diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,6151 @@
+{
+  "best_metric": 0.65735477,
+  "best_model_checkpoint": "/cpfs04/shared/rlproject/zhangqi/model_garden/0709_intern2b_v7-1-part15-19-resize-decay/v0-20250710-072707/checkpoint-3000",
+  "epoch": 0.9599616015359386,
+  "eval_steps": 250,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0003199872005119795,
+      "grad_norm": 0.06346331978668461,
+      "learning_rate": 2.0000000000000002e-07,
+      "loss": 0.4898327589035034,
+      "memory(GiB)": 37.39,
+      "step": 1,
+      "token_acc": 0.8589147286821706,
+      "train_speed(iter/s)": 0.017141
+    },
+    {
+      "epoch": 0.0015999360025598975,
+      "grad_norm": 0.06701772816278752,
+      "learning_rate": 1.0000000000000002e-06,
+      "loss": 0.479672372341156,
+      "memory(GiB)": 57.22,
+      "step": 5,
+      "token_acc": 0.8759901666211418,
+      "train_speed(iter/s)": 0.029212
+    },
+    {
+      "epoch": 0.003199872005119795,
+      "grad_norm": 0.0577242460197625,
+      "learning_rate": 2.0000000000000003e-06,
+      "loss": 0.4767899990081787,
+      "memory(GiB)": 57.22,
+      "step": 10,
+      "token_acc": 0.8920853213584058,
+      "train_speed(iter/s)": 0.030985
+    },
+    {
+      "epoch": 0.004799808007679693,
+      "grad_norm": 0.05378464715409511,
+      "learning_rate": 3e-06,
+      "loss": 0.46750531196594236,
+      "memory(GiB)": 57.22,
+      "step": 15,
+      "token_acc": 0.846137292877125,
+      "train_speed(iter/s)": 0.030693
+    },
+    {
+      "epoch": 0.00639974401023959,
+      "grad_norm": 0.05482863789544427,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.47017059326171873,
+      "memory(GiB)": 57.22,
+      "step": 20,
+      "token_acc": 0.8701638201463925,
+      "train_speed(iter/s)": 0.031839
+    },
+    {
+      "epoch": 0.007999680012799487,
+      "grad_norm": 0.05575025715512655,
+      "learning_rate": 5e-06,
+      "loss": 0.46974716186523435,
+      "memory(GiB)": 57.22,
+      "step": 25,
+      "token_acc": 0.8939419941650936,
+      "train_speed(iter/s)": 0.031203
+    },
+    {
+      "epoch": 0.009599616015359386,
+      "grad_norm": 0.05007200892621057,
+      "learning_rate": 6e-06,
+      "loss": 0.4700496196746826,
+      "memory(GiB)": 57.22,
+      "step": 30,
+      "token_acc": 0.872761844398669,
+      "train_speed(iter/s)": 0.030738
+    },
+    {
+      "epoch": 0.011199552017919284,
+      "grad_norm": 0.06284973201247342,
+      "learning_rate": 7e-06,
+      "loss": 0.477018404006958,
+      "memory(GiB)": 57.22,
+      "step": 35,
+      "token_acc": 0.8738273921200751,
+      "train_speed(iter/s)": 0.031491
+    },
+    {
+      "epoch": 0.01279948802047918,
+      "grad_norm": 0.051858307829802386,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.46584124565124513,
+      "memory(GiB)": 57.22,
+      "step": 40,
+      "token_acc": 0.8791312559017942,
+      "train_speed(iter/s)": 0.031035
+    },
+    {
+      "epoch": 0.014399424023039079,
+      "grad_norm": 0.05024484287076301,
+      "learning_rate": 9e-06,
+      "loss": 0.4685808658599854,
+      "memory(GiB)": 57.22,
+      "step": 45,
+      "token_acc": 0.8846325167037862,
+      "train_speed(iter/s)": 0.031503
+    },
+    {
+      "epoch": 0.015999360025598975,
+      "grad_norm": 0.06025612278216295,
+      "learning_rate": 1e-05,
+      "loss": 0.4697974681854248,
+      "memory(GiB)": 57.22,
+      "step": 50,
+      "token_acc": 0.854153041203401,
+      "train_speed(iter/s)": 0.031549
+    },
+    {
+      "epoch": 0.017599296028158875,
+      "grad_norm": 0.05171252494611451,
+      "learning_rate": 1.1000000000000001e-05,
+      "loss": 0.4622661113739014,
+      "memory(GiB)": 73.06,
+      "step": 55,
+      "token_acc": 0.8281821878812525,
+      "train_speed(iter/s)": 0.031181
+    },
+    {
+      "epoch": 0.01919923203071877,
+      "grad_norm": 0.06238294030934267,
+      "learning_rate": 1.2e-05,
+      "loss": 0.47000856399536134,
+      "memory(GiB)": 73.06,
+      "step": 60,
+      "token_acc": 0.8703465982028241,
+      "train_speed(iter/s)": 0.031617
+    },
+    {
+      "epoch": 0.020799168033278668,
+      "grad_norm": 0.05576663204628902,
+      "learning_rate": 1.3000000000000001e-05,
+      "loss": 0.4765446186065674,
+      "memory(GiB)": 73.06,
+      "step": 65,
+      "token_acc": 0.867019517036057,
+      "train_speed(iter/s)": 0.031138
+    },
+    {
+      "epoch": 0.022399104035838568,
+      "grad_norm": 0.054406694286476175,
+      "learning_rate": 1.4e-05,
+      "loss": 0.47959036827087403,
+      "memory(GiB)": 73.06,
+      "step": 70,
+      "token_acc": 0.8927940657011657,
+      "train_speed(iter/s)": 0.030807
+    },
+    {
+      "epoch": 0.023999040038398464,
+      "grad_norm": 0.05458413282309297,
+      "learning_rate": 1.5000000000000002e-05,
+      "loss": 0.45742173194885255,
+      "memory(GiB)": 73.06,
+      "step": 75,
+      "token_acc": 0.8916037316748112,
+      "train_speed(iter/s)": 0.031204
+    },
+    {
+      "epoch": 0.02559897604095836,
+      "grad_norm": 0.05986601718141533,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.46221466064453126,
+      "memory(GiB)": 73.06,
+      "step": 80,
+      "token_acc": 0.8513287970214405,
+      "train_speed(iter/s)": 0.030907
+    },
+    {
+      "epoch": 0.02719891204351826,
+      "grad_norm": 0.0610597448330617,
+      "learning_rate": 1.7e-05,
+      "loss": 0.4636178493499756,
+      "memory(GiB)": 73.06,
+      "step": 85,
+      "token_acc": 0.8755391250770179,
+      "train_speed(iter/s)": 0.030852
+    },
+    {
+      "epoch": 0.028798848046078157,
+      "grad_norm": 0.0537272421077024,
+      "learning_rate": 1.8e-05,
+      "loss": 0.4674212455749512,
+      "memory(GiB)": 73.06,
+      "step": 90,
+      "token_acc": 0.8330635370295427,
+      "train_speed(iter/s)": 0.03096
+    },
+    {
+      "epoch": 0.030398784048638054,
+      "grad_norm": 0.05874793146275866,
+      "learning_rate": 1.9e-05,
+      "loss": 0.459043550491333,
+      "memory(GiB)": 73.06,
+      "step": 95,
+      "token_acc": 0.8686557483379277,
+      "train_speed(iter/s)": 0.030744
+    },
+    {
+      "epoch": 0.03199872005119795,
+      "grad_norm": 0.0635697375742028,
+      "learning_rate": 2e-05,
+      "loss": 0.4696988582611084,
+      "memory(GiB)": 73.06,
+      "step": 100,
+      "token_acc": 0.8658624414836155,
+      "train_speed(iter/s)": 0.030967
+    },
+    {
+      "epoch": 0.03359865605375785,
+      "grad_norm": 0.05212754941929234,
+      "learning_rate": 1.9999865178850847e-05,
+      "loss": 0.47245235443115235,
+      "memory(GiB)": 73.06,
+      "step": 105,
+      "token_acc": 0.8773064185459707,
+      "train_speed(iter/s)": 0.030857
+    },
+    {
+      "epoch": 0.03519859205631775,
+      "grad_norm": 0.0584305183084298,
+      "learning_rate": 1.999946071903873e-05,
+      "loss": 0.4617309093475342,
+      "memory(GiB)": 73.06,
+      "step": 110,
+      "token_acc": 0.8634434872058634,
+      "train_speed(iter/s)": 0.030617
+    },
+    {
+      "epoch": 0.03679852805887764,
+      "grad_norm": 0.0594956422290189,
+      "learning_rate": 1.9998786631469602e-05,
+      "loss": 0.4737922191619873,
+      "memory(GiB)": 73.06,
+      "step": 115,
+      "token_acc": 0.8586429725363489,
+      "train_speed(iter/s)": 0.030917
+    },
+    {
+      "epoch": 0.03839846406143754,
+      "grad_norm": 0.05542732612540538,
+      "learning_rate": 1.999784293431971e-05,
+      "loss": 0.46903514862060547,
+      "memory(GiB)": 73.06,
+      "step": 120,
+      "token_acc": 0.8490970309152127,
+      "train_speed(iter/s)": 0.030707
+    },
+    {
+      "epoch": 0.03999840006399744,
+      "grad_norm": 0.06279644130355853,
+      "learning_rate": 1.9996629653035128e-05,
+      "loss": 0.47660508155822756,
+      "memory(GiB)": 73.06,
+      "step": 125,
+      "token_acc": 0.8797399783315276,
+      "train_speed(iter/s)": 0.03059
+    },
+    {
+      "epoch": 0.041598336066557336,
+      "grad_norm": 0.0589281224878424,
+      "learning_rate": 1.999514682033104e-05,
+      "loss": 0.47182955741882326,
+      "memory(GiB)": 73.06,
+      "step": 130,
+      "token_acc": 0.8641215106732348,
+      "train_speed(iter/s)": 0.030792
+    },
+    {
+      "epoch": 0.043198272069117236,
+      "grad_norm": 0.06201767062726238,
+      "learning_rate": 1.99933944761909e-05,
+      "loss": 0.4760914325714111,
+      "memory(GiB)": 73.06,
+      "step": 135,
+      "token_acc": 0.856929955290611,
+      "train_speed(iter/s)": 0.030572
+    },
+    {
+      "epoch": 0.044798208071677136,
+      "grad_norm": 0.06842512509591067,
+      "learning_rate": 1.999137266786531e-05,
+      "loss": 0.4673017501831055,
+      "memory(GiB)": 73.06,
+      "step": 140,
+      "token_acc": 0.8725868725868726,
+      "train_speed(iter/s)": 0.030583
+    },
+    {
+      "epoch": 0.04639814407423703,
+      "grad_norm": 0.06206860242026748,
+      "learning_rate": 1.998908144987078e-05,
+      "loss": 0.4650784969329834,
+      "memory(GiB)": 73.06,
+      "step": 145,
+      "token_acc": 0.8592991206195039,
+      "train_speed(iter/s)": 0.030589
+    },
+    {
+      "epoch": 0.04799808007679693,
+      "grad_norm": 0.07443005443822684,
+      "learning_rate": 1.9986520883988233e-05,
+      "loss": 0.46671695709228517,
+      "memory(GiB)": 73.06,
+      "step": 150,
+      "token_acc": 0.8422288642186165,
+      "train_speed(iter/s)": 0.030331
+    },
+    {
+      "epoch": 0.04959801607935683,
+      "grad_norm": 0.06105984389773331,
+      "learning_rate": 1.9983691039261358e-05,
+      "loss": 0.47356271743774414,
+      "memory(GiB)": 73.06,
+      "step": 155,
+      "token_acc": 0.8814565604591332,
+      "train_speed(iter/s)": 0.030476
+    },
+    {
+      "epoch": 0.05119795208191672,
+      "grad_norm": 0.08001789963695773,
+      "learning_rate": 1.998059199199474e-05,
+      "loss": 0.4710524559020996,
+      "memory(GiB)": 73.06,
+      "step": 160,
+      "token_acc": 0.8710629921259843,
+      "train_speed(iter/s)": 0.03042
+    },
+    {
+      "epoch": 0.05279788808447662,
+      "grad_norm": 0.05874066660842649,
+      "learning_rate": 1.9977223825751802e-05,
+      "loss": 0.46933708190917967,
+      "memory(GiB)": 73.06,
+      "step": 165,
+      "token_acc": 0.894688221709007,
+      "train_speed(iter/s)": 0.030265
+    },
+    {
+      "epoch": 0.05439782408703652,
+      "grad_norm": 0.06544476253513323,
+      "learning_rate": 1.997358663135255e-05,
+      "loss": 0.46097607612609864,
+      "memory(GiB)": 73.06,
+      "step": 170,
+      "token_acc": 0.8832285384319261,
+      "train_speed(iter/s)": 0.030391
+    },
+    {
+      "epoch": 0.055997760089596414,
+      "grad_norm": 0.052612734205809274,
+      "learning_rate": 1.9969680506871138e-05,
+      "loss": 0.4674376010894775,
+      "memory(GiB)": 73.06,
+      "step": 175,
+      "token_acc": 0.8822400558269365,
+      "train_speed(iter/s)": 0.030298
+    },
+    {
+      "epoch": 0.057597696092156314,
+      "grad_norm": 0.05831524002340728,
+      "learning_rate": 1.9965505557633188e-05,
