diff --git "a/trainer_state.json" "b/trainer_state.json"
--- "a/trainer_state.json"
+++ "b/trainer_state.json"
@@ -1,3251 +1,2005 @@
 {
-  "best_metric": 1.89783895,
-  "best_model_checkpoint": "D:\\_____NEW_NN\\LLM\\MiniCPM-V\\finetune\\output\\phi3-vision-128k-instruct\\v0-20240531-071942\\checkpoint-1476",
-  "epoch": 4.0,
+  "best_metric": 1.52509904,
+  "best_model_checkpoint": "D:\\_____NEW_NN\\LLM\\MiniCPM-V\\finetune\\output\\phi3-vision-128k-instruct\\v0-20240629-080216\\checkpoint-300",
+  "epoch": 3.5225048923679063,
   "eval_steps": 50,
-  "global_step": 1476,
+  "global_step": 900,
   "is_hyper_param_search": false,
   "is_local_process_zero": true,
   "is_world_process_zero": true,
   "log_history": [
     {
-      "acc": 0.47843874,
-      "epoch": 0.0027100271002710027,
-      "grad_norm": 0.921875,
-      "learning_rate": 2.027027027027027e-06,
-      "loss": 2.52092218,
-      "memory(GiB)": 12.33,
+      "acc": 0.49833804,
+      "epoch": 0.003913894324853229,
+      "grad_norm": 0.77734375,
+      "learning_rate": 2.745098039215686e-06,
+      "loss": 2.37747383,
+      "memory(GiB)": 17.35,
       "step": 1,
-      "train_speed(iter/s)": 0.066334
+      "train_speed(iter/s)": 0.076826
     },
     {
-      "acc": 0.51126236,
-      "epoch": 0.013550135501355014,
-      "grad_norm": 0.796875,
-      "learning_rate": 1.0135135135135135e-05,
-      "loss": 2.42214346,
-      "memory(GiB)": 13.38,
+      "acc": 0.50652587,
+      "epoch": 0.019569471624266144,
+      "grad_norm": 1.140625,
+      "learning_rate": 1.372549019607843e-05,
+      "loss": 2.29183841,
+      "memory(GiB)": 19.33,
       "step": 5,
-      "train_speed(iter/s)": 0.127981
+      "train_speed(iter/s)": 0.082188
     },
     {
-      "acc": 0.48636103,
-      "epoch": 0.02710027100271003,
-      "grad_norm": 1.09375,
-      "learning_rate": 2.027027027027027e-05,
-      "loss": 2.53938828,
-      "memory(GiB)": 13.38,
+      "acc": 0.52587533,
+      "epoch": 0.03913894324853229,
+      "grad_norm": 0.68359375,
+      "learning_rate": 2.745098039215686e-05,
+      "loss": 2.22724895,
+      "memory(GiB)": 19.89,
       "step": 10,
-      "train_speed(iter/s)": 0.144622
+      "train_speed(iter/s)": 0.082805
     },
     {
-      "acc": 0.47292571,
-      "epoch": 0.04065040650406504,
-      "grad_norm": 1.1328125,
-      "learning_rate": 3.0405405405405404e-05,
-      "loss": 2.50403137,
-      "memory(GiB)": 13.38,
+      "acc": 0.52128973,
+      "epoch": 0.05870841487279843,
+      "grad_norm": 0.8359375,
+      "learning_rate": 4.117647058823529e-05,
+      "loss": 2.27491264,
+      "memory(GiB)": 19.24,
       "step": 15,
-      "train_speed(iter/s)": 0.150329
+      "train_speed(iter/s)": 0.082482
     },
     {
-      "acc": 0.52069569,
-      "epoch": 0.05420054200542006,
-      "grad_norm": 0.83984375,
-      "learning_rate": 4.054054054054054e-05,
-      "loss": 2.28312988,
-      "memory(GiB)": 13.38,
+      "acc": 0.51135335,
+      "epoch": 0.07827788649706457,
+      "grad_norm": 0.66015625,
+      "learning_rate": 5.490196078431372e-05,
+      "loss": 2.32762127,
+      "memory(GiB)": 19.86,
       "step": 20,
-      "train_speed(iter/s)": 0.152832
+      "train_speed(iter/s)": 0.082557
     },
     {
-      "acc": 0.53072515,
-      "epoch": 0.06775067750677506,
-      "grad_norm": 0.59765625,
-      "learning_rate": 5.067567567567567e-05,
-      "loss": 2.20206394,
-      "memory(GiB)": 13.38,
+      "acc": 0.54442377,
+      "epoch": 0.09784735812133072,
+      "grad_norm": 0.65625,
+      "learning_rate": 6.862745098039214e-05,
+      "loss": 2.09772224,
+      "memory(GiB)": 19.05,
       "step": 25,
-      "train_speed(iter/s)": 0.154186
+      "train_speed(iter/s)": 0.082348
     },
     {
-      "acc": 0.52423077,
-      "epoch": 0.08130081300813008,
-      "grad_norm": 0.89453125,
-      "learning_rate": 6.081081081081081e-05,
-      "loss": 2.31712227,
-      "memory(GiB)": 14.44,
+      "acc": 0.5545311,
+      "epoch": 0.11741682974559686,
+      "grad_norm": 0.62109375,
+      "learning_rate": 8.235294117647058e-05,
+      "loss": 2.00072975,
+      "memory(GiB)": 19.89,
       "step": 30,
-      "train_speed(iter/s)": 0.155403
+      "train_speed(iter/s)": 0.082166
     },
     {
-      "acc": 0.54341574,
-      "epoch": 0.0948509485094851,
-      "grad_norm": 0.7109375,
-      "learning_rate": 7.094594594594594e-05,
-      "loss": 2.03825722,
-      "memory(GiB)": 14.44,
+      "acc": 0.57092514,
+      "epoch": 0.136986301369863,
+      "grad_norm": 0.9296875,
+      "learning_rate": 9.6078431372549e-05,
+      "loss": 1.94450474,
+      "memory(GiB)": 19.16,
       "step": 35,
-      "train_speed(iter/s)": 0.156383
+      "train_speed(iter/s)": 0.081966
     },
     {
-      "acc": 0.56128917,
-      "epoch": 0.10840108401084012,
-      "grad_norm": 0.734375,
-      "learning_rate": 8.108108108108108e-05,
-      "loss": 1.98176575,
-      "memory(GiB)": 14.44,
+      "acc": 0.56716595,
+      "epoch": 0.15655577299412915,
+      "grad_norm": 0.7734375,
+      "learning_rate": 0.00010980392156862745,
+      "loss": 1.90242462,
+      "memory(GiB)": 19.62,
       "step": 40,
-      "train_speed(iter/s)": 0.156973
+      "train_speed(iter/s)": 0.081987
     },
     {
-      "acc": 0.53511982,
-      "epoch": 0.12195121951219512,
-      "grad_norm": 1.1484375,
-      "learning_rate": 9.121621621621621e-05,
-      "loss": 2.07999401,
-      "memory(GiB)": 14.44,
+      "acc": 0.57822714,
+      "epoch": 0.1761252446183953,
+      "grad_norm": 0.74609375,
+      "learning_rate": 0.00012352941176470587,
+      "loss": 1.83147659,
+      "memory(GiB)": 19.99,
       "step": 45,
-      "train_speed(iter/s)": 0.157395
+      "train_speed(iter/s)": 0.081878
     },
     {
-      "acc": 0.56112757,
-      "epoch": 0.13550135501355012,
-      "grad_norm": 0.83984375,
-      "learning_rate": 0.00010135135135135135,
-      "loss": 1.92188244,
-      "memory(GiB)": 14.44,
+      "acc": 0.57696896,
+      "epoch": 0.19569471624266144,
+      "grad_norm": 0.85546875,
+      "learning_rate": 0.00013725490196078428,
+      "loss": 1.82299595,
+      "memory(GiB)": 19.11,
       "step": 50,
-      "train_speed(iter/s)": 0.157924
+      "train_speed(iter/s)": 0.081843
     },
     {
-      "epoch": 0.13550135501355012,
-      "eval_acc": 0.5618983279851376,
-      "eval_loss": 1.901513695716858,
-      "eval_runtime": 44.6741,
-      "eval_samples_per_second": 0.851,
-      "eval_steps_per_second": 0.851,
+      "epoch": 0.19569471624266144,
+      "eval_acc": 0.583503534956795,
+      "eval_loss": 1.8029242753982544,
+      "eval_runtime": 85.1254,
+      "eval_samples_per_second": 0.893,
+      "eval_steps_per_second": 0.446,
       "step": 50
     },
     {
-      "acc": 0.57467785,
-      "epoch": 0.14905149051490515,
-      "grad_norm": 0.9921875,
-      "learning_rate": 0.00011148648648648647,
-      "loss": 1.92710114,
-      "memory(GiB)": 15.21,
+      "acc": 0.59343066,
+      "epoch": 0.21526418786692758,
+      "grad_norm": 1.0,
+      "learning_rate": 0.0001399941138119636,
+      "loss": 1.82339039,
+      "memory(GiB)": 22.92,
       "step": 55,
-      "train_speed(iter/s)": 0.140365
+      "train_speed(iter/s)": 0.072544
     },
     {
-      "acc": 0.55041471,
-      "epoch": 0.16260162601626016,
-      "grad_norm": 1.1484375,
-      "learning_rate": 0.00012162162162162162,
-      "loss": 2.09731407,
-      "memory(GiB)": 15.21,
+      "acc": 0.58571839,
+      "epoch": 0.23483365949119372,
+      "grad_norm": 0.7734375,
+      "learning_rate": 0.00013997020286964757,
+      "loss": 1.80549526,
+      "memory(GiB)": 19.43,
       "step": 60,
-      "train_speed(iter/s)": 0.141894
+      "train_speed(iter/s)": 0.073269
     },
     {
-      "acc": 0.53853202,
-      "epoch": 0.17615176151761516,
-      "grad_norm": 1.03125,
-      "learning_rate": 0.00013175675675675675,
-      "loss": 2.01167774,
-      "memory(GiB)": 15.21,
+      "acc": 0.60369935,
+      "epoch": 0.25440313111545987,
+      "grad_norm": 0.99609375,
+      "learning_rate": 0.0001399279055646442,
+      "loss": 1.6768074,
+      "memory(GiB)": 19.57,
       "step": 65,
-      "train_speed(iter/s)": 0.143162
+      "train_speed(iter/s)": 0.073897
     },
     {
-      "acc": 0.57625084,
-      "epoch": 0.1897018970189702,
-      "grad_norm": 0.81640625,
-      "learning_rate": 0.00014189189189189188,
-      "loss": 1.77697067,
-      "memory(GiB)": 15.21,
+      "acc": 0.58763909,
+      "epoch": 0.273972602739726,
+      "grad_norm": 1.1640625,
+      "learning_rate": 0.00013986723301159307,
+      "loss": 1.79169483,
+      "memory(GiB)": 19.48,
       "step": 70,
-      "train_speed(iter/s)": 0.144318
+      "train_speed(iter/s)": 0.074533
     },
     {
-      "acc": 0.57468171,
-      "epoch": 0.2032520325203252,
-      "grad_norm": 0.96484375,
-      "learning_rate": 0.00014989300998573466,
-      "loss": 1.92596684,
-      "memory(GiB)": 15.21,
+      "acc": 0.58979025,
+      "epoch": 0.29354207436399216,
+      "grad_norm": 0.69140625,
+      "learning_rate": 0.00013978820115367462,
+      "loss": 1.72388344,
+      "memory(GiB)": 19.35,
       "step": 75,
-      "train_speed(iter/s)": 0.145344
+      "train_speed(iter/s)": 0.075045
     },
     {
-      "acc": 0.58551345,
-      "epoch": 0.21680216802168023,
-      "grad_norm": 1.0546875,
-      "learning_rate": 0.00014935805991440798,
-      "loss": 1.85387783,
-      "memory(GiB)": 15.21,
+      "acc": 0.59725327,
+      "epoch": 0.3131115459882583,
+      "grad_norm": 0.75,
+      "learning_rate": 0.00013969083075842048,
+      "loss": 1.70864868,
+      "memory(GiB)": 19.49,
       "step": 80,
-      "train_speed(iter/s)": 0.146049
+      "train_speed(iter/s)": 0.075523
     },
     {
-      "acc": 0.54782548,
-      "epoch": 0.23035230352303523,
-      "grad_norm": 1.0078125,
-      "learning_rate": 0.0001488231098430813,
-      "loss": 2.04980812,
-      "memory(GiB)": 15.21,
+      "acc": 0.60098982,
+      "epoch": 0.33268101761252444,
+      "grad_norm": 4.59375,
+      "learning_rate": 0.00013957514741225646,
+      "loss": 1.67311764,
+      "memory(GiB)": 20.01,
       "step": 85,
-      "train_speed(iter/s)": 0.146726
+      "train_speed(iter/s)": 0.075928
     },
     {
-      "acc": 0.56594706,
-      "epoch": 0.24390243902439024,
-      "grad_norm": 1.0,
-      "learning_rate": 0.0001482881597717546,
-      "loss": 1.93099308,
-      "memory(GiB)": 15.21,
+      "acc": 0.58315139,
+      "epoch": 0.3522504892367906,
+      "grad_norm": 0.8359375,
+      "learning_rate": 0.00013944118151377894,
+      "loss": 1.74437752,
+      "memory(GiB)": 20.14,
       "step": 90,
-      "train_speed(iter/s)": 0.147437
+      "train_speed(iter/s)": 0.076154
     },
     {
-      "acc": 0.57289362,
-      "epoch": 0.25745257452574527,
-      "grad_norm": 0.8046875,
-      "learning_rate": 0.00014775320970042795,
-      "loss": 1.84038963,
-      "memory(GiB)": 15.58,
+      "acc": 0.6138227,
+      "epoch": 0.37181996086105673,
+      "grad_norm": 0.75,
+      "learning_rate": 0.0001392889682657671,
+      "loss": 1.63750076,
+      "memory(GiB)": 19.59,
       "step": 95,
-      "train_speed(iter/s)": 0.147976
+      "train_speed(iter/s)": 0.076253
     },
     {
-      "acc": 0.5938961,
-      "epoch": 0.27100271002710025,
-      "grad_norm": 1.015625,
-      "learning_rate": 0.00014721825962910127,
-      "loss": 1.71151619,
-      "memory(GiB)": 15.58,
+      "acc": 0.63383026,
+      "epoch": 0.3913894324853229,
+      "grad_norm": 0.8515625,
+      "learning_rate": 0.00013911854766593233,
+      "loss": 1.56653557,
+      "memory(GiB)": 19.5,
       "step": 100,
-      "train_speed(iter/s)": 0.148573
+      "train_speed(iter/s)": 0.076386
     },
     {
-      "epoch": 0.27100271002710025,
-      "eval_acc": 0.587907448066205,
-      "eval_loss": 1.7437644004821777,
-      "eval_runtime": 44.775,
-      "eval_samples_per_second": 0.849,
-      "eval_steps_per_second": 0.849,
+      "epoch": 0.3913894324853229,
+      "eval_acc": 0.604241948153967,
+      "eval_loss": 1.6681365966796875,
+      "eval_runtime": 72.2811,
+      "eval_samples_per_second": 1.051,
+      "eval_steps_per_second": 0.526,
       "step": 100
     },
     {
-      "acc": 0.5828393,
-      "epoch": 0.2845528455284553,
-      "grad_norm": 1.078125,
-      "learning_rate": 0.00014668330955777461,
-      "loss": 1.90272522,
-      "memory(GiB)": 15.58,
+      "acc": 0.61646304,
+      "epoch": 0.410958904109589,
+      "grad_norm": 0.73046875,
+      "learning_rate": 0.00013892996449640807,
+      "loss": 1.59651537,
+      "memory(GiB)": 22.5,
       "step": 105,
-      "train_speed(iter/s)": 0.140241
+      "train_speed(iter/s)": 0.072857
     },
     {
-      "acc": 0.57981925,
-      "epoch": 0.2981029810298103,
-      "grad_norm": 0.859375,
-      "learning_rate": 0.00014614835948644793,
-      "loss": 1.88997154,
-      "memory(GiB)": 15.58,
+      "acc": 0.60897431,
+      "epoch": 0.43052837573385516,
+      "grad_norm": 0.83984375,
+      "learning_rate": 0.00013872326831198205,
+      "loss": 1.70257473,
+      "memory(GiB)": 19.42,
       "step": 110,
-      "train_speed(iter/s)": 0.141057
+      "train_speed(iter/s)": 0.073309
     },
     {
-      "acc": 0.56191244,
-      "epoch": 0.3116531165311653,
-      "grad_norm": 0.94140625,
-      "learning_rate": 0.00014561340941512125,
-      "loss": 1.92102909,
-      "memory(GiB)": 15.58,
+      "acc": 0.58328586,
+      "epoch": 0.4500978473581213,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00013849851342707462,
+      "loss": 1.71216717,
+      "memory(GiB)": 19.47,
       "step": 115,
-      "train_speed(iter/s)": 0.141824
+      "train_speed(iter/s)": 0.073753
     },
     {
-      "acc": 0.57420607,
-      "epoch": 0.3252032520325203,
-      "grad_norm": 1.171875,
-      "learning_rate": 0.00014507845934379456,
-      "loss": 1.80924416,
-      "memory(GiB)": 15.58,
+      "acc": 0.62397904,
+      "epoch": 0.46966731898238745,
+      "grad_norm": 0.80078125,
+      "learning_rate": 0.0001382557589014664,
+      "loss": 1.54239073,
+      "memory(GiB)": 19.33,
       "step": 120,
-      "train_speed(iter/s)": 0.14256
+      "train_speed(iter/s)": 0.074078
     },
     {
-      "acc": 0.57703466,
-      "epoch": 0.33875338753387535,
-      "grad_norm": 0.84765625,
-      "learning_rate": 0.00014454350927246788,
-      "loss": 1.85804043,
-      "memory(GiB)": 15.58,
+      "acc": 0.60271235,
+      "epoch": 0.4892367906066536,
+      "grad_norm": 1.171875,
+      "learning_rate": 0.0001379950685247788,
+      "loss": 1.72333088,
+      "memory(GiB)": 19.37,
       "step": 125,
-      "train_speed(iter/s)": 0.143173
+      "train_speed(iter/s)": 0.074428
     },
     {
-      "acc": 0.58581357,
-      "epoch": 0.3523035230352303,
-      "grad_norm": 0.96484375,
-      "learning_rate": 0.0001440085592011412,
-      "loss": 1.75103779,
-      "memory(GiB)": 15.58,
+      "acc": 0.5755064,
+      "epoch": 0.5088062622309197,
+      "grad_norm": 0.94921875,
+      "learning_rate": 0.00013771651079971182,
+      "loss": 1.81728477,
+      "memory(GiB)": 19.52,
       "step": 130,
-      "train_speed(iter/s)": 0.143759
+      "train_speed(iter/s)": 0.074768
     },
     {
-      "acc": 0.59179254,
-      "epoch": 0.36585365853658536,
-      "grad_norm": 0.93359375,
-      "learning_rate": 0.00014347360912981454,
-      "loss": 1.67790794,
-      "memory(GiB)": 15.58,
+      "acc": 0.5844254,
+      "epoch": 0.5283757338551859,
+      "grad_norm": 0.8515625,
+      "learning_rate": 0.00013742015892404325,
+      "loss": 1.77252998,
+      "memory(GiB)": 19.51,
       "step": 135,
-      "train_speed(iter/s)": 0.14428
+      "train_speed(iter/s)": 0.075066
     },
     {
-      "acc": 0.59449296,
-      "epoch": 0.3794037940379404,
-      "grad_norm": 0.74609375,
-      "learning_rate": 0.00014293865905848786,
-      "loss": 1.72561359,
-      "memory(GiB)": 15.58,
+      "acc": 0.5998323,
+      "epoch": 0.547945205479452,
+      "grad_norm": 0.8671875,
+      "learning_rate": 0.0001371060907713942,
+      "loss": 1.69012871,
+      "memory(GiB)": 19.54,
       "step": 140,
-      "train_speed(iter/s)": 0.144799
+      "train_speed(iter/s)": 0.07528
     },
     {
-      "acc": 0.57711525,
-      "epoch": 0.39295392953929537,
-      "grad_norm": 1.109375,
-      "learning_rate": 0.0001424037089871612,
-      "loss": 1.79506092,
-      "memory(GiB)": 15.58,
+      "acc": 0.62686119,
+      "epoch": 0.5675146771037182,
+      "grad_norm": 0.68359375,
+      "learning_rate": 0.00013677438887076603,
+      "loss": 1.66314449,
+      "memory(GiB)": 19.54,
       "step": 145,
-      "train_speed(iter/s)": 0.14532
+      "train_speed(iter/s)": 0.075467
