loss": 0.0169,
+      "loss": 0.0175,
       "num_input_tokens_seen": 505778176,
       "step": 7720,
-      "train_runtime": 1291.1392,
-      "train_tokens_per_second": 391730.18
+      "train_runtime": 3440.1328,
+      "train_tokens_per_second": 147022.862
     },
     {
       "epoch": 1.672436174816097,
-      "grad_norm": 0.375,
+      "grad_norm": 0.30859375,
       "learning_rate": 0.00014343157894736842,
-      "loss": 0.0189,
+      "loss": 0.0195,
       "num_input_tokens_seen": 506433536,
       "step": 7730,
-      "train_runtime": 1294.9437,
-      "train_tokens_per_second": 391085.381
+      "train_runtime": 3443.9218,
+      "train_tokens_per_second": 147051.402
     },
     {
       "epoch": 1.6745997403721333,
-      "grad_norm": 0.47265625,
+      "grad_norm": 0.5078125,
       "learning_rate": 0.00014279999999999997,
-      "loss": 0.0178,
+      "loss": 0.0192,
       "num_input_tokens_seen": 507088896,
       "step": 7740,
-      "train_runtime": 1298.7576,
-      "train_tokens_per_second": 390441.534
+      "train_runtime": 3447.7441,
+      "train_tokens_per_second": 147078.462
     },
     {
       "epoch": 1.6767633059281697,
-      "grad_norm": 0.55859375,
+      "grad_norm": 0.4375,
       "learning_rate": 0.00014216842105263156,
-      "loss": 0.0176,
+      "loss": 0.0175,
       "num_input_tokens_seen": 507744256,
       "step": 7750,
-      "train_runtime": 1302.5808,
-      "train_tokens_per_second": 389798.651
+      "train_runtime": 3451.5673,
+      "train_tokens_per_second": 147105.42
     },
     {
       "epoch": 1.6789268714842058,
-      "grad_norm": 0.4609375,
+      "grad_norm": 0.35546875,
       "learning_rate": 0.00014153684210526315,
-      "loss": 0.0187,
+      "loss": 0.0192,
       "num_input_tokens_seen": 508399616,
       "step": 7760,
-      "train_runtime": 1306.3997,
-      "train_tokens_per_second": 389160.854
+      "train_runtime": 3457.6351,
+      "train_tokens_per_second": 147036.804
     },
     {
       "epoch": 1.6810904370402424,
-      "grad_norm": 0.412109375,
+      "grad_norm": 0.46484375,
       "learning_rate": 0.00014090526315789473,
-      "loss": 0.0174,
+      "loss": 0.0178,
       "num_input_tokens_seen": 509054976,
       "step": 7770,
-      "train_runtime": 1310.226,
-      "train_tokens_per_second": 388524.563
+      "train_runtime": 3461.4502,
+      "train_tokens_per_second": 147064.076
     },
     {
       "epoch": 1.6832540025962786,
-      "grad_norm": 0.435546875,
+      "grad_norm": 0.421875,
       "learning_rate": 0.0001402736842105263,
-      "loss": 0.0186,
+      "loss": 0.0194,
       "num_input_tokens_seen": 509710336,
       "step": 7780,
-      "train_runtime": 1314.0427,
-      "train_tokens_per_second": 387894.794
+      "train_runtime": 3465.2701,
+      "train_tokens_per_second": 147091.085
     },
     {
       "epoch": 1.6854175681523151,
-      "grad_norm": 0.279296875,
+      "grad_norm": 0.33203125,
       "learning_rate": 0.00013964210526315788,
-      "loss": 0.0201,
+      "loss": 0.0207,
       "num_input_tokens_seen": 510365696,
       "step": 7790,
-      "train_runtime": 1317.8586,
-      "train_tokens_per_second": 387268.921
+      "train_runtime": 3469.0795,
+      "train_tokens_per_second": 147118.476
     },
     {
       "epoch": 1.6875811337083513,
-      "grad_norm": 0.55078125,
+      "grad_norm": 0.44140625,
       "learning_rate": 0.00013901052631578947,
-      "loss": 0.0205,
+      "loss": 0.0214,
       "num_input_tokens_seen": 511021056,
       "step": 7800,
-      "train_runtime": 1321.683,
-      "train_tokens_per_second": 386644.18
+      "train_runtime": 3472.8918,
+      "train_tokens_per_second": 147145.689
     },
     {
       "epoch": 1.6875811337083513,
-      "eval_loss": 0.013206811621785164,
-      "eval_runtime": 1.6732,
-      "eval_samples_per_second": 19.125,
-      "eval_steps_per_second": 0.598,
+      "eval_loss": 0.011290564201772213,
+      "eval_runtime": 2.1435,
+      "eval_samples_per_second": 14.929,
+      "eval_steps_per_second": 0.467,
       "num_input_tokens_seen": 511021056,
       "step": 7800
     },
     {
       "epoch": 1.6875811337083513,
-      "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.72750672761462,
-      "eval_sacrebleu": 96.96578008549908,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 511021056,
-      "perplexity": 1.0132944067510776,
+      "perplexity": 1.0113545431813067,
       "step": 7800
     },
     {
       "epoch": 1.6897446992643879,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.326171875,
       "learning_rate": 0.00013837894736842103,
-      "loss": 0.0202,
+      "loss": 0.0209,
       "num_input_tokens_seen": 511676416,
       "step": 7810,
-      "train_runtime": 1327.1865,
-      "train_tokens_per_second": 385534.678
+      "train_runtime": 3481.1094,
+      "train_tokens_per_second": 146986.596
     },
     {
       "epoch": 1.691908264820424,
-      "grad_norm": 0.470703125,
+      "grad_norm": 0.3671875,
       "learning_rate": 0.0001377473684210526,
-      "loss": 0.0194,
+      "loss": 0.0199,
       "num_input_tokens_seen": 512331776,
       "step": 7820,
-      "train_runtime": 1331.0028,
-      "train_tokens_per_second": 384921.632
+      "train_runtime": 3484.9263,
+      "train_tokens_per_second": 147013.66
     },
     {
       "epoch": 1.6940718303764604,
-      "grad_norm": 0.287109375,
+      "grad_norm": 0.37109375,
       "learning_rate": 0.0001371157894736842,
-      "loss": 0.0187,
+      "loss": 0.0195,
       "num_input_tokens_seen": 512987136,
       "step": 7830,
-      "train_runtime": 1334.8205,
-      "train_tokens_per_second": 384311.689
+      "train_runtime": 3488.7424,
+      "train_tokens_per_second": 147040.702
     },
     {
       "epoch": 1.6962353959324967,
-      "grad_norm": 0.29296875,
+      "grad_norm": 0.37109375,
       "learning_rate": 0.00013648421052631578,
-      "loss": 0.0179,
+      "loss": 0.0182,
       "num_input_tokens_seen": 513642496,
       "step": 7840,
-      "train_runtime": 1338.649,
-      "train_tokens_per_second": 383702.138
+      "train_runtime": 3492.5624,
+      "train_tokens_per_second": 147067.522
     },
     {
       "epoch": 1.698398961488533,
-      "grad_norm": 0.30859375,
+      "grad_norm": 0.330078125,
       "learning_rate": 0.00013585263157894734,
-      "loss": 0.0172,
+      "loss": 0.0184,
       "num_input_tokens_seen": 514297856,
       "step": 7850,
-      "train_runtime": 1342.4692,
-      "train_tokens_per_second": 383098.448
+      "train_runtime": 3496.381,
+      "train_tokens_per_second": 147094.34
     },
     {
       "epoch": 1.7005625270445694,
-      "grad_norm": 0.48828125,
+      "grad_norm": 0.462890625,
       "learning_rate": 0.00013522105263157893,
-      "loss": 0.0179,
+      "loss": 0.0188,
       "num_input_tokens_seen": 514953216,
       "step": 7860,
-      "train_runtime": 1346.2887,
-      "train_tokens_per_second": 382498.344
+      "train_runtime": 3502.2526,
+      "train_tokens_per_second": 147034.859
     },
     {
       "epoch": 1.7027260926006058,
-      "grad_norm": 0.5078125,
+      "grad_norm": 0.486328125,
       "learning_rate": 0.00013458947368421052,
-      "loss": 0.0171,
+      "loss": 0.0187,
       "num_input_tokens_seen": 515608576,
       "step": 7870,
-      "train_runtime": 1350.1111,
-      "train_tokens_per_second": 381900.851
+      "train_runtime": 3506.0684,
+      "train_tokens_per_second": 147061.755
     },
     {
       "epoch": 1.7048896581566422,
-      "grad_norm": 0.314453125,
+      "grad_norm": 0.3046875,
       "learning_rate": 0.00013395789473684208,
-      "loss": 0.0169,
+      "loss": 0.0185,
       "num_input_tokens_seen": 516263936,
       "step": 7880,
-      "train_runtime": 1353.9361,
-      "train_tokens_per_second": 381305.991
+      "train_runtime": 3509.8938,
+      "train_tokens_per_second": 147088.194
     },
     {
       "epoch": 1.7070532237126785,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.44921875,
       "learning_rate": 0.00013332631578947366,
-      "loss": 0.019,
+      "loss": 0.0201,
       "num_input_tokens_seen": 516919296,
       "step": 7890,
-      "train_runtime": 1357.7602,
-      "train_tokens_per_second": 380714.716
+      "train_runtime": 3513.7148,
+      "train_tokens_per_second": 147114.755
     },
     {
       "epoch": 1.709216789268715,
-      "grad_norm": 0.40234375,
+      "grad_norm": 0.458984375,
       "learning_rate": 0.00013269473684210525,
       "loss": 0.0178,
       "num_input_tokens_seen": 517574656,
       "step": 7900,
-      "train_runtime": 1361.5822,
-      "train_tokens_per_second": 380127.358
+      "train_runtime": 3517.5325,
+      "train_tokens_per_second": 147141.402
     },
     {
       "epoch": 1.709216789268715,
-      "eval_loss": 0.013531056232750416,
-      "eval_runtime": 1.689,
-      "eval_samples_per_second": 18.946,
-      "eval_steps_per_second": 0.592,
+      "eval_loss": 0.010886249132454395,
+      "eval_runtime": 2.4894,
+      "eval_samples_per_second": 12.855,
+      "eval_steps_per_second": 0.402,
       "num_input_tokens_seen": 517574656,
       "step": 7900
     },
     {
       "epoch": 1.709216789268715,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
+      "eval_byte_accuracy": 0.9970326409495549,
+      "eval_chrf": 98.32407535469744,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 517574656,
-      "perplexity": 1.0136230152736803,
+      "perplexity": 1.010945719951341,
       "step": 7900
     },
     {
       "epoch": 1.711380354824751,
-      "grad_norm": 0.47265625,
+      "grad_norm": 0.3671875,
       "learning_rate": 0.00013206315789473684,
-      "loss": 0.0179,
+      "loss": 0.0193,
       "num_input_tokens_seen": 518230016,
       "step": 7910,
-      "train_runtime": 1367.1045,
-      "train_tokens_per_second": 379071.261
+      "train_runtime": 3525.9923,
+      "train_tokens_per_second": 146974.235
     },
     {
       "epoch": 1.7135439203807876,
-      "grad_norm": 0.482421875,
+      "grad_norm": 0.337890625,
       "learning_rate": 0.0001314315789473684,
-      "loss": 0.0184,
+      "loss": 0.0201,
       "num_input_tokens_seen": 518885376,
       "step": 7920,
-      "train_runtime": 1370.9241,
-      "train_tokens_per_second": 378493.134
+      "train_runtime": 3529.8039,
+      "train_tokens_per_second": 147001.191
     },
     {
       "epoch": 1.7157074859368238,
-      "grad_norm": 0.5,
+      "grad_norm": 0.404296875,
       "learning_rate": 0.00013079999999999998,
-      "loss": 0.0167,
+      "loss": 0.017,
       "num_input_tokens_seen": 519540736,
       "step": 7930,
-      "train_runtime": 1374.742,
-      "train_tokens_per_second": 377918.727
+      "train_runtime": 3533.6263,
+      "train_tokens_per_second": 147027.639
     },
     {
       "epoch": 1.7178710514928603,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00013016842105263157,
-      "loss": 0.0185,
+      "loss": 0.0197,
       "num_input_tokens_seen": 520196096,
       "step": 7940,
-      "train_runtime": 1378.5609,
-      "train_tokens_per_second": 377347.193
+      "train_runtime": 3537.4337,
+      "train_tokens_per_second": 147054.655
     },
     {
       "epoch": 1.7200346170488965,
-      "grad_norm": 0.427734375,
+      "grad_norm": 0.357421875,
       "learning_rate": 0.00012953684210526313,
       "loss": 0.0199,
       "num_input_tokens_seen": 520851456,
       "step": 7950,
-      "train_runtime": 1382.3761,
-      "train_tokens_per_second": 376779.845
+      "train_runtime": 3541.2447,
+      "train_tokens_per_second": 147081.465
     },
     {
       "epoch": 1.722198182604933,
       "grad_norm": 0.310546875,
       "learning_rate": 0.0001289052631578947,
-      "loss": 0.0196,
+      "loss": 0.0204,
       "num_input_tokens_seen": 521506816,
       "step": 7960,
-      "train_runtime": 1386.191,
-      "train_tokens_per_second": 376215.708
+      "train_runtime": 3547.2565,
+      "train_tokens_per_second": 147016.947
     },
     {
       "epoch": 1.7243617481609692,
-      "grad_norm": 0.3125,
+      "grad_norm": 0.326171875,
       "learning_rate": 0.0001282736842105263,
-      "loss": 0.0172,
+      "loss": 0.0181,
       "num_input_tokens_seen": 522162176,
       "step": 7970,
-      "train_runtime": 1390.0057,
-      "train_tokens_per_second": 375654.689
+      "train_runtime": 3551.0783,
+      "train_tokens_per_second": 147043.272
     },
     {
       "epoch": 1.7265253137170056,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.474609375,
       "learning_rate": 0.00012764210526315789,
-      "loss": 0.019,
+      "loss": 0.0194,
       "num_input_tokens_seen": 522817536,
       "step": 7980,
-      "train_runtime": 1393.8333,
-      "train_tokens_per_second": 375093.309
+      "train_runtime": 3554.8984,
+      "train_tokens_per_second": 147069.614
     },
     {
       "epoch": 1.728688879273042,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.5625,
       "learning_rate": 0.00012701052631578945,
-      "loss": 0.0202,
+      "loss": 0.02,
       "num_input_tokens_seen": 523472896,
       "step": 7990,
-      "train_runtime": 1397.6593,
-      "train_tokens_per_second": 374535.397
+      "train_runtime": 3558.7128,
+      "train_tokens_per_second": 147096.132
     },
     {
       "epoch": 1.7308524448290783,
-      "grad_norm": 0.33203125,
+      "grad_norm": 0.421875,
       "learning_rate": 0.00012637894736842103,
-      "loss": 0.0185,
+      "loss": 0.0204,
       "num_input_tokens_seen": 524128256,
       "step": 8000,
-      "train_runtime": 1401.4815,
-      "train_tokens_per_second": 373981.569
+      "train_runtime": 3562.5131,
+      "train_tokens_per_second": 147123.179
     },
     {
       "epoch": 1.7308524448290783,
-      "eval_loss": 0.014470809139311314,
-      "eval_runtime": 1.7021,
-      "eval_samples_per_second": 18.801,
-      "eval_steps_per_second": 0.588,
+      "eval_loss": 0.011375652626156807,
+      "eval_runtime": 2.1623,
+      "eval_samples_per_second": 14.799,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 524128256,
       "step": 8000
     },
     {
       "epoch": 1.7308524448290783,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.36176305597374,
-      "eval_sacrebleu": 96.96578008549908,
-      "eval_word_accuracy": 0.9903846153846154,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