+      "loss": 0.47021942138671874,
+      "memory(GiB)": 73.06,
+      "step": 180,
+      "token_acc": 0.8729036501150937,
+      "train_speed(iter/s)": 0.030222
+    },
+    {
+      "epoch": 0.059197632094716214,
+      "grad_norm": 0.05558463298745032,
+      "learning_rate": 1.9961061896213006e-05,
+      "loss": 0.4707474708557129,
+      "memory(GiB)": 73.06,
+      "step": 185,
+      "token_acc": 0.8507278220269157,
+      "train_speed(iter/s)": 0.030322
+    },
+    {
+      "epoch": 0.06079756809727611,
+      "grad_norm": 0.05835065786169716,
+      "learning_rate": 1.9956349642430494e-05,
+      "loss": 0.4792951583862305,
+      "memory(GiB)": 73.06,
+      "step": 190,
+      "token_acc": 0.8596896665566194,
+      "train_speed(iter/s)": 0.030221
+    },
+    {
+      "epoch": 0.06239750409983601,
+      "grad_norm": 0.055138234334700054,
+      "learning_rate": 1.9951368923347945e-05,
+      "loss": 0.4755210876464844,
+      "memory(GiB)": 73.06,
+      "step": 195,
+      "token_acc": 0.8501170960187353,
+      "train_speed(iter/s)": 0.030259
+    },
+    {
+      "epoch": 0.0639974401023959,
+      "grad_norm": 0.06763946666899583,
+      "learning_rate": 1.9946119873266615e-05,
+      "loss": 0.4560092926025391,
+      "memory(GiB)": 73.06,
+      "step": 200,
+      "token_acc": 0.8891928864569083,
+      "train_speed(iter/s)": 0.030234
+    },
+    {
+      "epoch": 0.0655973761049558,
+      "grad_norm": 0.056758126149685124,
+      "learning_rate": 1.9940602633723097e-05,
+      "loss": 0.470977258682251,
+      "memory(GiB)": 73.06,
+      "step": 205,
+      "token_acc": 0.8635224424698109,
+      "train_speed(iter/s)": 0.030086
+    },
+    {
+      "epoch": 0.0671973121075157,
+      "grad_norm": 0.06089496841996932,
+      "learning_rate": 1.99348173534855e-05,
+      "loss": 0.4699739456176758,
+      "memory(GiB)": 73.06,
+      "step": 210,
+      "token_acc": 0.8621679827709978,
+      "train_speed(iter/s)": 0.03019
+    },
+    {
+      "epoch": 0.06879724811007559,
+      "grad_norm": 0.05802383504947049,
+      "learning_rate": 1.9928764188549462e-05,
+      "loss": 0.46386079788208007,
+      "memory(GiB)": 73.06,
+      "step": 215,
+      "token_acc": 0.8754250939681403,
+      "train_speed(iter/s)": 0.030114
+    },
+    {
+      "epoch": 0.0703971841126355,
+      "grad_norm": 0.05515703750577632,
+      "learning_rate": 1.9922443302133906e-05,
+      "loss": 0.4679898262023926,
+      "memory(GiB)": 73.06,
+      "step": 220,
+      "token_acc": 0.88738807102747,
+      "train_speed(iter/s)": 0.030007
+    },
+    {
+      "epoch": 0.07199712011519539,
+      "grad_norm": 0.05967474480415036,
+      "learning_rate": 1.9915854864676665e-05,
+      "loss": 0.47310919761657716,
+      "memory(GiB)": 73.06,
+      "step": 225,
+      "token_acc": 0.8383795309168444,
+      "train_speed(iter/s)": 0.030071
+    },
+    {
+      "epoch": 0.07359705611775529,
+      "grad_norm": 0.06076387420670948,
+      "learning_rate": 1.990899905382988e-05,
+      "loss": 0.4678232192993164,
+      "memory(GiB)": 73.06,
+      "step": 230,
+      "token_acc": 0.8767123287671232,
+      "train_speed(iter/s)": 0.029948
+    },
+    {
+      "epoch": 0.07519699212031519,
+      "grad_norm": 0.05635803811030448,
+      "learning_rate": 1.9901876054455217e-05,
+      "loss": 0.4821170330047607,
+      "memory(GiB)": 73.06,
+      "step": 235,
+      "token_acc": 0.8841222879684418,
+      "train_speed(iter/s)": 0.029913
+    },
+    {
+      "epoch": 0.07679692812287509,
+      "grad_norm": 0.053886114557468945,
+      "learning_rate": 1.9894486058618863e-05,
+      "loss": 0.46213107109069823,
+      "memory(GiB)": 73.06,
+      "step": 240,
+      "token_acc": 0.8886558627264061,
+      "train_speed(iter/s)": 0.02992
+    },
+    {
+      "epoch": 0.07839686412543498,
+      "grad_norm": 0.06048992108753748,
+      "learning_rate": 1.9886829265586368e-05,
+      "loss": 0.4749046802520752,
+      "memory(GiB)": 73.06,
+      "step": 245,
+      "token_acc": 0.8758281279575999,
+      "train_speed(iter/s)": 0.0298
+    },
+    {
+      "epoch": 0.07999680012799489,
+      "grad_norm": 0.06662896613700448,
+      "learning_rate": 1.9878905881817254e-05,
+      "loss": 0.47487664222717285,
+      "memory(GiB)": 73.06,
+      "step": 250,
+      "token_acc": 0.8239743295897318,
+      "train_speed(iter/s)": 0.029855
+    },
+    {
+      "epoch": 0.07999680012799489,
+      "eval_loss": 0.6802101731300354,
+      "eval_runtime": 108.8605,
+      "eval_samples_per_second": 184.53,
+      "eval_steps_per_second": 0.928,
+      "eval_token_acc": 0.8656411339267154,
+      "step": 250
+    },
+    {
+      "epoch": 0.08159673613055478,
+      "grad_norm": 0.058967589577093804,
+      "learning_rate": 1.9870716120959462e-05,
+      "loss": 0.4691306591033936,
+      "memory(GiB)": 73.24,
+      "step": 255,
+      "token_acc": 0.8612697569398327,
+      "train_speed(iter/s)": 0.029606
+    },
+    {
+      "epoch": 0.08319667213311467,
+      "grad_norm": 0.057353651690814994,
+      "learning_rate": 1.986226020384359e-05,
+      "loss": 0.46143622398376466,
+      "memory(GiB)": 73.24,
+      "step": 260,
+      "token_acc": 0.8685547371094742,
+      "train_speed(iter/s)": 0.029681
+    },
+    {
+      "epoch": 0.08479660813567458,
+      "grad_norm": 0.05409688809510523,
+      "learning_rate": 1.9853538358476933e-05,
+      "loss": 0.4704445838928223,
+      "memory(GiB)": 73.24,
+      "step": 265,
+      "token_acc": 0.8804637020144431,
+      "train_speed(iter/s)": 0.02979
+    },
+    {
+      "epoch": 0.08639654413823447,
+      "grad_norm": 0.06968473514476099,
+      "learning_rate": 1.9844550820037326e-05,
+      "loss": 0.4717890739440918,
+      "memory(GiB)": 73.24,
+      "step": 270,
+      "token_acc": 0.8638003174145145,
+      "train_speed(iter/s)": 0.029834
+    },
+    {
+      "epoch": 0.08799648014079436,
+      "grad_norm": 0.06009720175343309,
+      "learning_rate": 1.9835297830866827e-05,
+      "loss": 0.4709662437438965,
+      "memory(GiB)": 73.24,
+      "step": 275,
+      "token_acc": 0.8634590377113134,
+      "train_speed(iter/s)": 0.029835
+    },
+    {
+      "epoch": 0.08959641614335427,
+      "grad_norm": 0.058778539356308675,
+      "learning_rate": 1.9825779640465157e-05,
+      "loss": 0.47084336280822753,
+      "memory(GiB)": 73.24,
+      "step": 280,
+      "token_acc": 0.9203691779351793,
+      "train_speed(iter/s)": 0.029942
+    },
+    {
+      "epoch": 0.09119635214591416,
+      "grad_norm": 0.054325246749067864,
+      "learning_rate": 1.9815996505483e-05,
+      "loss": 0.4666774749755859,
+      "memory(GiB)": 73.24,
+      "step": 285,
+      "token_acc": 0.8521723454119344,
+      "train_speed(iter/s)": 0.029906
+    },
+    {
+      "epoch": 0.09279628814847406,
+      "grad_norm": 0.058698263071843435,
+      "learning_rate": 1.9805948689715043e-05,
+      "loss": 0.45826416015625,
+      "memory(GiB)": 73.24,
+      "step": 290,
+      "token_acc": 0.8421138211382114,
+      "train_speed(iter/s)": 0.029895
+    },
+    {
+      "epoch": 0.09439622415103396,
+      "grad_norm": 0.05517972536097747,
+      "learning_rate": 1.979563646409291e-05,
+      "loss": 0.47627692222595214,
+      "memory(GiB)": 73.24,
+      "step": 295,
+      "token_acc": 0.8784122999686226,
+      "train_speed(iter/s)": 0.029966
+    },
+    {
+      "epoch": 0.09599616015359386,
+      "grad_norm": 0.06223926082468345,
+      "learning_rate": 1.9785060106677818e-05,
+      "loss": 0.4711057186126709,
+      "memory(GiB)": 73.24,
+      "step": 300,
+      "token_acc": 0.876372039283651,
+      "train_speed(iter/s)": 0.029911
+    },
+    {
+      "epoch": 0.09759609615615375,
+      "grad_norm": 0.060678733702642235,
+      "learning_rate": 1.97742199026531e-05,
+      "loss": 0.46833024024963377,
+      "memory(GiB)": 73.24,
+      "step": 305,
+      "token_acc": 0.8586995355484102,
+      "train_speed(iter/s)": 0.029998
+    },
+    {
+      "epoch": 0.09919603215871366,
+      "grad_norm": 0.06117494885421727,
+      "learning_rate": 1.9763116144316506e-05,
+      "loss": 0.4692807197570801,
+      "memory(GiB)": 73.24,
+      "step": 310,
+      "token_acc": 0.8383072793304911,
+      "train_speed(iter/s)": 0.029983
+    },
+    {
+      "epoch": 0.10079596816127355,
+      "grad_norm": 0.059512004342169564,
+      "learning_rate": 1.9751749131072335e-05,
+      "loss": 0.462421178817749,
+      "memory(GiB)": 73.24,
+      "step": 315,
+      "token_acc": 0.865073787772312,
+      "train_speed(iter/s)": 0.029959
+    },
+    {
+      "epoch": 0.10239590416383344,
+      "grad_norm": 0.05759903892800583,
+      "learning_rate": 1.9740119169423337e-05,
+      "loss": 0.4749638080596924,
+      "memory(GiB)": 73.24,
+      "step": 320,
+      "token_acc": 0.8657438292194797,
+      "train_speed(iter/s)": 0.030064
+    },
+    {
+      "epoch": 0.10399584016639335,
+      "grad_norm": 0.05512670495542287,
+      "learning_rate": 1.9728226572962474e-05,
+      "loss": 0.48053979873657227,
+      "memory(GiB)": 73.24,
+      "step": 325,
+      "token_acc": 0.9068181818181819,
+      "train_speed(iter/s)": 0.03004
+    },
+    {
+      "epoch": 0.10559577616895324,
+      "grad_norm": 0.05723038100011267,
+      "learning_rate": 1.9716071662364454e-05,
+      "loss": 0.47551665306091306,
+      "memory(GiB)": 73.24,
+      "step": 330,
+      "token_acc": 0.8362432269717038,
+      "train_speed(iter/s)": 0.030003
+    },
+    {
+      "epoch": 0.10719571217151314,
+      "grad_norm": 0.057638605082885846,
+      "learning_rate": 1.970365476537707e-05,
+      "loss": 0.4652701854705811,
+      "memory(GiB)": 73.24,
+      "step": 335,
+      "token_acc": 0.8735049401976079,
+      "train_speed(iter/s)": 0.030082
+    },
+    {
+      "epoch": 0.10879564817407304,
+      "grad_norm": 0.05903871731521889,
+      "learning_rate": 1.9690976216812397e-05,
+      "loss": 0.4698742389678955,
+      "memory(GiB)": 73.24,
+      "step": 340,
+      "token_acc": 0.8620361560418649,
+      "train_speed(iter/s)": 0.030027
+    },
+    {
+      "epoch": 0.11039558417663294,
+      "grad_norm": 0.053856521964694516,
+      "learning_rate": 1.9678036358537726e-05,
+      "loss": 0.4701416015625,
+      "memory(GiB)": 73.24,
+      "step": 345,
+      "token_acc": 0.8708435421771089,
+      "train_speed(iter/s)": 0.03002
+    },
+    {
+      "epoch": 0.11199552017919283,
+      "grad_norm": 0.05586893539038131,
+      "learning_rate": 1.966483553946637e-05,
+      "loss": 0.47447028160095217,
+      "memory(GiB)": 73.24,
+      "step": 350,
+      "token_acc": 0.8617533718689788,
+      "train_speed(iter/s)": 0.030041
+    },
+    {
+      "epoch": 0.11359545618175274,
+      "grad_norm": 0.052599438001953325,
+      "learning_rate": 1.9651374115548255e-05,
+      "loss": 0.4637298583984375,
+      "memory(GiB)": 73.24,
+      "step": 355,
+      "token_acc": 0.8874341610233258,
+      "train_speed(iter/s)": 0.029967
+    },
+    {
+      "epoch": 0.11519539218431263,
+      "grad_norm": 0.05804143123407663,
+      "learning_rate": 1.9637652449760297e-05,