     },
     {
-      "acc": 0.57541366,
-      "epoch": 0.4065040650406504,
-      "grad_norm": 5.46875,
-      "learning_rate": 0.00014186875891583452,
-      "loss": 1.80377541,
-      "memory(GiB)": 15.58,
+      "acc": 0.59954901,
+      "epoch": 0.5870841487279843,
+      "grad_norm": 0.6328125,
+      "learning_rate": 0.00013642514038485367,
+      "loss": 1.67525444,
+      "memory(GiB)": 19.55,
       "step": 150,
-      "train_speed(iter/s)": 0.145757
+      "train_speed(iter/s)": 0.075722
     },
     {
-      "epoch": 0.4065040650406504,
-      "eval_acc": 0.5983786522546867,
-      "eval_loss": 1.676965594291687,
-      "eval_runtime": 44.3227,
-      "eval_samples_per_second": 0.857,
-      "eval_steps_per_second": 0.857,
+      "epoch": 0.5870841487279843,
+      "eval_acc": 0.6184603299293009,
+      "eval_loss": 1.5965631008148193,
+      "eval_runtime": 72.3005,
+      "eval_samples_per_second": 1.051,
+      "eval_steps_per_second": 0.526,
       "step": 150
     },
     {
-      "acc": 0.58702893,
-      "epoch": 0.42005420054200543,
-      "grad_norm": 1.0625,
-      "learning_rate": 0.00014133380884450783,
-      "loss": 1.86865826,
-      "memory(GiB)": 15.58,
+      "acc": 0.585955,
+      "epoch": 0.6066536203522505,
+      "grad_norm": 0.9375,
+      "learning_rate": 0.00013605843708714162,
+      "loss": 1.7486639,
+      "memory(GiB)": 23.22,
       "step": 155,
-      "train_speed(iter/s)": 0.140358
+      "train_speed(iter/s)": 0.073368
     },
     {
-      "acc": 0.58541198,
-      "epoch": 0.43360433604336046,
-      "grad_norm": 1.015625,
-      "learning_rate": 0.00014079885877318115,
-      "loss": 1.82644196,
-      "memory(GiB)": 15.58,
+      "acc": 0.62769904,
+      "epoch": 0.6262230919765166,
+      "grad_norm": 0.7265625,
+      "learning_rate": 0.00013567437533778826,
+      "loss": 1.55238762,
+      "memory(GiB)": 19.62,
       "step": 160,
-      "train_speed(iter/s)": 0.140936
+      "train_speed(iter/s)": 0.073628
     },
     {
-      "acc": 0.58856125,
-      "epoch": 0.44715447154471544,
-      "grad_norm": 0.9921875,
-      "learning_rate": 0.00014026390870185447,
-      "loss": 1.68714104,
-      "memory(GiB)": 15.58,
+      "acc": 0.63651643,
+      "epoch": 0.6457925636007827,
+      "grad_norm": 0.80078125,
+      "learning_rate": 0.00013527305605830488,
+      "loss": 1.54306393,
+      "memory(GiB)": 19.88,
       "step": 165,
-      "train_speed(iter/s)": 0.141495
+      "train_speed(iter/s)": 0.073903
     },
     {
-      "acc": 0.59458299,
-      "epoch": 0.46070460704607047,
-      "grad_norm": 0.91015625,
-      "learning_rate": 0.0001397289586305278,
-      "loss": 1.72063637,
-      "memory(GiB)": 15.58,
+      "acc": 0.59288979,
+      "epoch": 0.6653620352250489,
+      "grad_norm": 0.703125,
+      "learning_rate": 0.0001348545847050361,
+      "loss": 1.69727612,
+      "memory(GiB)": 19.58,
       "step": 170,
-      "train_speed(iter/s)": 0.142005
+      "train_speed(iter/s)": 0.074077
     },
     {
-      "acc": 0.57354069,
-      "epoch": 0.4742547425474255,
-      "grad_norm": 0.7734375,
-      "learning_rate": 0.00013919400855920113,
-      "loss": 1.83291931,
-      "memory(GiB)": 15.58,
+      "acc": 0.61248484,
+      "epoch": 0.684931506849315,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00013441907124144866,
+      "loss": 1.65900764,
+      "memory(GiB)": 19.49,
       "step": 175,
-      "train_speed(iter/s)": 0.142509
+      "train_speed(iter/s)": 0.074329
     },
     {
-      "acc": 0.58851786,
-      "epoch": 0.4878048780487805,
-      "grad_norm": 1.234375,
-      "learning_rate": 0.00013865905848787447,
-      "loss": 1.70540199,
-      "memory(GiB)": 15.58,
+      "acc": 0.61740661,
+      "epoch": 0.7045009784735812,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.0001339666301092358,
+      "loss": 1.6518961,
+      "memory(GiB)": 19.68,
       "step": 180,
-      "train_speed(iter/s)": 0.143002
+      "train_speed(iter/s)": 0.074558
     },
     {
-      "acc": 0.59666262,
-      "epoch": 0.5013550135501355,
-      "grad_norm": 0.87890625,
-      "learning_rate": 0.0001381241084165478,
-      "loss": 1.70772285,
-      "memory(GiB)": 15.58,
+      "acc": 0.62250223,
+      "epoch": 0.7240704500978473,
+      "grad_norm": 0.84765625,
+      "learning_rate": 0.00013349738019824512,
+      "loss": 1.55100412,
+      "memory(GiB)": 19.34,
       "step": 185,
-      "train_speed(iter/s)": 0.143463
+      "train_speed(iter/s)": 0.07477
     },
     {
-      "acc": 0.58001013,
-      "epoch": 0.5149051490514905,
-      "grad_norm": 0.890625,
-      "learning_rate": 0.0001375891583452211,
-      "loss": 1.73730106,
-      "memory(GiB)": 15.58,
+      "acc": 0.61055808,
+      "epoch": 0.7436399217221135,
+      "grad_norm": 0.90625,
+      "learning_rate": 0.00013301144481523718,
+      "loss": 1.67241592,
+      "memory(GiB)": 19.56,
       "step": 190,
-      "train_speed(iter/s)": 0.143883
+      "train_speed(iter/s)": 0.075006
     },
     {
-      "acc": 0.5934463,
-      "epoch": 0.5284552845528455,
-      "grad_norm": 1.421875,
-      "learning_rate": 0.00013705420827389442,
-      "loss": 1.68941402,
-      "memory(GiB)": 15.58,
+      "acc": 0.6389596,
+      "epoch": 0.7632093933463796,
+      "grad_norm": 0.83203125,
+      "learning_rate": 0.00013250895165148384,
+      "loss": 1.54227753,
+      "memory(GiB)": 19.29,
       "step": 195,
-      "train_speed(iter/s)": 0.144297
+      "train_speed(iter/s)": 0.075192
     },
     {
-      "acc": 0.58476119,
-      "epoch": 0.5420054200542005,
-      "grad_norm": 0.91015625,
-      "learning_rate": 0.00013651925820256774,
-      "loss": 1.74896088,
-      "memory(GiB)": 15.58,
+      "acc": 0.59149747,
+      "epoch": 0.7827788649706457,
+      "grad_norm": 0.68359375,
+      "learning_rate": 0.00013199003274921416,
+      "loss": 1.71190453,
+      "memory(GiB)": 19.35,
       "step": 200,
-      "train_speed(iter/s)": 0.144655
+      "train_speed(iter/s)": 0.075393
     },
     {
-      "epoch": 0.5420054200542005,
-      "eval_acc": 0.6105387603445364,
-      "eval_loss": 1.6297248601913452,
-      "eval_runtime": 44.3331,
-      "eval_samples_per_second": 0.857,
-      "eval_steps_per_second": 0.857,
+      "epoch": 0.7827788649706457,
+      "eval_acc": 0.6241162608012569,
+      "eval_loss": 1.5573129653930664,
+      "eval_runtime": 69.5471,
+      "eval_samples_per_second": 1.093,
+      "eval_steps_per_second": 0.546,
       "step": 200
     },
     {
-      "acc": 0.58323898,
-      "epoch": 0.5555555555555556,
-      "grad_norm": 0.79296875,
-      "learning_rate": 0.00013598430813124105,
-      "loss": 1.74196358,
-      "memory(GiB)": 15.58,
+      "acc": 0.62623324,
+      "epoch": 0.8023483365949119,
+      "grad_norm": 0.81640625,
+      "learning_rate": 0.00013145482446691724,
+      "loss": 1.55779324,
+      "memory(GiB)": 20.56,
       "step": 205,
-      "train_speed(iter/s)": 0.140612
+      "train_speed(iter/s)": 0.073671
     },
     {
-      "acc": 0.59854908,
-      "epoch": 0.5691056910569106,
-      "grad_norm": 1.109375,
-      "learning_rate": 0.0001354493580599144,
-      "loss": 1.6279623,
-      "memory(GiB)": 15.58,
+      "acc": 0.61495056,
+      "epoch": 0.821917808219178,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00013090346744351058,
+      "loss": 1.56424398,
+      "memory(GiB)": 19.48,
       "step": 210,
-      "train_speed(iter/s)": 0.141053
+      "train_speed(iter/s)": 0.073902
     },
     {
-      "acc": 0.58306313,
-      "epoch": 0.5826558265582655,
-      "grad_norm": 0.98828125,
-      "learning_rate": 0.00013491440798858771,
-      "loss": 1.85492191,
-      "memory(GiB)": 15.58,
+      "acc": 0.59643593,
+      "epoch": 0.8414872798434442,
+      "grad_norm": 1.0703125,
+      "learning_rate": 0.00013033610656138395,
+      "loss": 1.62190418,
+      "memory(GiB)": 19.5,
       "step": 215,
-      "train_speed(iter/s)": 0.141461
+      "train_speed(iter/s)": 0.074133
     },
     {
-      "acc": 0.58454275,
-      "epoch": 0.5962059620596206,
-      "grad_norm": 0.890625,
-      "learning_rate": 0.00013437945791726106,
-      "loss": 1.75104046,
-      "memory(GiB)": 15.58,
+      "acc": 0.63052382,
+      "epoch": 0.8610567514677103,
+      "grad_norm": 0.59765625,
+      "learning_rate": 0.00012975289090832792,
+      "loss": 1.53521852,
+      "memory(GiB)": 19.53,
       "step": 220,
-      "train_speed(iter/s)": 0.141869
+      "train_speed(iter/s)": 0.074334
     },
     {
-      "acc": 0.60898943,
-      "epoch": 0.6097560975609756,
-      "grad_norm": 0.828125,
-      "learning_rate": 0.00013384450784593437,
-      "loss": 1.57786808,
-      "memory(GiB)": 15.58,
+      "acc": 0.61408448,
+      "epoch": 0.8806262230919765,
+      "grad_norm": 0.7734375,
+      "learning_rate": 0.00012915397373835754,
+      "loss": 1.59712257,
+      "memory(GiB)": 19.52,
       "step": 225,
-      "train_speed(iter/s)": 0.142237
+      "train_speed(iter/s)": 0.074533
     },
     {
-      "acc": 0.58954048,
-      "epoch": 0.6233062330623306,
-      "grad_norm": 1.1640625,
-      "learning_rate": 0.0001333095577746077,
-      "loss": 1.72581081,
-      "memory(GiB)": 15.58,
+      "acc": 0.62307076,
+      "epoch": 0.9001956947162426,
+      "grad_norm": 0.66796875,
+      "learning_rate": 0.00012853951243144105,
+      "loss": 1.57903328,
+      "memory(GiB)": 19.49,
       "step": 230,
-      "train_speed(iter/s)": 0.142622
+      "train_speed(iter/s)": 0.074719
     },
     {
-      "acc": 0.59608021,
-      "epoch": 0.6368563685636857,
-      "grad_norm": 1.2265625,
-      "learning_rate": 0.000132774607703281,
-      "loss": 1.70160866,
-      "memory(GiB)": 15.58,
+      "acc": 0.61717134,
+      "epoch": 0.9197651663405088,
+      "grad_norm": 0.84375,
+      "learning_rate": 0.00012790966845214457,
+      "loss": 1.61422024,
+      "memory(GiB)": 19.25,
       "step": 235,
-      "train_speed(iter/s)": 0.142983
+      "train_speed(iter/s)": 0.074916
     },
     {
-      "acc": 0.57084417,
-      "epoch": 0.6504065040650406,
-      "grad_norm": 1.1640625,
-      "learning_rate": 0.00013223965763195432,
-      "loss": 1.81941319,
-      "memory(GiB)": 15.58,
+      "acc": 0.62549253,
+      "epoch": 0.9393346379647749,
+      "grad_norm": 0.8125,
+      "learning_rate": 0.0001272646073072033,
+      "loss": 1.62806015,
+      "memory(GiB)": 19.36,
       "step": 240,
-      "train_speed(iter/s)": 0.143329
+      "train_speed(iter/s)": 0.0751
     },
     {
-      "acc": 0.61476159,
-      "epoch": 0.6639566395663956,
-      "grad_norm": 0.9453125,
-      "learning_rate": 0.00013170470756062767,
-      "loss": 1.68505306,
-      "memory(GiB)": 15.58,
+      "acc": 0.61903515,
+      "epoch": 0.958904109589041,
+      "grad_norm": 0.74609375,
+      "learning_rate": 0.0001266044985020307,
+      "loss": 1.55927486,
+      "memory(GiB)": 19.36,
       "step": 245,
-      "train_speed(iter/s)": 0.143652
+      "train_speed(iter/s)": 0.075266
     },
     {
-      "acc": 0.5995626,
-      "epoch": 0.6775067750677507,
-      "grad_norm": 0.9140625,
-      "learning_rate": 0.00013116975748930098,
-      "loss": 1.77789631,
-      "memory(GiB)": 15.58,
+      "acc": 0.61238952,
+      "epoch": 0.9784735812133072,
+      "grad_norm": 0.87890625,
+      "learning_rate": 0.00012592951549617683,
+      "loss": 1.52888412,
+      "memory(GiB)": 19.33,
       "step": 250,
-      "train_speed(iter/s)": 0.144002
+      "train_speed(iter/s)": 0.075438
     },
     {
-      "epoch": 0.6775067750677507,
-      "eval_acc": 0.61087654112481,
-      "eval_loss": 1.6108067035675049,
-      "eval_runtime": 44.1094,
-      "eval_samples_per_second": 0.861,
-      "eval_steps_per_second": 0.861,
+      "epoch": 0.9784735812133072,
+      "eval_acc": 0.6267085624509033,
+      "eval_loss": 1.5281730890274048,
+      "eval_runtime": 69.069,
+      "eval_samples_per_second": 1.1,
+      "eval_steps_per_second": 0.55,
       "step": 250
     },
     {
-      "acc": 0.62165804,
-      "epoch": 0.6910569105691057,
-      "grad_norm": 1.6171875,
-      "learning_rate": 0.0001306348074179743,
-      "loss": 1.57510653,
-      "memory(GiB)": 15.58,
+      "acc": 0.63230977,
+      "epoch": 0.9980430528375733,
+      "grad_norm": 0.84765625,
+      "learning_rate": 0.00012523983565774753,
+      "loss": 1.53058205,
+      "memory(GiB)": 19.46,
       "step": 255,
-      "train_speed(iter/s)": 0.140805
+      "train_speed(iter/s)": 0.074081
     },
     {
-      "acc": 0.59346585,
-      "epoch": 0.7046070460704607,
-      "grad_norm": 1.046875,
-      "learning_rate": 0.00013009985734664764,
-      "loss": 1.76646061,
-      "memory(GiB)": 15.58,
+      "acc": 0.66042156,
+      "epoch": 1.0176125244618395,
+      "grad_norm": 0.76171875,
+      "learning_rate": 0.00012453564021679692,
+      "loss": 1.37123928,
+      "memory(GiB)": 20.18,
       "step": 260,
-      "train_speed(iter/s)": 0.141165
+      "train_speed(iter/s)": 0.074295
     },
     {
-      "acc": 0.60632119,
-      "epoch": 0.7181571815718157,
-      "grad_norm": 1.078125,
-      "learning_rate": 0.00012956490727532096,
-      "loss": 1.59940271,
-      "memory(GiB)": 15.95,
+      "acc": 0.67253222,
+      "epoch": 1.0371819960861057,
+      "grad_norm": 0.76953125,
+      "learning_rate": 0.00012381711421770455,
+      "loss": 1.28407507,
+      "memory(GiB)": 19.7,
       "step": 265,
-      "train_speed(iter/s)": 0.141502
+      "train_speed(iter/s)": 0.074448
     },
     {
-      "acc": 0.59642992,
-      "epoch": 0.7317073170731707,
-      "grad_norm": 1.203125,
-      "learning_rate": 0.00012902995720399428,
-      "loss": 1.6549921,
-      "memory(GiB)": 15.95,
+      "acc": 0.66850777,
+      "epoch": 1.0567514677103718,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.0001230844464705507,
+      "loss": 1.27961807,
+      "memory(GiB)": 19.58,
       "step": 270,
-      "train_speed(iter/s)": 0.141855
+      "train_speed(iter/s)": 0.07459
     },
     {
-      "acc": 0.60280037,
-      "epoch": 0.7452574525745257,
-      "grad_norm": 1.0625,
-      "learning_rate": 0.0001284950071326676,
-      "loss": 1.63179569,
-      "memory(GiB)": 15.95,
+      "acc": 0.67196817,
+      "epoch": 1.076320939334638,
+      "grad_norm": 0.9140625,
+      "learning_rate": 0.00012233782950150186,
+      "loss": 1.28494987,
+      "memory(GiB)": 19.61,
       "step": 275,
-      "train_speed(iter/s)": 0.142164
+      "train_speed(iter/s)": 0.074728
     },
     {
-      "acc": 0.594034,
-      "epoch": 0.7588075880758808,
-      "grad_norm": 1.1796875,
-      "learning_rate": 0.00012796005706134094,
-      "loss": 1.61152973,
-      "memory(GiB)": 15.95,
+      "acc": 0.67708378,
+      "epoch": 1.095890410958904,
+      "grad_norm": 0.87109375,
+      "learning_rate": 0.00012157745950221989,
+      "loss": 1.29551096,
+      "memory(GiB)": 19.63,
       "step": 280,
-      "train_speed(iter/s)": 0.142474
+      "train_speed(iter/s)": 0.074881
     },
     {
-      "acc": 0.59763761,
-      "epoch": 0.7723577235772358,
+      "acc": 0.66973438,
+      "epoch": 1.1154598825831703,
       "grad_norm": 1.0859375,
-      "learning_rate": 0.00012742510699001425,
-      "loss": 1.66443863,
-      "memory(GiB)": 15.95,
+      "learning_rate": 0.0001208035362783079,
+      "loss": 1.27705774,
+      "memory(GiB)": 19.49,
       "step": 285,
-      "train_speed(iter/s)": 0.142791
+      "train_speed(iter/s)": 0.075029
     },
     {
-      "acc": 0.58316121,
-      "epoch": 0.7859078590785907,
-      "grad_norm": 0.73828125,
-      "learning_rate": 0.00012689015691868757,
-      "loss": 1.77028389,
-      "memory(GiB)": 15.95,
+      "acc": 0.6750237,
+      "epoch": 1.1350293542074363,
+      "grad_norm": 1.0859375,
+      "learning_rate": 0.00012001626319680648,
+      "loss": 1.25660419,
+      "memory(GiB)": 19.55,
       "step": 290,
-      "train_speed(iter/s)": 0.143067
+      "train_speed(iter/s)": 0.07515
     },
     {
-      "acc": 0.60665183,
-      "epoch": 0.7994579945799458,
-      "grad_norm": 1.25,
-      "learning_rate": 0.00012635520684736091,
-      "loss": 1.64849663,
-      "memory(GiB)": 15.95,
+      "acc": 0.624368,
+      "epoch": 1.1545988258317026,
+      "grad_norm": 1.1953125,
+      "learning_rate": 0.00011921584713275411,
+      "loss": 1.5070508,
+      "memory(GiB)": 19.52,
       "step": 295,
-      "train_speed(iter/s)": 0.143366
+      "train_speed(iter/s)": 0.075278
     },
     {
-      "acc": 0.55526123,
-      "epoch": 0.8130081300813008,
-      "grad_norm": 1.203125,
-      "learning_rate": 0.00012582025677603423,
-      "loss": 1.86367321,
-      "memory(GiB)": 15.95,
+      "acc": 0.66252189,
+      "epoch": 1.1741682974559686,