+      "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 524128256,
-      "perplexity": 1.0145760181719266,
+      "perplexity": 1.0114406014071076,
       "step": 8000
     },
     {
       "epoch": 1.7330160103851147,
-      "grad_norm": 0.48046875,
+      "grad_norm": 0.478515625,
       "learning_rate": 0.00012574736842105262,
-      "loss": 0.018,
+      "loss": 0.0184,
       "num_input_tokens_seen": 524783616,
       "step": 8010,
-      "train_runtime": 1407.0677,
-      "train_tokens_per_second": 372962.592
+      "train_runtime": 3570.7811,
+      "train_tokens_per_second": 146966.057
     },
     {
       "epoch": 1.735179575941151,
-      "grad_norm": 0.29296875,
+      "grad_norm": 0.296875,
       "learning_rate": 0.0001251157894736842,
-      "loss": 0.0191,
+      "loss": 0.0207,
       "num_input_tokens_seen": 525438976,
       "step": 8020,
-      "train_runtime": 1410.8879,
-      "train_tokens_per_second": 372417.238
+      "train_runtime": 3575.036,
+      "train_tokens_per_second": 146974.456
     },
     {
       "epoch": 1.7373431414971874,
-      "grad_norm": 0.40625,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.00012448421052631576,
-      "loss": 0.0192,
+      "loss": 0.0203,
       "num_input_tokens_seen": 526094336,
       "step": 8030,
-      "train_runtime": 1414.7074,
-      "train_tokens_per_second": 371875.006
+      "train_runtime": 3578.8478,
+      "train_tokens_per_second": 147001.035
     },
     {
       "epoch": 1.7395067070532237,
-      "grad_norm": 0.44140625,
+      "grad_norm": 0.310546875,
       "learning_rate": 0.00012385263157894735,
-      "loss": 0.0186,
+      "loss": 0.0193,
       "num_input_tokens_seen": 526749696,
       "step": 8040,
-      "train_runtime": 1418.533,
-      "train_tokens_per_second": 371334.117
+      "train_runtime": 3582.657,
+      "train_tokens_per_second": 147027.666
     },
     {
       "epoch": 1.74167027260926,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.353515625,
       "learning_rate": 0.00012322105263157894,
-      "loss": 0.0181,
+      "loss": 0.0193,
       "num_input_tokens_seen": 527405056,
       "step": 8050,
-      "train_runtime": 1422.3527,
-      "train_tokens_per_second": 370797.665
+      "train_runtime": 3586.4702,
+      "train_tokens_per_second": 147054.076
     },
     {
       "epoch": 1.7438338381652962,
-      "grad_norm": 0.43359375,
+      "grad_norm": 0.337890625,
       "learning_rate": 0.00012258947368421052,
-      "loss": 0.0188,
+      "loss": 0.0207,
       "num_input_tokens_seen": 528060416,
       "step": 8060,
-      "train_runtime": 1426.1678,
-      "train_tokens_per_second": 370265.285
+      "train_runtime": 3592.3668,
+      "train_tokens_per_second": 146995.127
     },
     {
       "epoch": 1.7459974037213328,
-      "grad_norm": 0.349609375,
+      "grad_norm": 0.31640625,
       "learning_rate": 0.0001219578947368421,
-      "loss": 0.0158,
+      "loss": 0.0168,
       "num_input_tokens_seen": 528715776,
       "step": 8070,
-      "train_runtime": 1429.9887,
-      "train_tokens_per_second": 369734.225
+      "train_runtime": 3596.1775,
+      "train_tokens_per_second": 147021.603
     },
     {
       "epoch": 1.748160969277369,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.357421875,
       "learning_rate": 0.00012132631578947368,
-      "loss": 0.0175,
+      "loss": 0.0185,
       "num_input_tokens_seen": 529371136,
       "step": 8080,
-      "train_runtime": 1433.8098,
-      "train_tokens_per_second": 369205.976
+      "train_runtime": 3599.9884,
+      "train_tokens_per_second": 147048.01
     },
     {
       "epoch": 1.7503245348334056,
-      "grad_norm": 0.39453125,
+      "grad_norm": 0.373046875,
       "learning_rate": 0.00012069473684210526,
-      "loss": 0.018,
+      "loss": 0.019,
       "num_input_tokens_seen": 530026496,
       "step": 8090,
-      "train_runtime": 1437.6297,
-      "train_tokens_per_second": 368680.818
+      "train_runtime": 3603.7949,
+      "train_tokens_per_second": 147074.546
     },
     {
       "epoch": 1.7524881003894417,
-      "grad_norm": 0.62109375,
+      "grad_norm": 0.380859375,
       "learning_rate": 0.00012006315789473683,
-      "loss": 0.0203,
+      "loss": 0.0219,
       "num_input_tokens_seen": 530681856,
       "step": 8100,
-      "train_runtime": 1441.4439,
-      "train_tokens_per_second": 368159.91
+      "train_runtime": 3607.6029,
+      "train_tokens_per_second": 147100.961
     },
     {
       "epoch": 1.7524881003894417,
-      "eval_loss": 0.013024972751736641,
-      "eval_runtime": 1.7245,
-      "eval_samples_per_second": 18.556,
-      "eval_steps_per_second": 0.58,
+      "eval_loss": 0.010724018327891827,
+      "eval_runtime": 2.0748,
+      "eval_samples_per_second": 15.423,
+      "eval_steps_per_second": 0.482,
       "num_input_tokens_seen": 530681856,
       "step": 8100
     },
     {
       "epoch": 1.7524881003894417,
-      "eval_byte_accuracy": 0.9962908011869436,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
+      "eval_byte_accuracy": 0.9970326409495549,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 530681856,
-      "perplexity": 1.0131101671925917,
+      "perplexity": 1.010781726716561,
       "step": 8100
     },
     {
       "epoch": 1.7546516659454783,
-      "grad_norm": 0.33984375,
+      "grad_norm": 0.515625,
       "learning_rate": 0.00011943157894736841,
-      "loss": 0.0191,
+      "loss": 0.0195,
       "num_input_tokens_seen": 531337216,
       "step": 8110,
-      "train_runtime": 1447.0008,
-      "train_tokens_per_second": 367198.984
+      "train_runtime": 3615.6952,
+      "train_tokens_per_second": 146952.99
     },
     {
       "epoch": 1.7568152315015144,
-      "grad_norm": 0.40625,
+      "grad_norm": 0.40234375,
       "learning_rate": 0.0001188,
-      "loss": 0.0187,
+      "loss": 0.0194,
       "num_input_tokens_seen": 531992576,
       "step": 8120,
-      "train_runtime": 1450.8158,
-      "train_tokens_per_second": 366685.131
+      "train_runtime": 3619.5079,
+      "train_tokens_per_second": 146979.257
     },
     {
       "epoch": 1.758978797057551,
-      "grad_norm": 0.294921875,
+      "grad_norm": 0.4296875,
       "learning_rate": 0.00011816842105263156,
-      "loss": 0.0184,
+      "loss": 0.0193,
       "num_input_tokens_seen": 532647936,
       "step": 8130,
-      "train_runtime": 1454.6211,
-      "train_tokens_per_second": 366176.407
+      "train_runtime": 3623.3058,
+      "train_tokens_per_second": 147006.066
     },
     {
       "epoch": 1.7611423626135871,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.4296875,
       "learning_rate": 0.00011753684210526315,
-      "loss": 0.0182,
+      "loss": 0.0191,
       "num_input_tokens_seen": 533303296,
       "step": 8140,
-      "train_runtime": 1458.4408,
-      "train_tokens_per_second": 365666.738
+      "train_runtime": 3627.1278,
+      "train_tokens_per_second": 147031.844
     },
     {
       "epoch": 1.7633059281696235,
-      "grad_norm": 0.578125,
+      "grad_norm": 0.53515625,
       "learning_rate": 0.00011690526315789473,
-      "loss": 0.0208,
+      "loss": 0.0218,
       "num_input_tokens_seen": 533958656,
       "step": 8150,
-      "train_runtime": 1462.2518,
-      "train_tokens_per_second": 365161.91
+      "train_runtime": 3630.9307,
+      "train_tokens_per_second": 147058.345
     },
     {
       "epoch": 1.7654694937256599,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.5,
       "learning_rate": 0.00011627368421052632,
-      "loss": 0.0248,
+      "loss": 0.0247,
       "num_input_tokens_seen": 534614016,
       "step": 8160,
-      "train_runtime": 1466.0724,
-      "train_tokens_per_second": 364657.313
+      "train_runtime": 3636.9571,
+      "train_tokens_per_second": 146994.864
     },
     {
       "epoch": 1.7676330592816962,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.3359375,
       "learning_rate": 0.00011564210526315788,
-      "loss": 0.0193,
+      "loss": 0.0207,
       "num_input_tokens_seen": 535269376,
       "step": 8170,
-      "train_runtime": 1469.8985,
-      "train_tokens_per_second": 364153.975
+      "train_runtime": 3640.7675,
+      "train_tokens_per_second": 147021.025
     },
     {
       "epoch": 1.7697966248377326,
-      "grad_norm": 0.466796875,
+      "grad_norm": 0.515625,
       "learning_rate": 0.00011501052631578947,
-      "loss": 0.0216,
+      "loss": 0.023,
       "num_input_tokens_seen": 535924736,
       "step": 8180,
-      "train_runtime": 1473.719,
-      "train_tokens_per_second": 363654.622
+      "train_runtime": 3644.5823,
+      "train_tokens_per_second": 147046.959
     },
     {
       "epoch": 1.771960190393769,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.345703125,
       "learning_rate": 0.00011437894736842105,
-      "loss": 0.0169,
+      "loss": 0.018,
       "num_input_tokens_seen": 536580096,
       "step": 8190,
-      "train_runtime": 1477.5411,
-      "train_tokens_per_second": 363157.486
+      "train_runtime": 3648.3887,
+      "train_tokens_per_second": 147073.17
     },
     {
       "epoch": 1.7741237559498053,
-      "grad_norm": 0.298828125,
+      "grad_norm": 0.373046875,
       "learning_rate": 0.00011374736842105261,
-      "loss": 0.0163,
+      "loss": 0.0172,
       "num_input_tokens_seen": 537235456,
       "step": 8200,
-      "train_runtime": 1481.3719,
-      "train_tokens_per_second": 362660.748
+      "train_runtime": 3652.206,
+      "train_tokens_per_second": 147098.892
     },
     {
       "epoch": 1.7741237559498053,
-      "eval_loss": 0.013625239953398705,
-      "eval_runtime": 1.727,
-      "eval_samples_per_second": 18.529,
-      "eval_steps_per_second": 0.579,
+      "eval_loss": 0.01021627802401781,
+      "eval_runtime": 2.0679,
+      "eval_samples_per_second": 15.474,
+      "eval_steps_per_second": 0.484,
       "num_input_tokens_seen": 537235456,
       "step": 8200
     },
     {
       "epoch": 1.7741237559498053,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.80284314119724,
-      "eval_sacrebleu": 97.30034967398244,
+      "eval_chrf": 98.29630543849224,
+      "eval_sacrebleu": 97.56427984542309,
       "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 537235456,
-      "perplexity": 1.0137184865564433,
+      "perplexity": 1.0102686423633132,
       "step": 8200
     },
     {
       "epoch": 1.7762873215058417,
-      "grad_norm": 0.380859375,
+      "grad_norm": 0.5390625,
       "learning_rate": 0.0001131157894736842,
-      "loss": 0.0202,
+      "loss": 0.0215,
       "num_input_tokens_seen": 537890816,
       "step": 8210,
-      "train_runtime": 1486.9277,
-      "train_tokens_per_second": 361746.45
+      "train_runtime": 3660.4725,
+      "train_tokens_per_second": 146945.734
     },
     {
       "epoch": 1.778450887061878,
-      "grad_norm": 0.349609375,
+      "grad_norm": 0.396484375,
       "learning_rate": 0.00011248421052631578,
-      "loss": 0.017,
+      "loss": 0.018,
       "num_input_tokens_seen": 538546176,
       "step": 8220,
-      "train_runtime": 1490.7499,
-      "train_tokens_per_second": 361258.562
+      "train_runtime": 3664.28,
+      "train_tokens_per_second": 146971.895
     },
     {
       "epoch": 1.7806144526179142,
-      "grad_norm": 0.296875,
+      "grad_norm": 0.333984375,
       "learning_rate": 0.00011185263157894737,
-      "loss": 0.0203,
+      "loss": 0.0219,
       "num_input_tokens_seen": 539201536,
       "step": 8230,
-      "train_runtime": 1494.5707,
-      "train_tokens_per_second": 360773.533
+      "train_runtime": 3668.0854,
+      "train_tokens_per_second": 146998.086
     },
     {
       "epoch": 1.7827780181739508,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.412109375,
       "learning_rate": 0.00011122105263157893,
-      "loss": 0.0159,
+      "loss": 0.0178,
       "num_input_tokens_seen": 539856896,
       "step": 8240,
-      "train_runtime": 1498.3854,
-      "train_tokens_per_second": 360292.412
+      "train_runtime": 3671.8996,
+      "train_tokens_per_second": 147023.871
     },
     {
       "epoch": 1.784941583729987,
-      "grad_norm": 0.341796875,
+      "grad_norm": 0.365234375,
       "learning_rate": 0.00011058947368421052,
-      "loss": 0.0207,
+      "loss": 0.0218,
       "num_input_tokens_seen": 540512256,
       "step": 8250,
-      "train_runtime": 1502.2091,
-      "train_tokens_per_second": 359811.589
+      "train_runtime": 3675.7129,
+      "train_tokens_per_second": 147049.638
     },
     {
       "epoch": 1.7871051492860235,
-      "grad_norm": 0.3671875,
+      "grad_norm": 0.369140625,
       "learning_rate": 0.0001099578947368421,
-      "loss": 0.0173,
+      "loss": 0.0184,
       "num_input_tokens_seen": 541167616,
       "step": 8260,
-      "train_runtime": 1506.0281,
-      "train_tokens_per_second": 359334.348
+      "train_runtime": 3681.7547,
+      "train_tokens_per_second": 146986.332
     },
     {
       "epoch": 1.7892687148420596,
-      "grad_norm": 0.52734375,
+      "grad_norm": 0.353515625,
       "learning_rate": 0.00010932631578947366,
-      "loss": 0.0185,
+      "loss": 0.0189,
       "num_input_tokens_seen": 541822976,
       "step": 8270,
-      "train_runtime": 1509.8436,
-      "train_tokens_per_second": 358860.325
+      "train_runtime": 3685.564,
+      "train_tokens_per_second": 147012.228
     },
     {
       "epoch": 1.7914322803980962,
-      "grad_norm": 0.30859375,
+      "grad_norm": 0.384765625,
       "learning_rate": 0.00010869473684210525,
-      "loss": 0.0167,
+      "loss": 0.0179,
       "num_input_tokens_seen": 542478336,
       "step": 8280,
-      "train_runtime": 1513.6693,
-      "train_tokens_per_second": 358386.294
+      "train_runtime": 3689.3687,
+      "train_tokens_per_second": 147038.257
     },
     {
       "epoch": 1.7935958459541324,
-      "grad_norm": 0.3359375,
+      "grad_norm": 0.3671875,
       "learning_rate": 0.00010806315789473683,
-      "loss": 0.0203,
+      "loss": 0.0216,
       "num_input_tokens_seen": 543133696,
       "step": 8290,
-      "train_runtime": 1517.488,
-      "train_tokens_per_second": 357916.313
+      "train_runtime": 3693.18,
+      "train_tokens_per_second": 147063.966
     },
     {
       "epoch": 1.7957594115101687,
-      "grad_norm": 0.4765625,
+      "grad_norm": 0.341796875,