+      "loss": 0.4660144329071045,
+      "memory(GiB)": 73.24,
+      "step": 360,
+      "token_acc": 0.8349885408708938,
+      "train_speed(iter/s)": 0.030034
+    },
+    {
+      "epoch": 0.11679532818687252,
+      "grad_norm": 0.06055547849970778,
+      "learning_rate": 1.9623670912096656e-05,
+      "loss": 0.4716383934020996,
+      "memory(GiB)": 73.24,
+      "step": 365,
+      "token_acc": 0.8751012473675684,
+      "train_speed(iter/s)": 0.02998
+    },
+    {
+      "epoch": 0.11839526418943243,
+      "grad_norm": 0.058520598293842735,
+      "learning_rate": 1.9609429879558726e-05,
+      "loss": 0.46298699378967284,
+      "memory(GiB)": 73.24,
+      "step": 370,
+      "token_acc": 0.8553921568627451,
+      "train_speed(iter/s)": 0.029931
+    },
+    {
+      "epoch": 0.11999520019199232,
+      "grad_norm": 0.058584318589478955,
+      "learning_rate": 1.9594929736144978e-05,
+      "loss": 0.4756875514984131,
+      "memory(GiB)": 73.24,
+      "step": 375,
+      "token_acc": 0.8618346545866364,
+      "train_speed(iter/s)": 0.030006
+    },
+    {
+      "epoch": 0.12159513619455221,
+      "grad_norm": 0.05966533070217228,
+      "learning_rate": 1.958017087284061e-05,
+      "loss": 0.4596414089202881,
+      "memory(GiB)": 73.24,
+      "step": 380,
+      "token_acc": 0.8836156297165856,
+      "train_speed(iter/s)": 0.029968
+    },
+    {
+      "epoch": 0.12319507219711212,
+      "grad_norm": 0.06510894340277039,
+      "learning_rate": 1.9565153687607006e-05,
+      "loss": 0.4687026023864746,
+      "memory(GiB)": 73.24,
+      "step": 385,
+      "token_acc": 0.8694005270092227,
+      "train_speed(iter/s)": 0.029956
+    },
+    {
+      "epoch": 0.12479500819967201,
+      "grad_norm": 0.05180588304383506,
+      "learning_rate": 1.9549878585371006e-05,
+      "loss": 0.4649878978729248,
+      "memory(GiB)": 73.24,
+      "step": 390,
+      "token_acc": 0.8677233429394813,
+      "train_speed(iter/s)": 0.029999
+    },
+    {
+      "epoch": 0.1263949442022319,
+      "grad_norm": 0.060875424512666344,
+      "learning_rate": 1.9534345978013972e-05,
+      "loss": 0.47073874473571775,
+      "memory(GiB)": 73.24,
+      "step": 395,
+      "token_acc": 0.8484663512894858,
+      "train_speed(iter/s)": 0.029947
+    },
+    {
+      "epoch": 0.1279948802047918,
+      "grad_norm": 0.0571374353277554,
+      "learning_rate": 1.9518556284360696e-05,
+      "loss": 0.4666412353515625,
+      "memory(GiB)": 73.24,
+      "step": 400,
+      "token_acc": 0.8975701436434421,
+      "train_speed(iter/s)": 0.029998
+    },
+    {
+      "epoch": 0.1295948162073517,
+      "grad_norm": 0.05215050598306155,
+      "learning_rate": 1.9502509930168113e-05,
+      "loss": 0.4628121376037598,
+      "memory(GiB)": 73.24,
+      "step": 405,
+      "token_acc": 0.8816677696889477,
+      "train_speed(iter/s)": 0.029966
+    },
+    {
+      "epoch": 0.1311947522099116,
+      "grad_norm": 0.07947570193916972,
+      "learning_rate": 1.9486207348113803e-05,
+      "loss": 0.4593012809753418,
+      "memory(GiB)": 73.24,
+      "step": 410,
+      "token_acc": 0.8692473832862602,
+      "train_speed(iter/s)": 0.029911
+    },
+    {
+      "epoch": 0.1327946882124715,
+      "grad_norm": 0.07262611466641217,
+      "learning_rate": 1.946964897778433e-05,
+      "loss": 0.47004990577697753,
+      "memory(GiB)": 73.24,
+      "step": 415,
+      "token_acc": 0.8736337958983176,
+      "train_speed(iter/s)": 0.029969
+    },
+    {
+      "epoch": 0.1343946242150314,
+      "grad_norm": 0.053754461298334506,
+      "learning_rate": 1.9452835265663404e-05,
+      "loss": 0.4695271015167236,
+      "memory(GiB)": 73.24,
+      "step": 420,
+      "token_acc": 0.8747993579454254,
+      "train_speed(iter/s)": 0.029901
+    },
+    {
+      "epoch": 0.1359945602175913,
+      "grad_norm": 0.0742051800083311,
+      "learning_rate": 1.9435766665119823e-05,
+      "loss": 0.47011446952819824,
+      "memory(GiB)": 73.24,
+      "step": 425,
+      "token_acc": 0.8356736242884251,
+      "train_speed(iter/s)": 0.029856
+    },
+    {
+      "epoch": 0.13759449622015119,
+      "grad_norm": 0.06429200177825628,
+      "learning_rate": 1.941844363639525e-05,
+      "loss": 0.476796817779541,
+      "memory(GiB)": 73.24,
+      "step": 430,
+      "token_acc": 0.869019972131909,
+      "train_speed(iter/s)": 0.0299
+    },
+    {
+      "epoch": 0.13919443222271108,
+      "grad_norm": 0.06544854557851852,
+      "learning_rate": 1.9400866646591816e-05,
+      "loss": 0.4666853904724121,
+      "memory(GiB)": 73.24,
+      "step": 435,
+      "token_acc": 0.8204667863554758,
+      "train_speed(iter/s)": 0.029847
+    },
+    {
+      "epoch": 0.140794368225271,
+      "grad_norm": 0.0546565929911768,
+      "learning_rate": 1.9383036169659513e-05,
+      "loss": 0.4738778591156006,
+      "memory(GiB)": 73.24,
+      "step": 440,
+      "token_acc": 0.8605809128630706,
+      "train_speed(iter/s)": 0.029855
+    },
+    {
+      "epoch": 0.1423943042278309,
+      "grad_norm": 0.06789336848906298,
+      "learning_rate": 1.936495268638342e-05,
+      "loss": 0.47726120948791506,
+      "memory(GiB)": 73.24,
+      "step": 445,
+      "token_acc": 0.8404369243949454,
+      "train_speed(iter/s)": 0.029875
+    },
+    {
+      "epoch": 0.14399424023039079,
+      "grad_norm": 0.049909982274150465,
+      "learning_rate": 1.934661668437073e-05,
+      "loss": 0.47165632247924805,
+      "memory(GiB)": 73.24,
+      "step": 450,
+      "token_acc": 0.848471615720524,
+      "train_speed(iter/s)": 0.029826
+    },
+    {
+      "epoch": 0.14559417623295068,
+      "grad_norm": 0.057441474731933166,
+      "learning_rate": 1.932802865803763e-05,
+      "loss": 0.4703391075134277,
+      "memory(GiB)": 73.24,
+      "step": 455,
+      "token_acc": 0.8466442358774571,
+      "train_speed(iter/s)": 0.029877
+    },
+    {
+      "epoch": 0.14719411223551057,
+      "grad_norm": 0.07263904251491092,
+      "learning_rate": 1.930918910859592e-05,
+      "loss": 0.467697811126709,
+      "memory(GiB)": 73.24,
+      "step": 460,
+      "token_acc": 0.8491142333536957,
+      "train_speed(iter/s)": 0.02984
+    },
+    {
+      "epoch": 0.14879404823807046,
+      "grad_norm": 0.06769237623086669,
+      "learning_rate": 1.9290098544039546e-05,
+      "loss": 0.46541628837585447,
+      "memory(GiB)": 73.24,
+      "step": 465,
+      "token_acc": 0.8555353301340394,
+      "train_speed(iter/s)": 0.02978
+    },
+    {
+      "epoch": 0.15039398424063039,
+      "grad_norm": 0.06751583633556477,
+      "learning_rate": 1.927075747913088e-05,
+      "loss": 0.47134056091308596,
+      "memory(GiB)": 73.24,
+      "step": 470,
+      "token_acc": 0.8708000507163687,
+      "train_speed(iter/s)": 0.029832
+    },
+    {
+      "epoch": 0.15199392024319028,
+      "grad_norm": 0.0539492567012165,
+      "learning_rate": 1.9251166435386837e-05,
+      "loss": 0.4645866394042969,
+      "memory(GiB)": 73.24,
+      "step": 475,
+      "token_acc": 0.849832526981764,
+      "train_speed(iter/s)": 0.029779
+    },
+    {
+      "epoch": 0.15359385624575017,
+      "grad_norm": 0.06038706866556876,
+      "learning_rate": 1.923132594106483e-05,
+      "loss": 0.46890692710876464,
+      "memory(GiB)": 73.24,
+      "step": 480,
+      "token_acc": 0.8665925514174542,
+      "train_speed(iter/s)": 0.02976
+    },
+    {
+      "epoch": 0.15519379224831006,
+      "grad_norm": 0.05215840717634863,
+      "learning_rate": 1.92112365311485e-05,
+      "loss": 0.46829919815063475,
+      "memory(GiB)": 73.24,
+      "step": 485,
+      "token_acc": 0.861963565228023,
+      "train_speed(iter/s)": 0.029794
+    },
+    {
+      "epoch": 0.15679372825086996,
+      "grad_norm": 0.06554142579397569,
+      "learning_rate": 1.919089874733332e-05,
+      "loss": 0.4702622413635254,
+      "memory(GiB)": 73.24,
+      "step": 490,
+      "token_acc": 0.8809186723297153,
+      "train_speed(iter/s)": 0.029747
+    },
+    {
+      "epoch": 0.15839366425342985,
+      "grad_norm": 0.0601172563145885,
+      "learning_rate": 1.9170313138011964e-05,
+      "loss": 0.46490135192871096,
+      "memory(GiB)": 73.24,
+      "step": 495,
+      "token_acc": 0.8890911637025627,
+      "train_speed(iter/s)": 0.02977
+    },
+    {
+      "epoch": 0.15999360025598977,
+      "grad_norm": 0.05924399402367875,
+      "learning_rate": 1.9149480258259535e-05,
+      "loss": 0.46698894500732424,
+      "memory(GiB)": 73.24,
+      "step": 500,
+      "token_acc": 0.8781434114096853,
+      "train_speed(iter/s)": 0.029766
+    },
+    {
+      "epoch": 0.15999360025598977,
+      "eval_loss": 0.677643895149231,
+      "eval_runtime": 109.3458,
+      "eval_samples_per_second": 183.711,
+      "eval_steps_per_second": 0.924,
+      "eval_token_acc": 0.8661408286670019,
+      "step": 500
+    },
+    {
+      "epoch": 0.16159353625854966,
+      "grad_norm": 0.05118070522939682,
+      "learning_rate": 1.9128400669818586e-05,
+      "loss": 0.4606743812561035,
+      "memory(GiB)": 73.24,
+      "step": 505,
+      "token_acc": 0.8727327237295758,
+      "train_speed(iter/s)": 0.029628
+    },
+    {
+      "epoch": 0.16319347226110956,
+      "grad_norm": 0.05904937387674259,
+      "learning_rate": 1.9107074941083987e-05,
+      "loss": 0.47115492820739746,
+      "memory(GiB)": 73.24,
+      "step": 510,
+      "token_acc": 0.8801781737193763,
+      "train_speed(iter/s)": 0.029663
+    },
+    {
+      "epoch": 0.16479340826366945,
+      "grad_norm": 0.061211680590962145,
+      "learning_rate": 1.9085503647087588e-05,
+      "loss": 0.46154184341430665,
+      "memory(GiB)": 73.24,
+      "step": 515,
+      "token_acc": 0.8573438874230431,
+      "train_speed(iter/s)": 0.029714
+    },
+    {
+      "epoch": 0.16639334426622934,
+      "grad_norm": 0.05461804298242196,
+      "learning_rate": 1.906368736948272e-05,
+      "loss": 0.46891465187072756,
+      "memory(GiB)": 73.24,
+      "step": 520,
+      "token_acc": 0.8665508756694167,
+      "train_speed(iter/s)": 0.029721
+    },
+    {
+      "epoch": 0.16799328026878924,
+      "grad_norm": 0.059072521440841075,
+      "learning_rate": 1.9041626696528503e-05,
+      "loss": 0.4666083812713623,
+      "memory(GiB)": 73.24,
+      "step": 525,
+      "token_acc": 0.8742783835792175,
+      "train_speed(iter/s)": 0.029735
+    },
+    {
+      "epoch": 0.16959321627134916,
+      "grad_norm": 0.06762878495647719,
+      "learning_rate": 1.9019322223073997e-05,
+      "loss": 0.4684437274932861,
+      "memory(GiB)": 73.24,
+      "step": 530,
+      "token_acc": 0.8906074591493077,
+      "train_speed(iter/s)": 0.029782
+    },
+    {
+      "epoch": 0.17119315227390905,
+      "grad_norm": 0.05741557316745661,
+      "learning_rate": 1.899677455054215e-05,
+      "loss": 0.4690097332000732,
+      "memory(GiB)": 73.24,
+      "step": 535,
+      "token_acc": 0.8231878958479943,
+      "train_speed(iter/s)": 0.029785
+    },
+    {
+      "epoch": 0.17279308827646894,
+      "grad_norm": 0.049026865135578496,