+      "grad_norm": 0.828125,
+      "learning_rate": 0.0001184024984148257,
+      "loss": 1.32014723,
+      "memory(GiB)": 19.92,
       "step": 300,
-      "train_speed(iter/s)": 0.143649
+      "train_speed(iter/s)": 0.075433
     },
     {
-      "epoch": 0.8130081300813008,
-      "eval_acc": 0.6129032258064516,
-      "eval_loss": 1.5865528583526611,
-      "eval_runtime": 44.0975,
-      "eval_samples_per_second": 0.862,
-      "eval_steps_per_second": 0.862,
+      "epoch": 1.1741682974559686,
+      "eval_acc": 0.6282796543597801,
+      "eval_loss": 1.5250990390777588,
+      "eval_runtime": 70.3986,
+      "eval_samples_per_second": 1.08,
+      "eval_steps_per_second": 0.54,
       "step": 300
     },
     {
-      "acc": 0.6019598,
-      "epoch": 0.8265582655826558,
-      "grad_norm": 1.1328125,
-      "learning_rate": 0.00012528530670470755,
-      "loss": 1.71853142,
-      "memory(GiB)": 15.95,
+      "acc": 0.67028356,
+      "epoch": 1.1937377690802349,
+      "grad_norm": 1.7109375,
+      "learning_rate": 0.00011757643077006372,
+      "loss": 1.28037386,
+      "memory(GiB)": 22.6,
       "step": 305,
-      "train_speed(iter/s)": 0.14098
+      "train_speed(iter/s)": 0.074243
     },
     {
-      "acc": 0.6009192,
-      "epoch": 0.8401084010840109,
-      "grad_norm": 1.359375,
-      "learning_rate": 0.00012475035663338086,
-      "loss": 1.72751312,
-      "memory(GiB)": 16.34,
+      "acc": 0.655305,
+      "epoch": 1.213307240704501,
+      "grad_norm": 1.1015625,
+      "learning_rate": 0.00011673786126771617,
+      "loss": 1.31057158,
+      "memory(GiB)": 19.72,
       "step": 310,
-      "train_speed(iter/s)": 0.141284
+      "train_speed(iter/s)": 0.074392
     },
     {
-      "acc": 0.58907671,
-      "epoch": 0.8536585365853658,
-      "grad_norm": 0.98046875,
-      "learning_rate": 0.0001242154065620542,
-      "loss": 1.71860523,
-      "memory(GiB)": 16.34,
+      "acc": 0.66528535,
+      "epoch": 1.2328767123287672,
+      "grad_norm": 1.6171875,
+      "learning_rate": 0.0001158870102621965,
+      "loss": 1.29698696,
+      "memory(GiB)": 19.08,
       "step": 315,
-      "train_speed(iter/s)": 0.141562
+      "train_speed(iter/s)": 0.074534
     },
     {
-      "acc": 0.62246222,
-      "epoch": 0.8672086720867209,
-      "grad_norm": 1.0234375,
-      "learning_rate": 0.00012368045649072752,
-      "loss": 1.58847914,
-      "memory(GiB)": 16.34,
+      "acc": 0.66950455,
+      "epoch": 1.2524461839530332,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00011502410133517998,
+      "loss": 1.27706356,
+      "memory(GiB)": 19.87,
       "step": 320,
-      "train_speed(iter/s)": 0.141848
+      "train_speed(iter/s)": 0.074667
     },
     {
-      "acc": 0.61927052,
-      "epoch": 0.8807588075880759,
-      "grad_norm": 1.59375,
-      "learning_rate": 0.00012314550641940084,
-      "loss": 1.54207745,
-      "memory(GiB)": 16.34,
+      "acc": 0.65843534,
+      "epoch": 1.2720156555772995,
+      "grad_norm": 1.2265625,
+      "learning_rate": 0.0001141493612368524,
+      "loss": 1.30308371,
+      "memory(GiB)": 19.87,
       "step": 325,
-      "train_speed(iter/s)": 0.142133
+      "train_speed(iter/s)": 0.0748
     },
     {
-      "acc": 0.60672359,
-      "epoch": 0.8943089430894309,
-      "grad_norm": 2.140625,
-      "learning_rate": 0.00012261055634807416,
-      "loss": 1.71789646,
-      "memory(GiB)": 16.34,
+      "acc": 0.66441913,
+      "epoch": 1.2915851272015655,
+      "grad_norm": 1.2578125,
+      "learning_rate": 0.00011326301982632583,
+      "loss": 1.26109972,
+      "memory(GiB)": 19.09,
       "step": 330,
-      "train_speed(iter/s)": 0.142392
+      "train_speed(iter/s)": 0.074935
     },
     {
-      "acc": 0.63249903,
-      "epoch": 0.907859078590786,
-      "grad_norm": 1.1328125,
-      "learning_rate": 0.0001220756062767475,
-      "loss": 1.46934195,
-      "memory(GiB)": 16.34,
+      "acc": 0.68711085,
+      "epoch": 1.3111545988258317,
+      "grad_norm": 0.95703125,
+      "learning_rate": 0.00011236531001123771,
+      "loss": 1.19278584,
+      "memory(GiB)": 19.73,
       "step": 335,
-      "train_speed(iter/s)": 0.142658
+      "train_speed(iter/s)": 0.075053
     },
     {
-      "acc": 0.60973873,
-      "epoch": 0.9214092140921409,
-      "grad_norm": 0.84375,
-      "learning_rate": 0.00012154065620542082,
-      "loss": 1.67951736,
-      "memory(GiB)": 16.34,
+      "acc": 0.66676803,
+      "epoch": 1.3307240704500978,
+      "grad_norm": 1.96875,
+      "learning_rate": 0.0001114564676865486,
+      "loss": 1.3068346,
+      "memory(GiB)": 19.84,
       "step": 340,
-      "train_speed(iter/s)": 0.142914
+      "train_speed(iter/s)": 0.075151
     },
     {
-      "acc": 0.58657136,
-      "epoch": 0.9349593495934959,
-      "grad_norm": 1.2734375,
-      "learning_rate": 0.00012100570613409413,
-      "loss": 1.68692303,
-      "memory(GiB)": 16.34,
+      "acc": 0.66865935,
+      "epoch": 1.350293542074364,
+      "grad_norm": 1.2421875,
+      "learning_rate": 0.00011053673167255516,
+      "loss": 1.30573978,
+      "memory(GiB)": 19.66,
       "step": 345,
-      "train_speed(iter/s)": 0.143167
+      "train_speed(iter/s)": 0.075271
     },
     {
-      "acc": 0.59965162,
-      "epoch": 0.948509485094851,
-      "grad_norm": 1.421875,
-      "learning_rate": 0.00012047075606276746,
-      "loss": 1.71949959,
-      "memory(GiB)": 16.34,
+      "acc": 0.66606102,
+      "epoch": 1.36986301369863,
+      "grad_norm": 0.76171875,
+      "learning_rate": 0.00010960634365213437,
+      "loss": 1.26872787,
+      "memory(GiB)": 19.73,
       "step": 350,
-      "train_speed(iter/s)": 0.143387
+      "train_speed(iter/s)": 0.075377
     },
     {
-      "epoch": 0.948509485094851,
-      "eval_acc": 0.6171254855598717,
-      "eval_loss": 1.562721610069275,
-      "eval_runtime": 44.0566,
-      "eval_samples_per_second": 0.863,
-      "eval_steps_per_second": 0.863,
+      "epoch": 1.36986301369863,
+      "eval_acc": 0.6315003927729772,
+      "eval_loss": 1.5066882371902466,
+      "eval_runtime": 72.5685,
+      "eval_samples_per_second": 1.047,
+      "eval_steps_per_second": 0.524,
       "step": 350
     },
     {
-      "acc": 0.59322553,
-      "epoch": 0.962059620596206,
-      "grad_norm": 1.3046875,
-      "learning_rate": 0.0001199358059914408,
-      "loss": 1.66712799,
-      "memory(GiB)": 16.34,
+      "acc": 0.67307239,
+      "epoch": 1.3894324853228963,
+      "grad_norm": 1.1796875,
+      "learning_rate": 0.0001086655481072354,
+      "loss": 1.27917318,
+      "memory(GiB)": 22.92,
       "step": 355,
-      "train_speed(iter/s)": 0.14111
+      "train_speed(iter/s)": 0.074318
     },
     {
-      "acc": 0.59898105,
-      "epoch": 0.975609756097561,
-      "grad_norm": 1.265625,
-      "learning_rate": 0.00011940085592011411,
-      "loss": 1.66754684,
-      "memory(GiB)": 16.34,
+      "acc": 0.65870218,
+      "epoch": 1.4090019569471623,
+      "grad_norm": 3.609375,
+      "learning_rate": 0.00010771459225463617,
+      "loss": 1.33731461,
+      "memory(GiB)": 19.67,
       "step": 360,
-      "train_speed(iter/s)": 0.141373
+      "train_speed(iter/s)": 0.074416
     },
     {
-      "acc": 0.60211391,
-      "epoch": 0.989159891598916,
-      "grad_norm": 1.359375,
-      "learning_rate": 0.00011886590584878744,
-      "loss": 1.67542057,
-      "memory(GiB)": 16.34,
+      "acc": 0.68150563,
+      "epoch": 1.4285714285714286,
+      "grad_norm": 0.9296875,
+      "learning_rate": 0.00010675372598098113,
+      "loss": 1.20515957,
+      "memory(GiB)": 19.99,
       "step": 365,
-      "train_speed(iter/s)": 0.141638
+      "train_speed(iter/s)": 0.07451
     },
     {
-      "acc": 0.63076615,
-      "epoch": 1.002710027100271,
-      "grad_norm": 0.8515625,
-      "learning_rate": 0.00011833095577746076,
-      "loss": 1.51196938,
-      "memory(GiB)": 16.34,
+      "acc": 0.66793504,
+      "epoch": 1.4481409001956946,
+      "grad_norm": 1.03125,
+      "learning_rate": 0.00010578320177711743,
+      "loss": 1.31133595,
+      "memory(GiB)": 19.9,
       "step": 370,
-      "train_speed(iter/s)": 0.141355
+      "train_speed(iter/s)": 0.074613
     },
     {
-      "acc": 0.63542643,
-      "epoch": 1.016260162601626,
-      "grad_norm": 0.828125,
-      "learning_rate": 0.00011779600570613407,
-      "loss": 1.38132334,
-      "memory(GiB)": 16.34,
+      "acc": 0.66840873,
+      "epoch": 1.467710371819961,
+      "grad_norm": 0.9453125,
+      "learning_rate": 0.00010480327467174705,
+      "loss": 1.27730675,
+      "memory(GiB)": 19.91,
       "step": 375,
-      "train_speed(iter/s)": 0.141603
+      "train_speed(iter/s)": 0.074709
     },
     {
-      "acc": 0.66358423,
-      "epoch": 1.029810298102981,
-      "grad_norm": 0.859375,
-      "learning_rate": 0.00011726105563480742,
-      "loss": 1.41512909,
-      "memory(GiB)": 16.34,
+      "acc": 0.6621439,
+      "epoch": 1.487279843444227,
+      "grad_norm": 0.7890625,
+      "learning_rate": 0.00010381420216441152,
+      "loss": 1.29670372,
+      "memory(GiB)": 19.65,
       "step": 380,
-      "train_speed(iter/s)": 0.141845
+      "train_speed(iter/s)": 0.074824
     },
     {
-      "acc": 0.63692493,
-      "epoch": 1.043360433604336,
-      "grad_norm": 1.0859375,
-      "learning_rate": 0.00011672610556348073,
-      "loss": 1.38078823,
-      "memory(GiB)": 16.34,
+      "acc": 0.66805882,
+      "epoch": 1.5068493150684932,
+      "grad_norm": 0.8203125,
+      "learning_rate": 0.00010281624415782804,
+      "loss": 1.23922901,
+      "memory(GiB)": 19.77,
       "step": 385,
-      "train_speed(iter/s)": 0.142082
+      "train_speed(iter/s)": 0.074927
     },
     {
-      "acc": 0.6641499,
-      "epoch": 1.056910569105691,
-      "grad_norm": 1.125,
-      "learning_rate": 0.00011619115549215406,
-      "loss": 1.3458046,
-      "memory(GiB)": 16.34,
+      "acc": 0.66435666,
+      "epoch": 1.5264187866927594,
+      "grad_norm": 0.82421875,
+      "learning_rate": 0.0001018096628895935,
+      "loss": 1.27945633,
+      "memory(GiB)": 19.79,
       "step": 390,
-      "train_speed(iter/s)": 0.142305
+      "train_speed(iter/s)": 0.075033
     },
     {
-      "acc": 0.67276783,
-      "epoch": 1.070460704607046,
-      "grad_norm": 0.98828125,
-      "learning_rate": 0.00011565620542082738,
-      "loss": 1.2076004,
-      "memory(GiB)": 16.34,
+      "acc": 0.68444743,
+      "epoch": 1.5459882583170255,
+      "grad_norm": 0.98046875,
+      "learning_rate": 0.00010079472286327533,
+      "loss": 1.2325819,
+      "memory(GiB)": 19.55,
       "step": 395,
-      "train_speed(iter/s)": 0.142533
+      "train_speed(iter/s)": 0.075133
     },
     {
-      "acc": 0.65821433,
-      "epoch": 1.084010840108401,
-      "grad_norm": 1.15625,
-      "learning_rate": 0.0001151212553495007,
-      "loss": 1.38641891,
-      "memory(GiB)": 16.34,
+      "acc": 0.68633671,
+      "epoch": 1.5655577299412915,
+      "grad_norm": 1.171875,
+      "learning_rate": 9.977169077890672e-05,
+      "loss": 1.26248102,
+      "memory(GiB)": 19.79,
       "step": 400,
-      "train_speed(iter/s)": 0.142742
+      "train_speed(iter/s)": 0.075233
     },
     {
-      "epoch": 1.084010840108401,
-      "eval_acc": 0.6162810336091876,
-      "eval_loss": 1.5887514352798462,
-      "eval_runtime": 44.1624,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
+      "epoch": 1.5655577299412915,
+      "eval_acc": 0.6297721916732129,
+      "eval_loss": 1.5114485025405884,
+      "eval_runtime": 70.7985,
+      "eval_samples_per_second": 1.073,
+      "eval_steps_per_second": 0.537,
       "step": 400
     },
     {
-      "acc": 0.63352804,
-      "epoch": 1.0975609756097562,
-      "grad_norm": 1.2421875,
-      "learning_rate": 0.00011458630527817403,
-      "loss": 1.44696302,
-      "memory(GiB)": 16.34,
+      "acc": 0.67859097,
+      "epoch": 1.5851272015655578,
+      "grad_norm": 1.046875,
+      "learning_rate": 9.874083546290482e-05,
+      "loss": 1.2065486,
+      "memory(GiB)": 22.72,
       "step": 405,
-      "train_speed(iter/s)": 0.140759
+      "train_speed(iter/s)": 0.074347
     },
     {
-      "acc": 0.61955671,
-      "epoch": 1.1111111111111112,
-      "grad_norm": 1.5546875,
-      "learning_rate": 0.00011405135520684734,
-      "loss": 1.48915672,
-      "memory(GiB)": 16.34,
+      "acc": 0.66178751,
+      "epoch": 1.604696673189824,
+      "grad_norm": 0.96484375,
+      "learning_rate": 9.770242779743008e-05,
+      "loss": 1.30969448,
+      "memory(GiB)": 20.13,
       "step": 410,
-      "train_speed(iter/s)": 0.140987
+      "train_speed(iter/s)": 0.074453
     },
     {
-      "acc": 0.69259667,
-      "epoch": 1.1246612466124661,
-      "grad_norm": 0.796875,
-      "learning_rate": 0.00011351640513552069,
-      "loss": 1.22069292,
-      "memory(GiB)": 16.34,
+      "acc": 0.65872512,
+      "epoch": 1.62426614481409,
+      "grad_norm": 0.74609375,
+      "learning_rate": 9.665674064920533e-05,
+      "loss": 1.27483397,
+      "memory(GiB)": 20.17,
       "step": 415,
-      "train_speed(iter/s)": 0.141191
+      "train_speed(iter/s)": 0.074534
     },
     {
-      "acc": 0.67346158,
-      "epoch": 1.1382113821138211,
-      "grad_norm": 1.171875,
-      "learning_rate": 0.000112981455064194,
-      "loss": 1.36795778,
-      "memory(GiB)": 16.34,
+      "acc": 0.66567349,
+      "epoch": 1.643835616438356,
+      "grad_norm": 0.87109375,
+      "learning_rate": 9.560404879781353e-05,
+      "loss": 1.31585007,
+      "memory(GiB)": 20.07,
       "step": 420,
-      "train_speed(iter/s)": 0.141417
+      "train_speed(iter/s)": 0.074639
     },
     {
-      "acc": 0.67346702,
-      "epoch": 1.151761517615176,
-      "grad_norm": 1.0390625,
-      "learning_rate": 0.00011244650499286732,
-      "loss": 1.27111759,
-      "memory(GiB)": 16.34,
+      "acc": 0.66216898,
+      "epoch": 1.6634050880626223,
+      "grad_norm": 0.85546875,
+      "learning_rate": 9.454462886349281e-05,
+      "loss": 1.32738457,
+      "memory(GiB)": 19.43,
       "step": 425,
-      "train_speed(iter/s)": 0.141634
+      "train_speed(iter/s)": 0.074732
     },
     {
-      "acc": 0.61625972,
-      "epoch": 1.165311653116531,
-      "grad_norm": 1.2578125,
-      "learning_rate": 0.00011191155492154065,
-      "loss": 1.42474003,
-      "memory(GiB)": 16.34,
+      "acc": 0.6608973,
+      "epoch": 1.6829745596868886,
+      "grad_norm": 1.1328125,
+      "learning_rate": 9.347875923444772e-05,
+      "loss": 1.2792593,
+      "memory(GiB)": 20.05,
       "step": 430,
-      "train_speed(iter/s)": 0.141853
+      "train_speed(iter/s)": 0.074827
     },
     {
-      "acc": 0.65880041,
-      "epoch": 1.1788617886178863,
-      "grad_norm": 1.2578125,
-      "learning_rate": 0.00011137660485021397,
-      "loss": 1.33981524,
-      "memory(GiB)": 16.34,
+      "acc": 0.65830297,
+      "epoch": 1.7025440313111546,
+      "grad_norm": 0.94921875,
+      "learning_rate": 9.240671999369607e-05,
+      "loss": 1.34132614,
+      "memory(GiB)": 19.82,
       "step": 435,
-      "train_speed(iter/s)": 0.142048
+      "train_speed(iter/s)": 0.074914
     },
     {
-      "acc": 0.63503499,
-      "epoch": 1.1924119241192412,
-      "grad_norm": 1.2265625,
-      "learning_rate": 0.00011084165477888728,
-      "loss": 1.44122829,
-      "memory(GiB)": 16.34,
+      "acc": 0.68926673,
+      "epoch": 1.7221135029354206,
+      "grad_norm": 0.76953125,
+      "learning_rate": 9.132879284547038e-05,
+      "loss": 1.15266266,
+      "memory(GiB)": 19.28,
       "step": 440,
-      "train_speed(iter/s)": 0.142254
+      "train_speed(iter/s)": 0.074997
     },
     {
-      "acc": 0.61518383,
-      "epoch": 1.2059620596205962,
-      "grad_norm": 1.4765625,
-      "learning_rate": 0.00011030670470756061,
-      "loss": 1.54818697,
-      "memory(GiB)": 16.34,
+      "acc": 0.65699558,
+      "epoch": 1.741682974559687,
+      "grad_norm": 0.96484375,
+      "learning_rate": 9.024526104119312e-05,
+      "loss": 1.32417459,
+      "memory(GiB)": 19.29,
       "step": 445,
-      "train_speed(iter/s)": 0.142446
+      "train_speed(iter/s)": 0.075079
     },
     {
-      "acc": 0.68010144,
-      "epoch": 1.2195121951219512,
-      "grad_norm": 1.3828125,
-      "learning_rate": 0.00010977175463623394,
-      "loss": 1.28560524,
-      "memory(GiB)": 16.34,
+      "acc": 0.68860197,
+      "epoch": 1.7612524461839532,
+      "grad_norm": 0.8203125,
+      "learning_rate": 8.91564093050458e-05,
+      "loss": 1.20134068,
+      "memory(GiB)": 19.33,
       "step": 450,
-      "train_speed(iter/s)": 0.142639
+      "train_speed(iter/s)": 0.07515
     },
     {
-      "epoch": 1.2195121951219512,
-      "eval_acc": 0.6230366492146597,