       "learning_rate": 0.00010743157894736842,
-      "loss": 0.0189,
+      "loss": 0.0183,
       "num_input_tokens_seen": 543789056,
       "step": 8300,
-      "train_runtime": 1521.3068,
-      "train_tokens_per_second": 357448.64
+      "train_runtime": 3696.986,
+      "train_tokens_per_second": 147089.833
     },
     {
       "epoch": 1.7957594115101687,
-      "eval_loss": 0.013864190317690372,
-      "eval_runtime": 1.7263,
-      "eval_samples_per_second": 18.537,
-      "eval_steps_per_second": 0.579,
+      "eval_loss": 0.010247361846268177,
+      "eval_runtime": 2.1361,
+      "eval_samples_per_second": 14.981,
+      "eval_steps_per_second": 0.468,
       "num_input_tokens_seen": 543789056,
       "step": 8300
     },
     {
       "epoch": 1.7957594115101687,
-      "eval_byte_accuracy": 0.9974035608308606,
-      "eval_chrf": 98.29638986189566,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
+      "eval_word_accuracy": 0.9917582417582418,
       "num_input_tokens_seen": 543789056,
-      "perplexity": 1.0139607439006826,
+      "perplexity": 1.0103000458622853,
       "step": 8300
     },
     {
       "epoch": 1.797922977066205,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.427734375,
       "learning_rate": 0.00010679999999999998,
-      "loss": 0.0209,
+      "loss": 0.0217,
       "num_input_tokens_seen": 544444416,
       "step": 8310,
-      "train_runtime": 1526.8669,
-      "train_tokens_per_second": 356576.208
+      "train_runtime": 3705.3713,
+      "train_tokens_per_second": 146933.836
     },
     {
       "epoch": 1.8000865426222414,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.3046875,
       "learning_rate": 0.00010616842105263157,
-      "loss": 0.0185,
+      "loss": 0.0188,
       "num_input_tokens_seen": 545099776,
       "step": 8320,
-      "train_runtime": 1530.6897,
-      "train_tokens_per_second": 356113.831
+      "train_runtime": 3709.1839,
+      "train_tokens_per_second": 146959.49
     },
     {
       "epoch": 1.8022501081782778,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.298828125,
       "learning_rate": 0.00010553684210526315,
-      "loss": 0.0179,
+      "loss": 0.019,
       "num_input_tokens_seen": 545755136,
       "step": 8330,
-      "train_runtime": 1534.5113,
-      "train_tokens_per_second": 355654.039
+      "train_runtime": 3713.0027,
+      "train_tokens_per_second": 146984.848
     },
     {
       "epoch": 1.8044136737343142,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.3828125,
       "learning_rate": 0.00010490526315789473,
-      "loss": 0.0188,
+      "loss": 0.0198,
       "num_input_tokens_seen": 546410496,
       "step": 8340,
-      "train_runtime": 1538.3352,
-      "train_tokens_per_second": 355195.994
+      "train_runtime": 3716.8139,
+      "train_tokens_per_second": 147010.452
     },
     {
       "epoch": 1.8065772392903505,
-      "grad_norm": 0.322265625,
+      "grad_norm": 0.2490234375,
       "learning_rate": 0.0001042736842105263,
-      "loss": 0.0169,
+      "loss": 0.019,
       "num_input_tokens_seen": 547065856,
       "step": 8350,
-      "train_runtime": 1542.1463,
-      "train_tokens_per_second": 354743.156
+      "train_runtime": 3720.6281,
+      "train_tokens_per_second": 147035.889
     },
     {
       "epoch": 1.8087408048463869,
-      "grad_norm": 0.427734375,
+      "grad_norm": 0.345703125,
       "learning_rate": 0.00010364210526315789,
-      "loss": 0.0183,
+      "loss": 0.019,
       "num_input_tokens_seen": 547721216,
       "step": 8360,
-      "train_runtime": 1545.9682,
-      "train_tokens_per_second": 354290.09
+      "train_runtime": 3726.5016,
+      "train_tokens_per_second": 146980.002
     },
     {
       "epoch": 1.8109043704024232,
-      "grad_norm": 0.392578125,
+      "grad_norm": 0.439453125,
       "learning_rate": 0.00010301052631578947,
-      "loss": 0.019,
+      "loss": 0.0196,
       "num_input_tokens_seen": 548376576,
       "step": 8370,
-      "train_runtime": 1549.7896,
-      "train_tokens_per_second": 353839.38
+      "train_runtime": 3730.3164,
+      "train_tokens_per_second": 147005.379
     },
     {
       "epoch": 1.8130679359584594,
-      "grad_norm": 0.33984375,
+      "grad_norm": 0.3984375,
       "learning_rate": 0.00010237894736842104,
-      "loss": 0.0178,
+      "loss": 0.0187,
       "num_input_tokens_seen": 549031936,
       "step": 8380,
-      "train_runtime": 1553.6089,
-      "train_tokens_per_second": 353391.347
+      "train_runtime": 3734.1271,
+      "train_tokens_per_second": 147030.864
     },
     {
       "epoch": 1.815231501514496,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.39453125,
       "learning_rate": 0.00010174736842105262,
-      "loss": 0.0192,
+      "loss": 0.0202,
       "num_input_tokens_seen": 549687296,
       "step": 8390,
-      "train_runtime": 1557.4331,
-      "train_tokens_per_second": 352944.411
+      "train_runtime": 3737.9352,
+      "train_tokens_per_second": 147056.4
     },
     {
       "epoch": 1.817395067070532,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.3671875,
       "learning_rate": 0.0001011157894736842,
-      "loss": 0.0198,
+      "loss": 0.0208,
       "num_input_tokens_seen": 550342656,
       "step": 8400,
-      "train_runtime": 1561.2661,
-      "train_tokens_per_second": 352497.674
+      "train_runtime": 3741.7466,
+      "train_tokens_per_second": 147081.753
     },
     {
       "epoch": 1.817395067070532,
-      "eval_loss": 0.013265056535601616,
-      "eval_runtime": 1.7475,
-      "eval_samples_per_second": 18.312,
-      "eval_steps_per_second": 0.572,
+      "eval_loss": 0.010128295980393887,
+      "eval_runtime": 2.142,
+      "eval_samples_per_second": 14.939,
+      "eval_steps_per_second": 0.467,
       "num_input_tokens_seen": 550342656,
       "step": 8400
     },
     {
       "epoch": 1.817395067070532,
-      "eval_byte_accuracy": 0.995919881305638,
-      "eval_chrf": 97.15717684332516,
-      "eval_sacrebleu": 96.77994309150014,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 98.23489619471393,
+      "eval_sacrebleu": 97.07371933981932,
       "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 550342656,
-      "perplexity": 1.0133534277152885,
+      "perplexity": 1.010179760773598,
       "step": 8400
     },
     {
       "epoch": 1.8195586326265687,
-      "grad_norm": 0.51953125,
+      "grad_norm": 0.41796875,
       "learning_rate": 0.00010048421052631578,
-      "loss": 0.018,
+      "loss": 0.019,
       "num_input_tokens_seen": 550993920,
       "step": 8410,
-      "train_runtime": 1566.8283,
-      "train_tokens_per_second": 351661.961
+      "train_runtime": 3750.089,
+      "train_tokens_per_second": 146928.223
     },
     {
       "epoch": 1.8217221981826048,
-      "grad_norm": 0.318359375,
+      "grad_norm": 0.52734375,
       "learning_rate": 9.985263157894735e-05,
-      "loss": 0.016,
+      "loss": 0.0171,
       "num_input_tokens_seen": 551649280,
       "step": 8420,
-      "train_runtime": 1570.653,
-      "train_tokens_per_second": 351222.883
+      "train_runtime": 3753.8968,
+      "train_tokens_per_second": 146953.767
     },
     {
       "epoch": 1.8238857637386414,
-      "grad_norm": 0.349609375,
+      "grad_norm": 0.33203125,
       "learning_rate": 9.922105263157894e-05,
-      "loss": 0.0202,
+      "loss": 0.0212,
       "num_input_tokens_seen": 552304640,
       "step": 8430,
-      "train_runtime": 1574.475,
-      "train_tokens_per_second": 350786.55
+      "train_runtime": 3757.7027,
+      "train_tokens_per_second": 146979.333
     },
     {
       "epoch": 1.8260493292946776,
-      "grad_norm": 0.478515625,
+      "grad_norm": 0.41796875,
       "learning_rate": 9.858947368421052e-05,
-      "loss": 0.0191,
+      "loss": 0.0203,
       "num_input_tokens_seen": 552960000,
       "step": 8440,
-      "train_runtime": 1578.2914,
-      "train_tokens_per_second": 350353.564
+      "train_runtime": 3761.516,
+      "train_tokens_per_second": 147004.558
     },
     {
       "epoch": 1.8282128948507141,
-      "grad_norm": 0.341796875,
+      "grad_norm": 0.32421875,
       "learning_rate": 9.79578947368421e-05,
-      "loss": 0.0185,
+      "loss": 0.0195,
       "num_input_tokens_seen": 553615360,
       "step": 8450,
-      "train_runtime": 1582.1154,
-      "train_tokens_per_second": 349920.975
+      "train_runtime": 3765.3177,
+      "train_tokens_per_second": 147030.187
     },
     {
       "epoch": 1.8303764604067503,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.296875,
       "learning_rate": 9.732631578947367e-05,
-      "loss": 0.017,
+      "loss": 0.0182,
       "num_input_tokens_seen": 554270720,
       "step": 8460,
-      "train_runtime": 1585.9303,
-      "train_tokens_per_second": 349492.475
+      "train_runtime": 3771.3697,
+      "train_tokens_per_second": 146968.016
     },
     {
       "epoch": 1.8325400259627866,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.337890625,
       "learning_rate": 9.669473684210525e-05,
-      "loss": 0.016,
+      "loss": 0.0161,
       "num_input_tokens_seen": 554926080,
       "step": 8470,
-      "train_runtime": 1589.7507,
-      "train_tokens_per_second": 349064.849
+      "train_runtime": 3775.1871,
+      "train_tokens_per_second": 146993.0
     },
     {
       "epoch": 1.834703591518823,
-      "grad_norm": 0.359375,
+      "grad_norm": 0.318359375,
       "learning_rate": 9.606315789473684e-05,
-      "loss": 0.0166,
+      "loss": 0.0169,
       "num_input_tokens_seen": 555581440,
       "step": 8480,
-      "train_runtime": 1593.5726,
-      "train_tokens_per_second": 348638.922
+      "train_runtime": 3779.0067,
+      "train_tokens_per_second": 147017.849
     },
     {
       "epoch": 1.8368671570748594,
-      "grad_norm": 0.30859375,
+      "grad_norm": 0.380859375,
       "learning_rate": 9.543157894736841e-05,
-      "loss": 0.0194,
+      "loss": 0.0208,
       "num_input_tokens_seen": 556236800,
       "step": 8490,
-      "train_runtime": 1597.3881,
-      "train_tokens_per_second": 348216.432
+      "train_runtime": 3782.8155,
+      "train_tokens_per_second": 147043.07
     },
     {
       "epoch": 1.8390307226308957,
-      "grad_norm": 0.380859375,
+      "grad_norm": 0.474609375,
       "learning_rate": 9.479999999999999e-05,
-      "loss": 0.0194,
+      "loss": 0.0202,
       "num_input_tokens_seen": 556892160,
       "step": 8500,
-      "train_runtime": 1601.2204,
-      "train_tokens_per_second": 347792.313
+      "train_runtime": 3786.6204,
+      "train_tokens_per_second": 147068.389
     },
     {
       "epoch": 1.8390307226308957,
-      "eval_loss": 0.013057196512818336,
-      "eval_runtime": 1.7679,
-      "eval_samples_per_second": 18.101,
-      "eval_steps_per_second": 0.566,
+      "eval_loss": 0.01045873574912548,
+      "eval_runtime": 2.1368,
+      "eval_samples_per_second": 14.976,
+      "eval_steps_per_second": 0.468,
       "num_input_tokens_seen": 556892160,
       "step": 8500
     },
     {
       "epoch": 1.8390307226308957,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 556892160,
-      "perplexity": 1.0131428139385665,
+      "perplexity": 1.0105136194971873,
       "step": 8500
     },
     {
       "epoch": 1.841194288186932,
-      "grad_norm": 0.423828125,
+      "grad_norm": 0.462890625,
       "learning_rate": 9.416842105263157e-05,
-      "loss": 0.0176,
+      "loss": 0.02,
       "num_input_tokens_seen": 557547520,
       "step": 8510,
-      "train_runtime": 1606.8812,
-      "train_tokens_per_second": 346974.938
+      "train_runtime": 3794.8809,
+      "train_tokens_per_second": 146920.954
     },
     {
       "epoch": 1.8433578537429685,
-      "grad_norm": 0.37109375,
+      "grad_norm": 0.345703125,
       "learning_rate": 9.353684210526315e-05,
-      "loss": 0.0168,
+      "loss": 0.0176,
       "num_input_tokens_seen": 558202880,
       "step": 8520,
-      "train_runtime": 1610.6957,
-      "train_tokens_per_second": 346560.113
+      "train_runtime": 3798.6923,
+      "train_tokens_per_second": 146946.064
     },
     {
       "epoch": 1.8455214192990046,
-      "grad_norm": 0.333984375,
+      "grad_norm": 0.349609375,
       "learning_rate": 9.290526315789473e-05,
-      "loss": 0.0196,
+      "loss": 0.0197,
       "num_input_tokens_seen": 558858240,
       "step": 8530,
-      "train_runtime": 1614.5112,
-      "train_tokens_per_second": 346147.017
+      "train_runtime": 3802.5052,
+      "train_tokens_per_second": 146971.064
     },
     {
       "epoch": 1.8476849848550412,
-      "grad_norm": 0.38671875,
+      "grad_norm": 0.314453125,
       "learning_rate": 9.22736842105263e-05,
-      "loss": 0.0173,
+      "loss": 0.0176,
       "num_input_tokens_seen": 559513600,
       "step": 8540,
-      "train_runtime": 1618.3296,
-      "train_tokens_per_second": 345735.256
+      "train_runtime": 3806.3149,
+      "train_tokens_per_second": 146996.142
     },
     {
       "epoch": 1.8498485504110773,
-      "grad_norm": 0.306640625,
+      "grad_norm": 0.34375,
       "learning_rate": 9.164210526315789e-05,
-      "loss": 0.018,
+      "loss": 0.0177,
       "num_input_tokens_seen": 560168960,
       "step": 8550,
-      "train_runtime": 1622.1481,
-      "train_tokens_per_second": 345325.416
+      "train_runtime": 3810.127,
+      "train_tokens_per_second": 147021.073
     },
     {
       "epoch": 1.852012115967114,
-      "grad_norm": 0.267578125,
+      "grad_norm": 0.294921875,
       "learning_rate": 9.101052631578946e-05,
-      "loss": 0.0176,
+      "loss": 0.0187,
       "num_input_tokens_seen": 560824320,
       "step": 8560,
-      "train_runtime": 1625.9619,
-      "train_tokens_per_second": 344918.496
+      "train_runtime": 3816.3836,
+      "train_tokens_per_second": 146951.77
     },
     {
       "epoch": 1.85417568152315,
-      "grad_norm": 0.248046875,
+      "grad_norm": 0.298828125,
       "learning_rate": 9.037894736842105e-05,
-      "loss": 0.0173,
+      "loss": 0.018,
       "num_input_tokens_seen": 561479680,
       "step": 8570,
-      "train_runtime": 1629.7856,
-      "train_tokens_per_second": 344511.37
+      "train_runtime": 3820.1956,
+      "train_tokens_per_second": 146976.682
     },
     {
       "epoch": 1.8563392470791866,
-      "grad_norm": 0.37890625,
+      "grad_norm": 0.328125,