+      "learning_rate": 1.8973984286913584e-05,
+      "loss": 0.469140625,
+      "memory(GiB)": 73.24,
+      "step": 540,
+      "token_acc": 0.8849415539766216,
+      "train_speed(iter/s)": 0.029789
+    },
+    {
+      "epoch": 0.17439302427902884,
+      "grad_norm": 0.059746465018255104,
+      "learning_rate": 1.895095204671021e-05,
+      "loss": 0.4646149158477783,
+      "memory(GiB)": 73.24,
+      "step": 545,
+      "token_acc": 0.8944385405596883,
+      "train_speed(iter/s)": 0.029813
+    },
+    {
+      "epoch": 0.17599296028158873,
+      "grad_norm": 0.049833714934798115,
+      "learning_rate": 1.892767845097864e-05,
+      "loss": 0.47077240943908694,
+      "memory(GiB)": 73.24,
+      "step": 550,
+      "token_acc": 0.8640860961638605,
+      "train_speed(iter/s)": 0.029794
+    },
+    {
+      "epoch": 0.17759289628414862,
+      "grad_norm": 0.06593845007149325,
+      "learning_rate": 1.890416412727346e-05,
+      "loss": 0.46265759468078616,
+      "memory(GiB)": 73.24,
+      "step": 555,
+      "token_acc": 0.8249895412076419,
+      "train_speed(iter/s)": 0.02984
+    },
+    {
+      "epoch": 0.17919283228670854,
+      "grad_norm": 0.058254003445636866,
+      "learning_rate": 1.88804097096403e-05,
+      "loss": 0.459829044342041,
+      "memory(GiB)": 73.24,
+      "step": 560,
+      "token_acc": 0.8835873095178616,
+      "train_speed(iter/s)": 0.029842
+    },
+    {
+      "epoch": 0.18079276828926844,
+      "grad_norm": 0.07335953644753283,
+      "learning_rate": 1.8856415838598738e-05,
+      "loss": 0.45765042304992676,
+      "memory(GiB)": 73.24,
+      "step": 565,
+      "token_acc": 0.8755007210382951,
+      "train_speed(iter/s)": 0.029818
+    },
+    {
+      "epoch": 0.18239270429182833,
+      "grad_norm": 0.06659181547700674,
+      "learning_rate": 1.8832183161125026e-05,
+      "loss": 0.4609128475189209,
+      "memory(GiB)": 73.24,
+      "step": 570,
+      "token_acc": 0.8344311377245509,
+      "train_speed(iter/s)": 0.029871
+    },
+    {
+      "epoch": 0.18399264029438822,
+      "grad_norm": 0.05836437871791382,
+      "learning_rate": 1.8807712330634645e-05,
+      "loss": 0.4691438674926758,
+      "memory(GiB)": 73.24,
+      "step": 575,
+      "token_acc": 0.8848027659908848,
+      "train_speed(iter/s)": 0.029828
+    },
+    {
+      "epoch": 0.18559257629694811,
+      "grad_norm": 0.05735059462858394,
+      "learning_rate": 1.87830040069647e-05,
+      "loss": 0.4602513790130615,
+      "memory(GiB)": 73.24,
+      "step": 580,
+      "token_acc": 0.8959147903465012,
+      "train_speed(iter/s)": 0.029816
+    },
+    {
+      "epoch": 0.187192512299508,
+      "grad_norm": 0.05337219773586585,
+      "learning_rate": 1.87580588563561e-05,
+      "loss": 0.46318631172180175,
+      "memory(GiB)": 73.24,
+      "step": 585,
+      "token_acc": 0.8725881039706586,
+      "train_speed(iter/s)": 0.029851
+    },
+    {
+      "epoch": 0.18879244830206793,
+      "grad_norm": 0.05886716832883729,
+      "learning_rate": 1.873287755143563e-05,
+      "loss": 0.4604507923126221,
+      "memory(GiB)": 73.24,
+      "step": 590,
+      "token_acc": 0.9041755130927105,
+      "train_speed(iter/s)": 0.029822
+    },
+    {
+      "epoch": 0.19039238430462782,
+      "grad_norm": 0.053483810048332456,
+      "learning_rate": 1.8707460771197773e-05,
+      "loss": 0.46618080139160156,
+      "memory(GiB)": 73.24,
+      "step": 595,
+      "token_acc": 0.8785046728971962,
+      "train_speed(iter/s)": 0.029819
+    },
+    {
+      "epoch": 0.1919923203071877,
+      "grad_norm": 0.0518592001281956,
+      "learning_rate": 1.868180920098644e-05,
+      "loss": 0.4680916786193848,
+      "memory(GiB)": 73.24,
+      "step": 600,
+      "token_acc": 0.8467063770147162,
+      "train_speed(iter/s)": 0.029843
+    },
+    {
+      "epoch": 0.1935922563097476,
+      "grad_norm": 0.07018232236413237,
+      "learning_rate": 1.8655923532476463e-05,
+      "loss": 0.46170759201049805,
+      "memory(GiB)": 73.24,
+      "step": 605,
+      "token_acc": 0.889030612244898,
+      "train_speed(iter/s)": 0.02981
+    },
+    {
+      "epoch": 0.1951921923123075,
+      "grad_norm": 0.06030421269833889,
+      "learning_rate": 1.8629804463654956e-05,
+      "loss": 0.46511187553405764,
+      "memory(GiB)": 73.24,
+      "step": 610,
+      "token_acc": 0.8554680664916885,
+      "train_speed(iter/s)": 0.029852
+    },
+    {
+      "epoch": 0.1967921283148674,
+      "grad_norm": 0.056137765321266526,
+      "learning_rate": 1.8603452698802498e-05,
+      "loss": 0.47327299118041993,
+      "memory(GiB)": 76.61,
+      "step": 615,
+      "token_acc": 0.8645191852202747,
+      "train_speed(iter/s)": 0.029831
+    },
+    {
+      "epoch": 0.1983920643174273,
+      "grad_norm": 0.05458475201274465,
+      "learning_rate": 1.857686894847413e-05,
+      "loss": 0.45963249206542967,
+      "memory(GiB)": 76.61,
+      "step": 620,
+      "token_acc": 0.8517509197438343,
+      "train_speed(iter/s)": 0.029791
+    },
+    {
+      "epoch": 0.1999920003199872,
+      "grad_norm": 0.059902578480064236,
+      "learning_rate": 1.8550053929480202e-05,
+      "loss": 0.4687147617340088,
+      "memory(GiB)": 76.61,
+      "step": 625,
+      "token_acc": 0.8958185683912119,
+      "train_speed(iter/s)": 0.029833
+    },
+    {
+      "epoch": 0.2015919363225471,
+      "grad_norm": 0.0539478773118384,
+      "learning_rate": 1.8523008364867056e-05,
+      "loss": 0.4696544647216797,
+      "memory(GiB)": 76.61,
+      "step": 630,
+      "token_acc": 0.8439355385920272,
+      "train_speed(iter/s)": 0.029796
+    },
+    {
+      "epoch": 0.203191872325107,
+      "grad_norm": 0.05688926646164217,
+      "learning_rate": 1.8495732983897504e-05,
+      "loss": 0.4628334045410156,
+      "memory(GiB)": 76.61,
+      "step": 635,
+      "token_acc": 0.8406656465187452,
+      "train_speed(iter/s)": 0.029775
+    },
+    {
+      "epoch": 0.20479180832766689,
+      "grad_norm": 0.055104479428209605,
+      "learning_rate": 1.8468228522031197e-05,
+      "loss": 0.4559271812438965,
+      "memory(GiB)": 76.61,
+      "step": 640,
+      "token_acc": 0.8823529411764706,
+      "train_speed(iter/s)": 0.029794
+    },
+    {
+      "epoch": 0.20639174433022678,
+      "grad_norm": 0.058080447436547736,
+      "learning_rate": 1.8440495720904758e-05,
+      "loss": 0.4649765968322754,
+      "memory(GiB)": 76.61,
+      "step": 645,
+      "token_acc": 0.8708735027753433,
+      "train_speed(iter/s)": 0.029752
+    },
+    {
+      "epoch": 0.2079916803327867,
+      "grad_norm": 0.06300003986546152,
+      "learning_rate": 1.8412535328311813e-05,
+      "loss": 0.47095327377319335,
+      "memory(GiB)": 76.61,
+      "step": 650,
+      "token_acc": 0.8504976200778883,
+      "train_speed(iter/s)": 0.029755
+    },
+    {
+      "epoch": 0.2095916163353466,
+      "grad_norm": 0.06584526718748161,
+      "learning_rate": 1.8384348098182815e-05,
+      "loss": 0.46697392463684084,
+      "memory(GiB)": 76.61,
+      "step": 655,
+      "token_acc": 0.8224407171775593,
+      "train_speed(iter/s)": 0.029762
+    },
+    {
+      "epoch": 0.21119155233790649,
+      "grad_norm": 0.07147957728971413,
+      "learning_rate": 1.8355934790564718e-05,
+      "loss": 0.4684570789337158,
+      "memory(GiB)": 76.61,
+      "step": 660,
+      "token_acc": 0.8842165898617511,
+      "train_speed(iter/s)": 0.029723
+    },
+    {
+      "epoch": 0.21279148834046638,
+      "grad_norm": 0.06592046292925295,
+      "learning_rate": 1.832729617160047e-05,
+      "loss": 0.461454439163208,
+      "memory(GiB)": 76.61,
+      "step": 665,
+      "token_acc": 0.9114801444043321,
+      "train_speed(iter/s)": 0.02976
+    },
+    {
+      "epoch": 0.21439142434302627,
+      "grad_norm": 0.0656829490109071,
+      "learning_rate": 1.8298433013508384e-05,
+      "loss": 0.46404447555541994,
+      "memory(GiB)": 76.61,
+      "step": 670,
+      "token_acc": 0.8516549891278087,
+      "train_speed(iter/s)": 0.029736
+    },
+    {
+      "epoch": 0.21599136034558616,
+      "grad_norm": 0.05417998837874903,
+      "learning_rate": 1.826934609456129e-05,
+      "loss": 0.47208566665649415,
+      "memory(GiB)": 76.61,
+      "step": 675,
+      "token_acc": 0.8798815733822078,
+      "train_speed(iter/s)": 0.029718
+    },
+    {
+      "epoch": 0.21759129634814608,
+      "grad_norm": 0.06917195844649823,
+      "learning_rate": 1.8240036199065546e-05,
+      "loss": 0.4724391460418701,
+      "memory(GiB)": 76.61,
+      "step": 680,
+      "token_acc": 0.875845675626257,
+      "train_speed(iter/s)": 0.029745
+    },
+    {
+      "epoch": 0.21919123235070598,
+      "grad_norm": 0.055849189404917746,
+      "learning_rate": 1.8210504117339917e-05,
+      "loss": 0.463816499710083,
+      "memory(GiB)": 76.61,
+      "step": 685,
+      "token_acc": 0.8841904379268782,
+      "train_speed(iter/s)": 0.029711
+    },
+    {
+      "epoch": 0.22079116835326587,
+      "grad_norm": 0.059563786969142496,
+      "learning_rate": 1.8180750645694236e-05,
+      "loss": 0.4678086757659912,
+      "memory(GiB)": 76.61,
+      "step": 690,
+      "token_acc": 0.8675231977159172,
+      "train_speed(iter/s)": 0.029714
+    },
+    {
+      "epoch": 0.22239110435582576,
+      "grad_norm": 0.05908606421708839,
+      "learning_rate": 1.8150776586407957e-05,
+      "loss": 0.46315860748291016,
+      "memory(GiB)": 76.61,
+      "step": 695,
+      "token_acc": 0.8914956011730205,
+      "train_speed(iter/s)": 0.029731
+    },
+    {
+      "epoch": 0.22399104035838566,
+      "grad_norm": 0.05617530731492468,
+      "learning_rate": 1.8120582747708503e-05,
+      "loss": 0.46682062149047854,
+      "memory(GiB)": 76.61,
+      "step": 700,
+      "token_acc": 0.8805088596092685,
+      "train_speed(iter/s)": 0.029689
+    },
+    {
+      "epoch": 0.22559097636094555,
+      "grad_norm": 0.06138477303861948,
+      "learning_rate": 1.8090169943749477e-05,
+      "loss": 0.47155141830444336,
+      "memory(GiB)": 76.61,
+      "step": 705,
+      "token_acc": 0.8753766681015928,
+      "train_speed(iter/s)": 0.029703
+    },
+    {
+      "epoch": 0.22719091236350547,
+      "grad_norm": 0.07073141016351848,
+      "learning_rate": 1.8059538994588715e-05,
+      "loss": 0.45953845977783203,
+      "memory(GiB)": 76.61,
+      "step": 710,
+      "token_acc": 0.8449233877757198,
+      "train_speed(iter/s)": 0.0297
+    },
+    {
+      "epoch": 0.22879084836606536,
+      "grad_norm": 0.06266619359839708,
+      "learning_rate": 1.8028690726166172e-05,
+      "loss": 0.4604049205780029,
+      "memory(GiB)": 76.61,
+      "step": 715,
+      "token_acc": 0.8688032048072108,
+      "train_speed(iter/s)": 0.02966
+    },
+    {
+      "epoch": 0.23039078436862526,
+      "grad_norm": 0.0563660774004587,
+      "learning_rate": 1.7997625970281652e-05,
+      "loss": 0.4622708797454834,
+      "memory(GiB)": 76.61,
+      "step": 720,
+      "token_acc": 0.8698216735253772,
+      "train_speed(iter/s)": 0.029685
+    },
+    {
+      "epoch": 0.23199072037118515,