-      "eval_loss": 1.5483555793762207,
-      "eval_runtime": 44.0588,
-      "eval_samples_per_second": 0.862,
-      "eval_steps_per_second": 0.862,
+      "epoch": 1.7612524461839532,
+      "eval_acc": 0.6351924587588373,
+      "eval_loss": 1.4908838272094727,
+      "eval_runtime": 71.5161,
+      "eval_samples_per_second": 1.063,
+      "eval_steps_per_second": 0.531,
       "step": 450
     },
     {
-      "acc": 0.64718337,
-      "epoch": 1.2330623306233062,
-      "grad_norm": 1.4375,
-      "learning_rate": 0.00010923680456490727,
-      "loss": 1.3778326,
-      "memory(GiB)": 16.34,
+      "acc": 0.65404687,
+      "epoch": 1.7808219178082192,
+      "grad_norm": 1.0078125,
+      "learning_rate": 8.806252375915052e-05,
+      "loss": 1.31502724,
+      "memory(GiB)": 19.13,
       "step": 455,
-      "train_speed(iter/s)": 0.140879
+      "train_speed(iter/s)": 0.074358
     },
     {
-      "acc": 0.68637676,
-      "epoch": 1.2466124661246614,
-      "grad_norm": 1.453125,
-      "learning_rate": 0.00010870185449358059,
-      "loss": 1.21860657,
-      "memory(GiB)": 16.34,
+      "acc": 0.69379678,
+      "epoch": 1.8003913894324852,
+      "grad_norm": 1.1015625,
+      "learning_rate": 8.696389184838471e-05,
+      "loss": 1.1870966,
+      "memory(GiB)": 20.18,
       "step": 460,
-      "train_speed(iter/s)": 0.141097
+      "train_speed(iter/s)": 0.074437
     },
     {
-      "acc": 0.63144946,
-      "epoch": 1.2601626016260163,
-      "grad_norm": 1.0703125,
-      "learning_rate": 0.00010816690442225392,
-      "loss": 1.38577023,
-      "memory(GiB)": 16.34,
+      "acc": 0.67447538,
+      "epoch": 1.8199608610567515,
+      "grad_norm": 1.2890625,
+      "learning_rate": 8.586080226484789e-05,
+      "loss": 1.19511604,
+      "memory(GiB)": 20.09,
       "step": 465,
-      "train_speed(iter/s)": 0.141294
+      "train_speed(iter/s)": 0.074531
     },
     {
-      "acc": 0.66895781,
-      "epoch": 1.2737127371273713,
-      "grad_norm": 1.296875,
-      "learning_rate": 0.00010763195435092724,
-      "loss": 1.29167385,
-      "memory(GiB)": 16.34,
+      "acc": 0.67230067,
+      "epoch": 1.8395303326810177,
+      "grad_norm": 1.0390625,
+      "learning_rate": 8.475354487200092e-05,
+      "loss": 1.30591021,
+      "memory(GiB)": 19.29,
       "step": 470,
-      "train_speed(iter/s)": 0.141488
+      "train_speed(iter/s)": 0.074608
     },
     {
-      "acc": 0.6553885,
-      "epoch": 1.2872628726287263,
-      "grad_norm": 1.1484375,
-      "learning_rate": 0.00010709700427960055,
-      "loss": 1.35011473,
-      "memory(GiB)": 16.34,
+      "acc": 0.65006552,
+      "epoch": 1.8590998043052838,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.364241062849732e-05,
+      "loss": 1.35613279,
+      "memory(GiB)": 19.51,
       "step": 475,
-      "train_speed(iter/s)": 0.141687
+      "train_speed(iter/s)": 0.07469
     },
     {
-      "acc": 0.65576148,
-      "epoch": 1.3008130081300813,
-      "grad_norm": 1.1796875,
-      "learning_rate": 0.00010656205420827388,
-      "loss": 1.39116755,
-      "memory(GiB)": 16.34,
+      "acc": 0.66248426,
+      "epoch": 1.8786692759295498,
+      "grad_norm": 1.0703125,
+      "learning_rate": 8.252769151172682e-05,
+      "loss": 1.34706697,
+      "memory(GiB)": 19.16,
       "step": 480,
-      "train_speed(iter/s)": 0.141863
+      "train_speed(iter/s)": 0.074779
     },
     {
-      "acc": 0.62025633,
-      "epoch": 1.3143631436314362,
-      "grad_norm": 1.375,
-      "learning_rate": 0.00010602710413694721,
-      "loss": 1.55190315,
-      "memory(GiB)": 16.34,
+      "acc": 0.66462736,
+      "epoch": 1.898238747553816,
+      "grad_norm": 0.8515625,
+      "learning_rate": 8.140968044109134e-05,
+      "loss": 1.31343336,
+      "memory(GiB)": 19.17,
       "step": 485,
-      "train_speed(iter/s)": 0.142058
+      "train_speed(iter/s)": 0.07486
     },
     {
-      "acc": 0.65667233,
-      "epoch": 1.3279132791327912,
-      "grad_norm": 1.1484375,
-      "learning_rate": 0.00010549215406562054,
-      "loss": 1.30497828,
-      "memory(GiB)": 16.34,
+      "acc": 0.65373287,
+      "epoch": 1.9178082191780823,
+      "grad_norm": 1.078125,
+      "learning_rate": 8.028867120103326e-05,
+      "loss": 1.31145601,
+      "memory(GiB)": 19.46,
       "step": 490,
-      "train_speed(iter/s)": 0.142237
+      "train_speed(iter/s)": 0.074941
     },
     {
-      "acc": 0.6458045,
-      "epoch": 1.3414634146341464,
-      "grad_norm": 2.421875,
-      "learning_rate": 0.00010495720399429386,
-      "loss": 1.40175285,
-      "memory(GiB)": 16.34,
+      "acc": 0.6731041,
+      "epoch": 1.9373776908023483,
+      "grad_norm": 0.89453125,
+      "learning_rate": 7.916495836383648e-05,
+      "loss": 1.24272699,
+      "memory(GiB)": 19.45,
       "step": 495,
-      "train_speed(iter/s)": 0.14241
+      "train_speed(iter/s)": 0.075011
     },
     {
-      "acc": 0.65004997,
-      "epoch": 1.3550135501355014,
-      "grad_norm": 1.265625,
-      "learning_rate": 0.00010442225392296718,
-      "loss": 1.31625471,
-      "memory(GiB)": 16.34,
+      "acc": 0.66485052,
+      "epoch": 1.9569471624266144,
+      "grad_norm": 1.03125,
+      "learning_rate": 7.80388372122204e-05,
+      "loss": 1.28164721,
+      "memory(GiB)": 19.24,
       "step": 500,
-      "train_speed(iter/s)": 0.142583
+      "train_speed(iter/s)": 0.07509
     },
     {
-      "epoch": 1.3550135501355014,
-      "eval_acc": 0.61932106063165,
-      "eval_loss": 1.535814642906189,
-      "eval_runtime": 44.07,
-      "eval_samples_per_second": 0.862,
-      "eval_steps_per_second": 0.862,
+      "epoch": 1.9569471624266144,
+      "eval_acc": 0.6349567949725059,
+      "eval_loss": 1.483258843421936,
+      "eval_runtime": 72.4797,
+      "eval_samples_per_second": 1.049,
+      "eval_steps_per_second": 0.524,
       "step": 500
     },
     {
-      "acc": 0.65512385,
-      "epoch": 1.3685636856368564,
-      "grad_norm": 0.9375,
-      "learning_rate": 0.00010388730385164051,
-      "loss": 1.2989893,
-      "memory(GiB)": 16.34,
+      "acc": 0.68325486,
+      "epoch": 1.9765166340508806,
+      "grad_norm": 1.2890625,
+      "learning_rate": 7.691060366174728e-05,
+      "loss": 1.2257865,
+      "memory(GiB)": 22.98,
       "step": 505,
-      "train_speed(iter/s)": 0.140995
+      "train_speed(iter/s)": 0.074371
     },
     {
-      "acc": 0.65444102,
-      "epoch": 1.3821138211382114,
-      "grad_norm": 1.5625,
-      "learning_rate": 0.00010335235378031382,
-      "loss": 1.32140932,
-      "memory(GiB)": 16.34,
+      "acc": 0.68977013,
+      "epoch": 1.9960861056751469,
+      "grad_norm": 1.0234375,
+      "learning_rate": 7.578055418306327e-05,
+      "loss": 1.25723343,
+      "memory(GiB)": 19.56,
       "step": 510,
-      "train_speed(iter/s)": 0.141172
+      "train_speed(iter/s)": 0.074471
     },
     {
-      "acc": 0.66242504,
-      "epoch": 1.3956639566395663,
-      "grad_norm": 0.9140625,
-      "learning_rate": 0.00010281740370898714,
-      "loss": 1.34259834,
-      "memory(GiB)": 16.34,
+      "acc": 0.72185702,
+      "epoch": 2.015655577299413,
+      "grad_norm": 0.7890625,
+      "learning_rate": 7.464898572399353e-05,
+      "loss": 1.01715631,
+      "memory(GiB)": 20.07,
       "step": 515,
-      "train_speed(iter/s)": 0.141336
+      "train_speed(iter/s)": 0.074591
     },
     {
-      "acc": 0.62716827,
-      "epoch": 1.4092140921409215,
-      "grad_norm": 1.4296875,
-      "learning_rate": 0.00010228245363766048,
-      "loss": 1.46508284,
-      "memory(GiB)": 16.34,
+      "acc": 0.71889682,
+      "epoch": 2.035225048923679,
+      "grad_norm": 1.0625,
+      "learning_rate": 7.351619563151208e-05,
+      "loss": 1.03077154,
+      "memory(GiB)": 19.92,
       "step": 520,
-      "train_speed(iter/s)": 0.141508
+      "train_speed(iter/s)": 0.074683
     },
     {
-      "acc": 0.61702657,
-      "epoch": 1.4227642276422765,
-      "grad_norm": 1.5546875,
-      "learning_rate": 0.0001017475035663338,
-      "loss": 1.43188276,
-      "memory(GiB)": 16.34,
+      "acc": 0.7505311,
+      "epoch": 2.0547945205479454,
+      "grad_norm": 1.9609375,
+      "learning_rate": 7.238248157360663e-05,
+      "loss": 0.93218956,
+      "memory(GiB)": 19.85,
       "step": 525,
-      "train_speed(iter/s)": 0.141681
+      "train_speed(iter/s)": 0.07477
     },
     {
-      "acc": 0.63238263,
-      "epoch": 1.4363143631436315,
-      "grad_norm": 1.3671875,
-      "learning_rate": 0.00010121255349500713,
-      "loss": 1.52214499,
-      "memory(GiB)": 16.34,
+      "acc": 0.7315311,
+      "epoch": 2.0743639921722115,
+      "grad_norm": 1.1875,
+      "learning_rate": 7.124814146105921e-05,
+      "loss": 0.96330833,
+      "memory(GiB)": 19.87,
       "step": 530,
-      "train_speed(iter/s)": 0.141854
+      "train_speed(iter/s)": 0.074853
     },
     {
-      "acc": 0.6323854,
-      "epoch": 1.4498644986449865,
-      "grad_norm": 1.375,
-      "learning_rate": 0.00010067760342368045,
-      "loss": 1.38910236,
-      "memory(GiB)": 16.34,
+      "acc": 0.75555606,
+      "epoch": 2.0939334637964775,
+      "grad_norm": 1.3515625,
+      "learning_rate": 7.011347336916277e-05,
+      "loss": 0.86877937,
+      "memory(GiB)": 18.46,
       "step": 535,
-      "train_speed(iter/s)": 0.142019
+      "train_speed(iter/s)": 0.074938
     },
     {
-      "acc": 0.60772176,
-      "epoch": 1.4634146341463414,
-      "grad_norm": 1.21875,
-      "learning_rate": 0.00010014265335235376,
-      "loss": 1.46768923,
-      "memory(GiB)": 16.34,
+      "acc": 0.74034052,
+      "epoch": 2.1135029354207435,
+      "grad_norm": 1.546875,
+      "learning_rate": 6.897877545939475e-05,
+      "loss": 0.90922012,
+      "memory(GiB)": 19.89,
       "step": 540,
-      "train_speed(iter/s)": 0.14218
+      "train_speed(iter/s)": 0.075027
     },
     {
-      "acc": 0.67315965,
-      "epoch": 1.4769647696476964,
-      "grad_norm": 1.34375,
-      "learning_rate": 9.96077032810271e-05,
-      "loss": 1.25136938,
-      "memory(GiB)": 16.34,
+      "acc": 0.72400937,
+      "epoch": 2.1330724070450096,
+      "grad_norm": 1.90625,
+      "learning_rate": 6.784434590106808e-05,
+      "loss": 0.98424711,
+      "memory(GiB)": 19.11,
       "step": 545,
-      "train_speed(iter/s)": 0.142338
+      "train_speed(iter/s)": 0.075114
     },
     {
-      "acc": 0.66408758,
-      "epoch": 1.4905149051490514,
-      "grad_norm": 2.046875,
-      "learning_rate": 9.907275320970041e-05,
-      "loss": 1.25655928,
-      "memory(GiB)": 16.34,
+      "acc": 0.77706275,
+      "epoch": 2.152641878669276,
+      "grad_norm": 1.359375,
+      "learning_rate": 6.671048279297972e-05,
+      "loss": 0.80820856,
+      "memory(GiB)": 19.86,
       "step": 550,
-      "train_speed(iter/s)": 0.142495
+      "train_speed(iter/s)": 0.075193
     },
     {
-      "epoch": 1.4905149051490514,
-      "eval_acc": 0.6254011146765749,
-      "eval_loss": 1.5346177816390991,
-      "eval_runtime": 44.1805,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
+      "epoch": 2.152641878669276,
+      "eval_acc": 0.6260015710919089,
+      "eval_loss": 1.6081812381744385,
+      "eval_runtime": 68.6973,
+      "eval_samples_per_second": 1.106,
+      "eval_steps_per_second": 0.553,
       "step": 550
     },
     {
-      "acc": 0.68655195,
-      "epoch": 1.5040650406504064,
-      "grad_norm": 1.125,
-      "learning_rate": 9.853780313837375e-05,
-      "loss": 1.08906736,
-      "memory(GiB)": 16.34,
+      "acc": 0.75351696,
+      "epoch": 2.172211350293542,
+      "grad_norm": 2.015625,
+      "learning_rate": 6.55774840850782e-05,
+      "loss": 0.86192131,
+      "memory(GiB)": 22.21,
       "step": 555,
-      "train_speed(iter/s)": 0.141037
+      "train_speed(iter/s)": 0.074578
     },
     {
-      "acc": 0.65898876,
-      "epoch": 1.5176151761517616,
-      "grad_norm": 1.4296875,
-      "learning_rate": 9.800285306704707e-05,
-      "loss": 1.26696377,
-      "memory(GiB)": 16.34,
+      "acc": 0.74249997,
+      "epoch": 2.191780821917808,
+      "grad_norm": 1.4609375,
+      "learning_rate": 6.444564750017003e-05,
+      "loss": 0.91982813,
+      "memory(GiB)": 19.87,
       "step": 560,
-      "train_speed(iter/s)": 0.14119
+      "train_speed(iter/s)": 0.074665
     },
     {
-      "acc": 0.63674688,
-      "epoch": 1.5311653116531165,
-      "grad_norm": 1.5390625,
-      "learning_rate": 9.746790299572039e-05,
-      "loss": 1.38277016,
-      "memory(GiB)": 16.34,
+      "acc": 0.73636398,
+      "epoch": 2.2113502935420746,
+      "grad_norm": 1.9375,
+      "learning_rate": 6.331527045568573e-05,
+      "loss": 0.93448582,
+      "memory(GiB)": 19.33,
       "step": 565,
-      "train_speed(iter/s)": 0.141349
+      "train_speed(iter/s)": 0.074752
     },
     {
-      "acc": 0.67204466,
-      "epoch": 1.5447154471544715,
-      "grad_norm": 1.234375,
-      "learning_rate": 9.693295292439372e-05,
-      "loss": 1.25243311,
-      "memory(GiB)": 16.34,
+      "acc": 0.74081583,
+      "epoch": 2.2309197651663406,
+      "grad_norm": 2.21875,
+      "learning_rate": 6.218664998552634e-05,
+      "loss": 0.94956303,
+      "memory(GiB)": 19.8,
       "step": 570,
-      "train_speed(iter/s)": 0.141502
+      "train_speed(iter/s)": 0.074842
     },
     {
-      "acc": 0.67878027,
-      "epoch": 1.5582655826558267,
-      "grad_norm": 1.3828125,
-      "learning_rate": 9.639800285306703e-05,
-      "loss": 1.18253031,
-      "memory(GiB)": 16.34,
+      "acc": 0.74573116,
+      "epoch": 2.2504892367906066,
+      "grad_norm": 2.546875,
+      "learning_rate": 6.106008266201046e-05,
+      "loss": 0.88486786,
+      "memory(GiB)": 19.92,
       "step": 575,
-      "train_speed(iter/s)": 0.141663
+      "train_speed(iter/s)": 0.074925
     },
     {
-      "acc": 0.64645357,
-      "epoch": 1.5718157181571817,
-      "grad_norm": 1.671875,
-      "learning_rate": 9.586305278174036e-05,
-      "loss": 1.39789114,
-      "memory(GiB)": 16.34,
+      "acc": 0.75495067,
+      "epoch": 2.2700587084148727,
+      "grad_norm": 2.09375,
+      "learning_rate": 5.9935864517942844e-05,
+      "loss": 0.84776802,
+      "memory(GiB)": 19.89,
       "step": 580,
-      "train_speed(iter/s)": 0.141812
+      "train_speed(iter/s)": 0.075
     },
     {
-      "acc": 0.63896599,
-      "epoch": 1.5853658536585367,
-      "grad_norm": 1.5546875,
-      "learning_rate": 9.532810271041368e-05,
-      "loss": 1.37434454,
-      "memory(GiB)": 16.34,
+      "acc": 0.74743519,
+      "epoch": 2.2896281800391387,
+      "grad_norm": 1.5859375,
+      "learning_rate": 5.881429096882449e-05,
+      "loss": 0.92330503,
+      "memory(GiB)": 19.03,
       "step": 585,
-      "train_speed(iter/s)": 0.141961
+      "train_speed(iter/s)": 0.075076
     },
     {
-      "acc": 0.67930498,
-      "epoch": 1.5989159891598916,
-      "grad_norm": 1.3671875,
-      "learning_rate": 9.479315263908701e-05,
-      "loss": 1.24675446,
-      "memory(GiB)": 16.34,
+      "acc": 0.74913769,
+      "epoch": 2.309197651663405,
+      "grad_norm": 1.6640625,
+      "learning_rate": 5.769565673522515e-05,
+      "loss": 0.92942295,
+      "memory(GiB)": 20.04,
       "step": 590,
-      "train_speed(iter/s)": 0.142113
+      "train_speed(iter/s)": 0.075149
     },
     {
-      "acc": 0.66582651,
-      "epoch": 1.6124661246612466,
-      "grad_norm": 1.1171875,
-      "learning_rate": 9.425820256776034e-05,
-      "loss": 1.33937092,
-      "memory(GiB)": 16.34,
+      "acc": 0.74875064,
+      "epoch": 2.328767123287671,
+      "grad_norm": 1.25,
+      "learning_rate": 5.658025576533832e-05,
+      "loss": 0.90142069,
+      "memory(GiB)": 19.96,
       "step": 595,
-      "train_speed(iter/s)": 0.142256
+      "train_speed(iter/s)": 0.075215
     },
     {
-      "acc": 0.65553112,
-      "epoch": 1.6260162601626016,
-      "grad_norm": 1.5859375,
-      "learning_rate": 9.372325249643366e-05,
-      "loss": 1.25127707,
-      "memory(GiB)": 16.34,
+      "acc": 0.74648356,
+      "epoch": 2.3483365949119372,
+      "grad_norm": 1.65625,
+      "learning_rate": 5.546838115773929e-05,
+      "loss": 0.91528139,
+      "memory(GiB)": 19.84,
       "step": 600,
-      "train_speed(iter/s)": 0.142394
+      "train_speed(iter/s)": 0.075292
     },
     {
-      "epoch": 1.6260162601626016,
-      "eval_acc": 0.6233744299949333,
-      "eval_loss": 1.5426762104034424,
-      "eval_runtime": 44.3564,
-      "eval_samples_per_second": 0.857,
-      "eval_steps_per_second": 0.857,
+      "epoch": 2.3483365949119372,
+      "eval_acc": 0.6284367635506677,
+      "eval_loss": 1.593437910079956,
+      "eval_runtime": 68.9856,
+      "eval_samples_per_second": 1.102,