       "learning_rate": 8.974736842105262e-05,
-      "loss": 0.0166,
+      "loss": 0.0181,
       "num_input_tokens_seen": 562135040,
       "step": 8580,
-      "train_runtime": 1633.6031,
-      "train_tokens_per_second": 344107.479
+      "train_runtime": 3824.0034,
+      "train_tokens_per_second": 147001.709
     },
     {
       "epoch": 1.8585028126352228,
-      "grad_norm": 0.39453125,
+      "grad_norm": 0.384765625,
       "learning_rate": 8.91157894736842e-05,
-      "loss": 0.0186,
+      "loss": 0.0198,
       "num_input_tokens_seen": 562790400,
       "step": 8590,
-      "train_runtime": 1637.4214,
-      "train_tokens_per_second": 343705.288
+      "train_runtime": 3827.8147,
+      "train_tokens_per_second": 147026.553
     },
     {
       "epoch": 1.8606663781912594,
-      "grad_norm": 0.4140625,
+      "grad_norm": 0.353515625,
       "learning_rate": 8.848421052631578e-05,
-      "loss": 0.0196,
+      "loss": 0.0206,
       "num_input_tokens_seen": 563445760,
       "step": 8600,
-      "train_runtime": 1641.2424,
-      "train_tokens_per_second": 343304.424
+      "train_runtime": 3831.6313,
+      "train_tokens_per_second": 147051.143
     },
     {
       "epoch": 1.8606663781912594,
-      "eval_loss": 0.012696781195700169,
-      "eval_runtime": 1.759,
-      "eval_samples_per_second": 18.192,
-      "eval_steps_per_second": 0.568,
+      "eval_loss": 0.010154013521969318,
+      "eval_runtime": 2.1562,
+      "eval_samples_per_second": 14.841,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 563445760,
       "step": 8600
     },
     {
       "epoch": 1.8606663781912594,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.70457548553854,
-      "eval_sacrebleu": 97.11458584551553,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 98.32407535469744,
+      "eval_sacrebleu": 97.07371933981932,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 563445760,
-      "perplexity": 1.012777727545311,
+      "perplexity": 1.0102057404476597,
       "step": 8600
     },
     {
       "epoch": 1.8628299437472955,
-      "grad_norm": 0.337890625,
+      "grad_norm": 0.388671875,
       "learning_rate": 8.785263157894737e-05,
-      "loss": 0.0167,
+      "loss": 0.0183,
       "num_input_tokens_seen": 564101120,
       "step": 8610,
-      "train_runtime": 1646.8299,
-      "train_tokens_per_second": 342537.572
+      "train_runtime": 3839.8519,
+      "train_tokens_per_second": 146906.999
     },
     {
       "epoch": 1.8649935093033319,
       "grad_norm": 0.322265625,
       "learning_rate": 8.722105263157894e-05,
-      "loss": 0.0172,
+      "loss": 0.0187,
       "num_input_tokens_seen": 564756480,
       "step": 8620,
-      "train_runtime": 1650.639,
-      "train_tokens_per_second": 342144.146
+      "train_runtime": 3843.6679,
+      "train_tokens_per_second": 146931.653
     },
     {
       "epoch": 1.8671570748593682,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.314453125,
       "learning_rate": 8.658947368421052e-05,
-      "loss": 0.0163,
+      "loss": 0.0165,
       "num_input_tokens_seen": 565411840,
       "step": 8630,
-      "train_runtime": 1654.4626,
-      "train_tokens_per_second": 341749.55
+      "train_runtime": 3847.4757,
+      "train_tokens_per_second": 146956.572
     },
     {
       "epoch": 1.8693206404154046,
-      "grad_norm": 0.38671875,
+      "grad_norm": 0.330078125,
       "learning_rate": 8.59578947368421e-05,
-      "loss": 0.0186,
+      "loss": 0.0191,
       "num_input_tokens_seen": 566063104,
       "step": 8640,
-      "train_runtime": 1658.2639,
-      "train_tokens_per_second": 341358.878
+      "train_runtime": 3851.2684,
+      "train_tokens_per_second": 146980.955
     },
     {
       "epoch": 1.871484205971441,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.353515625,
       "learning_rate": 8.532631578947369e-05,
-      "loss": 0.018,
+      "loss": 0.0185,
       "num_input_tokens_seen": 566718464,
       "step": 8650,
-      "train_runtime": 1662.0809,
-      "train_tokens_per_second": 340969.238
+      "train_runtime": 3855.0844,
+      "train_tokens_per_second": 147005.461
     },
     {
       "epoch": 1.8736477715274773,
-      "grad_norm": 0.390625,
+      "grad_norm": 0.30859375,
       "learning_rate": 8.469473684210525e-05,
-      "loss": 0.0189,
+      "loss": 0.0203,
       "num_input_tokens_seen": 567373824,
       "step": 8660,
-      "train_runtime": 1665.898,
-      "train_tokens_per_second": 340581.365
+      "train_runtime": 3861.1356,
+      "train_tokens_per_second": 146944.805
     },
     {
       "epoch": 1.8758113370835137,
-      "grad_norm": 0.45703125,
+      "grad_norm": 0.3671875,
       "learning_rate": 8.406315789473683e-05,
-      "loss": 0.0372,
+      "loss": 0.0365,
       "num_input_tokens_seen": 568029184,
       "step": 8670,
-      "train_runtime": 1669.7217,
-      "train_tokens_per_second": 340193.919
+      "train_runtime": 3864.9452,
+      "train_tokens_per_second": 146969.53
     },
     {
       "epoch": 1.87797490263955,
-      "grad_norm": 0.41796875,
+      "grad_norm": 0.35546875,
       "learning_rate": 8.343157894736842e-05,
-      "loss": 0.0174,
+      "loss": 0.0185,
       "num_input_tokens_seen": 568684544,
       "step": 8680,
-      "train_runtime": 1673.5403,
-      "train_tokens_per_second": 339809.298
+      "train_runtime": 3868.7494,
+      "train_tokens_per_second": 146994.411
     },
     {
       "epoch": 1.8801384681955864,
-      "grad_norm": 0.400390625,
+      "grad_norm": 0.3671875,
       "learning_rate": 8.28e-05,
-      "loss": 0.0188,
+      "loss": 0.0189,
       "num_input_tokens_seen": 569339904,
       "step": 8690,
-      "train_runtime": 1677.3658,
-      "train_tokens_per_second": 339425.007
+      "train_runtime": 3872.5576,
+      "train_tokens_per_second": 147019.093
     },
     {
       "epoch": 1.8823020337516225,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.41015625,
       "learning_rate": 8.216842105263157e-05,
-      "loss": 0.0155,
+      "loss": 0.0163,
       "num_input_tokens_seen": 569995264,
       "step": 8700,
-      "train_runtime": 1681.1869,
-      "train_tokens_per_second": 339043.374
+      "train_runtime": 3876.3574,
+      "train_tokens_per_second": 147044.043
     },
     {
       "epoch": 1.8823020337516225,
-      "eval_loss": 0.012751961126923561,
-      "eval_runtime": 1.7169,
-      "eval_samples_per_second": 18.638,
-      "eval_steps_per_second": 0.582,
+      "eval_loss": 0.010016790591180325,
+      "eval_runtime": 2.1432,
+      "eval_samples_per_second": 14.931,
+      "eval_steps_per_second": 0.467,
       "num_input_tokens_seen": 569995264,
       "step": 8700
     },
     {
       "epoch": 1.8823020337516225,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.93277340374337,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 569995264,
-      "perplexity": 1.0128336140925551,
+      "perplexity": 1.0100671265659746,
       "step": 8700
     },
     {
       "epoch": 1.8844655993076591,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.2734375,
       "learning_rate": 8.153684210526315e-05,
       "loss": 0.0182,
       "num_input_tokens_seen": 570650624,
       "step": 8710,
-      "train_runtime": 1686.7192,
-      "train_tokens_per_second": 338319.878
+      "train_runtime": 3884.7322,
+      "train_tokens_per_second": 146895.744
     },
     {
       "epoch": 1.8866291648636953,
-      "grad_norm": 0.38671875,
+      "grad_norm": 0.380859375,
       "learning_rate": 8.090526315789474e-05,
-      "loss": 0.017,
+      "loss": 0.0173,
       "num_input_tokens_seen": 571305984,
       "step": 8720,
-      "train_runtime": 1690.52,
-      "train_tokens_per_second": 337946.886
+      "train_runtime": 3888.5094,
+      "train_tokens_per_second": 146921.59
     },
     {
       "epoch": 1.8887927304197318,
-      "grad_norm": 0.26953125,
+      "grad_norm": 0.302734375,
       "learning_rate": 8.02736842105263e-05,
-      "loss": 0.0167,
+      "loss": 0.0171,
       "num_input_tokens_seen": 571961344,
       "step": 8730,
-      "train_runtime": 1694.3384,
-      "train_tokens_per_second": 337572.087
+      "train_runtime": 3892.3271,
+      "train_tokens_per_second": 146945.86
     },
     {
       "epoch": 1.890956295975768,
-      "grad_norm": 0.365234375,
+      "grad_norm": 0.423828125,
       "learning_rate": 7.964210526315788e-05,
-      "loss": 0.0187,
+      "loss": 0.0193,
       "num_input_tokens_seen": 572616704,
       "step": 8740,
-      "train_runtime": 1698.1568,
-      "train_tokens_per_second": 337198.951
+      "train_runtime": 3896.1376,
+      "train_tokens_per_second": 146970.351
     },
     {
       "epoch": 1.8931198615318046,
-      "grad_norm": 0.26171875,
+      "grad_norm": 0.28125,
       "learning_rate": 7.901052631578947e-05,
-      "loss": 0.0187,
+      "loss": 0.0189,
       "num_input_tokens_seen": 573272064,
       "step": 8750,
-      "train_runtime": 1701.9803,
-      "train_tokens_per_second": 336826.499
+      "train_runtime": 3899.9488,
+      "train_tokens_per_second": 146994.766
     },
     {
       "epoch": 1.8952834270878407,
-      "grad_norm": 0.302734375,
+      "grad_norm": 0.330078125,
       "learning_rate": 7.837894736842106e-05,
-      "loss": 0.0168,
+      "loss": 0.0174,
       "num_input_tokens_seen": 573927424,
       "step": 8760,
-      "train_runtime": 1705.81,
-      "train_tokens_per_second": 336454.491
+      "train_runtime": 3906.0027,
+      "train_tokens_per_second": 146934.724
     },
     {
       "epoch": 1.897446992643877,
-      "grad_norm": 0.29296875,
+      "grad_norm": 0.318359375,
       "learning_rate": 7.774736842105262e-05,
-      "loss": 0.0194,
+      "loss": 0.0197,
       "num_input_tokens_seen": 574582784,
       "step": 8770,
-      "train_runtime": 1709.6347,
-      "train_tokens_per_second": 336085.119
+      "train_runtime": 3909.8136,
+      "train_tokens_per_second": 146959.123
     },
     {
       "epoch": 1.8996105581999134,
-      "grad_norm": 0.30078125,
+      "grad_norm": 0.4453125,
       "learning_rate": 7.71157894736842e-05,
-      "loss": 0.0167,
+      "loss": 0.0175,
       "num_input_tokens_seen": 575238144,
       "step": 8780,
-      "train_runtime": 1713.4531,
-      "train_tokens_per_second": 335718.631
+      "train_runtime": 3913.6229,
+      "train_tokens_per_second": 146983.538
     },
     {
       "epoch": 1.9017741237559498,
-      "grad_norm": 0.439453125,
+      "grad_norm": 0.421875,
       "learning_rate": 7.648421052631579e-05,
-      "loss": 0.0195,
+      "loss": 0.0204,
       "num_input_tokens_seen": 575893504,
       "step": 8790,
-      "train_runtime": 1717.2793,
-      "train_tokens_per_second": 335352.256
+      "train_runtime": 3917.4364,
+      "train_tokens_per_second": 147007.747
     },
     {
       "epoch": 1.9039376893119861,
-      "grad_norm": 0.341796875,
+      "grad_norm": 0.392578125,
       "learning_rate": 7.585263157894735e-05,
-      "loss": 0.0195,
+      "loss": 0.0205,
       "num_input_tokens_seen": 576548864,
       "step": 8800,
-      "train_runtime": 1721.0838,
-      "train_tokens_per_second": 334991.742
+      "train_runtime": 3921.2307,
+      "train_tokens_per_second": 147032.631
     },
     {
       "epoch": 1.9039376893119861,
-      "eval_loss": 0.01311762910336256,
-      "eval_runtime": 1.7389,
-      "eval_samples_per_second": 18.402,
-      "eval_steps_per_second": 0.575,
+      "eval_loss": 0.010483089834451675,
+      "eval_runtime": 2.1523,
+      "eval_samples_per_second": 14.868,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 576548864,
       "step": 8800
     },
     {
       "epoch": 1.9039376893119861,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.87702445919373,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 576548864,
-      "perplexity": 1.0132040426334898,
+      "perplexity": 1.0105382299317809,
       "step": 8800
     },
     {
       "epoch": 1.9061012548680225,
-      "grad_norm": 0.310546875,
+      "grad_norm": 0.326171875,
       "learning_rate": 7.522105263157894e-05,
-      "loss": 0.0172,
+      "loss": 0.0181,
       "num_input_tokens_seen": 577204224,
       "step": 8810,
-      "train_runtime": 1726.6618,
-      "train_tokens_per_second": 334289.099
+      "train_runtime": 3929.4565,
+      "train_tokens_per_second": 146891.618
     },
     {
       "epoch": 1.9082648204240589,
-      "grad_norm": 0.5859375,
+      "grad_norm": 0.357421875,
       "learning_rate": 7.458947368421052e-05,
-      "loss": 0.0173,
+      "loss": 0.0184,
       "num_input_tokens_seen": 577859584,
       "step": 8820,
-      "train_runtime": 1730.4841,
-      "train_tokens_per_second": 333929.433
+      "train_runtime": 3933.2587,
+      "train_tokens_per_second": 146916.243
     },
     {
       "epoch": 1.9104283859800952,
-      "grad_norm": 0.314453125,
+      "grad_norm": 0.365234375,
       "learning_rate": 7.39578947368421e-05,
-      "loss": 0.0176,
+      "loss": 0.0177,
       "num_input_tokens_seen": 578514944,
       "step": 8830,
-      "train_runtime": 1734.3033,
-      "train_tokens_per_second": 333571.965
+      "train_runtime": 3937.0791,
+      "train_tokens_per_second": 146940.136
     },
     {
       "epoch": 1.9125919515361316,
-      "grad_norm": 0.4296875,
+      "grad_norm": 0.376953125,
       "learning_rate": 7.332631578947368e-05,
-      "loss": 0.0178,
+      "loss": 0.0181,
       "num_input_tokens_seen": 579170304,
       "step": 8840,
-      "train_runtime": 1738.1235,
-      "train_tokens_per_second": 333215.847
+      "train_runtime": 3940.8873,
+      "train_tokens_per_second": 146964.442
     },
     {
       "epoch": 1.9147555170921677,
-      "grad_norm": 0.34375,
+      "grad_norm": 0.390625,
       "learning_rate": 7.269473684210525e-05,
-      "loss": 0.0177,
+      "loss": 0.0181,
       "num_input_tokens_seen": 579821568,
       "step": 8850,
-      "train_runtime": 1741.9307,
-      "train_tokens_per_second": 332861.452
+      "train_runtime": 3944.6768,
+      "train_tokens_per_second": 146988.357
     },
     {
       "epoch": 1.9169190826482043,
-      "grad_norm": 0.263671875,
+      "grad_norm": 0.419921875,
       "learning_rate": 7.206315789473684e-05,
-      "loss": 0.0173,
+      "loss": 0.0174,
       "num_input_tokens_seen": 580476928,
       "step": 8860,
-      "train_runtime": 1745.7567,