+      "grad_norm": 0.06596213612143108,
+      "learning_rate": 1.796634556457236e-05,
+      "loss": 0.4681892395019531,
+      "memory(GiB)": 76.61,
+      "step": 725,
+      "token_acc": 0.8842619184376795,
+      "train_speed(iter/s)": 0.029661
+    },
+    {
+      "epoch": 0.23359065637374504,
+      "grad_norm": 0.05364579438678848,
+      "learning_rate": 1.793485035249036e-05,
+      "loss": 0.46258745193481443,
+      "memory(GiB)": 76.61,
+      "step": 730,
+      "token_acc": 0.8599531615925059,
+      "train_speed(iter/s)": 0.02965
+    },
+    {
+      "epoch": 0.23519059237630494,
+      "grad_norm": 0.07509450433159735,
+      "learning_rate": 1.7903141183279776e-05,
+      "loss": 0.47242441177368166,
+      "memory(GiB)": 76.61,
+      "step": 735,
+      "token_acc": 0.8404958677685951,
+      "train_speed(iter/s)": 0.029665
+    },
+    {
+      "epoch": 0.23679052837886486,
+      "grad_norm": 0.06478313540282635,
+      "learning_rate": 1.7871218911953942e-05,
+      "loss": 0.4565444469451904,
+      "memory(GiB)": 76.61,
+      "step": 740,
+      "token_acc": 0.8338650865998177,
+      "train_speed(iter/s)": 0.029634
+    },
+    {
+      "epoch": 0.23839046438142475,
+      "grad_norm": 0.06348939893307848,
+      "learning_rate": 1.7839084399272317e-05,
+      "loss": 0.4670473575592041,
+      "memory(GiB)": 76.61,
+      "step": 745,
+      "token_acc": 0.8652410477034038,
+      "train_speed(iter/s)": 0.029638
+    },
+    {
+      "epoch": 0.23999040038398464,
+      "grad_norm": 0.07434587030241245,
+      "learning_rate": 1.780673851171728e-05,
+      "loss": 0.47047910690307615,
+      "memory(GiB)": 76.61,
+      "step": 750,
+      "token_acc": 0.88801504530689,
+      "train_speed(iter/s)": 0.029638
+    },
+    {
+      "epoch": 0.23999040038398464,
+      "eval_loss": 0.6746003031730652,
+      "eval_runtime": 113.2223,
+      "eval_samples_per_second": 177.421,
+      "eval_steps_per_second": 0.892,
+      "eval_token_acc": 0.8668385651547512,
+      "step": 750
+    },
+    {
+      "epoch": 0.24159033638654454,
+      "grad_norm": 0.06732795706859432,
+      "learning_rate": 1.777418212147079e-05,
+      "loss": 0.46190509796142576,
+      "memory(GiB)": 76.61,
+      "step": 755,
+      "token_acc": 0.8881346728210697,
+      "train_speed(iter/s)": 0.029543
+    },
+    {
+      "epoch": 0.24319027238910443,
+      "grad_norm": 0.06836940989947664,
+      "learning_rate": 1.7741416106390828e-05,
+      "loss": 0.46283302307128904,
+      "memory(GiB)": 76.61,
+      "step": 760,
+      "token_acc": 0.8831443688586545,
+      "train_speed(iter/s)": 0.029566
+    },
+    {
+      "epoch": 0.24479020839166432,
+      "grad_norm": 0.07072489516219096,
+      "learning_rate": 1.7708441349987753e-05,
+      "loss": 0.4619740962982178,
+      "memory(GiB)": 76.61,
+      "step": 765,
+      "token_acc": 0.8610668789808917,
+      "train_speed(iter/s)": 0.0296
+    },
+    {
+      "epoch": 0.24639014439422424,
+      "grad_norm": 0.07152232857362027,
+      "learning_rate": 1.767525874140048e-05,
+      "loss": 0.46694121360778806,
+      "memory(GiB)": 76.61,
+      "step": 770,
+      "token_acc": 0.8397869022869023,
+      "train_speed(iter/s)": 0.029606
+    },
+    {
+      "epoch": 0.24799008039678413,
+      "grad_norm": 0.059354056163304685,
+      "learning_rate": 1.7641869175372493e-05,
+      "loss": 0.4596868991851807,
+      "memory(GiB)": 76.61,
+      "step": 775,
+      "token_acc": 0.8582827406764961,
+      "train_speed(iter/s)": 0.029599
+    },
+    {
+      "epoch": 0.24959001639934403,
+      "grad_norm": 0.0629690289705531,
+      "learning_rate": 1.7608273552227723e-05,
+      "loss": 0.4583168029785156,
+      "memory(GiB)": 76.61,
+      "step": 780,
+      "token_acc": 0.8841532106646639,
+      "train_speed(iter/s)": 0.029639
+    },
+    {
+      "epoch": 0.25118995240190395,
+      "grad_norm": 0.05810355160479093,
+      "learning_rate": 1.7574472777846276e-05,
+      "loss": 0.47337069511413576,
+      "memory(GiB)": 76.61,
+      "step": 785,
+      "token_acc": 0.8676557863501484,
+      "train_speed(iter/s)": 0.029632
+    },
+    {
+      "epoch": 0.2527898884044638,
+      "grad_norm": 0.05365185572887828,
+      "learning_rate": 1.7540467763639994e-05,
+      "loss": 0.46567063331604003,
+      "memory(GiB)": 76.61,
+      "step": 790,
+      "token_acc": 0.8745288099084545,
+      "train_speed(iter/s)": 0.029629
+    },
+    {
+      "epoch": 0.25438982440702373,
+      "grad_norm": 0.054672322658953366,
+      "learning_rate": 1.7506259426527903e-05,
+      "loss": 0.47023472785949705,
+      "memory(GiB)": 76.61,
+      "step": 795,
+      "token_acc": 0.874407844001322,
+      "train_speed(iter/s)": 0.02965
+    },
+    {
+      "epoch": 0.2559897604095836,
+      "grad_norm": 0.057060955079149434,
+      "learning_rate": 1.7471848688911465e-05,
+      "loss": 0.4684537410736084,
+      "memory(GiB)": 76.61,
+      "step": 800,
+      "token_acc": 0.8839382448537378,
+      "train_speed(iter/s)": 0.029634
+    },
+    {
+      "epoch": 0.2575896964121435,
+      "grad_norm": 0.06051290772323595,
+      "learning_rate": 1.7437236478649718e-05,
+      "loss": 0.46199979782104494,
+      "memory(GiB)": 76.61,
+      "step": 805,
+      "token_acc": 0.8673650919153983,
+      "train_speed(iter/s)": 0.02966
+    },
+    {
+      "epoch": 0.2591896324147034,
+      "grad_norm": 0.0643397562387603,
+      "learning_rate": 1.7402423729034252e-05,
+      "loss": 0.4548381805419922,
+      "memory(GiB)": 76.61,
+      "step": 810,
+      "token_acc": 0.83125,
+      "train_speed(iter/s)": 0.029652
+    },
+    {
+      "epoch": 0.2607895684172633,
+      "grad_norm": 0.065624934571794,
+      "learning_rate": 1.736741137876405e-05,
+      "loss": 0.46353764533996583,
+      "memory(GiB)": 76.61,
+      "step": 815,
+      "token_acc": 0.8907202528787537,
+      "train_speed(iter/s)": 0.029628
+    },
+    {
+      "epoch": 0.2623895044198232,
+      "grad_norm": 0.053961693017135055,
+      "learning_rate": 1.7332200371920173e-05,
+      "loss": 0.46685361862182617,
+      "memory(GiB)": 76.61,
+      "step": 820,
+      "token_acc": 0.8522188711762172,
+      "train_speed(iter/s)": 0.029672
+    },
+    {
+      "epoch": 0.2639894404223831,
+      "grad_norm": 0.054388550053431586,
+      "learning_rate": 1.72967916579403e-05,
+      "loss": 0.46024084091186523,
+      "memory(GiB)": 76.61,
+      "step": 825,
+      "token_acc": 0.8684630384683567,
+      "train_speed(iter/s)": 0.02966
+    },
+    {
+      "epoch": 0.265589376424943,
+      "grad_norm": 0.0583019332597641,
+      "learning_rate": 1.7261186191593135e-05,
+      "loss": 0.47214059829711913,
+      "memory(GiB)": 76.61,
+      "step": 830,
+      "token_acc": 0.8717123935666982,
+      "train_speed(iter/s)": 0.029645
+    },
+    {
+      "epoch": 0.2671893124275029,
+      "grad_norm": 0.06004272220759217,
+      "learning_rate": 1.7225384932952655e-05,
+      "loss": 0.4626835823059082,
+      "memory(GiB)": 76.61,
+      "step": 835,
+      "token_acc": 0.8737211788059246,
+      "train_speed(iter/s)": 0.02967
+    },
+    {
+      "epoch": 0.2687892484300628,
+      "grad_norm": 0.05611993161069816,
+      "learning_rate": 1.7189388847372227e-05,
+      "loss": 0.46799750328063966,
+      "memory(GiB)": 76.61,
+      "step": 840,
+      "token_acc": 0.8781684382665577,
+      "train_speed(iter/s)": 0.029642
+    },
+    {
+      "epoch": 0.2703891844326227,
+      "grad_norm": 0.06345947319153013,
+      "learning_rate": 1.715319890545857e-05,
+      "loss": 0.4568619728088379,
+      "memory(GiB)": 76.61,
+      "step": 845,
+      "token_acc": 0.860916969527537,
+      "train_speed(iter/s)": 0.029655
+    },
+    {
+      "epoch": 0.2719891204351826,
+      "grad_norm": 0.0592531603954309,
+      "learning_rate": 1.7116816083045603e-05,
+      "loss": 0.46942729949951173,
+      "memory(GiB)": 76.61,
+      "step": 850,
+      "token_acc": 0.8726317245194303,
+      "train_speed(iter/s)": 0.029655
+    },
+    {
+      "epoch": 0.2735890564377425,
+      "grad_norm": 0.05711267065318382,
+      "learning_rate": 1.7080241361168108e-05,
+      "loss": 0.45801239013671874,
+      "memory(GiB)": 76.61,
+      "step": 855,
+      "token_acc": 0.8834167608590344,
+      "train_speed(iter/s)": 0.02963
+    },
+    {
+      "epoch": 0.27518899244030237,
+      "grad_norm": 0.05715792257951623,
+      "learning_rate": 1.704347572603529e-05,
+      "loss": 0.4675910472869873,
+      "memory(GiB)": 76.61,
+      "step": 860,
+      "token_acc": 0.8361073624231519,
+      "train_speed(iter/s)": 0.029659
+    },
+    {
+      "epoch": 0.2767889284428623,
+      "grad_norm": 0.056617536923221766,
+      "learning_rate": 1.700652016900419e-05,
+      "loss": 0.467483377456665,
+      "memory(GiB)": 76.61,
+      "step": 865,
+      "token_acc": 0.8753590807532716,
+      "train_speed(iter/s)": 0.029639
+    },
+    {
+      "epoch": 0.27838886444542216,
+      "grad_norm": 0.060433939578350394,
+      "learning_rate": 1.696937568655294e-05,
+      "loss": 0.46129570007324217,
+      "memory(GiB)": 76.61,
+      "step": 870,
+      "token_acc": 0.8700755748512623,
+      "train_speed(iter/s)": 0.029622
+    },
+    {
+      "epoch": 0.2799888004479821,
+      "grad_norm": 0.06826391103956585,
+      "learning_rate": 1.6932043280253892e-05,
+      "loss": 0.47449960708618166,
+      "memory(GiB)": 76.61,
+      "step": 875,
+      "token_acc": 0.8767408356010885,
+      "train_speed(iter/s)": 0.02965
+    },
+    {
+      "epoch": 0.281588736450542,
+      "grad_norm": 0.060978189753072065,
+      "learning_rate": 1.689452395674664e-05,
+      "loss": 0.464243745803833,
+      "memory(GiB)": 76.61,
+      "step": 880,
+      "token_acc": 0.8622170179547228,
+      "train_speed(iter/s)": 0.029624
+    },
+    {
+      "epoch": 0.28318867245310186,
+      "grad_norm": 0.0760276206328267,
+      "learning_rate": 1.6856818727710847e-05,
+      "loss": 0.4566212177276611,
+      "memory(GiB)": 76.61,
+      "step": 885,
+      "token_acc": 0.8465499485066942,
+      "train_speed(iter/s)": 0.029618
+    },
+    {
+      "epoch": 0.2847886084556618,
+      "grad_norm": 0.05693121191664627,
+      "learning_rate": 1.6818928609838967e-05,
+      "loss": 0.46042599678039553,
+      "memory(GiB)": 76.61,
+      "step": 890,
+      "token_acc": 0.8798391728891441,
+      "train_speed(iter/s)": 0.029627
+    },
+    {
+      "epoch": 0.28638854445822165,
+      "grad_norm": 0.05744826995499506,
+      "learning_rate": 1.678085462480885e-05,
+      "loss": 0.4604465961456299,
+      "memory(GiB)": 76.61,
+      "step": 895,
+      "token_acc": 0.8780676542118063,
+      "train_speed(iter/s)": 0.029599
+    },
+    {
+      "epoch": 0.28798848046078157,
+      "grad_norm": 0.06271464886952488,
+      "learning_rate": 1.6742597799256182e-05,
+      "loss": 0.46231966018676757,
+      "memory(GiB)": 76.61,
+      "step": 900,
+      "token_acc": 0.8866765515780555,
+      "train_speed(iter/s)": 0.029611
+    },
+    {
+      "epoch": 0.2895884164633415,