+      "eval_steps_per_second": 0.551,
       "step": 600
     },
     {
-      "acc": 0.64805899,
-      "epoch": 1.6395663956639566,
-      "grad_norm": 1.078125,
-      "learning_rate": 9.318830242510699e-05,
-      "loss": 1.30654621,
-      "memory(GiB)": 16.34,
+      "acc": 0.75246172,
+      "epoch": 2.3679060665362037,
+      "grad_norm": 1.2109375,
+      "learning_rate": 5.4360325084366416e-05,
+      "loss": 0.87402363,
+      "memory(GiB)": 22.69,
       "step": 605,
-      "train_speed(iter/s)": 0.141073
+      "train_speed(iter/s)": 0.074706
     },
     {
-      "acc": 0.67441335,
-      "epoch": 1.6531165311653115,
-      "grad_norm": 1.34375,
-      "learning_rate": 9.26533523537803e-05,
-      "loss": 1.20785751,
-      "memory(GiB)": 16.34,
+      "acc": 0.74078665,
+      "epoch": 2.3874755381604698,
+      "grad_norm": 1.0390625,
+      "learning_rate": 5.3256378713745815e-05,
+      "loss": 0.91142588,
+      "memory(GiB)": 20.15,
       "step": 610,
-      "train_speed(iter/s)": 0.141227
+      "train_speed(iter/s)": 0.074788
     },
     {
-      "acc": 0.64989614,
-      "epoch": 1.6666666666666665,
-      "grad_norm": 2.25,
-      "learning_rate": 9.211840228245362e-05,
-      "loss": 1.41231976,
-      "memory(GiB)": 16.34,
+      "acc": 0.75772052,
+      "epoch": 2.407045009784736,
+      "grad_norm": 2.03125,
+      "learning_rate": 5.21568321344799e-05,
+      "loss": 0.85517597,
+      "memory(GiB)": 19.37,
       "step": 615,
-      "train_speed(iter/s)": 0.141374
+      "train_speed(iter/s)": 0.074857
     },
     {
-      "acc": 0.65118213,
-      "epoch": 1.6802168021680217,
-      "grad_norm": 1.34375,
-      "learning_rate": 9.158345221112695e-05,
-      "loss": 1.33156195,
-      "memory(GiB)": 16.34,
+      "acc": 0.75341692,
+      "epoch": 2.426614481409002,
+      "grad_norm": 1.40625,
+      "learning_rate": 5.10619742790194e-05,
+      "loss": 0.87981377,
+      "memory(GiB)": 18.91,
       "step": 620,
-      "train_speed(iter/s)": 0.141527
+      "train_speed(iter/s)": 0.074925
     },
     {
-      "acc": 0.68491917,
-      "epoch": 1.6937669376693767,
-      "grad_norm": 2.078125,
-      "learning_rate": 9.104850213980028e-05,
-      "loss": 1.15608921,
-      "memory(GiB)": 16.34,
+      "acc": 0.76221485,
+      "epoch": 2.446183953033268,
+      "grad_norm": 5.5625,
+      "learning_rate": 4.9972092847739603e-05,
+      "loss": 0.89623175,
+      "memory(GiB)": 20.27,
       "step": 625,
-      "train_speed(iter/s)": 0.141656
+      "train_speed(iter/s)": 0.074994
     },
     {
-      "acc": 0.6636075,
-      "epoch": 1.7073170731707317,
-      "grad_norm": 1.2890625,
-      "learning_rate": 9.051355206847361e-05,
-      "loss": 1.3377409,
-      "memory(GiB)": 16.34,
+      "acc": 0.74322577,
+      "epoch": 2.4657534246575343,
+      "grad_norm": 1.6796875,
+      "learning_rate": 4.8887474233339963e-05,
+      "loss": 0.89493027,
+      "memory(GiB)": 19.38,
       "step": 630,
-      "train_speed(iter/s)": 0.141795
+      "train_speed(iter/s)": 0.075068
     },
     {
-      "acc": 0.66115265,
-      "epoch": 1.7208672086720869,
-      "grad_norm": 1.796875,
-      "learning_rate": 8.997860199714693e-05,
-      "loss": 1.41666918,
-      "memory(GiB)": 16.34,
+      "acc": 0.74455509,
+      "epoch": 2.4853228962818004,
+      "grad_norm": 1.3046875,
+      "learning_rate": 4.780840344558753e-05,
+      "loss": 0.92399101,
+      "memory(GiB)": 19.32,
       "step": 635,
-      "train_speed(iter/s)": 0.141928
+      "train_speed(iter/s)": 0.075143
     },
     {
-      "acc": 0.65316691,
-      "epoch": 1.7344173441734418,
-      "grad_norm": 1.875,
-      "learning_rate": 8.944365192582024e-05,
-      "loss": 1.37885714,
-      "memory(GiB)": 16.34,
+      "acc": 0.75597148,
+      "epoch": 2.5048923679060664,
+      "grad_norm": 1.65625,
+      "learning_rate": 4.673516403642383e-05,
+      "loss": 0.86396818,
+      "memory(GiB)": 19.52,
       "step": 640,
-      "train_speed(iter/s)": 0.14206
+      "train_speed(iter/s)": 0.075214
     },
     {
-      "acc": 0.63365035,
-      "epoch": 1.7479674796747968,
-      "grad_norm": 2.03125,
-      "learning_rate": 8.890870185449357e-05,
-      "loss": 1.43597651,
-      "memory(GiB)": 16.34,
+      "acc": 0.75100412,
+      "epoch": 2.524461839530333,
+      "grad_norm": 1.5390625,
+      "learning_rate": 4.5668038025454554e-05,
+      "loss": 0.89630232,
+      "memory(GiB)": 19.54,
       "step": 645,
-      "train_speed(iter/s)": 0.142195
+      "train_speed(iter/s)": 0.07528
     },
     {
-      "acc": 0.6461009,
-      "epoch": 1.7615176151761518,
-      "grad_norm": 1.390625,
-      "learning_rate": 8.837375178316689e-05,
-      "loss": 1.40720987,
-      "memory(GiB)": 16.34,
+      "acc": 0.74814,
+      "epoch": 2.544031311154599,
+      "grad_norm": 1.7265625,
+      "learning_rate": 4.460730582584228e-05,
+      "loss": 0.90660105,
+      "memory(GiB)": 19.46,
       "step": 650,
-      "train_speed(iter/s)": 0.142326
+      "train_speed(iter/s)": 0.075343
     },
     {
-      "epoch": 1.7615176151761518,
-      "eval_acc": 0.6223610876541125,
-      "eval_loss": 1.5279418230056763,
-      "eval_runtime": 44.3195,
-      "eval_samples_per_second": 0.857,
-      "eval_steps_per_second": 0.857,
+      "epoch": 2.544031311154599,
+      "eval_acc": 0.6304006284367636,
+      "eval_loss": 1.6207610368728638,
+      "eval_runtime": 68.9365,
+      "eval_samples_per_second": 1.102,
+      "eval_steps_per_second": 0.551,
       "step": 650
     },
     {
-      "acc": 0.68952079,
-      "epoch": 1.7750677506775068,
-      "grad_norm": 1.5859375,
-      "learning_rate": 8.783880171184023e-05,
-      "loss": 1.1804883,
-      "memory(GiB)": 16.34,
+      "acc": 0.74153934,
+      "epoch": 2.563600782778865,
+      "grad_norm": 2.328125,
+      "learning_rate": 4.3553246170621e-05,
+      "loss": 0.90404129,
+      "memory(GiB)": 19.38,
       "step": 655,
-      "train_speed(iter/s)": 0.141093
+      "train_speed(iter/s)": 0.074813
     },
     {
-      "acc": 0.63278737,
-      "epoch": 1.7886178861788617,
-      "grad_norm": 1.3125,
-      "learning_rate": 8.730385164051355e-05,
-      "loss": 1.45864544,
-      "memory(GiB)": 16.34,
+      "acc": 0.76082869,
+      "epoch": 2.583170254403131,
+      "grad_norm": 1.5390625,
+      "learning_rate": 4.2506136039452357e-05,
+      "loss": 0.90251627,
+      "memory(GiB)": 20.24,
       "step": 660,
-      "train_speed(iter/s)": 0.141227
+      "train_speed(iter/s)": 0.074877
     },
     {
-      "acc": 0.6499536,
-      "epoch": 1.8021680216802167,
-      "grad_norm": 1.0859375,
-      "learning_rate": 8.676890156918687e-05,
-      "loss": 1.36629667,
-      "memory(GiB)": 16.34,
+      "acc": 0.76424356,
+      "epoch": 2.602739726027397,
+      "grad_norm": 1.109375,
+      "learning_rate": 4.146625058584251e-05,
+      "loss": 0.85076065,
+      "memory(GiB)": 19.4,
       "step": 665,
-      "train_speed(iter/s)": 0.14135
+      "train_speed(iter/s)": 0.07494
     },
     {
-      "acc": 0.66262636,
-      "epoch": 1.8157181571815717,
-      "grad_norm": 2.8125,
-      "learning_rate": 8.62339514978602e-05,
-      "loss": 1.28030796,
-      "memory(GiB)": 16.34,
+      "acc": 0.75788155,
+      "epoch": 2.6223091976516635,
+      "grad_norm": 1.828125,
+      "learning_rate": 4.043386306483886e-05,
+      "loss": 0.8638917,
+      "memory(GiB)": 18.71,
       "step": 670,
-      "train_speed(iter/s)": 0.141486
+      "train_speed(iter/s)": 0.075
     },
     {
-      "acc": 0.6478013,
-      "epoch": 1.8292682926829267,
-      "grad_norm": 1.3203125,
-      "learning_rate": 8.569900142653351e-05,
-      "loss": 1.37910089,
-      "memory(GiB)": 16.34,
+      "acc": 0.74567804,
+      "epoch": 2.6418786692759295,
+      "grad_norm": 1.5078125,
+      "learning_rate": 3.940924476122573e-05,
+      "loss": 0.91406345,
+      "memory(GiB)": 19.53,
       "step": 675,
-      "train_speed(iter/s)": 0.141617
+      "train_speed(iter/s)": 0.075062
     },
     {
-      "acc": 0.65061078,
-      "epoch": 1.8428184281842819,
-      "grad_norm": 1.859375,
-      "learning_rate": 8.516405135520683e-05,
-      "loss": 1.24110394,
-      "memory(GiB)": 16.34,
+      "acc": 0.77229648,
+      "epoch": 2.6614481409001955,
+      "grad_norm": 1.3984375,
+      "learning_rate": 3.839266491823776e-05,
+      "loss": 0.79556112,
+      "memory(GiB)": 19.59,
       "step": 680,
-      "train_speed(iter/s)": 0.141749
+      "train_speed(iter/s)": 0.075125
     },
     {
-      "acc": 0.66720443,
-      "epoch": 1.8563685636856369,
-      "grad_norm": 1.2734375,
-      "learning_rate": 8.462910128388016e-05,
-      "loss": 1.36949673,
-      "memory(GiB)": 16.34,
+      "acc": 0.7331708,
+      "epoch": 2.681017612524462,
+      "grad_norm": 1.6015625,
+      "learning_rate": 3.73843906668096e-05,
+      "loss": 0.95133247,
+      "memory(GiB)": 19.69,
       "step": 685,
-      "train_speed(iter/s)": 0.141882
+      "train_speed(iter/s)": 0.075185
     },
     {
-      "acc": 0.65051932,
-      "epoch": 1.8699186991869918,
-      "grad_norm": 1.796875,
-      "learning_rate": 8.409415121255349e-05,
-      "loss": 1.3470686,
-      "memory(GiB)": 16.34,
+      "acc": 0.76955137,
+      "epoch": 2.700587084148728,
+      "grad_norm": 1.4140625,
+      "learning_rate": 3.6384686955380996e-05,
+      "loss": 0.82770052,
+      "memory(GiB)": 19.53,
       "step": 690,
-      "train_speed(iter/s)": 0.142007
+      "train_speed(iter/s)": 0.075245
     },
     {
-      "acc": 0.64647999,
-      "epoch": 1.883468834688347,
-      "grad_norm": 1.0546875,
-      "learning_rate": 8.355920114122682e-05,
-      "loss": 1.27561255,
-      "memory(GiB)": 16.34,
+      "acc": 0.73245583,
+      "epoch": 2.720156555772994,
+      "grad_norm": 1.59375,
+      "learning_rate": 3.539381648027495e-05,
+      "loss": 0.93347349,
+      "memory(GiB)": 19.38,
       "step": 695,
-      "train_speed(iter/s)": 0.142124
+      "train_speed(iter/s)": 0.075313
     },
     {
-      "acc": 0.64771528,
-      "epoch": 1.897018970189702,
-      "grad_norm": 2.5,
-      "learning_rate": 8.302425106990014e-05,
-      "loss": 1.3874403,
-      "memory(GiB)": 16.34,
+      "acc": 0.7664053,
+      "epoch": 2.73972602739726,
+      "grad_norm": 1.4296875,
+      "learning_rate": 3.441203961666818e-05,
+      "loss": 0.84118309,
+      "memory(GiB)": 19.55,
       "step": 700,
-      "train_speed(iter/s)": 0.142242
+      "train_speed(iter/s)": 0.075373
     },
     {
-      "epoch": 1.897018970189702,
-      "eval_acc": 0.6274277993582165,
-      "eval_loss": 1.5236802101135254,
-      "eval_runtime": 44.4112,
-      "eval_samples_per_second": 0.856,
-      "eval_steps_per_second": 0.856,
+      "epoch": 2.73972602739726,
+      "eval_acc": 0.628750981932443,
+      "eval_loss": 1.5982366800308228,
+      "eval_runtime": 69.1268,
+      "eval_samples_per_second": 1.099,
+      "eval_steps_per_second": 0.55,
       "step": 700
     },
     {
-      "acc": 0.65817127,
-      "epoch": 1.910569105691057,
-      "grad_norm": 1.3203125,
-      "learning_rate": 8.248930099857345e-05,
-      "loss": 1.29632025,
-      "memory(GiB)": 16.34,
+      "acc": 0.74386759,
+      "epoch": 2.759295499021526,
+      "grad_norm": 2.21875,
+      "learning_rate": 3.343961435017094e-05,
+      "loss": 0.92712116,
+      "memory(GiB)": 23.1,
       "step": 705,
-      "train_speed(iter/s)": 0.141099
+      "train_speed(iter/s)": 0.074881
     },
     {
-      "acc": 0.64647436,
-      "epoch": 1.924119241192412,
-      "grad_norm": 1.234375,
-      "learning_rate": 8.195435092724678e-05,
-      "loss": 1.39382238,
-      "memory(GiB)": 16.34,
+      "acc": 0.75352135,
+      "epoch": 2.7788649706457926,
+      "grad_norm": 1.5625,
+      "learning_rate": 3.247679620903533e-05,
+      "loss": 0.90610752,
+      "memory(GiB)": 19.56,
       "step": 710,
-      "train_speed(iter/s)": 0.141217
+      "train_speed(iter/s)": 0.074934
     },
     {
-      "acc": 0.65741391,
-      "epoch": 1.937669376693767,
-      "grad_norm": 0.98828125,
-      "learning_rate": 8.14194008559201e-05,
-      "loss": 1.32606802,
-      "memory(GiB)": 16.34,
+      "acc": 0.75765467,
+      "epoch": 2.7984344422700587,
+      "grad_norm": 4.4375,
+      "learning_rate": 3.1523838197008956e-05,
+      "loss": 0.88628139,
+      "memory(GiB)": 19.44,
       "step": 715,
-      "train_speed(iter/s)": 0.141337
+      "train_speed(iter/s)": 0.074999
     },
     {
-      "acc": 0.65078535,
-      "epoch": 1.951219512195122,
-      "grad_norm": 1.6171875,
-      "learning_rate": 8.088445078459343e-05,
-      "loss": 1.28092451,
-      "memory(GiB)": 16.34,
+      "acc": 0.763375,
+      "epoch": 2.8180039138943247,
+      "grad_norm": 1.1640625,
+      "learning_rate": 3.058099072685204e-05,
+      "loss": 0.86159172,
+      "memory(GiB)": 19.5,
       "step": 720,
-      "train_speed(iter/s)": 0.141457
+      "train_speed(iter/s)": 0.075059
     },
     {
-      "acc": 0.64983764,
-      "epoch": 1.9647696476964769,
-      "grad_norm": 1.7890625,
-      "learning_rate": 8.034950071326676e-05,
-      "loss": 1.35295801,
-      "memory(GiB)": 16.34,
+      "acc": 0.75694184,
+      "epoch": 2.837573385518591,
+      "grad_norm": 1.6171875,
+      "learning_rate": 2.964850155453543e-05,
+      "loss": 0.85433092,
+      "memory(GiB)": 19.38,
       "step": 725,
-      "train_speed(iter/s)": 0.141585
+      "train_speed(iter/s)": 0.075121
     },
     {
-      "acc": 0.64880919,
-      "epoch": 1.9783197831978319,
-      "grad_norm": 1.6640625,
-      "learning_rate": 7.981455064194009e-05,
-      "loss": 1.38945732,
-      "memory(GiB)": 16.34,
+      "acc": 0.76086893,
+      "epoch": 2.857142857142857,
+      "grad_norm": 1.5859375,
+      "learning_rate": 2.8726615714136827e-05,
+      "loss": 0.8608798,
+      "memory(GiB)": 19.58,
       "step": 730,
-      "train_speed(iter/s)": 0.141703
+      "train_speed(iter/s)": 0.075181
     },
     {
-      "acc": 0.64617214,
-      "epoch": 1.9918699186991868,
-      "grad_norm": 1.3671875,
-      "learning_rate": 7.927960057061341e-05,
-      "loss": 1.31189098,
-      "memory(GiB)": 16.34,
+      "acc": 0.74008894,
+      "epoch": 2.8767123287671232,
+      "grad_norm": 1.4375,
+      "learning_rate": 2.7815575453452058e-05,
+      "loss": 0.98413734,
+      "memory(GiB)": 19.59,
       "step": 735,
-      "train_speed(iter/s)": 0.141817
+      "train_speed(iter/s)": 0.075242
     },
     {
-      "acc": 0.67255039,
-      "epoch": 2.005420054200542,
-      "grad_norm": 1.3359375,
-      "learning_rate": 7.874465049928672e-05,
-      "loss": 1.17746153,
-      "memory(GiB)": 16.34,
+      "acc": 0.75941825,
+      "epoch": 2.8962818003913893,
+      "grad_norm": 1.7734375,
+      "learning_rate": 2.6915620170338612e-05,
+      "loss": 0.85438929,
+      "memory(GiB)": 19.39,
       "step": 740,
-      "train_speed(iter/s)": 0.141669
+      "train_speed(iter/s)": 0.075307
     },
     {
-      "acc": 0.69897633,
-      "epoch": 2.0189701897018972,
-      "grad_norm": 1.1875,
-      "learning_rate": 7.820970042796005e-05,
-      "loss": 1.03704672,
-      "memory(GiB)": 16.34,
+      "acc": 0.77891464,
+      "epoch": 2.9158512720156553,
+      "grad_norm": 1.7265625,
+      "learning_rate": 2.6026986349808058e-05,
+      "loss": 0.79716868,
+      "memory(GiB)": 19.61,
       "step": 745,
-      "train_speed(iter/s)": 0.141785
+      "train_speed(iter/s)": 0.075361
     },
     {
-      "acc": 0.71377811,
-      "epoch": 2.032520325203252,
-      "grad_norm": 1.640625,
-      "learning_rate": 7.767475035663337e-05,
-      "loss": 1.04516726,
-      "memory(GiB)": 16.34,
+      "acc": 0.75023217,
+      "epoch": 2.935420743639922,
+      "grad_norm": 1.28125,
+      "learning_rate": 2.514990750188399e-05,
+      "loss": 0.85774508,
+      "memory(GiB)": 18.86,
       "step": 750,
-      "train_speed(iter/s)": 0.141905
+      "train_speed(iter/s)": 0.075417
     },
     {
-      "epoch": 2.032520325203252,
-      "eval_acc": 0.6208410741428813,
-      "eval_loss": 1.603255271911621,
-      "eval_runtime": 44.4359,
-      "eval_samples_per_second": 0.855,
-      "eval_steps_per_second": 0.855,
+      "epoch": 2.935420743639922,
+      "eval_acc": 0.6324430479183032,
+      "eval_loss": 1.5986852645874023,
+      "eval_runtime": 69.3348,
+      "eval_samples_per_second": 1.096,
+      "eval_steps_per_second": 0.548,
       "step": 750
     },
     {
-      "acc": 0.73705945,
-      "epoch": 2.046070460704607,
-      "grad_norm": 1.8203125,
-      "learning_rate": 7.713980028530669e-05,
-      "loss": 0.9383255,
-      "memory(GiB)": 16.34,
+      "acc": 0.74531512,
+      "epoch": 2.954990215264188,
+      "grad_norm": 1.5625,