-      "train_tokens_per_second": 332507.342
+      "train_runtime": 3950.7316,
+      "train_tokens_per_second": 146928.971
     },
     {
       "epoch": 1.9190826482042405,
-      "grad_norm": 0.283203125,
+      "grad_norm": 0.330078125,
       "learning_rate": 7.143157894736841e-05,
-      "loss": 0.0396,
+      "loss": 0.0412,
       "num_input_tokens_seen": 581132288,
       "step": 8870,
-      "train_runtime": 1749.5799,
-      "train_tokens_per_second": 332155.333
+      "train_runtime": 3954.5446,
+      "train_tokens_per_second": 146953.023
     },
     {
       "epoch": 1.921246213760277,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.3125,
       "learning_rate": 7.079999999999999e-05,
       "loss": 0.0205,
       "num_input_tokens_seen": 581787648,
       "step": 8880,
-      "train_runtime": 1753.4027,
-      "train_tokens_per_second": 331804.921
+      "train_runtime": 3958.3515,
+      "train_tokens_per_second": 146977.258
     },
     {
       "epoch": 1.9234097793163132,
-      "grad_norm": 0.38671875,
+      "grad_norm": 0.29296875,
       "learning_rate": 7.016842105263157e-05,
-      "loss": 0.0184,
+      "loss": 0.0192,
       "num_input_tokens_seen": 582443008,
       "step": 8890,
-      "train_runtime": 1757.2196,
-      "train_tokens_per_second": 331457.158
+      "train_runtime": 3962.164,
+      "train_tokens_per_second": 147001.235
     },
     {
       "epoch": 1.9255733448723498,
-      "grad_norm": 0.416015625,
+      "grad_norm": 0.40234375,
       "learning_rate": 6.953684210526315e-05,
-      "loss": 0.0197,
+      "loss": 0.0214,
       "num_input_tokens_seen": 583098368,
       "step": 8900,
-      "train_runtime": 1761.037,
-      "train_tokens_per_second": 331110.803
+      "train_runtime": 3965.9697,
+      "train_tokens_per_second": 147025.422
     },
     {
       "epoch": 1.9255733448723498,
-      "eval_loss": 0.013076567091047764,
-      "eval_runtime": 1.7602,
-      "eval_samples_per_second": 18.18,
-      "eval_steps_per_second": 0.568,
+      "eval_loss": 0.01048352662473917,
+      "eval_runtime": 2.1294,
+      "eval_samples_per_second": 15.028,
+      "eval_steps_per_second": 0.47,
       "num_input_tokens_seen": 583098368,
       "step": 8900
     },
     {
       "epoch": 1.9255733448723498,
-      "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 98.33791810109832,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_byte_accuracy": 0.9966617210682492,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 583098368,
-      "perplexity": 1.013162439290778,
+      "perplexity": 1.0105386713251614,
       "step": 8900
     },
     {
       "epoch": 1.927736910428386,
-      "grad_norm": 0.419921875,
+      "grad_norm": 0.392578125,
       "learning_rate": 6.890526315789473e-05,
-      "loss": 0.0175,
+      "loss": 0.019,
       "num_input_tokens_seen": 583753728,
       "step": 8910,
-      "train_runtime": 1766.6312,
-      "train_tokens_per_second": 330433.265
+      "train_runtime": 3974.3517,
+      "train_tokens_per_second": 146880.241
     },
     {
       "epoch": 1.9299004759844225,
-      "grad_norm": 0.2734375,
+      "grad_norm": 0.35546875,
       "learning_rate": 6.82736842105263e-05,
-      "loss": 0.0169,
+      "loss": 0.0192,
       "num_input_tokens_seen": 584409088,
       "step": 8920,
-      "train_runtime": 1770.4611,
-      "train_tokens_per_second": 330088.638
+      "train_runtime": 3978.1649,
+      "train_tokens_per_second": 146904.189
     },
     {
       "epoch": 1.9320640415404586,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.3125,
       "learning_rate": 6.764210526315789e-05,
-      "loss": 0.0171,
+      "loss": 0.0174,
       "num_input_tokens_seen": 585064448,
       "step": 8930,
-      "train_runtime": 1774.2831,
-      "train_tokens_per_second": 329746.95
+      "train_runtime": 3981.9831,
+      "train_tokens_per_second": 146927.909
     },
     {
       "epoch": 1.934227607096495,
-      "grad_norm": 0.48046875,
+      "grad_norm": 0.318359375,
       "learning_rate": 6.701052631578946e-05,
-      "loss": 0.0171,
+      "loss": 0.0183,
       "num_input_tokens_seen": 585719808,
       "step": 8940,
-      "train_runtime": 1778.1038,
-      "train_tokens_per_second": 329406.981
+      "train_runtime": 3985.7936,
+      "train_tokens_per_second": 146951.868
     },
     {
       "epoch": 1.9363911726525314,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.376953125,
       "learning_rate": 6.637894736842104e-05,
-      "loss": 0.0159,
+      "loss": 0.0169,
       "num_input_tokens_seen": 586375168,
       "step": 8950,
-      "train_runtime": 1781.9085,
-      "train_tokens_per_second": 329071.433
+      "train_runtime": 3989.5895,
+      "train_tokens_per_second": 146976.314
     },
     {
       "epoch": 1.9385547382085677,
-      "grad_norm": 0.333984375,
+      "grad_norm": 0.3359375,
       "learning_rate": 6.574736842105262e-05,
-      "loss": 0.019,
+      "loss": 0.0195,
       "num_input_tokens_seen": 587030528,
       "step": 8960,
-      "train_runtime": 1785.7304,
-      "train_tokens_per_second": 328734.137
+      "train_runtime": 3995.6481,
+      "train_tokens_per_second": 146917.474
     },
     {
       "epoch": 1.940718303764604,
-      "grad_norm": 0.337890625,
+      "grad_norm": 0.359375,
       "learning_rate": 6.51157894736842e-05,
-      "loss": 0.0182,
+      "loss": 0.0188,
       "num_input_tokens_seen": 587685888,
       "step": 8970,
-      "train_runtime": 1789.5563,
-      "train_tokens_per_second": 328397.543
+      "train_runtime": 3999.455,
+      "train_tokens_per_second": 146941.492
     },
     {
       "epoch": 1.9428818693206404,
-      "grad_norm": 0.55859375,
+      "grad_norm": 0.400390625,
       "learning_rate": 6.448421052631578e-05,
-      "loss": 0.0203,
+      "loss": 0.021,
       "num_input_tokens_seen": 588341248,
       "step": 8980,
-      "train_runtime": 1793.385,
-      "train_tokens_per_second": 328061.874
+      "train_runtime": 4003.2606,
+      "train_tokens_per_second": 146965.513
     },
     {
       "epoch": 1.9450454348766768,
-      "grad_norm": 0.4609375,
+      "grad_norm": 0.296875,
       "learning_rate": 6.385263157894736e-05,
-      "loss": 0.0177,
+      "loss": 0.019,
       "num_input_tokens_seen": 588996608,
       "step": 8990,
-      "train_runtime": 1797.2072,
-      "train_tokens_per_second": 327728.829
+      "train_runtime": 4007.0732,
+      "train_tokens_per_second": 146989.231
     },
     {
       "epoch": 1.9472090004327132,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.380859375,
       "learning_rate": 6.322105263157894e-05,
-      "loss": 0.0186,
+      "loss": 0.019,
       "num_input_tokens_seen": 589651968,
       "step": 9000,
-      "train_runtime": 1801.0244,
-      "train_tokens_per_second": 327398.099
+      "train_runtime": 4010.88,
+      "train_tokens_per_second": 147013.117
     },
     {
       "epoch": 1.9472090004327132,
-      "eval_loss": 0.013011611066758633,
-      "eval_runtime": 1.7305,
-      "eval_samples_per_second": 18.492,
-      "eval_steps_per_second": 0.578,
+      "eval_loss": 0.010117894969880581,
+      "eval_runtime": 2.1514,
+      "eval_samples_per_second": 14.874,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 589651968,
       "step": 9000
     },
     {
       "epoch": 1.9472090004327132,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.8443713803999,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 589651968,
-      "perplexity": 1.013096630424127,
+      "perplexity": 1.010169253937927,
       "step": 9000
     },
     {
       "epoch": 1.9493725659887495,
-      "grad_norm": 0.41796875,
+      "grad_norm": 0.361328125,
       "learning_rate": 6.258947368421051e-05,
-      "loss": 0.0176,
+      "loss": 0.0174,
       "num_input_tokens_seen": 590307328,
       "step": 9010,
-      "train_runtime": 1806.6431,
-      "train_tokens_per_second": 326742.641
+      "train_runtime": 4019.172,
+      "train_tokens_per_second": 146872.872
     },
     {
       "epoch": 1.9515361315447857,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.32421875,
       "learning_rate": 6.19578947368421e-05,
-      "loss": 0.0168,
+      "loss": 0.0177,
       "num_input_tokens_seen": 590962688,
       "step": 9020,
-      "train_runtime": 1810.4672,
-      "train_tokens_per_second": 326414.463
+      "train_runtime": 4022.9902,
+      "train_tokens_per_second": 146896.377
     },
     {
       "epoch": 1.9536996971008223,
-      "grad_norm": 0.35546875,
+      "grad_norm": 0.302734375,
       "learning_rate": 6.132631578947367e-05,
-      "loss": 0.0178,
+      "loss": 0.0182,
       "num_input_tokens_seen": 591618048,
       "step": 9030,
-      "train_runtime": 1814.284,
-      "train_tokens_per_second": 326088.998
+      "train_runtime": 4026.7959,
+      "train_tokens_per_second": 146920.297
     },
     {
       "epoch": 1.9558632626568584,
-      "grad_norm": 0.375,
+      "grad_norm": 0.359375,
       "learning_rate": 6.0694736842105254e-05,
-      "loss": 0.0142,
+      "loss": 0.0158,
       "num_input_tokens_seen": 592273408,
       "step": 9040,
-      "train_runtime": 1818.0992,
-      "train_tokens_per_second": 325765.181
+      "train_runtime": 4030.5968,
+      "train_tokens_per_second": 146944.346
     },
     {
       "epoch": 1.958026828212895,
-      "grad_norm": 0.3671875,
+      "grad_norm": 0.349609375,
       "learning_rate": 6.006315789473684e-05,
-      "loss": 0.0167,
+      "loss": 0.0172,
       "num_input_tokens_seen": 592928768,
       "step": 9050,
-      "train_runtime": 1821.9234,
-      "train_tokens_per_second": 325441.099
+      "train_runtime": 4034.4087,
+      "train_tokens_per_second": 146967.95
     },
     {
       "epoch": 1.9601903937689311,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.337890625,
       "learning_rate": 5.943157894736841e-05,
-      "loss": 0.0179,
+      "loss": 0.0184,
       "num_input_tokens_seen": 593580032,
       "step": 9060,
-      "train_runtime": 1825.7215,
-      "train_tokens_per_second": 325120.802
+      "train_runtime": 4040.6374,
+      "train_tokens_per_second": 146902.572
     },
     {
       "epoch": 1.9623539593249677,
-      "grad_norm": 0.51171875,
+      "grad_norm": 0.390625,
       "learning_rate": 5.88e-05,
-      "loss": 0.0164,
+      "loss": 0.0173,
       "num_input_tokens_seen": 594235392,
       "step": 9070,
-      "train_runtime": 1829.5267,
-      "train_tokens_per_second": 324802.796
+      "train_runtime": 4044.4294,
+      "train_tokens_per_second": 146926.88
     },
     {
       "epoch": 1.9645175248810038,
-      "grad_norm": 0.294921875,
+      "grad_norm": 0.3125,
       "learning_rate": 5.816842105263157e-05,
-      "loss": 0.0171,
+      "loss": 0.0181,
       "num_input_tokens_seen": 594890752,
       "step": 9080,
-      "train_runtime": 1833.3509,
-      "train_tokens_per_second": 324482.76
+      "train_runtime": 4048.2382,
+      "train_tokens_per_second": 146950.53
     },
     {
       "epoch": 1.9666810904370402,
-      "grad_norm": 0.3203125,
+      "grad_norm": 0.296875,
       "learning_rate": 5.753684210526316e-05,
-      "loss": 0.0158,
+      "loss": 0.0167,
       "num_input_tokens_seen": 595546112,
       "step": 9090,
-      "train_runtime": 1837.1771,
-      "train_tokens_per_second": 324163.685
+      "train_runtime": 4052.0487,
+      "train_tokens_per_second": 146974.077
     },
     {
       "epoch": 1.9688446559930766,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.447265625,
       "learning_rate": 5.690526315789473e-05,
-      "loss": 0.0175,
+      "loss": 0.0189,
       "num_input_tokens_seen": 596201472,
       "step": 9100,
-      "train_runtime": 1841.0114,
-      "train_tokens_per_second": 323844.539
+      "train_runtime": 4055.8497,
+      "train_tokens_per_second": 146997.921
     },
     {
       "epoch": 1.9688446559930766,
-      "eval_loss": 0.012902812100946903,
-      "eval_runtime": 1.721,
-      "eval_samples_per_second": 18.594,
-      "eval_steps_per_second": 0.581,
+      "eval_loss": 0.010281143710017204,
+      "eval_runtime": 2.1304,
+      "eval_samples_per_second": 15.021,
+      "eval_steps_per_second": 0.469,
       "num_input_tokens_seen": 596201472,
       "step": 9100
     },
     {
       "epoch": 1.9688446559930766,
-      "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 596201472,
-      "perplexity": 1.0129864125543733,
+      "perplexity": 1.0103341762572713,
       "step": 9100
     },
     {
       "epoch": 1.971008221549113,
-      "grad_norm": 0.375,
+      "grad_norm": 0.41796875,
       "learning_rate": 5.6273684210526304e-05,
-      "loss": 0.021,
+      "loss": 0.0219,
       "num_input_tokens_seen": 596856832,
       "step": 9110,
-      "train_runtime": 1846.5729,
-      "train_tokens_per_second": 323224.081
+      "train_runtime": 4064.0593,
+      "train_tokens_per_second": 146862.236
     },
     {
       "epoch": 1.9731717871051493,
-      "grad_norm": 0.408203125,
+      "grad_norm": 0.267578125,
       "learning_rate": 5.564210526315789e-05,
-      "loss": 0.0165,
+      "loss": 0.0175,
       "num_input_tokens_seen": 597512192,
       "step": 9120,
-      "train_runtime": 1850.3877,
-      "train_tokens_per_second": 322911.89
+      "train_runtime": 4067.8568,
+      "train_tokens_per_second": 146886.241
     },
     {
       "epoch": 1.9753353526611857,
-      "grad_norm": 0.275390625,
+      "grad_norm": 0.263671875,
       "learning_rate": 5.5010526315789464e-05,
-      "loss": 0.0182,
+      "loss": 0.0195,
       "num_input_tokens_seen": 598167552,
       "step": 9130,
-      "train_runtime": 1854.2112,
-      "train_tokens_per_second": 322599.478
+      "train_runtime": 4071.6745,
+      "train_tokens_per_second": 146909.473
     },
     {
       "epoch": 1.977498918217222,
-      "grad_norm": 0.2431640625,
+      "grad_norm": 0.326171875,
       "learning_rate": 5.437894736842105e-05,
-      "loss": 0.0161,
+      "loss": 0.0175,
       "num_input_tokens_seen": 598822912,
       "step": 9140,
-      "train_runtime": 1858.0463,
-      "train_tokens_per_second": 322286.325
+      "train_runtime": 4075.4996,
+      "train_tokens_per_second": 146932.393
     },
     {
       "epoch": 1.9796624837732584,
-      "grad_norm": 0.404296875,
+      "grad_norm": 0.412109375,