+      "grad_norm": 0.06044356676681803,
+      "learning_rate": 1.6704159164746797e-05,
+      "loss": 0.47655544281005857,
+      "memory(GiB)": 76.61,
+      "step": 905,
+      "token_acc": 0.8872944211544663,
+      "train_speed(iter/s)": 0.029601
+    },
+    {
+      "epoch": 0.29118835246590136,
+      "grad_norm": 0.05103569816400521,
+      "learning_rate": 1.6665539757748866e-05,
+      "loss": 0.4603917121887207,
+      "memory(GiB)": 76.61,
+      "step": 910,
+      "token_acc": 0.8611705475141599,
+      "train_speed(iter/s)": 0.029574
+    },
+    {
+      "epoch": 0.2927882884684613,
+      "grad_norm": 0.055811472748585486,
+      "learning_rate": 1.6626740619604967e-05,
+      "loss": 0.46213679313659667,
+      "memory(GiB)": 76.61,
+      "step": 915,
+      "token_acc": 0.8148507643775783,
+      "train_speed(iter/s)": 0.029594
+    },
+    {
+      "epoch": 0.29438822447102114,
+      "grad_norm": 0.05463929857953068,
+      "learning_rate": 1.658776279650397e-05,
+      "loss": 0.4658839702606201,
+      "memory(GiB)": 76.61,
+      "step": 920,
+      "token_acc": 0.8766637089618456,
+      "train_speed(iter/s)": 0.029577
+    },
+    {
+      "epoch": 0.29598816047358106,
+      "grad_norm": 0.06343067949686905,
+      "learning_rate": 1.6548607339452853e-05,
+      "loss": 0.46423888206481934,
+      "memory(GiB)": 76.61,
+      "step": 925,
+      "token_acc": 0.8785782119115453,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.29758809647614093,
+      "grad_norm": 0.052431934937864355,
+      "learning_rate": 1.6509275304248366e-05,
+      "loss": 0.46324734687805175,
+      "memory(GiB)": 76.61,
+      "step": 930,
+      "token_acc": 0.8571011956838729,
+      "train_speed(iter/s)": 0.02958
+    },
+    {
+      "epoch": 0.29918803247870085,
+      "grad_norm": 0.059009943510604755,
+      "learning_rate": 1.6469767751448538e-05,
+      "loss": 0.46290836334228513,
+      "memory(GiB)": 76.61,
+      "step": 935,
+      "token_acc": 0.8388616290480864,
+      "train_speed(iter/s)": 0.029556
+    },
+    {
+      "epoch": 0.30078796848126077,
+      "grad_norm": 0.05160057372757322,
+      "learning_rate": 1.6430085746344107e-05,
+      "loss": 0.45898871421813964,
+      "memory(GiB)": 76.61,
+      "step": 940,
+      "token_acc": 0.8690580344123651,
+      "train_speed(iter/s)": 0.029556
+    },
+    {
+      "epoch": 0.30238790448382064,
+      "grad_norm": 0.05612231994140208,
+      "learning_rate": 1.639023035892978e-05,
+      "loss": 0.4546724796295166,
+      "memory(GiB)": 76.61,
+      "step": 945,
+      "token_acc": 0.876509544215037,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.30398784048638056,
+      "grad_norm": 0.06733149115024578,
+      "learning_rate": 1.6350202663875385e-05,
+      "loss": 0.4598522663116455,
+      "memory(GiB)": 76.61,
+      "step": 950,
+      "token_acc": 0.8623452294246177,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.3055877764889404,
+      "grad_norm": 0.05450569676621943,
+      "learning_rate": 1.6310003740496887e-05,
+      "loss": 0.4602477550506592,
+      "memory(GiB)": 76.61,
+      "step": 955,
+      "token_acc": 0.8647700701480904,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.30718771249150034,
+      "grad_norm": 0.06736921151917717,
+      "learning_rate": 1.6269634672727296e-05,
+      "loss": 0.4589672565460205,
+      "memory(GiB)": 76.61,
+      "step": 960,
+      "token_acc": 0.877502001601281,
+      "train_speed(iter/s)": 0.029536
+    },
+    {
+      "epoch": 0.30878764849406026,
+      "grad_norm": 0.06166660436042404,
+      "learning_rate": 1.6229096549087434e-05,
+      "loss": 0.4601268291473389,
+      "memory(GiB)": 76.61,
+      "step": 965,
+      "token_acc": 0.8723534201954397,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.31038758449662013,
+      "grad_norm": 0.055128746386822226,
+      "learning_rate": 1.618839046265658e-05,
+      "loss": 0.4666788101196289,
+      "memory(GiB)": 76.61,
+      "step": 970,
+      "token_acc": 0.8550563360689943,
+      "train_speed(iter/s)": 0.029541
+    },
+    {
+      "epoch": 0.31198752049918005,
+      "grad_norm": 0.056867326711030626,
+      "learning_rate": 1.614751751104301e-05,
+      "loss": 0.4646125793457031,
+      "memory(GiB)": 76.61,
+      "step": 975,
+      "token_acc": 0.8651571964234208,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.3135874565017399,
+      "grad_norm": 0.05501107287069041,
+      "learning_rate": 1.6106478796354382e-05,
+      "loss": 0.4588280200958252,
+      "memory(GiB)": 76.61,
+      "step": 980,
+      "token_acc": 0.8767766331985918,
+      "train_speed(iter/s)": 0.029517
+    },
+    {
+      "epoch": 0.31518739250429983,
+      "grad_norm": 0.08099201898186387,
+      "learning_rate": 1.6065275425168034e-05,
+      "loss": 0.4589373111724854,
+      "memory(GiB)": 76.61,
+      "step": 985,
+      "token_acc": 0.8917890157694399,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.3167873285068597,
+      "grad_norm": 0.0522899382710734,
+      "learning_rate": 1.602390850850113e-05,
+      "loss": 0.46761279106140136,
+      "memory(GiB)": 76.61,
+      "step": 990,
+      "token_acc": 0.8461229409401366,
+      "train_speed(iter/s)": 0.029505
+    },
+    {
+      "epoch": 0.3183872645094196,
+      "grad_norm": 0.05838858698011934,
+      "learning_rate": 1.5982379161780722e-05,
+      "loss": 0.44941887855529783,
+      "memory(GiB)": 76.61,
+      "step": 995,
+      "token_acc": 0.8547228871294421,
+      "train_speed(iter/s)": 0.029511
+    },
+    {
+      "epoch": 0.31998720051197954,
+      "grad_norm": 0.054930484370324516,
+      "learning_rate": 1.5940688504813664e-05,
+      "loss": 0.4591392517089844,
+      "memory(GiB)": 76.61,
+      "step": 1000,
+      "token_acc": 0.8995555555555556,
+      "train_speed(iter/s)": 0.029505
+    },
+    {
+      "epoch": 0.31998720051197954,
+      "eval_loss": 0.671963095664978,
+      "eval_runtime": 110.8694,
+      "eval_samples_per_second": 181.186,
+      "eval_steps_per_second": 0.911,
+      "eval_token_acc": 0.8676077802864524,
+      "step": 1000
+    },
+    {
+      "epoch": 0.3215871365145394,
+      "grad_norm": 0.0578985798516978,
+      "learning_rate": 1.5898837661756405e-05,
+      "loss": 0.46222972869873047,
+      "memory(GiB)": 76.61,
+      "step": 1005,
+      "token_acc": 0.8840002569208042,
+      "train_speed(iter/s)": 0.029425
+    },
+    {
+      "epoch": 0.3231870725170993,
+      "grad_norm": 0.05872050053297838,
+      "learning_rate": 1.5856827761084698e-05,
+      "loss": 0.45543718338012695,
+      "memory(GiB)": 76.61,
+      "step": 1010,
+      "token_acc": 0.8753668220265838,
+      "train_speed(iter/s)": 0.02945
+    },
+    {
+      "epoch": 0.3247870085196592,
+      "grad_norm": 0.05268695066428434,
+      "learning_rate": 1.5814659935563165e-05,
+      "loss": 0.46614727973937986,
+      "memory(GiB)": 76.61,
+      "step": 1015,
+      "token_acc": 0.8792250035355678,
+      "train_speed(iter/s)": 0.029474
+    },
+    {
+      "epoch": 0.3263869445222191,
+      "grad_norm": 0.059454673806441594,
+      "learning_rate": 1.577233532221474e-05,
+      "loss": 0.45902605056762696,
+      "memory(GiB)": 76.61,
+      "step": 1020,
+      "token_acc": 0.86709886547812,
+      "train_speed(iter/s)": 0.029475
+    },
+    {
+      "epoch": 0.32798688052477903,
+      "grad_norm": 0.053728974295076275,
+      "learning_rate": 1.5729855062290024e-05,
+      "loss": 0.46491541862487795,
+      "memory(GiB)": 76.61,
+      "step": 1025,
+      "token_acc": 0.8708870261478794,
+      "train_speed(iter/s)": 0.029469
+    },
+    {
+      "epoch": 0.3295868165273389,
+      "grad_norm": 0.07030309576814114,
+      "learning_rate": 1.568722030123651e-05,
+      "loss": 0.453840970993042,
+      "memory(GiB)": 76.61,
+      "step": 1030,
+      "token_acc": 0.8568111455108359,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.3311867525298988,
+      "grad_norm": 0.07385415365022158,
+      "learning_rate": 1.5644432188667695e-05,
+      "loss": 0.45582828521728513,
+      "memory(GiB)": 76.61,
+      "step": 1035,
+      "token_acc": 0.8800162140251317,
+      "train_speed(iter/s)": 0.029488
+    },
+    {
+      "epoch": 0.3327866885324587,
+      "grad_norm": 0.05407863995123405,
+      "learning_rate": 1.5601491878332077e-05,
+      "loss": 0.4665637969970703,
+      "memory(GiB)": 76.61,
+      "step": 1040,
+      "token_acc": 0.8628481345244351,
+      "train_speed(iter/s)": 0.029487
+    },
+    {
+      "epoch": 0.3343866245350186,
+      "grad_norm": 0.05879461372080454,
+      "learning_rate": 1.5558400528082057e-05,
+      "loss": 0.4657593250274658,
+      "memory(GiB)": 76.61,
+      "step": 1045,
+      "token_acc": 0.879185119574845,
+      "train_speed(iter/s)": 0.02951
+    },
+    {
+      "epoch": 0.33598656053757847,
+      "grad_norm": 0.06618244368029796,
+      "learning_rate": 1.551515929984271e-05,
+      "loss": 0.45760574340820315,
+      "memory(GiB)": 76.61,
+      "step": 1050,
+      "token_acc": 0.8899380348185305,
+      "train_speed(iter/s)": 0.029502
+    },
+    {
+      "epoch": 0.3375864965401384,
+      "grad_norm": 0.06388796415692906,
+      "learning_rate": 1.547176935958044e-05,
+      "loss": 0.46065597534179686,
+      "memory(GiB)": 76.61,
+      "step": 1055,
+      "token_acc": 0.8536853685368537,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.3391864325426983,
+      "grad_norm": 0.05811152365312673,
+      "learning_rate": 1.5428231877271584e-05,
+      "loss": 0.46312780380249025,
+      "memory(GiB)": 76.61,
+      "step": 1060,
+      "token_acc": 0.8520375161707633,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.3407863685452582,
+      "grad_norm": 0.05545936328508829,
+      "learning_rate": 1.538454802687081e-05,
+      "loss": 0.4615220546722412,
+      "memory(GiB)": 76.61,
+      "step": 1065,
+      "token_acc": 0.8744265080713679,
+      "train_speed(iter/s)": 0.029504
+    },
+    {
+      "epoch": 0.3423863045478181,
+      "grad_norm": 0.05964362984731802,
+      "learning_rate": 1.5340718986279505e-05,
+      "loss": 0.46706466674804686,
+      "memory(GiB)": 76.61,
+      "step": 1070,
+      "token_acc": 0.8592233009708737,
+      "train_speed(iter/s)": 0.029536
+    },
+    {
+      "epoch": 0.34398624055037796,
+      "grad_norm": 0.05356886450328198,
+      "learning_rate": 1.529674593731399e-05,
+      "loss": 0.45301499366760256,
+      "memory(GiB)": 76.61,
+      "step": 1075,
+      "token_acc": 0.8575192096597146,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.3455861765529379,
+      "grad_norm": 0.05995962073425321,
+      "learning_rate": 1.5252630065673662e-05,
+      "loss": 0.46819314956665037,
+      "memory(GiB)": 76.61,
+      "step": 1080,
+      "token_acc": 0.8875031814711123,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.3471861125554978,
+      "grad_norm": 0.05389432634852101,
+      "learning_rate": 1.5208372560909031e-05,
+      "loss": 0.46298394203186033,
+      "memory(GiB)": 76.61,
+      "step": 1085,