+      "learning_rate": 2.4284614100241538e-05,
+      "loss": 0.93483381,
+      "memory(GiB)": 23.14,
       "step": 755,
-      "train_speed(iter/s)": 0.14084
+      "train_speed(iter/s)": 0.074953
     },
     {
-      "acc": 0.73054934,
-      "epoch": 2.059620596205962,
-      "grad_norm": 1.8203125,
-      "learning_rate": 7.660485021398003e-05,
-      "loss": 0.91379232,
-      "memory(GiB)": 16.34,
+      "acc": 0.76761031,
+      "epoch": 2.974559686888454,
+      "grad_norm": 1.6171875,
+      "learning_rate": 2.343133352164477e-05,
+      "loss": 0.84630623,
+      "memory(GiB)": 19.36,
       "step": 760,
-      "train_speed(iter/s)": 0.140958
+      "train_speed(iter/s)": 0.075015
     },
     {
-      "acc": 0.73154573,
-      "epoch": 2.073170731707317,
-      "grad_norm": 1.6796875,
-      "learning_rate": 7.606990014265335e-05,
-      "loss": 0.97700481,
-      "memory(GiB)": 16.34,
+      "acc": 0.75018072,
+      "epoch": 2.9941291585127203,
+      "grad_norm": 1.5703125,
+      "learning_rate": 2.2590289986198136e-05,
+      "loss": 0.89352074,
+      "memory(GiB)": 19.6,
       "step": 765,
-      "train_speed(iter/s)": 0.141076
+      "train_speed(iter/s)": 0.075072
     },
     {
-      "acc": 0.72998781,
-      "epoch": 2.086720867208672,
-      "grad_norm": 1.4140625,
-      "learning_rate": 7.553495007132668e-05,
-      "loss": 0.93853807,
-      "memory(GiB)": 16.34,
+      "acc": 0.80383377,
+      "epoch": 3.0136986301369864,
+      "grad_norm": 1.453125,
+      "learning_rate": 2.1761704498427003e-05,
+      "loss": 0.68276234,
+      "memory(GiB)": 19.62,
       "step": 770,
-      "train_speed(iter/s)": 0.141186
+      "train_speed(iter/s)": 0.075153
     },
     {
-      "acc": 0.73213534,
-      "epoch": 2.100271002710027,
-      "grad_norm": 1.78125,
-      "learning_rate": 7.5e-05,
-      "loss": 0.94589176,
-      "memory(GiB)": 16.34,
+      "acc": 0.82252359,
+      "epoch": 3.0332681017612524,
+      "grad_norm": 1.328125,
+      "learning_rate": 2.094579478920358e-05,
+      "loss": 0.64008789,
+      "memory(GiB)": 19.76,
       "step": 775,
-      "train_speed(iter/s)": 0.141303
+      "train_speed(iter/s)": 0.075213
     },
     {
-      "acc": 0.73035493,
-      "epoch": 2.113821138211382,
-      "grad_norm": 1.6953125,
-      "learning_rate": 7.446504992867331e-05,
-      "loss": 0.95597754,
-      "memory(GiB)": 16.34,
+      "acc": 0.83448801,
+      "epoch": 3.0528375733855184,
+      "grad_norm": 1.8828125,
+      "learning_rate": 2.0142775258532654e-05,
+      "loss": 0.61610913,
+      "memory(GiB)": 19.59,
       "step": 780,
-      "train_speed(iter/s)": 0.141422
+      "train_speed(iter/s)": 0.075271
     },
     {
-      "acc": 0.73428354,
-      "epoch": 2.127371273712737,
-      "grad_norm": 1.484375,
-      "learning_rate": 7.393009985734664e-05,
-      "loss": 0.98254423,
-      "memory(GiB)": 16.34,
+      "acc": 0.83116817,
+      "epoch": 3.072407045009785,
+      "grad_norm": 1.5546875,
+      "learning_rate": 1.9352856919212994e-05,
+      "loss": 0.58688097,
+      "memory(GiB)": 19.53,
       "step": 785,
-      "train_speed(iter/s)": 0.141543
+      "train_speed(iter/s)": 0.075323
     },
     {
-      "acc": 0.69750729,
-      "epoch": 2.140921409214092,
-      "grad_norm": 1.828125,
-      "learning_rate": 7.339514978601997e-05,
-      "loss": 1.10206041,
-      "memory(GiB)": 16.34,
+      "acc": 0.82525949,
+      "epoch": 3.091976516634051,
+      "grad_norm": 1.4375,
+      "learning_rate": 1.8576247341388544e-05,
+      "loss": 0.62312498,
+      "memory(GiB)": 19.85,
       "step": 790,
-      "train_speed(iter/s)": 0.141651
+      "train_speed(iter/s)": 0.07537
     },
     {
-      "acc": 0.73764381,
-      "epoch": 2.154471544715447,
-      "grad_norm": 1.296875,
-      "learning_rate": 7.286019971469329e-05,
-      "loss": 0.84075432,
-      "memory(GiB)": 16.34,
+      "acc": 0.81645441,
+      "epoch": 3.111545988258317,
+      "grad_norm": 1.65625,
+      "learning_rate": 1.7813150598004313e-05,
+      "loss": 0.62203112,
+      "memory(GiB)": 19.79,
       "step": 795,
-      "train_speed(iter/s)": 0.141765
+      "train_speed(iter/s)": 0.075423
     },
     {
-      "acc": 0.72124152,
-      "epoch": 2.168021680216802,
-      "grad_norm": 1.4375,
-      "learning_rate": 7.23252496433666e-05,
-      "loss": 1.05863771,
-      "memory(GiB)": 16.34,
+      "acc": 0.83432789,
+      "epoch": 3.131115459882583,
+      "grad_norm": 1.5859375,
+      "learning_rate": 1.7063767211181333e-05,
+      "loss": 0.60077624,
+      "memory(GiB)": 19.52,
       "step": 800,
-      "train_speed(iter/s)": 0.141874
+      "train_speed(iter/s)": 0.07548
     },
     {
-      "epoch": 2.168021680216802,
-      "eval_acc": 0.6240499915554805,
-      "eval_loss": 1.61227285861969,
-      "eval_runtime": 44.3145,
-      "eval_samples_per_second": 0.858,
-      "eval_steps_per_second": 0.858,
+      "epoch": 3.131115459882583,
+      "eval_acc": 0.6209740769835035,
+      "eval_loss": 1.7955598831176758,
+      "eval_runtime": 69.0109,
+      "eval_samples_per_second": 1.101,
+      "eval_steps_per_second": 0.551,
       "step": 800
     },
     {
-      "acc": 0.73518519,
-      "epoch": 2.181571815718157,
-      "grad_norm": 1.546875,
-      "learning_rate": 7.179029957203993e-05,
-      "loss": 0.96204119,
-      "memory(GiB)": 16.34,
+      "acc": 0.82124023,
+      "epoch": 3.1506849315068495,
+      "grad_norm": 1.7578125,
+      "learning_rate": 1.6328294099524644e-05,
+      "loss": 0.60847788,
+      "memory(GiB)": 22.65,
       "step": 805,
-      "train_speed(iter/s)": 0.140881
+      "train_speed(iter/s)": 0.075043
     },
     {
-      "acc": 0.74290161,
-      "epoch": 2.1951219512195124,
-      "grad_norm": 2.015625,
-      "learning_rate": 7.125534950071326e-05,
-      "loss": 0.92387733,
-      "memory(GiB)": 16.34,
+      "acc": 0.83265171,
+      "epoch": 3.1702544031311155,
+      "grad_norm": 4.09375,
+      "learning_rate": 1.5606924526378136e-05,
+      "loss": 0.57863126,
+      "memory(GiB)": 18.89,
       "step": 810,
-      "train_speed(iter/s)": 0.14021
+      "train_speed(iter/s)": 0.07509
     },
     {
-      "acc": 0.73101602,
-      "epoch": 2.2086720867208673,
-      "grad_norm": 1.6015625,
-      "learning_rate": 7.072039942938658e-05,
-      "loss": 0.96248655,
-      "memory(GiB)": 16.34,
+      "acc": 0.8407362,
+      "epoch": 3.1898238747553815,
+      "grad_norm": 1.1796875,
+      "learning_rate": 1.4899848049039881e-05,
+      "loss": 0.53706379,
+      "memory(GiB)": 19.37,
       "step": 815,
-      "train_speed(iter/s)": 0.140323
+      "train_speed(iter/s)": 0.075142
     },
     {
-      "acc": 0.73887796,
-      "epoch": 2.2222222222222223,
-      "grad_norm": 1.796875,
-      "learning_rate": 7.018544935805991e-05,
-      "loss": 0.94751673,
-      "memory(GiB)": 16.34,
+      "acc": 0.82116756,
+      "epoch": 3.2093933463796476,
+      "grad_norm": 1.859375,
+      "learning_rate": 1.4207250468951426e-05,
+      "loss": 0.64039102,
+      "memory(GiB)": 19.52,
       "step": 820,
-      "train_speed(iter/s)": 0.140441
+      "train_speed(iter/s)": 0.075197
     },
     {
-      "acc": 0.75793715,
-      "epoch": 2.2357723577235773,
-      "grad_norm": 2.015625,
-      "learning_rate": 6.965049928673323e-05,
-      "loss": 0.84263477,
-      "memory(GiB)": 16.34,
+      "acc": 0.85004549,
+      "epoch": 3.228962818003914,
+      "grad_norm": 1.0390625,
+      "learning_rate": 1.3529313782874023e-05,
+      "loss": 0.53315983,
+      "memory(GiB)": 19.52,
       "step": 825,
-      "train_speed(iter/s)": 0.140555
+      "train_speed(iter/s)": 0.07525
     },
     {
-      "acc": 0.75873909,
-      "epoch": 2.2493224932249323,
-      "grad_norm": 1.953125,
-      "learning_rate": 6.911554921540656e-05,
-      "loss": 0.89048252,
-      "memory(GiB)": 16.34,
+      "acc": 0.83273296,
+      "epoch": 3.24853228962818,
+      "grad_norm": 1.578125,
+      "learning_rate": 1.2866216135064487e-05,
+      "loss": 0.58545351,
+      "memory(GiB)": 19.36,
       "step": 830,
-      "train_speed(iter/s)": 0.14067
+      "train_speed(iter/s)": 0.075303
     },
     {
-      "acc": 0.74456077,
-      "epoch": 2.2628726287262872,
-      "grad_norm": 1.7890625,
-      "learning_rate": 6.858059914407987e-05,
-      "loss": 0.90777779,
-      "memory(GiB)": 16.34,
+      "acc": 0.80788403,
+      "epoch": 3.268101761252446,
+      "grad_norm": 2.296875,
+      "learning_rate": 1.2218131770463487e-05,
+      "loss": 0.67468171,
+      "memory(GiB)": 19.28,
       "step": 835,
-      "train_speed(iter/s)": 0.140775
+      "train_speed(iter/s)": 0.075356
     },
     {
-      "acc": 0.75809846,
-      "epoch": 2.2764227642276422,
-      "grad_norm": 2.0625,
-      "learning_rate": 6.80456490727532e-05,
-      "loss": 0.87556753,
-      "memory(GiB)": 16.34,
+      "acc": 0.8440134,
+      "epoch": 3.287671232876712,
+      "grad_norm": 1.21875,
+      "learning_rate": 1.1585230988908576e-05,
+      "loss": 0.55293651,
+      "memory(GiB)": 19.37,
       "step": 840,
-      "train_speed(iter/s)": 0.14088
+      "train_speed(iter/s)": 0.07541
     },
     {
-      "acc": 0.73194971,
-      "epoch": 2.289972899728997,
-      "grad_norm": 2.765625,
-      "learning_rate": 6.751069900142653e-05,
-      "loss": 0.9232769,
-      "memory(GiB)": 16.34,
+      "acc": 0.81569691,
+      "epoch": 3.3072407045009786,
+      "grad_norm": 1.671875,
+      "learning_rate": 1.0967680100383645e-05,
+      "loss": 0.61190109,
+      "memory(GiB)": 18.09,
       "step": 845,
-      "train_speed(iter/s)": 0.140986
+      "train_speed(iter/s)": 0.075466
     },
     {
-      "acc": 0.74470835,
-      "epoch": 2.303523035230352,
-      "grad_norm": 1.8515625,
-      "learning_rate": 6.697574893009985e-05,
-      "loss": 0.88600664,
-      "memory(GiB)": 16.34,
+      "acc": 0.84766483,
+      "epoch": 3.3268101761252447,
+      "grad_norm": 1.8046875,
+      "learning_rate": 1.0365641381317113e-05,
+      "loss": 0.52525816,
+      "memory(GiB)": 19.31,
       "step": 850,
-      "train_speed(iter/s)": 0.141086
+      "train_speed(iter/s)": 0.075523
     },
     {
-      "epoch": 2.303523035230352,
-      "eval_acc": 0.6174632663401453,
-      "eval_loss": 1.660492181777954,
-      "eval_runtime": 44.1393,
-      "eval_samples_per_second": 0.861,
-      "eval_steps_per_second": 0.861,
+      "epoch": 3.3268101761252447,
+      "eval_acc": 0.6203456402199529,
+      "eval_loss": 1.7881730794906616,
+      "eval_runtime": 69.1552,
+      "eval_samples_per_second": 1.099,
+      "eval_steps_per_second": 0.549,
       "step": 850
     },
     {
-      "acc": 0.76063514,
-      "epoch": 2.317073170731707,
-      "grad_norm": 2.40625,
-      "learning_rate": 6.644079885877318e-05,
-      "loss": 0.80236473,
-      "memory(GiB)": 16.34,
+      "acc": 0.84491625,
+      "epoch": 3.3463796477495107,
+      "grad_norm": 1.8046875,
+      "learning_rate": 9.779273031939692e-06,
+      "loss": 0.56272998,
+      "memory(GiB)": 23.04,
       "step": 855,
-      "train_speed(iter/s)": 0.140168
+      "train_speed(iter/s)": 0.07511
     },
     {
-      "acc": 0.71505499,
-      "epoch": 2.330623306233062,
-      "grad_norm": 2.671875,
-      "learning_rate": 6.59058487874465e-05,
-      "loss": 1.01317081,
-      "memory(GiB)": 16.34,
+      "acc": 0.84104662,
+      "epoch": 3.3659491193737767,
+      "grad_norm": 1.796875,
+      "learning_rate": 9.20872913471363e-06,
+      "loss": 0.57019663,
+      "memory(GiB)": 19.42,
       "step": 860,
-      "train_speed(iter/s)": 0.140281
+      "train_speed(iter/s)": 0.075157
     },
     {
-      "acc": 0.73396034,
-      "epoch": 2.3441734417344176,
-      "grad_norm": 1.703125,
-      "learning_rate": 6.537089871611983e-05,
-      "loss": 0.96496754,
-      "memory(GiB)": 16.34,
+      "acc": 0.84433002,
+      "epoch": 3.385518590998043,
+      "grad_norm": 1.6484375,
+      "learning_rate": 8.654159613843715e-06,
+      "loss": 0.55449514,
+      "memory(GiB)": 19.59,
       "step": 865,
-      "train_speed(iter/s)": 0.140387
+      "train_speed(iter/s)": 0.07521
     },
     {
-      "acc": 0.74145699,
-      "epoch": 2.3577235772357725,
-      "grad_norm": 1.9453125,
-      "learning_rate": 6.483594864479316e-05,
-      "loss": 1.00490999,
-      "memory(GiB)": 16.34,
+      "acc": 0.80005312,
+      "epoch": 3.4050880626223092,
+      "grad_norm": 1.46875,
+      "learning_rate": 8.115710195881068e-06,
+      "loss": 0.73595409,
+      "memory(GiB)": 19.36,
       "step": 870,
-      "train_speed(iter/s)": 0.140492
+      "train_speed(iter/s)": 0.075258
     },
     {
-      "acc": 0.76035104,
-      "epoch": 2.3712737127371275,
-      "grad_norm": 1.7265625,
-      "learning_rate": 6.430099857346647e-05,
-      "loss": 0.93969469,
-      "memory(GiB)": 16.34,
+      "acc": 0.83217945,
+      "epoch": 3.4246575342465753,
+      "grad_norm": 3.328125,
+      "learning_rate": 7.593522371429972e-06,
+      "loss": 0.58270836,
+      "memory(GiB)": 19.58,
       "step": 875,
-      "train_speed(iter/s)": 0.140599
+      "train_speed(iter/s)": 0.075306
     },
     {
-      "acc": 0.75214877,
-      "epoch": 2.3848238482384825,
-      "grad_norm": 1.9296875,
-      "learning_rate": 6.376604850213979e-05,
-      "loss": 0.9385232,
-      "memory(GiB)": 16.34,
+      "acc": 0.82742786,
+      "epoch": 3.4442270058708413,
+      "grad_norm": 1.234375,
+      "learning_rate": 7.0877333579678585e-06,
+      "loss": 0.59052157,
+      "memory(GiB)": 19.6,
       "step": 880,
-      "train_speed(iter/s)": 0.140705
+      "train_speed(iter/s)": 0.075358
     },
     {
-      "acc": 0.73846526,
-      "epoch": 2.3983739837398375,
-      "grad_norm": 2.421875,
-      "learning_rate": 6.323109843081312e-05,
-      "loss": 0.95887814,
-      "memory(GiB)": 16.34,
+      "acc": 0.81994705,
+      "epoch": 3.4637964774951078,
+      "grad_norm": 1.7578125,
+      "learning_rate": 6.598476063788036e-06,
+      "loss": 0.62256751,
+      "memory(GiB)": 19.56,
       "step": 885,
-      "train_speed(iter/s)": 0.140812
+      "train_speed(iter/s)": 0.075405
     },
     {
-      "acc": 0.72576594,
-      "epoch": 2.4119241192411924,
-      "grad_norm": 1.46875,
-      "learning_rate": 6.269614835948645e-05,
-      "loss": 0.95252094,
-      "memory(GiB)": 16.34,
+      "acc": 0.8157341,
+      "epoch": 3.483365949119374,
+      "grad_norm": 1.8203125,
+      "learning_rate": 6.12587905307477e-06,
+      "loss": 0.66806622,
+      "memory(GiB)": 19.49,
       "step": 890,
-      "train_speed(iter/s)": 0.140912
+      "train_speed(iter/s)": 0.075454
     },
     {
-      "acc": 0.74866586,
-      "epoch": 2.4254742547425474,
-      "grad_norm": 1.5546875,
-      "learning_rate": 6.216119828815977e-05,
-      "loss": 0.91533632,
-      "memory(GiB)": 16.34,
+      "acc": 0.82838688,
+      "epoch": 3.50293542074364,
+      "grad_norm": 1.515625,
+      "learning_rate": 5.67006651212008e-06,
+      "loss": 0.63044977,
+      "memory(GiB)": 19.54,
       "step": 895,
-      "train_speed(iter/s)": 0.141011
+      "train_speed(iter/s)": 0.075497
     },
     {
-      "acc": 0.72289066,
-      "epoch": 2.4390243902439024,
-      "grad_norm": 1.375,
-      "learning_rate": 6.162624821683308e-05,
-      "loss": 0.996418,
-      "memory(GiB)": 16.34,
+      "acc": 0.79130597,
+      "epoch": 3.5225048923679063,
+      "grad_norm": 1.640625,
+      "learning_rate": 5.2311582166906605e-06,
+      "loss": 0.7558567,
+      "memory(GiB)": 19.28,
       "step": 900,
-      "train_speed(iter/s)": 0.13966
+      "train_speed(iter/s)": 0.07555
     },
     {
-      "epoch": 2.4390243902439024,
-      "eval_acc": 0.6152676912683668,
-      "eval_loss": 1.6503233909606934,
-      "eval_runtime": 44.1939,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
+      "epoch": 3.5225048923679063,
+      "eval_acc": 0.6211311861743912,
+      "eval_loss": 1.7854998111724854,
+      "eval_runtime": 69.2434,
+      "eval_samples_per_second": 1.098,
+      "eval_steps_per_second": 0.549,
       "step": 900
-    },
-    {
-      "acc": 0.73535914,
-      "epoch": 2.4525745257452574,
-      "grad_norm": 1.3125,
-      "learning_rate": 6.109129814550641e-05,
-      "loss": 1.00612879,
-      "memory(GiB)": 16.34,
-      "step": 905,
-      "train_speed(iter/s)": 0.138805
-    },
-    {
-      "acc": 0.7277998,
-      "epoch": 2.4661246612466123,
-      "grad_norm": 2.328125,
-      "learning_rate": 6.0556348074179737e-05,
-      "loss": 1.00636635,
-      "memory(GiB)": 16.34,
-      "step": 910,
-      "train_speed(iter/s)": 0.138904
-    },
-    {
-      "acc": 0.70795035,
-      "epoch": 2.4796747967479673,
-      "grad_norm": 1.6484375,
-      "learning_rate": 6.002139800285306e-05,