       "learning_rate": 5.374736842105262e-05,
-      "loss": 0.0169,
+      "loss": 0.0181,
       "num_input_tokens_seen": 599478272,
       "step": 9150,
-      "train_runtime": 1861.8665,
-      "train_tokens_per_second": 321977.04
+      "train_runtime": 4079.3078,
+      "train_tokens_per_second": 146955.88
     },
     {
       "epoch": 1.9818260493292947,
-      "grad_norm": 0.462890625,
+      "grad_norm": 0.322265625,
       "learning_rate": 5.311578947368421e-05,
-      "loss": 0.017,
+      "loss": 0.0182,
       "num_input_tokens_seen": 600133632,
       "step": 9160,
-      "train_runtime": 1865.6883,
-      "train_tokens_per_second": 321668.749
+      "train_runtime": 4085.3923,
+      "train_tokens_per_second": 146897.429
     },
     {
       "epoch": 1.9839896148853309,
-      "grad_norm": 0.416015625,
+      "grad_norm": 0.396484375,
       "learning_rate": 5.248421052631578e-05,
-      "loss": 0.0189,
+      "loss": 0.0194,
       "num_input_tokens_seen": 600788992,
       "step": 9170,
-      "train_runtime": 1869.5058,
-      "train_tokens_per_second": 321362.465
+      "train_runtime": 4089.2065,
+      "train_tokens_per_second": 146920.679
     },
     {
       "epoch": 1.9861531804413675,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.25390625,
       "learning_rate": 5.185263157894737e-05,
-      "loss": 0.0188,
+      "loss": 0.02,
       "num_input_tokens_seen": 601444352,
       "step": 9180,
-      "train_runtime": 1873.3305,
-      "train_tokens_per_second": 321056.196
+      "train_runtime": 4093.02,
+      "train_tokens_per_second": 146943.909
     },
     {
       "epoch": 1.9883167459974036,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.357421875,
       "learning_rate": 5.122105263157894e-05,
-      "loss": 0.0179,
+      "loss": 0.019,
       "num_input_tokens_seen": 602099712,
       "step": 9190,
-      "train_runtime": 1877.1578,
-      "train_tokens_per_second": 320750.71
+      "train_runtime": 4096.8265,
+      "train_tokens_per_second": 146967.346
     },
     {
       "epoch": 1.9904803115534402,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.384765625,
       "learning_rate": 5.058947368421052e-05,
-      "loss": 0.0154,
+      "loss": 0.0176,
       "num_input_tokens_seen": 602755072,
       "step": 9200,
-      "train_runtime": 1880.981,
-      "train_tokens_per_second": 320447.182
+      "train_runtime": 4100.6457,
+      "train_tokens_per_second": 146990.281
     },
     {
       "epoch": 1.9904803115534402,
-      "eval_loss": 0.012645396403968334,
-      "eval_runtime": 1.7339,
-      "eval_samples_per_second": 18.456,
-      "eval_steps_per_second": 0.577,
+      "eval_loss": 0.010350430384278297,
+      "eval_runtime": 2.1379,
+      "eval_samples_per_second": 14.968,
+      "eval_steps_per_second": 0.468,
       "num_input_tokens_seen": 602755072,
       "step": 9200
     },
     {
       "epoch": 1.9904803115534402,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 602755072,
-      "perplexity": 1.012725687509755,
+      "perplexity": 1.0104041813774194,
       "step": 9200
     },
     {
       "epoch": 1.9926438771094763,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.287109375,
       "learning_rate": 4.99578947368421e-05,
-      "loss": 0.0175,
+      "loss": 0.0174,
       "num_input_tokens_seen": 603410432,
       "step": 9210,
-      "train_runtime": 1886.5591,
-      "train_tokens_per_second": 319847.091
+      "train_runtime": 4108.873,
+      "train_tokens_per_second": 146855.461
     },
     {
       "epoch": 1.994807442665513,
-      "grad_norm": 0.4765625,
+      "grad_norm": 0.38671875,
       "learning_rate": 4.932631578947368e-05,
-      "loss": 0.0177,
+      "loss": 0.0195,
       "num_input_tokens_seen": 604065792,
       "step": 9220,
-      "train_runtime": 1890.3607,
-      "train_tokens_per_second": 319550.55
+      "train_runtime": 4112.6559,
+      "train_tokens_per_second": 146879.73
     },
     {
       "epoch": 1.996971008221549,
-      "grad_norm": 0.4296875,
+      "grad_norm": 0.3046875,
       "learning_rate": 4.869473684210526e-05,
-      "loss": 0.0184,
+      "loss": 0.0187,
       "num_input_tokens_seen": 604721152,
       "step": 9230,
-      "train_runtime": 1894.1637,
-      "train_tokens_per_second": 319254.96
+      "train_runtime": 4116.4529,
+      "train_tokens_per_second": 146903.456
     },
     {
       "epoch": 1.9991345737775854,
-      "grad_norm": 0.25,
+      "grad_norm": 0.34375,
       "learning_rate": 4.806315789473684e-05,
-      "loss": 0.0185,
+      "loss": 0.0197,
       "num_input_tokens_seen": 605372416,
       "step": 9240,
-      "train_runtime": 1897.9386,
-      "train_tokens_per_second": 318963.116
+      "train_runtime": 4120.2142,
+      "train_tokens_per_second": 146927.414
     },
     {
       "epoch": 2.0012981393336218,
-      "grad_norm": 0.421875,
+      "grad_norm": 0.359375,
       "learning_rate": 4.743157894736842e-05,
-      "loss": 0.0183,
+      "loss": 0.0193,
       "num_input_tokens_seen": 606015488,
       "step": 9250,
-      "train_runtime": 1901.7544,
-      "train_tokens_per_second": 318661.277
+      "train_runtime": 4123.9949,
+      "train_tokens_per_second": 146948.652
     },
     {
       "epoch": 2.0034617048896584,
-      "grad_norm": 0.322265625,
+      "grad_norm": 0.291015625,
       "learning_rate": 4.68e-05,
-      "loss": 0.0151,
+      "loss": 0.0166,
       "num_input_tokens_seen": 606670848,
       "step": 9260,
-      "train_runtime": 1905.5735,
-      "train_tokens_per_second": 318366.535
+      "train_runtime": 4130.0553,
+      "train_tokens_per_second": 146891.703
     },
     {
       "epoch": 2.0056252704456945,
-      "grad_norm": 0.365234375,
+      "grad_norm": 0.291015625,
       "learning_rate": 4.616842105263157e-05,
-      "loss": 0.019,
+      "loss": 0.0204,
       "num_input_tokens_seen": 607326208,
       "step": 9270,
-      "train_runtime": 1909.3949,
-      "train_tokens_per_second": 318072.6
+      "train_runtime": 4133.8681,
+      "train_tokens_per_second": 146914.751
     },
     {
       "epoch": 2.0077888360017306,
-      "grad_norm": 0.3203125,
+      "grad_norm": 0.41015625,
       "learning_rate": 4.553684210526315e-05,
-      "loss": 0.0176,
+      "loss": 0.0183,
       "num_input_tokens_seen": 607981568,
       "step": 9280,
-      "train_runtime": 1913.2182,
-      "train_tokens_per_second": 317779.529
+      "train_runtime": 4137.6785,
+      "train_tokens_per_second": 146937.847
     },
     {
       "epoch": 2.009952401557767,
-      "grad_norm": 0.322265625,
+      "grad_norm": 0.3671875,
       "learning_rate": 4.490526315789473e-05,
-      "loss": 0.0163,
+      "loss": 0.0169,
       "num_input_tokens_seen": 608636928,
       "step": 9290,
-      "train_runtime": 1917.0468,
-      "train_tokens_per_second": 317486.738
+      "train_runtime": 4141.487,
+      "train_tokens_per_second": 146960.966
     },
     {
       "epoch": 2.0121159671138034,
-      "grad_norm": 0.27734375,
+      "grad_norm": 0.396484375,
       "learning_rate": 4.427368421052631e-05,
-      "loss": 0.0164,
+      "loss": 0.0183,
       "num_input_tokens_seen": 609292288,
       "step": 9300,
-      "train_runtime": 1920.8707,
-      "train_tokens_per_second": 317195.894
+      "train_runtime": 4145.2997,
+      "train_tokens_per_second": 146983.894
     },
     {
       "epoch": 2.0121159671138034,
-      "eval_loss": 0.012494292110204697,
-      "eval_runtime": 1.7403,
-      "eval_samples_per_second": 18.387,
-      "eval_steps_per_second": 0.575,
+      "eval_loss": 0.010261913761496544,
+      "eval_runtime": 2.1393,
+      "eval_samples_per_second": 14.958,
+      "eval_steps_per_second": 0.467,
       "num_input_tokens_seen": 609292288,
       "step": 9300
     },
     {
       "epoch": 2.0121159671138034,
-      "eval_byte_accuracy": 0.9974035608308606,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 609292288,
-      "perplexity": 1.0125726718709187,
+      "perplexity": 1.0103147477698782,
       "step": 9300
     },
     {
       "epoch": 2.01427953266984,
-      "grad_norm": 0.357421875,
+      "grad_norm": 0.298828125,
       "learning_rate": 4.364210526315789e-05,
-      "loss": 0.0166,
+      "loss": 0.0177,
       "num_input_tokens_seen": 609947648,
       "step": 9310,
-      "train_runtime": 1926.4536,
-      "train_tokens_per_second": 316616.841
+      "train_runtime": 4153.5205,
+      "train_tokens_per_second": 146850.76
     },
     {
       "epoch": 2.016443098225876,
-      "grad_norm": 0.353515625,
+      "grad_norm": 0.3203125,
       "learning_rate": 4.301052631578947e-05,
-      "loss": 0.0153,
+      "loss": 0.016,
       "num_input_tokens_seen": 610603008,
       "step": 9320,
-      "train_runtime": 1930.2731,
-      "train_tokens_per_second": 316329.85
+      "train_runtime": 4157.3317,
+      "train_tokens_per_second": 146873.777
     },
     {
       "epoch": 2.0186066637819127,
-      "grad_norm": 0.271484375,
+      "grad_norm": 0.283203125,
       "learning_rate": 4.237894736842105e-05,
-      "loss": 0.0165,
+      "loss": 0.0186,
       "num_input_tokens_seen": 611258368,
       "step": 9330,
-      "train_runtime": 1934.0809,
-      "train_tokens_per_second": 316045.916
+      "train_runtime": 4161.1347,
+      "train_tokens_per_second": 146897.041
     },
     {
       "epoch": 2.020770229337949,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.3828125,
       "learning_rate": 4.174736842105262e-05,
-      "loss": 0.0158,
+      "loss": 0.0163,
       "num_input_tokens_seen": 611913728,
       "step": 9340,
-      "train_runtime": 1937.9052,
-      "train_tokens_per_second": 315760.403
+      "train_runtime": 4164.9462,
+      "train_tokens_per_second": 146919.959
     },
     {
       "epoch": 2.0229337948939854,
-      "grad_norm": 0.328125,
+      "grad_norm": 0.353515625,
       "learning_rate": 4.111578947368421e-05,
-      "loss": 0.0171,
+      "loss": 0.0182,
       "num_input_tokens_seen": 612569088,
       "step": 9350,
-      "train_runtime": 1941.7238,
-      "train_tokens_per_second": 315476.947
+      "train_runtime": 4168.7566,
+      "train_tokens_per_second": 146942.876
     },
     {
       "epoch": 2.0250973604500215,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.314453125,
       "learning_rate": 4.048421052631578e-05,
-      "loss": 0.0167,
+      "loss": 0.0176,
       "num_input_tokens_seen": 613224448,
       "step": 9360,
-      "train_runtime": 1945.5442,
-      "train_tokens_per_second": 315194.305
+      "train_runtime": 4175.0303,
+      "train_tokens_per_second": 146879.041
     },
     {
       "epoch": 2.027260926006058,
-      "grad_norm": 0.32421875,
+      "grad_norm": 0.310546875,
       "learning_rate": 3.985263157894737e-05,
-      "loss": 0.0162,
+      "loss": 0.0168,
       "num_input_tokens_seen": 613879808,
       "step": 9370,
-      "train_runtime": 1949.3723,
-      "train_tokens_per_second": 314911.526
+      "train_runtime": 4178.8392,
+      "train_tokens_per_second": 146901.994
     },
     {
       "epoch": 2.0294244915620943,
-      "grad_norm": 0.306640625,
+      "grad_norm": 0.34375,
       "learning_rate": 3.922105263157894e-05,
-      "loss": 0.0176,
+      "loss": 0.0184,
       "num_input_tokens_seen": 614535168,
       "step": 9380,
-      "train_runtime": 1953.1986,
-      "train_tokens_per_second": 314630.146
+      "train_runtime": 4182.6568,
+      "train_tokens_per_second": 146924.597
     },
     {
       "epoch": 2.031588057118131,
-      "grad_norm": 0.3203125,
+      "grad_norm": 0.30078125,
       "learning_rate": 3.858947368421053e-05,
-      "loss": 0.0199,
+      "loss": 0.02,
       "num_input_tokens_seen": 615190528,
       "step": 9390,
-      "train_runtime": 1957.0185,
-      "train_tokens_per_second": 314350.899
+      "train_runtime": 4186.4682,
+      "train_tokens_per_second": 146947.38
     },
     {
       "epoch": 2.033751622674167,
-      "grad_norm": 0.333984375,
+      "grad_norm": 0.31640625,
       "learning_rate": 3.79578947368421e-05,
-      "loss": 0.0182,
+      "loss": 0.0194,
       "num_input_tokens_seen": 615845888,
       "step": 9400,
-      "train_runtime": 1960.8436,
-      "train_tokens_per_second": 314071.907
+      "train_runtime": 4190.2936,
+      "train_tokens_per_second": 146969.629
     },
     {
       "epoch": 2.033751622674167,
-      "eval_loss": 0.01256804633885622,
-      "eval_runtime": 1.7572,
-      "eval_samples_per_second": 18.211,
-      "eval_steps_per_second": 0.569,
+      "eval_loss": 0.01002554688602686,
+      "eval_runtime": 2.1448,
+      "eval_samples_per_second": 14.92,
+      "eval_steps_per_second": 0.466,
       "num_input_tokens_seen": 615845888,
       "step": 9400
     },
     {
       "epoch": 2.033751622674167,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 615845888,
-      "perplexity": 1.0126473561413927,
+      "perplexity": 1.010075971050272,
       "step": 9400
     },
     {
       "epoch": 2.0359151882302036,
-      "grad_norm": 0.443359375,
+      "grad_norm": 0.376953125,
       "learning_rate": 3.732631578947368e-05,
-      "loss": 0.0167,
+      "loss": 0.0169,
       "num_input_tokens_seen": 616501248,
       "step": 9410,
-      "train_runtime": 1966.4282,
-      "train_tokens_per_second": 313513.226
+      "train_runtime": 4198.5084,
+      "train_tokens_per_second": 146838.16
     },
     {
       "epoch": 2.0380787537862397,
-      "grad_norm": 0.384765625,
+      "grad_norm": 0.373046875,
       "learning_rate": 3.669473684210526e-05,
-      "loss": 0.0167,
+      "loss": 0.0177,
       "num_input_tokens_seen": 617156608,
       "step": 9420,
-      "train_runtime": 1970.2537,
-      "train_tokens_per_second": 313237.124
+      "train_runtime": 4202.3302,
+      "train_tokens_per_second": 146860.568
     },
     {
       "epoch": 2.0402423193422763,
-      "grad_norm": 0.32421875,
+      "grad_norm": 0.337890625,
       "learning_rate": 3.606315789473684e-05,
-      "loss": 0.0187,
+      "loss": 0.0201,
       "num_input_tokens_seen": 617811968,
       "step": 9430,
-      "train_runtime": 1974.0803,
-      "train_tokens_per_second": 312961.927
+      "train_runtime": 4206.1464,
+      "train_tokens_per_second": 146883.135
     },
     {
       "epoch": 2.0424058848983124,
-      "grad_norm": 0.34765625,
+      "grad_norm": 0.390625,