+      "token_acc": 0.8872426699937617,
+      "train_speed(iter/s)": 0.029543
+    },
+    {
+      "epoch": 0.34878604855805767,
+      "grad_norm": 0.06642390255342462,
+      "learning_rate": 1.5163974616389621e-05,
+      "loss": 0.45978522300720215,
+      "memory(GiB)": 76.61,
+      "step": 1090,
+      "token_acc": 0.8246258860593332,
+      "train_speed(iter/s)": 0.029525
+    },
+    {
+      "epoch": 0.3503859845606176,
+      "grad_norm": 0.06115184110491886,
+      "learning_rate": 1.5119437429271813e-05,
+      "loss": 0.4637304782867432,
+      "memory(GiB)": 76.61,
+      "step": 1095,
+      "token_acc": 0.8666082895504962,
+      "train_speed(iter/s)": 0.029534
+    },
+    {
+      "epoch": 0.35198592056317746,
+      "grad_norm": 0.060865150660591956,
+      "learning_rate": 1.5074762200466557e-05,
+      "loss": 0.4542848587036133,
+      "memory(GiB)": 76.61,
+      "step": 1100,
+      "token_acc": 0.8913602663035255,
+      "train_speed(iter/s)": 0.029544
+    },
+    {
+      "epoch": 0.3535858565657374,
+      "grad_norm": 0.057666943430007674,
+      "learning_rate": 1.5029950134606991e-05,
+      "loss": 0.4574248790740967,
+      "memory(GiB)": 76.61,
+      "step": 1105,
+      "token_acc": 0.8634470336597996,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.35518579256829724,
+      "grad_norm": 0.054034554153381265,
+      "learning_rate": 1.4985002440015959e-05,
+      "loss": 0.4520272254943848,
+      "memory(GiB)": 76.61,
+      "step": 1110,
+      "token_acc": 0.8674898358680921,
+      "train_speed(iter/s)": 0.029551
+    },
+    {
+      "epoch": 0.35678572857085716,
+      "grad_norm": 0.06416854479766453,
+      "learning_rate": 1.4939920328673422e-05,
+      "loss": 0.4668846130371094,
+      "memory(GiB)": 76.61,
+      "step": 1115,
+      "token_acc": 0.9170854271356784,
+      "train_speed(iter/s)": 0.029541
+    },
+    {
+      "epoch": 0.3583856645734171,
+      "grad_norm": 0.05775941336987237,
+      "learning_rate": 1.4894705016183803e-05,
+      "loss": 0.4518620491027832,
+      "memory(GiB)": 76.61,
+      "step": 1120,
+      "token_acc": 0.8672782874617737,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.35998560057597695,
+      "grad_norm": 0.0625175589581686,
+      "learning_rate": 1.4849357721743169e-05,
+      "loss": 0.4566941738128662,
+      "memory(GiB)": 76.61,
+      "step": 1125,
+      "token_acc": 0.8505491793163026,
+      "train_speed(iter/s)": 0.029557
+    },
+    {
+      "epoch": 0.36158553657853687,
+      "grad_norm": 0.05911529293553411,
+      "learning_rate": 1.4803879668106393e-05,
+      "loss": 0.4640664577484131,
+      "memory(GiB)": 76.61,
+      "step": 1130,
+      "token_acc": 0.8772325625117503,
+      "train_speed(iter/s)": 0.029544
+    },
+    {
+      "epoch": 0.36318547258109674,
+      "grad_norm": 0.06483783687935218,
+      "learning_rate": 1.4758272081554168e-05,
+      "loss": 0.45419878959655763,
+      "memory(GiB)": 76.61,
+      "step": 1135,
+      "token_acc": 0.8594914930223667,
+      "train_speed(iter/s)": 0.029539
+    },
+    {
+      "epoch": 0.36478540858365666,
+      "grad_norm": 0.06032730304497941,
+      "learning_rate": 1.4712536191859934e-05,
+      "loss": 0.45779004096984866,
+      "memory(GiB)": 76.61,
+      "step": 1140,
+      "token_acc": 0.8938053097345132,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.3663853445862166,
+      "grad_norm": 0.0637380940226065,
+      "learning_rate": 1.4666673232256738e-05,
+      "loss": 0.46385722160339354,
+      "memory(GiB)": 76.61,
+      "step": 1145,
+      "token_acc": 0.8621830209481808,
+      "train_speed(iter/s)": 0.029544
+    },
+    {
+      "epoch": 0.36798528058877644,
+      "grad_norm": 0.057006770373085346,
+      "learning_rate": 1.4620684439403962e-05,
+      "loss": 0.4613553524017334,
+      "memory(GiB)": 76.61,
+      "step": 1150,
+      "token_acc": 0.8831837819873712,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.36958521659133636,
+      "grad_norm": 0.057569299635009126,
+      "learning_rate": 1.4574571053353987e-05,
+      "loss": 0.4598341464996338,
+      "memory(GiB)": 76.61,
+      "step": 1155,
+      "token_acc": 0.8825154371140721,
+      "train_speed(iter/s)": 0.029557
+    },
+    {
+      "epoch": 0.37118515259389623,
+      "grad_norm": 0.06747695219063263,
+      "learning_rate": 1.452833431751875e-05,
+      "loss": 0.4570640563964844,
+      "memory(GiB)": 76.61,
+      "step": 1160,
+      "token_acc": 0.8726823238566132,
+      "train_speed(iter/s)": 0.029543
+    },
+    {
+      "epoch": 0.37278508859645615,
+      "grad_norm": 0.05405367649749466,
+      "learning_rate": 1.448197547863622e-05,
+      "loss": 0.4516812801361084,
+      "memory(GiB)": 76.61,
+      "step": 1165,
+      "token_acc": 0.8704696273608984,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.374385024599016,
+      "grad_norm": 0.06041157710672601,
+      "learning_rate": 1.4435495786736796e-05,
+      "loss": 0.465837287902832,
+      "memory(GiB)": 76.61,
+      "step": 1170,
+      "token_acc": 0.8673412029229904,
+      "train_speed(iter/s)": 0.029554
+    },
+    {
+      "epoch": 0.37598496060157593,
+      "grad_norm": 0.05229585247228306,
+      "learning_rate": 1.438889649510956e-05,
+      "loss": 0.4427653789520264,
+      "memory(GiB)": 76.61,
+      "step": 1175,
+      "token_acc": 0.8558139534883721,
+      "train_speed(iter/s)": 0.02954
+    },
+    {
+      "epoch": 0.37758489660413586,
+      "grad_norm": 0.0547875272797444,
+      "learning_rate": 1.4342178860268523e-05,
+      "loss": 0.45673260688781736,
+      "memory(GiB)": 76.61,
+      "step": 1180,
+      "token_acc": 0.880563238622077,
+      "train_speed(iter/s)": 0.029563
+    },
+    {
+      "epoch": 0.3791848326066957,
+      "grad_norm": 0.0565328006493161,
+      "learning_rate": 1.4295344141918734e-05,
+      "loss": 0.46208748817443845,
+      "memory(GiB)": 76.61,
+      "step": 1185,
+      "token_acc": 0.8671328671328671,
+      "train_speed(iter/s)": 0.029544
+    },
+    {
+      "epoch": 0.38078476860925564,
+      "grad_norm": 0.062473905403265834,
+      "learning_rate": 1.4248393602922299e-05,
+      "loss": 0.46883163452148435,
+      "memory(GiB)": 76.61,
+      "step": 1190,
+      "token_acc": 0.8412252145605209,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.3823847046118155,
+      "grad_norm": 0.05646151042315891,
+      "learning_rate": 1.420132850926434e-05,
+      "loss": 0.45732822418212893,
+      "memory(GiB)": 76.61,
+      "step": 1195,
+      "token_acc": 0.8820655966503839,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.3839846406143754,
+      "grad_norm": 0.052981558367052706,
+      "learning_rate": 1.4154150130018867e-05,
+      "loss": 0.45579113960266116,
+      "memory(GiB)": 76.61,
+      "step": 1200,
+      "token_acc": 0.8677085226240233,
+      "train_speed(iter/s)": 0.029546
+    },
+    {
+      "epoch": 0.38558457661693535,
+      "grad_norm": 0.052315204322432474,
+      "learning_rate": 1.4106859737314532e-05,
+      "loss": 0.45348801612854006,
+      "memory(GiB)": 76.61,
+      "step": 1205,
+      "token_acc": 0.8616187989556136,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.3871845126194952,
+      "grad_norm": 0.05319888084520812,
+      "learning_rate": 1.4059458606300358e-05,
+      "loss": 0.45279593467712403,
+      "memory(GiB)": 76.61,
+      "step": 1210,
+      "token_acc": 0.86090645233311,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.38878444862205513,
+      "grad_norm": 0.054475973938428034,
+      "learning_rate": 1.4011948015111334e-05,
+      "loss": 0.4616706848144531,
+      "memory(GiB)": 76.61,
+      "step": 1215,
+      "token_acc": 0.8390133684805121,
+      "train_speed(iter/s)": 0.029549
+    },
+    {
+      "epoch": 0.390384384624615,
+      "grad_norm": 0.054891067059900926,
+      "learning_rate": 1.396432924483396e-05,
+      "loss": 0.4553243637084961,
+      "memory(GiB)": 76.61,
+      "step": 1220,
+      "token_acc": 0.8715350793347353,
+      "train_speed(iter/s)": 0.029571
+    },
+    {
+      "epoch": 0.3919843206271749,
+      "grad_norm": 0.06058246643434403,
+      "learning_rate": 1.3916603579471705e-05,
+      "loss": 0.47067904472351074,
+      "memory(GiB)": 76.61,
+      "step": 1225,
+      "token_acc": 0.8662144337667232,
+      "train_speed(iter/s)": 0.029556
+    },
+    {
+      "epoch": 0.3935842566297348,
+      "grad_norm": 0.05715510214651738,
+      "learning_rate": 1.3868772305910376e-05,
+      "loss": 0.46147928237915037,
+      "memory(GiB)": 76.61,
+      "step": 1230,
+      "token_acc": 0.868918375552875,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.3951841926322947,
+      "grad_norm": 0.06593047910666934,
+      "learning_rate": 1.3820836713883424e-05,
+      "loss": 0.45935769081115724,
+      "memory(GiB)": 76.61,
+      "step": 1235,
+      "token_acc": 0.8596291476903057,
+      "train_speed(iter/s)": 0.02957
+    },
+    {
+      "epoch": 0.3967841286348546,
+      "grad_norm": 0.056071042953882384,
+      "learning_rate": 1.3772798095937172e-05,
+      "loss": 0.4495890140533447,
+      "memory(GiB)": 76.61,
+      "step": 1240,
+      "token_acc": 0.8471917163476623,
+      "train_speed(iter/s)": 0.029553
+    },
+    {
+      "epoch": 0.3983840646374145,
+      "grad_norm": 0.05810589720196263,
+      "learning_rate": 1.3724657747395957e-05,
+      "loss": 0.4619898319244385,
+      "memory(GiB)": 76.61,
+      "step": 1245,
+      "token_acc": 0.8691186216037111,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.3999840006399744,
+      "grad_norm": 0.055604926632171425,
+      "learning_rate": 1.3676416966327201e-05,
+      "loss": 0.4587514400482178,
+      "memory(GiB)": 76.61,
+      "step": 1250,
+      "token_acc": 0.8369355461211887,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.3999840006399744,
+      "eval_loss": 0.6690404415130615,
+      "eval_runtime": 106.3444,
+      "eval_samples_per_second": 188.896,
+      "eval_steps_per_second": 0.95,
+      "eval_token_acc": 0.8683678146748934,
+      "step": 1250
+    },
+    {
+      "epoch": 0.4015839366425343,
+      "grad_norm": 0.04782987834900457,
+      "learning_rate": 1.362807705350641e-05,
+      "loss": 0.46315851211547854,
+      "memory(GiB)": 76.61,
+      "step": 1255,
+      "token_acc": 0.8767961498796838,
+      "train_speed(iter/s)": 0.029512
+    },
+    {
+      "epoch": 0.4031838726450942,
+      "grad_norm": 0.05995996443795485,
+      "learning_rate": 1.3579639312382105e-05,
+      "loss": 0.46349530220031737,
+      "memory(GiB)": 76.61,
+      "step": 1260,
+      "token_acc": 0.8588617886178862,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.4047838086476541,
+      "grad_norm": 0.06488882353036057,
+      "learning_rate": 1.3531105049040667e-05,
+      "loss": 0.45726447105407714,
+      "memory(GiB)": 76.61,
+      "step": 1265,
+      "token_acc": 0.8802249582003344,
+      "train_speed(iter/s)": 0.029543
+    },
+    {
+      "epoch": 0.406383744650214,
+      "grad_norm": 0.05350128050935312,