-      "loss": 0.98440151,
-      "memory(GiB)": 16.34,
-      "step": 915,
-      "train_speed(iter/s)": 0.139013
-    },
-    {
-      "acc": 0.7650095,
-      "epoch": 2.4932249322493227,
-      "grad_norm": 2.25,
-      "learning_rate": 5.948644793152638e-05,
-      "loss": 0.77239523,
-      "memory(GiB)": 16.34,
-      "step": 920,
-      "train_speed(iter/s)": 0.139126
-    },
-    {
-      "acc": 0.71504927,
-      "epoch": 2.5067750677506773,
-      "grad_norm": 1.6328125,
-      "learning_rate": 5.895149786019971e-05,
-      "loss": 1.01518288,
-      "memory(GiB)": 16.34,
-      "step": 925,
-      "train_speed(iter/s)": 0.139233
-    },
-    {
-      "acc": 0.71582847,
-      "epoch": 2.5203252032520327,
-      "grad_norm": 2.28125,
-      "learning_rate": 5.841654778887303e-05,
-      "loss": 0.97862291,
-      "memory(GiB)": 16.34,
-      "step": 930,
-      "train_speed(iter/s)": 0.13934
-    },
-    {
-      "acc": 0.74984369,
-      "epoch": 2.5338753387533877,
-      "grad_norm": 1.8515625,
-      "learning_rate": 5.788159771754635e-05,
-      "loss": 0.85021992,
-      "memory(GiB)": 16.34,
-      "step": 935,
-      "train_speed(iter/s)": 0.139448
-    },
-    {
-      "acc": 0.74142261,
-      "epoch": 2.5474254742547426,
-      "grad_norm": 1.796875,
-      "learning_rate": 5.734664764621968e-05,
-      "loss": 0.91843338,
-      "memory(GiB)": 16.34,
-      "step": 940,
-      "train_speed(iter/s)": 0.139564
-    },
-    {
-      "acc": 0.71266222,
-      "epoch": 2.5609756097560976,
-      "grad_norm": 1.875,
-      "learning_rate": 5.6811697574893007e-05,
-      "loss": 1.14017658,
-      "memory(GiB)": 16.34,
-      "step": 945,
-      "train_speed(iter/s)": 0.139667
-    },
-    {
-      "acc": 0.73374839,
-      "epoch": 2.5745257452574526,
-      "grad_norm": 1.671875,
-      "learning_rate": 5.627674750356633e-05,
-      "loss": 0.99979248,
-      "memory(GiB)": 16.34,
-      "step": 950,
-      "train_speed(iter/s)": 0.139769
-    },
-    {
-      "epoch": 2.5745257452574526,
-      "eval_acc": 0.6161121432190508,
-      "eval_loss": 1.6511973142623901,
-      "eval_runtime": 44.1881,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
-      "step": 950
-    },
-    {
-      "acc": 0.74381332,
-      "epoch": 2.5880758807588076,
-      "grad_norm": 1.7734375,
-      "learning_rate": 5.5741797432239646e-05,
-      "loss": 0.93189411,
-      "memory(GiB)": 16.34,
-      "step": 955,
-      "train_speed(iter/s)": 0.138972
-    },
-    {
-      "acc": 0.73305674,
-      "epoch": 2.6016260162601625,
-      "grad_norm": 2.4375,
-      "learning_rate": 5.5206847360912977e-05,
-      "loss": 0.93946905,
-      "memory(GiB)": 16.34,
-      "step": 960,
-      "train_speed(iter/s)": 0.139079
-    },
-    {
-      "acc": 0.72961435,
-      "epoch": 2.6151761517615175,
-      "grad_norm": 1.84375,
-      "learning_rate": 5.46718972895863e-05,
-      "loss": 0.99850683,
-      "memory(GiB)": 16.34,
-      "step": 965,
-      "train_speed(iter/s)": 0.139175
-    },
-    {
-      "acc": 0.7133184,
-      "epoch": 2.6287262872628725,
-      "grad_norm": 1.96875,
-      "learning_rate": 5.413694721825962e-05,
-      "loss": 1.03378878,
-      "memory(GiB)": 16.34,
-      "step": 970,
-      "train_speed(iter/s)": 0.139282
-    },
-    {
-      "acc": 0.71164918,
-      "epoch": 2.642276422764228,
-      "grad_norm": 1.3828125,
-      "learning_rate": 5.360199714693295e-05,
-      "loss": 1.02268944,
-      "memory(GiB)": 16.34,
-      "step": 975,
-      "train_speed(iter/s)": 0.139382
-    },
-    {
-      "acc": 0.71824646,
-      "epoch": 2.6558265582655824,
-      "grad_norm": 1.84375,
-      "learning_rate": 5.306704707560627e-05,
-      "loss": 1.06014738,
-      "memory(GiB)": 16.34,
-      "step": 980,
-      "train_speed(iter/s)": 0.139479
-    },
-    {
-      "acc": 0.75523286,
-      "epoch": 2.669376693766938,
-      "grad_norm": 1.7734375,
-      "learning_rate": 5.253209700427959e-05,
-      "loss": 0.80533791,
-      "memory(GiB)": 16.34,
-      "step": 985,
-      "train_speed(iter/s)": 0.139581
-    },
-    {
-      "acc": 0.72265592,
-      "epoch": 2.682926829268293,
-      "grad_norm": 1.5546875,
-      "learning_rate": 5.199714693295292e-05,
-      "loss": 1.0223958,
-      "memory(GiB)": 16.34,
-      "step": 990,
-      "train_speed(iter/s)": 0.139675
-    },
-    {
-      "acc": 0.74640193,
-      "epoch": 2.696476964769648,
-      "grad_norm": 1.5546875,
-      "learning_rate": 5.1462196861626247e-05,
-      "loss": 0.8896265,
-      "memory(GiB)": 16.34,
-      "step": 995,
-      "train_speed(iter/s)": 0.139767
-    },
-    {
-      "acc": 0.75235152,
-      "epoch": 2.710027100271003,
-      "grad_norm": 1.6484375,
-      "learning_rate": 5.092724679029957e-05,
-      "loss": 0.87937946,
-      "memory(GiB)": 16.34,
-      "step": 1000,
-      "train_speed(iter/s)": 0.139867
-    },
-    {
-      "epoch": 2.710027100271003,
-      "eval_acc": 0.6123965546360413,
-      "eval_loss": 1.6541699171066284,
-      "eval_runtime": 44.2136,
-      "eval_samples_per_second": 0.859,
-      "eval_steps_per_second": 0.859,
-      "step": 1000
-    },
-    {
-      "acc": 0.73264089,
-      "epoch": 2.7235772357723578,
-      "grad_norm": 1.96875,
-      "learning_rate": 5.0392296718972886e-05,
-      "loss": 0.95834293,
-      "memory(GiB)": 16.34,
-      "step": 1005,
-      "train_speed(iter/s)": 0.139102
-    },
-    {
-      "acc": 0.76931543,
-      "epoch": 2.7371273712737128,
-      "grad_norm": 1.7890625,
-      "learning_rate": 4.9857346647646217e-05,
-      "loss": 0.82889891,
-      "memory(GiB)": 16.34,
-      "step": 1010,
-      "train_speed(iter/s)": 0.139199
-    },
-    {
-      "acc": 0.74891815,
-      "epoch": 2.7506775067750677,
-      "grad_norm": 2.65625,
-      "learning_rate": 4.932239657631954e-05,
-      "loss": 0.93503094,
-      "memory(GiB)": 16.34,
-      "step": 1015,
-      "train_speed(iter/s)": 0.139298
-    },
-    {
-      "acc": 0.72551074,
-      "epoch": 2.7642276422764227,
-      "grad_norm": 1.5703125,
-      "learning_rate": 4.878744650499286e-05,
-      "loss": 1.06086788,
-      "memory(GiB)": 16.34,
-      "step": 1020,
-      "train_speed(iter/s)": 0.139383
-    },
-    {
-      "acc": 0.75321589,
-      "epoch": 2.7777777777777777,
-      "grad_norm": 1.7734375,
-      "learning_rate": 4.825249643366619e-05,
-      "loss": 0.83553734,
-      "memory(GiB)": 16.34,
-      "step": 1025,
-      "train_speed(iter/s)": 0.139478
-    },
-    {
-      "acc": 0.73832054,
-      "epoch": 2.7913279132791327,
-      "grad_norm": 1.53125,
-      "learning_rate": 4.771754636233951e-05,
-      "loss": 0.94093418,
-      "memory(GiB)": 16.34,
-      "step": 1030,
-      "train_speed(iter/s)": 0.139572
-    },
-    {
-      "acc": 0.76351671,
-      "epoch": 2.8048780487804876,
-      "grad_norm": 1.2890625,
-      "learning_rate": 4.718259629101283e-05,
-      "loss": 0.84423561,
-      "memory(GiB)": 16.34,
-      "step": 1035,
-      "train_speed(iter/s)": 0.139664
-    },
-    {
-      "acc": 0.75100279,
-      "epoch": 2.818428184281843,
-      "grad_norm": 1.890625,
-      "learning_rate": 4.6647646219686156e-05,
-      "loss": 0.91757078,
-      "memory(GiB)": 16.34,
-      "step": 1040,
-      "train_speed(iter/s)": 0.139754
-    },
-    {
-      "acc": 0.73302913,
-      "epoch": 2.8319783197831976,
-      "grad_norm": 1.3671875,
-      "learning_rate": 4.6112696148359487e-05,
-      "loss": 0.94331446,
-      "memory(GiB)": 16.34,
-      "step": 1045,
-      "train_speed(iter/s)": 0.139841
-    },
-    {
-      "acc": 0.73944206,
-      "epoch": 2.845528455284553,
-      "grad_norm": 1.59375,
-      "learning_rate": 4.55777460770328e-05,
-      "loss": 0.98726377,
-      "memory(GiB)": 16.34,
-      "step": 1050,
-      "train_speed(iter/s)": 0.139931
-    },
-    {
-      "epoch": 2.845528455284553,
-      "eval_acc": 0.6145921297078196,
-      "eval_loss": 1.6571751832962036,
-      "eval_runtime": 44.3705,
-      "eval_samples_per_second": 0.856,
-      "eval_steps_per_second": 0.856,
-      "step": 1050
-    },
-    {
-      "acc": 0.73031135,
-      "epoch": 2.859078590785908,
-      "grad_norm": 2.25,
-      "learning_rate": 4.5042796005706126e-05,
-      "loss": 0.95681438,
-      "memory(GiB)": 16.34,
-      "step": 1055,
-      "train_speed(iter/s)": 0.139201
-    },
-    {
-      "acc": 0.71694627,
-      "epoch": 2.872628726287263,
-      "grad_norm": 1.6328125,
-      "learning_rate": 4.4507845934379456e-05,
-      "loss": 1.02266083,
-      "memory(GiB)": 16.34,
-      "step": 1060,
-      "train_speed(iter/s)": 0.139294
-    },
-    {
-      "acc": 0.72791638,
-      "epoch": 2.886178861788618,
-      "grad_norm": 1.7578125,
-      "learning_rate": 4.397289586305278e-05,
-      "loss": 0.99341927,
-      "memory(GiB)": 16.34,
-      "step": 1065,
-      "train_speed(iter/s)": 0.139385
-    },
-    {
-      "acc": 0.74939594,
-      "epoch": 2.899728997289973,
-      "grad_norm": 1.921875,
-      "learning_rate": 4.34379457917261e-05,
-      "loss": 0.91077061,
-      "memory(GiB)": 16.34,
-      "step": 1070,
-      "train_speed(iter/s)": 0.139478
-    },
-    {
-      "acc": 0.72694654,
-      "epoch": 2.913279132791328,
-      "grad_norm": 1.7265625,
-      "learning_rate": 4.290299572039942e-05,
-      "loss": 0.98774853,
-      "memory(GiB)": 16.34,
-      "step": 1075,
-      "train_speed(iter/s)": 0.139564
-    },
-    {
-      "acc": 0.70588508,
-      "epoch": 2.926829268292683,
-      "grad_norm": 2.15625,
-      "learning_rate": 4.236804564907275e-05,
-      "loss": 1.07887812,
-      "memory(GiB)": 16.34,
-      "step": 1080,
-      "train_speed(iter/s)": 0.139657
-    },
-    {
-      "acc": 0.74654303,
-      "epoch": 2.940379403794038,
-      "grad_norm": 1.7109375,
-      "learning_rate": 4.183309557774607e-05,
-      "loss": 0.91062069,
-      "memory(GiB)": 16.34,
-      "step": 1085,
-      "train_speed(iter/s)": 0.139743
-    },
-    {
-      "acc": 0.73493595,
-      "epoch": 2.953929539295393,
-      "grad_norm": 2.03125,
-      "learning_rate": 4.1298145506419396e-05,
-      "loss": 0.92819033,
-      "memory(GiB)": 16.34,
-      "step": 1090,
-      "train_speed(iter/s)": 0.139829
-    },
-    {
-      "acc": 0.71466756,
-      "epoch": 2.9674796747967482,
-      "grad_norm": 2.109375,
-      "learning_rate": 4.0763195435092727e-05,
-      "loss": 1.01220913,
-      "memory(GiB)": 16.34,
-      "step": 1095,
-      "train_speed(iter/s)": 0.139916
-    },
-    {
-      "acc": 0.7607831,
-      "epoch": 2.9810298102981028,
-      "grad_norm": 2.375,
-      "learning_rate": 4.022824536376604e-05,
-      "loss": 0.8505785,
-      "memory(GiB)": 16.34,
-      "step": 1100,
-      "train_speed(iter/s)": 0.140004
-    },
-    {
-      "epoch": 2.9810298102981028,
-      "eval_acc": 0.6206721837527445,
-      "eval_loss": 1.6525288820266724,
-      "eval_runtime": 44.2124,
-      "eval_samples_per_second": 0.859,
-      "eval_steps_per_second": 0.859,
-      "step": 1100
-    },
-    {
-      "acc": 0.74116454,
-      "epoch": 2.994579945799458,
-      "grad_norm": 2.65625,
-      "learning_rate": 3.9693295292439366e-05,
-      "loss": 0.92264805,
-      "memory(GiB)": 16.34,
-      "step": 1105,
-      "train_speed(iter/s)": 0.139311
-    },
-    {
-      "acc": 0.8182869,
-      "epoch": 3.008130081300813,
-      "grad_norm": 1.3671875,
-      "learning_rate": 3.915834522111269e-05,
-      "loss": 0.6927434,
-      "memory(GiB)": 16.34,
-      "step": 1110,
-      "train_speed(iter/s)": 0.13923
-    },
-    {
-      "acc": 0.83515587,
-      "epoch": 3.021680216802168,
-      "grad_norm": 1.9140625,
-      "learning_rate": 3.862339514978602e-05,
-      "loss": 0.58284421,
-      "memory(GiB)": 16.34,
-      "step": 1115,
-      "train_speed(iter/s)": 0.139319
-    },
-    {
-      "acc": 0.84008894,
-      "epoch": 3.035230352303523,
-      "grad_norm": 1.8828125,
-      "learning_rate": 3.808844507845934e-05,
-      "loss": 0.59128432,
-      "memory(GiB)": 16.34,
-      "step": 1120,
-      "train_speed(iter/s)": 0.139405
-    },
-    {
-      "acc": 0.84298267,
-      "epoch": 3.048780487804878,
-      "grad_norm": 2.4375,
-      "learning_rate": 3.755349500713266e-05,
-      "loss": 0.56685705,
-      "memory(GiB)": 16.34,
-      "step": 1125,
-      "train_speed(iter/s)": 0.139493
-    },
-    {
-      "acc": 0.83680372,
-      "epoch": 3.062330623306233,
-      "grad_norm": 3.171875,
-      "learning_rate": 3.701854493580599e-05,
-      "loss": 0.55599914,
-      "memory(GiB)": 16.34,
-      "step": 1130,
-      "train_speed(iter/s)": 0.139571
-    },
-    {
-      "acc": 0.81269236,
-      "epoch": 3.075880758807588,
-      "grad_norm": 2.40625,
-      "learning_rate": 3.648359486447931e-05,
-      "loss": 0.63413367,
-      "memory(GiB)": 16.34,
-      "step": 1135,
-      "train_speed(iter/s)": 0.139659
-    },
-    {
-      "acc": 0.8300642,
-      "epoch": 3.089430894308943,
-      "grad_norm": 2.71875,
-      "learning_rate": 3.5948644793152636e-05,
-      "loss": 0.59288554,
-      "memory(GiB)": 16.34,
-      "step": 1140,
-      "train_speed(iter/s)": 0.139738
-    },
-    {
-      "acc": 0.84225779,
-      "epoch": 3.102981029810298,
-      "grad_norm": 2.703125,
-      "learning_rate": 3.541369472182596e-05,
-      "loss": 0.57104263,
-      "memory(GiB)": 16.34,
-      "step": 1145,
-      "train_speed(iter/s)": 0.13982
-    },
-    {
-      "acc": 0.86128368,
-      "epoch": 3.116531165311653,
-      "grad_norm": 3.0,
-      "learning_rate": 3.487874465049928e-05,
-      "loss": 0.47776127,
-      "memory(GiB)": 16.34,
-      "step": 1150,
-      "train_speed(iter/s)": 0.139903
-    },
-    {
-      "epoch": 3.116531165311653,
-      "eval_acc": 0.6091876372234419,
-      "eval_loss": 1.9067459106445312,
-      "eval_runtime": 44.213,
-      "eval_samples_per_second": 0.859,
-      "eval_steps_per_second": 0.859,
-      "step": 1150
-    },
-    {
-      "acc": 0.81826935,
-      "epoch": 3.130081300813008,
-      "grad_norm": 2.234375,
-      "learning_rate": 3.4343794579172606e-05,
-      "loss": 0.64635777,
-      "memory(GiB)": 16.34,
-      "step": 1155,
-      "train_speed(iter/s)": 0.13924
-    },
-    {
-      "acc": 0.79423141,
-      "epoch": 3.1436314363143634,
-      "grad_norm": 2.859375,
-      "learning_rate": 3.380884450784593e-05,
-      "loss": 0.71202464,
-      "memory(GiB)": 16.34,
-      "step": 1160,
-      "train_speed(iter/s)": 0.139319
-    },
-    {
-      "acc": 0.81948729,
-      "epoch": 3.1571815718157183,
-      "grad_norm": 2.046875,
-      "learning_rate": 3.327389443651925e-05,
-      "loss": 0.62771091,
-      "memory(GiB)": 16.34,
-      "step": 1165,
-      "train_speed(iter/s)": 0.1394
-    },
-    {
-      "acc": 0.85081501,
-      "epoch": 3.1707317073170733,
-      "grad_norm": 2.125,
-      "learning_rate": 3.2738944365192576e-05,
-      "loss": 0.55290155,
-      "memory(GiB)": 16.34,
-      "step": 1170,
-      "train_speed(iter/s)": 0.139484
-    },
-    {
-      "acc": 0.84410248,
-      "epoch": 3.1842818428184283,
-      "grad_norm": 2.140625,
-      "learning_rate": 3.2203994293865906e-05,
-      "loss": 0.57252893,
-      "memory(GiB)": 16.34,
-      "step": 1175,
-      "train_speed(iter/s)": 0.139567
-    },
-    {
-      "acc": 0.78565025,
-      "epoch": 3.1978319783197833,
-      "grad_norm": 3.203125,
-      "learning_rate": 3.166904422253922e-05,
-      "loss": 0.73501797,
-      "memory(GiB)": 16.34,
-      "step": 1180,
-      "train_speed(iter/s)": 0.13965
-    },
-    {
-      "acc": 0.78738356,
-      "epoch": 3.2113821138211383,
-      "grad_norm": 2.953125,
-      "learning_rate": 3.113409415121255e-05,
-      "loss": 0.72265806,
-      "memory(GiB)": 16.34,
-      "step": 1185,
-      "train_speed(iter/s)": 0.139732
-    },
-    {
-      "acc": 0.818221,
-      "epoch": 3.2249322493224932,
-      "grad_norm": 2.21875,
-      "learning_rate": 3.0599144079885876e-05,
-      "loss": 0.61062384,
-      "memory(GiB)": 16.34,
-      "step": 1190,
-      "train_speed(iter/s)": 0.139813
-    },
-    {
-      "acc": 0.82556448,
-      "epoch": 3.238482384823848,
-      "grad_norm": 2.3125,
-      "learning_rate": 3.00641940085592e-05,
-      "loss": 0.61080799,
-      "memory(GiB)": 16.34,
-      "step": 1195,
-      "train_speed(iter/s)": 0.139893
-    },
-    {
-      "acc": 0.82586126,
-      "epoch": 3.252032520325203,
-      "grad_norm": 2.34375,
-      "learning_rate": 2.952924393723252e-05,
-      "loss": 0.6164794,
-      "memory(GiB)": 16.34,
-      "step": 1200,
-      "train_speed(iter/s)": 0.139974
-    },
-    {
-      "epoch": 3.252032520325203,
-      "eval_acc": 0.6044587062996115,
-      "eval_loss": 1.8586076498031616,