       "learning_rate": 3.543157894736842e-05,
-      "loss": 0.0164,
+      "loss": 0.0176,
       "num_input_tokens_seen": 618467328,
       "step": 9440,
-      "train_runtime": 1977.9041,
-      "train_tokens_per_second": 312688.224
+      "train_runtime": 4209.9584,
+      "train_tokens_per_second": 146905.804
     },
     {
       "epoch": 2.0445694504543486,
-      "grad_norm": 0.33984375,
+      "grad_norm": 0.36328125,
       "learning_rate": 3.48e-05,
-      "loss": 0.0156,
+      "loss": 0.0162,
       "num_input_tokens_seen": 619122688,
       "step": 9450,
-      "train_runtime": 1981.7289,
-      "train_tokens_per_second": 312415.428
+      "train_runtime": 4213.7813,
+      "train_tokens_per_second": 146928.053
     },
     {
       "epoch": 2.046733016010385,
-      "grad_norm": 0.3046875,
+      "grad_norm": 0.27734375,
       "learning_rate": 3.416842105263157e-05,
-      "loss": 0.0161,
+      "loss": 0.0168,
       "num_input_tokens_seen": 619778048,
       "step": 9460,
-      "train_runtime": 1985.5506,
-      "train_tokens_per_second": 312144.165
+      "train_runtime": 4219.8691,
+      "train_tokens_per_second": 146871.392
     },
     {
       "epoch": 2.0488965815664213,
-      "grad_norm": 0.30078125,
+      "grad_norm": 0.384765625,
       "learning_rate": 3.353684210526315e-05,
-      "loss": 0.0154,
+      "loss": 0.0171,
       "num_input_tokens_seen": 620433408,
       "step": 9470,
-      "train_runtime": 1989.3722,
-      "train_tokens_per_second": 311873.977
+      "train_runtime": 4223.6982,
+      "train_tokens_per_second": 146893.405
     },
     {
       "epoch": 2.051060147122458,
-      "grad_norm": 0.2890625,
+      "grad_norm": 0.349609375,
       "learning_rate": 3.290526315789473e-05,
-      "loss": 0.0151,
+      "loss": 0.0165,
       "num_input_tokens_seen": 621088768,
       "step": 9480,
-      "train_runtime": 1993.1753,
-      "train_tokens_per_second": 311607.691
+      "train_runtime": 4227.494,
+      "train_tokens_per_second": 146916.535
     },
     {
       "epoch": 2.053223712678494,
-      "grad_norm": 0.375,
+      "grad_norm": 0.2890625,
       "learning_rate": 3.227368421052631e-05,
-      "loss": 0.015,
+      "loss": 0.0161,
       "num_input_tokens_seen": 621744128,
       "step": 9490,
-      "train_runtime": 1996.9919,
-      "train_tokens_per_second": 311340.336
+      "train_runtime": 4231.3049,
+      "train_tokens_per_second": 146939.097
     },
     {
       "epoch": 2.0553872782345306,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.353515625,
       "learning_rate": 3.164210526315789e-05,
-      "loss": 0.0138,
+      "loss": 0.0158,
       "num_input_tokens_seen": 622399488,
       "step": 9500,
-      "train_runtime": 2000.8102,
-      "train_tokens_per_second": 311073.728
+      "train_runtime": 4235.1168,
+      "train_tokens_per_second": 146961.587
     },
     {
       "epoch": 2.0553872782345306,
-      "eval_loss": 0.01297488622367382,
-      "eval_runtime": 1.7539,
-      "eval_samples_per_second": 18.245,
-      "eval_steps_per_second": 0.57,
+      "eval_loss": 0.010010459460318089,
+      "eval_runtime": 2.1507,
+      "eval_samples_per_second": 14.879,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 622399488,
       "step": 9500
     },
     {
       "epoch": 2.0553872782345306,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 622399488,
-      "perplexity": 1.0130594252925254,
+      "perplexity": 1.01006073171906,
       "step": 9500
     },
     {
       "epoch": 2.0575508437905667,
-      "grad_norm": 0.3515625,
+      "grad_norm": 0.337890625,
       "learning_rate": 3.101052631578947e-05,
-      "loss": 0.0176,
+      "loss": 0.0186,
       "num_input_tokens_seen": 623054848,
       "step": 9510,
-      "train_runtime": 2006.4558,
-      "train_tokens_per_second": 310525.073
+      "train_runtime": 4243.3995,
+      "train_tokens_per_second": 146829.175
     },
     {
       "epoch": 2.0597144093466033,
-      "grad_norm": 0.287109375,
+      "grad_norm": 0.5,
       "learning_rate": 3.0378947368421053e-05,
-      "loss": 0.0153,
+      "loss": 0.0174,
       "num_input_tokens_seen": 623710208,
       "step": 9520,
-      "train_runtime": 2010.2876,
-      "train_tokens_per_second": 310259.192
+      "train_runtime": 4247.2255,
+      "train_tokens_per_second": 146851.211
     },
     {
       "epoch": 2.0618779749026395,
-      "grad_norm": 0.30859375,
+      "grad_norm": 0.32421875,
       "learning_rate": 2.974736842105263e-05,
-      "loss": 0.0151,
+      "loss": 0.0158,
       "num_input_tokens_seen": 624365568,
       "step": 9530,
-      "train_runtime": 2014.1246,
-      "train_tokens_per_second": 309993.522
+      "train_runtime": 4251.0383,
+      "train_tokens_per_second": 146873.663
     },
     {
       "epoch": 2.064041540458676,
-      "grad_norm": 0.453125,
+      "grad_norm": 0.345703125,
       "learning_rate": 2.9115789473684205e-05,
-      "loss": 0.0161,
+      "loss": 0.0173,
       "num_input_tokens_seen": 625020928,
       "step": 9540,
-      "train_runtime": 2017.9547,
-      "train_tokens_per_second": 309729.918
+      "train_runtime": 4254.8535,
+      "train_tokens_per_second": 146895.993
     },
     {
       "epoch": 2.066205106014712,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.373046875,
       "learning_rate": 2.8484210526315785e-05,
-      "loss": 0.0155,
+      "loss": 0.0165,
       "num_input_tokens_seen": 625676288,
       "step": 9550,
-      "train_runtime": 2021.7816,
-      "train_tokens_per_second": 309467.785
+      "train_runtime": 4258.673,
+      "train_tokens_per_second": 146918.134
     },
     {
       "epoch": 2.0683686715707488,
-      "grad_norm": 0.294921875,
+      "grad_norm": 0.28125,
       "learning_rate": 2.7852631578947365e-05,
-      "loss": 0.0147,
+      "loss": 0.0159,
       "num_input_tokens_seen": 626331648,
       "step": 9560,
-      "train_runtime": 2025.6074,
-      "train_tokens_per_second": 309206.828
+      "train_runtime": 4264.758,
+      "train_tokens_per_second": 146862.179
     },
     {
       "epoch": 2.070532237126785,
-      "grad_norm": 0.326171875,
+      "grad_norm": 0.3359375,
       "learning_rate": 2.7221052631578944e-05,
-      "loss": 0.0148,
+      "loss": 0.0158,
       "num_input_tokens_seen": 626987008,
       "step": 9570,
-      "train_runtime": 2029.4297,
-      "train_tokens_per_second": 308947.395
+      "train_runtime": 4268.5813,
+      "train_tokens_per_second": 146884.168
     },
     {
       "epoch": 2.0726958026828215,
-      "grad_norm": 0.37890625,
+      "grad_norm": 0.396484375,
       "learning_rate": 2.6589473684210524e-05,
-      "loss": 0.0173,
+      "loss": 0.0184,
       "num_input_tokens_seen": 627642368,
       "step": 9580,
-      "train_runtime": 2033.2601,
-      "train_tokens_per_second": 308687.686
+      "train_runtime": 4272.405,
+      "train_tokens_per_second": 146906.103
     },
     {
       "epoch": 2.0748593682388576,
-      "grad_norm": 0.326171875,
+      "grad_norm": 0.296875,
       "learning_rate": 2.5957894736842104e-05,
-      "loss": 0.0169,
+      "loss": 0.0183,
       "num_input_tokens_seen": 628297728,
       "step": 9590,
-      "train_runtime": 2037.0979,
-      "train_tokens_per_second": 308427.848
+      "train_runtime": 4276.2131,
+      "train_tokens_per_second": 146928.537
     },
     {
       "epoch": 2.0770229337948938,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.31640625,
       "learning_rate": 2.5326315789473683e-05,
-      "loss": 0.017,
+      "loss": 0.0171,
       "num_input_tokens_seen": 628953088,
       "step": 9600,
-      "train_runtime": 2040.9189,
-      "train_tokens_per_second": 308171.527
+      "train_runtime": 4280.0277,
+      "train_tokens_per_second": 146950.704
     },
     {
       "epoch": 2.0770229337948938,
-      "eval_loss": 0.01264923345297575,
-      "eval_runtime": 1.7454,
-      "eval_samples_per_second": 18.334,
-      "eval_steps_per_second": 0.573,
+      "eval_loss": 0.010010439902544022,
+      "eval_runtime": 2.1518,
+      "eval_samples_per_second": 14.871,
+      "eval_steps_per_second": 0.465,
       "num_input_tokens_seen": 628953088,
       "step": 9600
     },
     {
       "epoch": 2.0770229337948938,
-      "eval_byte_accuracy": 0.9970326409495549,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9931318681318682,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 628953088,
-      "perplexity": 1.0127295733953043,
+      "perplexity": 1.0100607119645206,
       "step": 9600
     },
     {
       "epoch": 2.0791864993509304,
-      "grad_norm": 0.310546875,
+      "grad_norm": 0.287109375,
       "learning_rate": 2.4694736842105263e-05,
-      "loss": 0.015,
+      "loss": 0.0167,
       "num_input_tokens_seen": 629608448,
       "step": 9610,
-      "train_runtime": 2046.5064,
-      "train_tokens_per_second": 307650.374
+      "train_runtime": 4288.0929,
+      "train_tokens_per_second": 146827.146
     },
     {
       "epoch": 2.0813500649069665,
-      "grad_norm": 0.259765625,
+      "grad_norm": 0.296875,
       "learning_rate": 2.406315789473684e-05,
-      "loss": 0.0162,
+      "loss": 0.0178,
       "num_input_tokens_seen": 630263808,
       "step": 9620,
-      "train_runtime": 2050.3255,
-      "train_tokens_per_second": 307396.946
+      "train_runtime": 4291.9127,
+      "train_tokens_per_second": 146849.169
     },
     {
       "epoch": 2.083513630463003,
-      "grad_norm": 0.298828125,
+      "grad_norm": 0.34375,
       "learning_rate": 2.343157894736842e-05,
-      "loss": 0.0151,
+      "loss": 0.0161,
       "num_input_tokens_seen": 630919168,
       "step": 9630,
-      "train_runtime": 2054.1477,
-      "train_tokens_per_second": 307144.006
+      "train_runtime": 4295.7199,
+      "train_tokens_per_second": 146871.579
     },
     {
       "epoch": 2.0856771960190392,
-      "grad_norm": 0.3125,
+      "grad_norm": 0.287109375,
       "learning_rate": 2.28e-05,
-      "loss": 0.0173,
+      "loss": 0.0186,
       "num_input_tokens_seen": 631574528,
       "step": 9640,
-      "train_runtime": 2057.9652,
-      "train_tokens_per_second": 306892.717
+      "train_runtime": 4299.5368,
+      "train_tokens_per_second": 146893.619
     },
     {
       "epoch": 2.087840761575076,
-      "grad_norm": 0.3515625,
+      "grad_norm": 0.337890625,
       "learning_rate": 2.2168421052631578e-05,
-      "loss": 0.0173,
+      "loss": 0.0188,
       "num_input_tokens_seen": 632229888,
       "step": 9650,
-      "train_runtime": 2061.7866,
-      "train_tokens_per_second": 306641.771
+      "train_runtime": 4303.3438,
+      "train_tokens_per_second": 146915.96
     },
     {
       "epoch": 2.090004327131112,
-      "grad_norm": 0.33203125,
+      "grad_norm": 0.318359375,
       "learning_rate": 2.1536842105263158e-05,
-      "loss": 0.0155,
+      "loss": 0.0164,
       "num_input_tokens_seen": 632885248,
       "step": 9660,
-      "train_runtime": 2065.6095,
-      "train_tokens_per_second": 306391.525
+      "train_runtime": 4309.5545,
+      "train_tokens_per_second": 146856.303
     },
     {
       "epoch": 2.0921678926871485,
       "grad_norm": 0.390625,
       "learning_rate": 2.0905263157894737e-05,
-      "loss": 0.0155,
+      "loss": 0.0173,
       "num_input_tokens_seen": 633540608,
       "step": 9670,
-      "train_runtime": 2069.4392,
-      "train_tokens_per_second": 306141.211
+      "train_runtime": 4313.3662,
+      "train_tokens_per_second": 146878.465
     },
     {
       "epoch": 2.0943314582431847,
-      "grad_norm": 0.330078125,
+      "grad_norm": 0.373046875,
       "learning_rate": 2.0273684210526317e-05,
-      "loss": 0.0186,
+      "loss": 0.0193,
       "num_input_tokens_seen": 634195968,
       "step": 9680,
-      "train_runtime": 2073.2575,
-      "train_tokens_per_second": 305893.487
+      "train_runtime": 4317.1835,
+      "train_tokens_per_second": 146900.396
     },
     {
       "epoch": 2.0964950237992213,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.30078125,
       "learning_rate": 1.964210526315789e-05,
       "loss": 0.017,
       "num_input_tokens_seen": 634851328,
       "step": 9690,
-      "train_runtime": 2077.0824,
-      "train_tokens_per_second": 305645.707
+      "train_runtime": 4320.9989,
+      "train_tokens_per_second": 146922.353
     },
     {
       "epoch": 2.0986585893552574,
-      "grad_norm": 0.388671875,
+      "grad_norm": 0.423828125,
       "learning_rate": 1.901052631578947e-05,
-      "loss": 0.0176,
+      "loss": 0.0196,
       "num_input_tokens_seen": 635506688,
       "step": 9700,
-      "train_runtime": 2080.9071,
-      "train_tokens_per_second": 305398.871
+      "train_runtime": 4324.807,
+      "train_tokens_per_second": 146944.519
     },
     {
       "epoch": 2.0986585893552574,
-      "eval_loss": 0.01279271300882101,
-      "eval_runtime": 1.7482,
-      "eval_samples_per_second": 18.305,
-      "eval_steps_per_second": 0.572,
+      "eval_loss": 0.009999026544392109,
+      "eval_runtime": 2.1556,
+      "eval_samples_per_second": 14.845,
+      "eval_steps_per_second": 0.464,
       "num_input_tokens_seen": 635506688,
       "step": 9700
     },
     {
       "epoch": 2.0986585893552574,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 635506688,
-      "perplexity": 1.0128748898094044,
+      "perplexity": 1.0100491838456471,
       "step": 9700
     },
     {
       "epoch": 2.100822154911294,
-      "grad_norm": 0.318359375,
+      "grad_norm": 0.3125,
       "learning_rate": 1.837894736842105e-05,
-      "loss": 0.0184,
+      "loss": 0.0193,
       "num_input_tokens_seen": 636162048,
       "step": 9710,
-      "train_runtime": 2086.5013,
-      "train_tokens_per_second": 304894.156
+      "train_runtime": 4333.0188,
+      "train_tokens_per_second": 146817.282
     },
     {
       "epoch": 2.10298572046733,
-      "grad_norm": 0.33984375,
+      "grad_norm": 0.3515625,
       "learning_rate": 1.774736842105263e-05,
-      "loss": 0.0165,
+      "loss": 0.0176,
       "num_input_tokens_seen": 636817408,
       "step": 9720,
-      "train_runtime": 2090.2846,
-      "train_tokens_per_second": 304655.832
+      "train_runtime": 4336.7843,
+      "train_tokens_per_second": 146840.924
     },
     {
       "epoch": 2.1051492860233667,
-      "grad_norm": 0.3359375,
+      "grad_norm": 0.384765625,