-      "eval_runtime": 44.1645,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
-      "step": 1200
-    },
-    {
-      "acc": 0.82835131,
-      "epoch": 3.265582655826558,
-      "grad_norm": 6.46875,
-      "learning_rate": 2.8994293865905846e-05,
-      "loss": 0.61753302,
-      "memory(GiB)": 16.34,
-      "step": 1205,
-      "train_speed(iter/s)": 0.139339
-    },
-    {
-      "acc": 0.82603951,
-      "epoch": 3.279132791327913,
-      "grad_norm": 2.90625,
-      "learning_rate": 2.8459343794579173e-05,
-      "loss": 0.58360481,
-      "memory(GiB)": 16.34,
-      "step": 1210,
-      "train_speed(iter/s)": 0.139423
-    },
-    {
-      "acc": 0.79192553,
-      "epoch": 3.292682926829268,
-      "grad_norm": 2.5,
-      "learning_rate": 2.7924393723252493e-05,
-      "loss": 0.75090132,
-      "memory(GiB)": 16.34,
-      "step": 1215,
-      "train_speed(iter/s)": 0.1395
-    },
-    {
-      "acc": 0.83583679,
-      "epoch": 3.306233062330623,
-      "grad_norm": 2.15625,
-      "learning_rate": 2.738944365192582e-05,
-      "loss": 0.56217456,
-      "memory(GiB)": 16.34,
-      "step": 1220,
-      "train_speed(iter/s)": 0.139581
-    },
-    {
-      "acc": 0.80481911,
-      "epoch": 3.3197831978319785,
-      "grad_norm": 4.125,
-      "learning_rate": 2.685449358059914e-05,
-      "loss": 0.73073688,
-      "memory(GiB)": 16.34,
-      "step": 1225,
-      "train_speed(iter/s)": 0.139659
-    },
-    {
-      "acc": 0.8173975,
-      "epoch": 3.3333333333333335,
-      "grad_norm": 3.1875,
-      "learning_rate": 2.6319543509272466e-05,
-      "loss": 0.59854188,
-      "memory(GiB)": 16.34,
-      "step": 1230,
-      "train_speed(iter/s)": 0.139735
-    },
-    {
-      "acc": 0.88763266,
-      "epoch": 3.3468834688346885,
-      "grad_norm": 2.078125,
-      "learning_rate": 2.5784593437945793e-05,
-      "loss": 0.3878314,
-      "memory(GiB)": 16.34,
-      "step": 1235,
-      "train_speed(iter/s)": 0.139826
-    },
-    {
-      "acc": 0.82199507,
-      "epoch": 3.3604336043360434,
-      "grad_norm": 2.28125,
-      "learning_rate": 2.5249643366619113e-05,
-      "loss": 0.65350094,
-      "memory(GiB)": 16.34,
-      "step": 1240,
-      "train_speed(iter/s)": 0.139901
-    },
-    {
-      "acc": 0.83115234,
-      "epoch": 3.3739837398373984,
-      "grad_norm": 2.484375,
-      "learning_rate": 2.471469329529244e-05,
-      "loss": 0.61815515,
-      "memory(GiB)": 16.34,
-      "step": 1245,
-      "train_speed(iter/s)": 0.13998
-    },
-    {
-      "acc": 0.81266117,
-      "epoch": 3.3875338753387534,
-      "grad_norm": 1.78125,
-      "learning_rate": 2.417974322396576e-05,
-      "loss": 0.65289016,
-      "memory(GiB)": 16.34,
-      "step": 1250,
-      "train_speed(iter/s)": 0.140052
-    },
-    {
-      "epoch": 3.3875338753387534,
-      "eval_acc": 0.6037831447390644,
-      "eval_loss": 1.893083095550537,
-      "eval_runtime": 44.1639,
-      "eval_samples_per_second": 0.86,
-      "eval_steps_per_second": 0.86,
-      "step": 1250
-    },
-    {
-      "acc": 0.84776039,
-      "epoch": 3.4010840108401084,
-      "grad_norm": 2.046875,
-      "learning_rate": 2.3644793152639086e-05,
-      "loss": 0.53162961,
-      "memory(GiB)": 16.34,
-      "step": 1255,
-      "train_speed(iter/s)": 0.139443
-    },
-    {
-      "acc": 0.84318447,
-      "epoch": 3.4146341463414633,
-      "grad_norm": 2.234375,
-      "learning_rate": 2.3109843081312406e-05,
-      "loss": 0.52226324,
-      "memory(GiB)": 16.34,
-      "step": 1260,
-      "train_speed(iter/s)": 0.139515
-    },
-    {
-      "acc": 0.80521078,
-      "epoch": 3.4281842818428183,
-      "grad_norm": 2.25,
-      "learning_rate": 2.2574893009985733e-05,
-      "loss": 0.62338996,
-      "memory(GiB)": 16.34,
-      "step": 1265,
-      "train_speed(iter/s)": 0.13959
-    },
-    {
-      "acc": 0.83542995,
-      "epoch": 3.4417344173441733,
-      "grad_norm": 3.0625,
-      "learning_rate": 2.203994293865906e-05,
-      "loss": 0.60965805,
-      "memory(GiB)": 16.34,
-      "step": 1270,
-      "train_speed(iter/s)": 0.139638
-    },
-    {
-      "acc": 0.8333952,
-      "epoch": 3.4552845528455283,
-      "grad_norm": 2.625,
-      "learning_rate": 2.150499286733238e-05,
-      "loss": 0.60031776,
-      "memory(GiB)": 16.34,
-      "step": 1275,
-      "train_speed(iter/s)": 0.139705
-    },
-    {
-      "acc": 0.84367867,
-      "epoch": 3.4688346883468837,
-      "grad_norm": 1.859375,
-      "learning_rate": 2.0970042796005706e-05,
-      "loss": 0.54981174,
-      "memory(GiB)": 16.34,
-      "step": 1280,
-      "train_speed(iter/s)": 0.13978
-    },
-    {
-      "acc": 0.8121892,
-      "epoch": 3.4823848238482387,
-      "grad_norm": 2.046875,
-      "learning_rate": 2.0435092724679026e-05,
-      "loss": 0.67483497,
-      "memory(GiB)": 16.34,
-      "step": 1285,
-      "train_speed(iter/s)": 0.139851
-    },
-    {
-      "acc": 0.87883768,
-      "epoch": 3.4959349593495936,
-      "grad_norm": 1.6875,
-      "learning_rate": 1.9900142653352353e-05,
-      "loss": 0.45111437,
-      "memory(GiB)": 16.34,
-      "step": 1290,
-      "train_speed(iter/s)": 0.139924
-    },
-    {
-      "acc": 0.81952734,
-      "epoch": 3.5094850948509486,
-      "grad_norm": 2.640625,
-      "learning_rate": 1.9365192582025676e-05,
-      "loss": 0.66096244,
-      "memory(GiB)": 16.34,
-      "step": 1295,
-      "train_speed(iter/s)": 0.139995
-    },
-    {
-      "acc": 0.82709446,
-      "epoch": 3.5230352303523036,
-      "grad_norm": 2.140625,
-      "learning_rate": 1.8830242510699e-05,
-      "loss": 0.58278303,
-      "memory(GiB)": 16.34,
-      "step": 1300,
-      "train_speed(iter/s)": 0.140068
-    },
-    {
-      "epoch": 3.5230352303523036,
-      "eval_acc": 0.603276473568654,
-      "eval_loss": 1.8986655473709106,
-      "eval_runtime": 44.3091,
-      "eval_samples_per_second": 0.858,
-      "eval_steps_per_second": 0.858,
-      "step": 1300
-    },
-    {
-      "acc": 0.83689537,
-      "epoch": 3.5365853658536586,
-      "grad_norm": 1.9453125,
-      "learning_rate": 1.8295292439372323e-05,
-      "loss": 0.55435977,
-      "memory(GiB)": 16.34,
-      "step": 1305,
-      "train_speed(iter/s)": 0.139476
-    },
-    {
-      "acc": 0.8525279,
-      "epoch": 3.5501355013550135,
-      "grad_norm": 2.21875,
-      "learning_rate": 1.776034236804565e-05,
-      "loss": 0.51831579,
-      "memory(GiB)": 16.34,
-      "step": 1310,
-      "train_speed(iter/s)": 0.139549
-    },
-    {
-      "acc": 0.83099527,
-      "epoch": 3.5636856368563685,
-      "grad_norm": 1.8828125,
-      "learning_rate": 1.7225392296718973e-05,
-      "loss": 0.61766572,
-      "memory(GiB)": 16.34,
-      "step": 1315,
-      "train_speed(iter/s)": 0.139617
-    },
-    {
-      "acc": 0.82115297,
-      "epoch": 3.5772357723577235,
-      "grad_norm": 2.765625,
-      "learning_rate": 1.6690442225392296e-05,
-      "loss": 0.62109137,
-      "memory(GiB)": 16.34,
-      "step": 1320,
-      "train_speed(iter/s)": 0.13969
-    },
-    {
-      "acc": 0.85132771,
-      "epoch": 3.5907859078590785,
-      "grad_norm": 2.421875,
-      "learning_rate": 1.615549215406562e-05,
-      "loss": 0.55242,
-      "memory(GiB)": 16.34,
-      "step": 1325,
-      "train_speed(iter/s)": 0.139763
-    },
-    {
-      "acc": 0.82085381,
-      "epoch": 3.6043360433604335,
-      "grad_norm": 2.8125,
-      "learning_rate": 1.5620542082738943e-05,
-      "loss": 0.62965093,
-      "memory(GiB)": 16.34,
-      "step": 1330,
-      "train_speed(iter/s)": 0.139837
-    },
-    {
-      "acc": 0.85592861,
-      "epoch": 3.617886178861789,
-      "grad_norm": 1.875,
-      "learning_rate": 1.5085592011412266e-05,
-      "loss": 0.54980545,
-      "memory(GiB)": 16.34,
-      "step": 1335,
-      "train_speed(iter/s)": 0.139907
-    },
-    {
-      "acc": 0.80786858,
-      "epoch": 3.6314363143631434,
-      "grad_norm": 2.328125,
-      "learning_rate": 1.4550641940085591e-05,
-      "loss": 0.67223382,
-      "memory(GiB)": 16.34,
-      "step": 1340,
-      "train_speed(iter/s)": 0.139974
-    },
-    {
-      "acc": 0.85045481,
-      "epoch": 3.644986449864499,
-      "grad_norm": 2.078125,
-      "learning_rate": 1.4015691868758914e-05,
-      "loss": 0.50758219,
-      "memory(GiB)": 16.34,
-      "step": 1345,
-      "train_speed(iter/s)": 0.140054
-    },
-    {
-      "acc": 0.85536547,
-      "epoch": 3.658536585365854,
-      "grad_norm": 1.7734375,
-      "learning_rate": 1.348074179743224e-05,
-      "loss": 0.53356686,
-      "memory(GiB)": 16.34,
-      "step": 1350,
-      "train_speed(iter/s)": 0.140124
-    },
-    {
-      "epoch": 3.658536585365854,
-      "eval_acc": 0.6010808984968755,
-      "eval_loss": 1.8972593545913696,
-      "eval_runtime": 44.2915,
-      "eval_samples_per_second": 0.858,
-      "eval_steps_per_second": 0.858,
-      "step": 1350
-    },
-    {
-      "acc": 0.8181016,
-      "epoch": 3.6720867208672088,
-      "grad_norm": 2.3125,
-      "learning_rate": 1.2945791726105563e-05,
-      "loss": 0.59544349,
-      "memory(GiB)": 16.34,
-      "step": 1355,
-      "train_speed(iter/s)": 0.139555
-    },
-    {
-      "acc": 0.87799835,
-      "epoch": 3.6856368563685638,
-      "grad_norm": 2.03125,
-      "learning_rate": 1.2410841654778886e-05,
-      "loss": 0.42857742,
-      "memory(GiB)": 16.34,
-      "step": 1360,
-      "train_speed(iter/s)": 0.139631
-    },
-    {
-      "acc": 0.81485729,
-      "epoch": 3.6991869918699187,
-      "grad_norm": 1.65625,
-      "learning_rate": 1.187589158345221e-05,
-      "loss": 0.64030995,
-      "memory(GiB)": 16.34,
-      "step": 1365,
-      "train_speed(iter/s)": 0.139703
-    },
-    {
-      "acc": 0.82011805,
-      "epoch": 3.7127371273712737,
-      "grad_norm": 2.125,
-      "learning_rate": 1.1340941512125534e-05,
-      "loss": 0.63428659,
-      "memory(GiB)": 16.34,
-      "step": 1370,
-      "train_speed(iter/s)": 0.139771
-    },
-    {
-      "acc": 0.78548884,
-      "epoch": 3.7262872628726287,
-      "grad_norm": 3.984375,
-      "learning_rate": 1.0805991440798858e-05,
-      "loss": 0.74558139,
-      "memory(GiB)": 16.34,
-      "step": 1375,
-      "train_speed(iter/s)": 0.139836
-    },
-    {
-      "acc": 0.85822716,
-      "epoch": 3.7398373983739837,
-      "grad_norm": 2.328125,
-      "learning_rate": 1.0271041369472183e-05,
-      "loss": 0.47486768,
-      "memory(GiB)": 16.34,
-      "step": 1380,
-      "train_speed(iter/s)": 0.139902
-    },
-    {
-      "acc": 0.83469105,
-      "epoch": 3.7533875338753386,
-      "grad_norm": 2.53125,
-      "learning_rate": 9.736091298145506e-06,
-      "loss": 0.56590233,
-      "memory(GiB)": 16.34,
-      "step": 1385,
-      "train_speed(iter/s)": 0.139971
-    },
-    {
-      "acc": 0.80486965,
-      "epoch": 3.7669376693766936,
-      "grad_norm": 3.28125,
-      "learning_rate": 9.20114122681883e-06,
-      "loss": 0.68021841,
-      "memory(GiB)": 16.34,
-      "step": 1390,
-      "train_speed(iter/s)": 0.140036
-    },
-    {
-      "acc": 0.84560328,
-      "epoch": 3.7804878048780486,
-      "grad_norm": 2.421875,
-      "learning_rate": 8.666191155492154e-06,
-      "loss": 0.52926989,
-      "memory(GiB)": 16.34,
-      "step": 1395,
-      "train_speed(iter/s)": 0.140102
-    },
-    {
-      "acc": 0.83154497,
-      "epoch": 3.794037940379404,
-      "grad_norm": 2.671875,
-      "learning_rate": 8.131241084165478e-06,
-      "loss": 0.5866107,
-      "memory(GiB)": 16.34,
-      "step": 1400,
-      "train_speed(iter/s)": 0.14017
-    },
-    {
-      "epoch": 3.794037940379404,
-      "eval_acc": 0.6036142543489276,
-      "eval_loss": 1.8983793258666992,
-      "eval_runtime": 44.4268,
-      "eval_samples_per_second": 0.855,
-      "eval_steps_per_second": 0.855,
-      "step": 1400
-    },
-    {
-      "acc": 0.84980259,
-      "epoch": 3.8075880758807585,
-      "grad_norm": 2.859375,
-      "learning_rate": 7.596291012838801e-06,
-      "loss": 0.56505013,
-      "memory(GiB)": 16.34,
-      "step": 1405,
-      "train_speed(iter/s)": 0.139621
-    },
-    {
-      "acc": 0.80126667,
-      "epoch": 3.821138211382114,
-      "grad_norm": 2.21875,
-      "learning_rate": 7.061340941512125e-06,
-      "loss": 0.67650762,
-      "memory(GiB)": 16.34,
-      "step": 1410,
-      "train_speed(iter/s)": 0.139688
-    },
-    {
-      "acc": 0.79974804,
-      "epoch": 3.834688346883469,
-      "grad_norm": 2.3125,
-      "learning_rate": 6.5263908701854486e-06,
-      "loss": 0.743855,
-      "memory(GiB)": 16.34,
-      "step": 1415,
-      "train_speed(iter/s)": 0.139749
-    },
-    {
-      "acc": 0.84023552,
-      "epoch": 3.848238482384824,
-      "grad_norm": 2.234375,
-      "learning_rate": 5.991440798858773e-06,
-      "loss": 0.57722979,
-      "memory(GiB)": 16.34,
-      "step": 1420,
-      "train_speed(iter/s)": 0.139813
-    },
-    {
-      "acc": 0.82472792,
-      "epoch": 3.861788617886179,
-      "grad_norm": 2.484375,
-      "learning_rate": 5.456490727532097e-06,
-      "loss": 0.56691737,
-      "memory(GiB)": 16.34,
-      "step": 1425,
-      "train_speed(iter/s)": 0.139877
-    },
-    {
-      "acc": 0.83573132,
-      "epoch": 3.875338753387534,
-      "grad_norm": 1.9765625,
-      "learning_rate": 4.92154065620542e-06,
-      "loss": 0.60298834,
-      "memory(GiB)": 16.34,
-      "step": 1430,
-      "train_speed(iter/s)": 0.13994
-    },
-    {
-      "acc": 0.83143454,
-      "epoch": 3.888888888888889,
-      "grad_norm": 1.9453125,
-      "learning_rate": 4.386590584878744e-06,
-      "loss": 0.61832671,
-      "memory(GiB)": 16.34,
-      "step": 1435,
-      "train_speed(iter/s)": 0.140003
-    },
-    {
-      "acc": 0.83209105,
-      "epoch": 3.902439024390244,
-      "grad_norm": 3.265625,
-      "learning_rate": 3.851640513552068e-06,
-      "loss": 0.6038147,
-      "memory(GiB)": 16.34,
-      "step": 1440,
-      "train_speed(iter/s)": 0.140065
-    },
-    {
-      "acc": 0.79715924,
-      "epoch": 3.915989159891599,
-      "grad_norm": 2.015625,
-      "learning_rate": 3.316690442225392e-06,
-      "loss": 0.70867634,
-      "memory(GiB)": 16.34,
-      "step": 1445,
-      "train_speed(iter/s)": 0.140125
-    },
-    {
-      "acc": 0.85556803,
-      "epoch": 3.9295392953929538,
-      "grad_norm": 1.7734375,
-      "learning_rate": 2.781740370898716e-06,
-      "loss": 0.52723417,
-      "memory(GiB)": 16.34,
-      "step": 1450,
-      "train_speed(iter/s)": 0.140182
-    },
-    {
-      "epoch": 3.9295392953929538,
-      "eval_acc": 0.6029386927883803,
-      "eval_loss": 1.8965932130813599,
-      "eval_runtime": 44.4125,
-      "eval_samples_per_second": 0.856,
-      "eval_steps_per_second": 0.856,
-      "step": 1450
-    },
-    {
-      "acc": 0.83306837,
-      "epoch": 3.943089430894309,
-      "grad_norm": 2.25,
-      "learning_rate": 2.2467902995720398e-06,
-      "loss": 0.55906692,
-      "memory(GiB)": 16.34,
-      "step": 1455,
-      "train_speed(iter/s)": 0.139645
-    },
-    {
-      "acc": 0.81574478,
-      "epoch": 3.9566395663956637,
-      "grad_norm": 2.6875,
-      "learning_rate": 1.7118402282453637e-06,
-      "loss": 0.62746248,
-      "memory(GiB)": 16.34,
-      "step": 1460,
-      "train_speed(iter/s)": 0.139708
-    },
-    {
-      "acc": 0.83905201,
-      "epoch": 3.970189701897019,
-      "grad_norm": 2.46875,
-      "learning_rate": 1.1768901569186875e-06,
-      "loss": 0.56978045,
-      "memory(GiB)": 16.34,
-      "step": 1465,
-      "train_speed(iter/s)": 0.139771
-    },
-    {
-      "acc": 0.85569115,
-      "epoch": 3.983739837398374,
-      "grad_norm": 1.90625,
-      "learning_rate": 6.419400855920114e-07,
-      "loss": 0.48882861,
-      "memory(GiB)": 16.34,
-      "step": 1470,
-      "train_speed(iter/s)": 0.139834
-    },
-    {
-      "acc": 0.82850809,
-      "epoch": 3.997289972899729,
-      "grad_norm": 1.828125,
-      "learning_rate": 1.0699001426533523e-07,
-      "loss": 0.56869435,
-      "memory(GiB)": 16.34,
-      "step": 1475,
-      "train_speed(iter/s)": 0.139901
-    },
-    {
-      "epoch": 4.0,
-      "eval_acc": 0.6037831447390644,
-      "eval_loss": 1.8978389501571655,
-      "eval_runtime": 44.2826,
-      "eval_samples_per_second": 0.858,
-      "eval_steps_per_second": 0.858,
-      "step": 1476
     }
   ],
   "logging_steps": 5,
-  "max_steps": 1476,
+  "max_steps": 1020,
   "num_input_tokens_seen": 0,
   "num_train_epochs": 4,
-  "save_steps": 369,
-  "total_flos": 1.991427194376192e+17,
-  "train_batch_size": 1,
+  "save_steps": 300,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.605539502350213e+17,
+  "train_batch_size": 2,
   "trial_name": null,
   "trial_params": null
 }