       "learning_rate": 1.711578947368421e-05,
-      "loss": 0.017,
+      "loss": 0.0181,
       "num_input_tokens_seen": 637472768,
       "step": 9730,
-      "train_runtime": 2094.1096,
-      "train_tokens_per_second": 304412.325
+      "train_runtime": 4340.605,
+      "train_tokens_per_second": 146862.653
     },
     {
       "epoch": 2.107312851579403,
-      "grad_norm": 0.375,
+      "grad_norm": 0.373046875,
       "learning_rate": 1.6484210526315788e-05,
-      "loss": 0.0166,
+      "loss": 0.0184,
       "num_input_tokens_seen": 638128128,
       "step": 9740,
-      "train_runtime": 2097.9304,
-      "train_tokens_per_second": 304170.301
+      "train_runtime": 4344.4225,
+      "train_tokens_per_second": 146884.455
     },
     {
       "epoch": 2.109476417135439,
-      "grad_norm": 0.287109375,
+      "grad_norm": 0.27734375,
       "learning_rate": 1.5852631578947364e-05,
-      "loss": 0.0147,
+      "loss": 0.0157,
       "num_input_tokens_seen": 638783488,
       "step": 9750,
-      "train_runtime": 2101.7387,
-      "train_tokens_per_second": 303930.978
+      "train_runtime": 4348.2274,
+      "train_tokens_per_second": 146906.644
     },
     {
       "epoch": 2.1116399826914756,
-      "grad_norm": 0.31640625,
+      "grad_norm": 0.275390625,
       "learning_rate": 1.5221052631578946e-05,
-      "loss": 0.0157,
+      "loss": 0.0171,
       "num_input_tokens_seen": 639438848,
       "step": 9760,
-      "train_runtime": 2105.5587,
-      "train_tokens_per_second": 303690.82
+      "train_runtime": 4354.2945,
+      "train_tokens_per_second": 146852.457
     },
     {
       "epoch": 2.1138035482475117,
-      "grad_norm": 0.34765625,
+      "grad_norm": 0.357421875,
       "learning_rate": 1.4589473684210525e-05,
-      "loss": 0.0156,
+      "loss": 0.0168,
       "num_input_tokens_seen": 640094208,
       "step": 9770,
-      "train_runtime": 2109.3849,
-      "train_tokens_per_second": 303450.638
+      "train_runtime": 4358.1147,
+      "train_tokens_per_second": 146874.107
     },
     {
       "epoch": 2.1159671138035483,
-      "grad_norm": 0.341796875,
+      "grad_norm": 0.283203125,
       "learning_rate": 1.3957894736842105e-05,
-      "loss": 0.0164,
+      "loss": 0.0178,
       "num_input_tokens_seen": 640749568,
       "step": 9780,
-      "train_runtime": 2113.2091,
-      "train_tokens_per_second": 303211.626
+      "train_runtime": 4361.9276,
+      "train_tokens_per_second": 146895.964
     },
     {
       "epoch": 2.1181306793595844,
-      "grad_norm": 0.298828125,
+      "grad_norm": 0.353515625,
       "learning_rate": 1.3326315789473681e-05,
-      "loss": 0.016,
+      "loss": 0.0178,
       "num_input_tokens_seen": 641404928,
       "step": 9790,
-      "train_runtime": 2117.0346,
-      "train_tokens_per_second": 302973.28
+      "train_runtime": 4365.7451,
+      "train_tokens_per_second": 146917.63
     },
     {
       "epoch": 2.120294244915621,
-      "grad_norm": 0.283203125,
+      "grad_norm": 0.462890625,
       "learning_rate": 1.2694736842105261e-05,
-      "loss": 0.0178,
+      "loss": 0.0184,
       "num_input_tokens_seen": 642060288,
       "step": 9800,
-      "train_runtime": 2120.8617,
-      "train_tokens_per_second": 302735.575
+      "train_runtime": 4369.5567,
+      "train_tokens_per_second": 146939.456
     },
     {
       "epoch": 2.120294244915621,
-      "eval_loss": 0.012742544524371624,
-      "eval_runtime": 1.7509,
-      "eval_samples_per_second": 18.276,
-      "eval_steps_per_second": 0.571,
+      "eval_loss": 0.009999697096645832,
+      "eval_runtime": 2.1661,
+      "eval_samples_per_second": 14.773,
+      "eval_steps_per_second": 0.462,
       "num_input_tokens_seen": 642060288,
       "step": 9800
     },
     {
       "epoch": 2.120294244915621,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 642060288,
-      "perplexity": 1.0128240766858652,
+      "perplexity": 1.0100498611366309,
       "step": 9800
     },
     {
       "epoch": 2.122457810471657,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.4140625,
       "learning_rate": 1.206315789473684e-05,
-      "loss": 0.0161,
+      "loss": 0.0172,
       "num_input_tokens_seen": 642715648,
       "step": 9810,
-      "train_runtime": 2126.4483,
-      "train_tokens_per_second": 302248.418
+      "train_runtime": 4377.9693,
+      "train_tokens_per_second": 146806.798
     },
     {
       "epoch": 2.1246213760276937,
-      "grad_norm": 0.34765625,
+      "grad_norm": 0.314453125,
       "learning_rate": 1.143157894736842e-05,
-      "loss": 0.0138,
+      "loss": 0.0155,
       "num_input_tokens_seen": 643371008,
       "step": 9820,
-      "train_runtime": 2130.2665,
-      "train_tokens_per_second": 302014.328
+      "train_runtime": 4381.7799,
+      "train_tokens_per_second": 146828.693
     },
     {
       "epoch": 2.12678494158373,
-      "grad_norm": 0.35546875,
+      "grad_norm": 0.357421875,
       "learning_rate": 1.0799999999999998e-05,
-      "loss": 0.0149,
+      "loss": 0.0159,
       "num_input_tokens_seen": 644026368,
       "step": 9830,
-      "train_runtime": 2134.0896,
-      "train_tokens_per_second": 301780.376
+      "train_runtime": 4385.602,
+      "train_tokens_per_second": 146850.163
     },
     {
       "epoch": 2.1289485071397665,
-      "grad_norm": 0.361328125,
+      "grad_norm": 0.396484375,
       "learning_rate": 1.0168421052631578e-05,
-      "loss": 0.0162,
+      "loss": 0.0168,
       "num_input_tokens_seen": 644681728,
       "step": 9840,
-      "train_runtime": 2137.916,
-      "train_tokens_per_second": 301546.8
+      "train_runtime": 4389.4172,
+      "train_tokens_per_second": 146871.827
     },
     {
       "epoch": 2.1311120726958026,
-      "grad_norm": 0.37890625,
+      "grad_norm": 0.482421875,
       "learning_rate": 9.536842105263158e-06,
-      "loss": 0.0171,
+      "loss": 0.0189,
       "num_input_tokens_seen": 645337088,
       "step": 9850,
-      "train_runtime": 2141.7343,
-      "train_tokens_per_second": 301315.201
+      "train_runtime": 4393.2316,
+      "train_tokens_per_second": 146893.482
     },
     {
       "epoch": 2.133275638251839,
-      "grad_norm": 0.29296875,
+      "grad_norm": 0.3046875,
       "learning_rate": 8.905263157894735e-06,
-      "loss": 0.0177,
+      "loss": 0.0183,
       "num_input_tokens_seen": 645992448,
       "step": 9860,
-      "train_runtime": 2145.5672,
-      "train_tokens_per_second": 301082.361
+      "train_runtime": 4399.285,
+      "train_tokens_per_second": 146840.326
     },
     {
       "epoch": 2.1354392038078753,
-      "grad_norm": 0.3828125,
+      "grad_norm": 0.462890625,
       "learning_rate": 8.273684210526315e-06,
-      "loss": 0.0144,
+      "loss": 0.0168,
       "num_input_tokens_seen": 646643712,
       "step": 9870,
-      "train_runtime": 2149.3524,
-      "train_tokens_per_second": 300855.133
+      "train_runtime": 4403.0615,
+      "train_tokens_per_second": 146862.295
     },
     {
       "epoch": 2.137602769363912,
-      "grad_norm": 0.25390625,
+      "grad_norm": 0.30078125,
       "learning_rate": 7.642105263157893e-06,
-      "loss": 0.0152,
+      "loss": 0.0167,
       "num_input_tokens_seen": 647299072,
       "step": 9880,
-      "train_runtime": 2153.1607,
-      "train_tokens_per_second": 300627.389
+      "train_runtime": 4406.8613,
+      "train_tokens_per_second": 146884.377
     },
     {
       "epoch": 2.139766334919948,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.37109375,
       "learning_rate": 7.0105263157894736e-06,
-      "loss": 0.0176,
+      "loss": 0.0189,
       "num_input_tokens_seen": 647954432,
       "step": 9890,
-      "train_runtime": 2156.9971,
-      "train_tokens_per_second": 300396.521
+      "train_runtime": 4410.6797,
+      "train_tokens_per_second": 146905.8
     },
     {
       "epoch": 2.1419299004759846,
       "grad_norm": 0.263671875,
       "learning_rate": 6.3789473684210515e-06,
-      "loss": 0.0167,
+      "loss": 0.0172,
       "num_input_tokens_seen": 648605696,
       "step": 9900,
-      "train_runtime": 2160.8044,
-      "train_tokens_per_second": 300168.632
+      "train_runtime": 4414.4742,
+      "train_tokens_per_second": 146927.055
     },
     {
       "epoch": 2.1419299004759846,
-      "eval_loss": 0.012824327684938908,
-      "eval_runtime": 1.755,
-      "eval_samples_per_second": 18.233,
-      "eval_steps_per_second": 0.57,
+      "eval_loss": 0.010031883604824543,
+      "eval_runtime": 2.146,
+      "eval_samples_per_second": 14.912,
+      "eval_steps_per_second": 0.466,
       "num_input_tokens_seen": 648605696,
       "step": 9900
     },
     {
       "epoch": 2.1419299004759846,
       "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 98.31023260829656,
-      "eval_sacrebleu": 97.89877731533007,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_chrf": 97.51098357963586,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 648605696,
-      "perplexity": 1.012906912027177,
+      "perplexity": 1.0100823716379443,
       "step": 9900
     },
     {
       "epoch": 2.144093466032021,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.365234375,
       "learning_rate": 5.747368421052631e-06,
-      "loss": 0.0179,
+      "loss": 0.0188,
       "num_input_tokens_seen": 649261056,
       "step": 9910,
-      "train_runtime": 2166.3917,
-      "train_tokens_per_second": 299696.981
+      "train_runtime": 4422.6956,
+      "train_tokens_per_second": 146802.113
     },
     {
       "epoch": 2.146257031588057,
-      "grad_norm": 0.369140625,
+      "grad_norm": 0.412109375,
       "learning_rate": 5.11578947368421e-06,
-      "loss": 0.0172,
+      "loss": 0.0184,
       "num_input_tokens_seen": 649916416,
       "step": 9920,
-      "train_runtime": 2170.2133,
-      "train_tokens_per_second": 299471.215
+      "train_runtime": 4426.5062,
+      "train_tokens_per_second": 146823.79
     },
     {
       "epoch": 2.1484205971440935,
-      "grad_norm": 0.36328125,
+      "grad_norm": 0.341796875,
       "learning_rate": 4.484210526315789e-06,
-      "loss": 0.0157,
+      "loss": 0.0164,
       "num_input_tokens_seen": 650571776,
       "step": 9930,
-      "train_runtime": 2174.0376,
-      "train_tokens_per_second": 299245.87
+      "train_runtime": 4430.3233,
+      "train_tokens_per_second": 146845.215
     },
     {
       "epoch": 2.1505841627001296,
-      "grad_norm": 0.30078125,
+      "grad_norm": 0.365234375,
       "learning_rate": 3.8526315789473676e-06,
-      "loss": 0.0155,
+      "loss": 0.0168,
       "num_input_tokens_seen": 651227136,
       "step": 9940,
-      "train_runtime": 2177.8684,
-      "train_tokens_per_second": 299020.427
+      "train_runtime": 4434.1482,
+      "train_tokens_per_second": 146866.345
     },
     {
       "epoch": 2.1527477282561662,
-      "grad_norm": 0.47265625,
+      "grad_norm": 0.3515625,
       "learning_rate": 3.2210526315789468e-06,
-      "loss": 0.0382,
+      "loss": 0.0406,
       "num_input_tokens_seen": 651882496,
       "step": 9950,
-      "train_runtime": 2181.6807,
-      "train_tokens_per_second": 298798.299
+      "train_runtime": 4437.9717,
+      "train_tokens_per_second": 146887.484
     },
     {
       "epoch": 2.1549112938122024,
-      "grad_norm": 0.341796875,
+      "grad_norm": 0.419921875,
       "learning_rate": 2.589473684210526e-06,
-      "loss": 0.0165,
+      "loss": 0.0179,
       "num_input_tokens_seen": 652537856,
       "step": 9960,
-      "train_runtime": 2185.4931,
-      "train_tokens_per_second": 298576.95
+      "train_runtime": 4444.2083,
+      "train_tokens_per_second": 146828.818
     },
     {
       "epoch": 2.157074859368239,
-      "grad_norm": 0.365234375,
+      "grad_norm": 0.44140625,
       "learning_rate": 1.957894736842105e-06,
-      "loss": 0.0161,
+      "loss": 0.0176,
       "num_input_tokens_seen": 653193216,
       "step": 9970,
-      "train_runtime": 2189.3203,
-      "train_tokens_per_second": 298354.34
+      "train_runtime": 4448.0238,
+      "train_tokens_per_second": 146850.208
     },
     {
       "epoch": 2.159238424924275,
-      "grad_norm": 0.32421875,
+      "grad_norm": 0.318359375,
       "learning_rate": 1.326315789473684e-06,
-      "loss": 0.0164,
+      "loss": 0.0181,
       "num_input_tokens_seen": 653848576,
       "step": 9980,
-      "train_runtime": 2193.1448,
-      "train_tokens_per_second": 298132.884
+      "train_runtime": 4451.8424,
+      "train_tokens_per_second": 146871.457
     },
     {
       "epoch": 2.1614019904803117,
-      "grad_norm": 0.3515625,
+      "grad_norm": 0.357421875,
       "learning_rate": 6.947368421052631e-07,
-      "loss": 0.0159,
+      "loss": 0.0172,
       "num_input_tokens_seen": 654503936,
       "step": 9990,
-      "train_runtime": 2196.9649,
-      "train_tokens_per_second": 297912.793
+      "train_runtime": 4455.6546,
+      "train_tokens_per_second": 146892.881
     },
     {
       "epoch": 2.163565556036348,
-      "grad_norm": 0.345703125,
+      "grad_norm": 0.294921875,
       "learning_rate": 6.31578947368421e-08,
-      "loss": 0.0166,
+      "loss": 0.0175,
       "num_input_tokens_seen": 655159296,
       "step": 10000,
-      "train_runtime": 2200.7835,
-      "train_tokens_per_second": 297693.656
+      "train_runtime": 4459.4692,
+      "train_tokens_per_second": 146914.188
     },
     {
       "epoch": 2.163565556036348,
-      "eval_loss": 0.012828610837459564,
-      "eval_runtime": 1.7704,
-      "eval_samples_per_second": 18.075,
-      "eval_steps_per_second": 0.565,
+      "eval_loss": 0.010019873268902302,
+      "eval_runtime": 2.1457,
+      "eval_samples_per_second": 14.914,
+      "eval_steps_per_second": 0.466,
       "num_input_tokens_seen": 655159296,
       "step": 10000
     },
     {
       "epoch": 2.163565556036348,
-      "eval_byte_accuracy": 0.9966617210682492,
-      "eval_chrf": 97.81668588759814,
-      "eval_sacrebleu": 97.30034967398244,
-      "eval_word_accuracy": 0.9917582417582418,
+      "eval_byte_accuracy": 0.9962908011869436,
+      "eval_chrf": 97.60016273961939,
+      "eval_sacrebleu": 96.29188240283864,
+      "eval_word_accuracy": 0.9903846153846154,
       "num_input_tokens_seen": 655159296,
-      "perplexity": 1.0129112504712616,
+      "perplexity": 1.0100702402822026,
       "step": 10000
     }
   ],