+      "learning_rate": 1.3482475572171132e-05,
+      "loss": 0.4516806125640869,
+      "memory(GiB)": 76.61,
+      "step": 1270,
+      "token_acc": 0.8560765550239234,
+      "train_speed(iter/s)": 0.029549
+    },
+    {
+      "epoch": 0.4079836806527739,
+      "grad_norm": 0.05672697687392494,
+      "learning_rate": 1.3433752193029888e-05,
+      "loss": 0.46581568717956545,
+      "memory(GiB)": 76.61,
+      "step": 1275,
+      "token_acc": 0.8881742738589211,
+      "train_speed(iter/s)": 0.029547
+    },
+    {
+      "epoch": 0.40958361665533377,
+      "grad_norm": 0.0598115330947421,
+      "learning_rate": 1.3384936225405326e-05,
+      "loss": 0.46333680152893064,
+      "memory(GiB)": 76.61,
+      "step": 1280,
+      "token_acc": 0.8608710985716804,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.4111835526578937,
+      "grad_norm": 0.05384417907735887,
+      "learning_rate": 1.333602898558242e-05,
+      "loss": 0.4611030578613281,
+      "memory(GiB)": 76.61,
+      "step": 1285,
+      "token_acc": 0.8845689770746749,
+      "train_speed(iter/s)": 0.029567
+    },
+    {
+      "epoch": 0.41278348866045356,
+      "grad_norm": 0.06043637267465684,
+      "learning_rate": 1.3287031792307226e-05,
+      "loss": 0.46013875007629396,
+      "memory(GiB)": 76.61,
+      "step": 1290,
+      "token_acc": 0.870195210303884,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.4143834246630135,
+      "grad_norm": 0.06140603532631629,
+      "learning_rate": 1.323794596675132e-05,
+      "loss": 0.45681238174438477,
+      "memory(GiB)": 76.61,
+      "step": 1295,
+      "token_acc": 0.8450012281994596,
+      "train_speed(iter/s)": 0.029583
+    },
+    {
+      "epoch": 0.4159833606655734,
+      "grad_norm": 0.062077229851937275,
+      "learning_rate": 1.318877283247619e-05,
+      "loss": 0.4490199565887451,
+      "memory(GiB)": 76.61,
+      "step": 1300,
+      "token_acc": 0.89259877573734,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.41758329666813326,
+      "grad_norm": 0.05216177276902916,
+      "learning_rate": 1.3139513715397521e-05,
+      "loss": 0.45108351707458494,
+      "memory(GiB)": 76.61,
+      "step": 1305,
+      "token_acc": 0.8547701815372731,
+      "train_speed(iter/s)": 0.029594
+    },
+    {
+      "epoch": 0.4191832326706932,
+      "grad_norm": 0.05738628087610287,
+      "learning_rate": 1.3090169943749475e-05,
+      "loss": 0.4558550834655762,
+      "memory(GiB)": 76.61,
+      "step": 1310,
+      "token_acc": 0.840696686491079,
+      "train_speed(iter/s)": 0.029587
+    },
+    {
+      "epoch": 0.42078316867325305,
+      "grad_norm": 0.05518036740697275,
+      "learning_rate": 1.304074284804885e-05,
+      "loss": 0.4631648063659668,
+      "memory(GiB)": 76.61,
+      "step": 1315,
+      "token_acc": 0.8788111708941839,
+      "train_speed(iter/s)": 0.029578
+    },
+    {
+      "epoch": 0.42238310467581297,
+      "grad_norm": 0.05902492258138098,
+      "learning_rate": 1.2991233761059214e-05,
+      "loss": 0.45921921730041504,
+      "memory(GiB)": 76.61,
+      "step": 1320,
+      "token_acc": 0.866059646344682,
+      "train_speed(iter/s)": 0.029604
+    },
+    {
+      "epoch": 0.4239830406783729,
+      "grad_norm": 0.059535437419073044,
+      "learning_rate": 1.2941644017754964e-05,
+      "loss": 0.46445517539978026,
+      "memory(GiB)": 76.61,
+      "step": 1325,
+      "token_acc": 0.8831345826235094,
+      "train_speed(iter/s)": 0.029591
+    },
+    {
+      "epoch": 0.42558297668093276,
+      "grad_norm": 0.04863893443696892,
+      "learning_rate": 1.289197495528534e-05,
+      "loss": 0.45836362838745115,
+      "memory(GiB)": 76.61,
+      "step": 1330,
+      "token_acc": 0.8986429177268872,
+      "train_speed(iter/s)": 0.029582
+    },
+    {
+      "epoch": 0.4271829126834927,
+      "grad_norm": 0.05945822860509985,
+      "learning_rate": 1.284222791293836e-05,
+      "loss": 0.45783252716064454,
+      "memory(GiB)": 76.61,
+      "step": 1335,
+      "token_acc": 0.8500874125874126,
+      "train_speed(iter/s)": 0.029596
+    },
+    {
+      "epoch": 0.42878284868605254,
+      "grad_norm": 0.05989616737178823,
+      "learning_rate": 1.2792404232104699e-05,
+      "loss": 0.45293269157409666,
+      "memory(GiB)": 76.61,
+      "step": 1340,
+      "token_acc": 0.8637480798771121,
+      "train_speed(iter/s)": 0.029584
+    },
+    {
+      "epoch": 0.43038278468861246,
+      "grad_norm": 0.0586629819404024,
+      "learning_rate": 1.2742505256241543e-05,
+      "loss": 0.45876450538635255,
+      "memory(GiB)": 76.61,
+      "step": 1345,
+      "token_acc": 0.8296499119890475,
+      "train_speed(iter/s)": 0.029588
+    },
+    {
+      "epoch": 0.43198272069117233,
+      "grad_norm": 0.052924904785980484,
+      "learning_rate": 1.2692532330836346e-05,
+      "loss": 0.45821080207824705,
+      "memory(GiB)": 76.61,
+      "step": 1350,
+      "token_acc": 0.8636084374360025,
+      "train_speed(iter/s)": 0.029594
+    },
+    {
+      "epoch": 0.43358265669373225,
+      "grad_norm": 0.059304249814977644,
+      "learning_rate": 1.2642486803370553e-05,
+      "loss": 0.45485148429870603,
+      "memory(GiB)": 76.61,
+      "step": 1355,
+      "token_acc": 0.8686690223792697,
+      "train_speed(iter/s)": 0.029579
+    },
+    {
+      "epoch": 0.43518259269629217,
+      "grad_norm": 0.06253442360689314,
+      "learning_rate": 1.2592370023283268e-05,
+      "loss": 0.45198469161987304,
+      "memory(GiB)": 76.61,
+      "step": 1360,
+      "token_acc": 0.8737075332348597,
+      "train_speed(iter/s)": 0.0296
+    },
+    {
+      "epoch": 0.43678252869885204,
+      "grad_norm": 0.05314091037792793,
+      "learning_rate": 1.2542183341934873e-05,
+      "loss": 0.4516898155212402,
+      "memory(GiB)": 76.61,
+      "step": 1365,
+      "token_acc": 0.8714476021314387,
+      "train_speed(iter/s)": 0.029596
+    },
+    {
+      "epoch": 0.43838246470141196,
+      "grad_norm": 0.06014404788689081,
+      "learning_rate": 1.2491928112570568e-05,
+      "loss": 0.45399184226989747,
+      "memory(GiB)": 76.61,
+      "step": 1370,
+      "token_acc": 0.8657097288676237,
+      "train_speed(iter/s)": 0.029583
+    },
+    {
+      "epoch": 0.4399824007039718,
+      "grad_norm": 0.05910144328100835,
+      "learning_rate": 1.2441605690283915e-05,
+      "loss": 0.4607128143310547,
+      "memory(GiB)": 76.61,
+      "step": 1375,
+      "token_acc": 0.8990952307928232,
+      "train_speed(iter/s)": 0.029603
+    },
+    {
+      "epoch": 0.44158233670653174,
+      "grad_norm": 0.059073628736854025,
+      "learning_rate": 1.2391217431980273e-05,
+      "loss": 0.4515543937683105,
+      "memory(GiB)": 76.61,
+      "step": 1380,
+      "token_acc": 0.9016349860428021,
+      "train_speed(iter/s)": 0.029591
+    },
+    {
+      "epoch": 0.44318227270909166,
+      "grad_norm": 0.058358968679540275,
+      "learning_rate": 1.234076469634022e-05,
+      "loss": 0.45762925148010253,
+      "memory(GiB)": 76.61,
+      "step": 1385,
+      "token_acc": 0.8919261822376009,
+      "train_speed(iter/s)": 0.029584
+    },
+    {
+      "epoch": 0.4447822087116515,
+      "grad_norm": 0.0672513399669503,
+      "learning_rate": 1.2290248843782915e-05,
+      "loss": 0.44803729057312014,
+      "memory(GiB)": 76.61,
+      "step": 1390,
+      "token_acc": 0.8975998070196599,
+      "train_speed(iter/s)": 0.029597
+    },
+    {
+      "epoch": 0.44638214471421145,
+      "grad_norm": 0.05793114375836921,
+      "learning_rate": 1.2239671236429413e-05,
+      "loss": 0.4537235736846924,
+      "memory(GiB)": 76.61,
+      "step": 1395,
+      "token_acc": 0.8839514422541486,
+      "train_speed(iter/s)": 0.02958
+    },
+    {
+      "epoch": 0.4479820807167713,
+      "grad_norm": 0.05955306099185102,
+      "learning_rate": 1.218903323806595e-05,
+      "loss": 0.4573692798614502,
+      "memory(GiB)": 76.61,
+      "step": 1400,
+      "token_acc": 0.8418099547511312,
+      "train_speed(iter/s)": 0.029594
+    },
+    {
+      "epoch": 0.44958201671933123,
+      "grad_norm": 0.058484796569864064,
+      "learning_rate": 1.2138336214107148e-05,
+      "loss": 0.44894704818725584,
+      "memory(GiB)": 76.61,
+      "step": 1405,
+      "token_acc": 0.8525200458190149,
+      "train_speed(iter/s)": 0.029594
+    },
+    {
+      "epoch": 0.4511819527218911,
+      "grad_norm": 0.05092836798588581,
+      "learning_rate": 1.2087581531559208e-05,
+      "loss": 0.45393967628479004,
+      "memory(GiB)": 76.61,
+      "step": 1410,
+      "token_acc": 0.8791390728476821,
+      "train_speed(iter/s)": 0.02958
+    },
+    {
+      "epoch": 0.452781888724451,
+      "grad_norm": 0.07033477253264378,
+      "learning_rate": 1.2036770558983067e-05,
+      "loss": 0.45307221412658694,
+      "memory(GiB)": 76.61,
+      "step": 1415,
+      "token_acc": 0.8387482900136799,
+      "train_speed(iter/s)": 0.029599
+    },
+    {
+      "epoch": 0.45438182472701094,
+      "grad_norm": 0.05966547548288182,
+      "learning_rate": 1.1985904666457455e-05,
+      "loss": 0.455959415435791,
+      "memory(GiB)": 76.61,
+      "step": 1420,
+      "token_acc": 0.9042096902303416,
+      "train_speed(iter/s)": 0.029583
+    },
+    {
+      "epoch": 0.4559817607295708,
+      "grad_norm": 0.08159145764722696,
+      "learning_rate": 1.1934985225541998e-05,
+      "loss": 0.462065601348877,
+      "memory(GiB)": 76.61,
+      "step": 1425,
+      "token_acc": 0.885252444621832,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.4575816967321307,
+      "grad_norm": 0.05540814227664117,
+      "learning_rate": 1.18840136092402e-05,
+      "loss": 0.4551572322845459,
+      "memory(GiB)": 76.61,
+      "step": 1430,
+      "token_acc": 0.8559651934966797,
+      "train_speed(iter/s)": 0.029592
+    },
+    {
+      "epoch": 0.4591816327346906,
+      "grad_norm": 0.05534004007067895,
+      "learning_rate": 1.1832991191962435e-05,
+      "loss": 0.4455368995666504,
+      "memory(GiB)": 76.61,
+      "step": 1435,
+      "token_acc": 0.875560538116592,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.4607815687372505,
+      "grad_norm": 0.058276771895487044,
+      "learning_rate": 1.1781919349488894e-05,
+      "loss": 0.4590908527374268,
+      "memory(GiB)": 76.61,
+      "step": 1440,
+      "token_acc": 0.8510418460478733,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.46238150473981043,
+      "grad_norm": 0.05839975543902795,
+      "learning_rate": 1.1730799458932473e-05,
+      "loss": 0.462816858291626,
+      "memory(GiB)": 76.61,
+      "step": 1445,
+      "token_acc": 0.9052378085490669,
+      "train_speed(iter/s)": 0.029586
+    },
+    {
+      "epoch": 0.4639814407423703,
+      "grad_norm": 0.07084434546926481,
+      "learning_rate": 1.1679632898701649e-05,
+      "loss": 0.4550295829772949,
+      "memory(GiB)": 76.61,
+      "step": 1450,
+      "token_acc": 0.8805626598465474,
+      "train_speed(iter/s)": 0.029572
+    },
+    {
+      "epoch": 0.4655813767449302,
+      "grad_norm": 0.06519996046237972,
+      "learning_rate": 1.1628421048463315e-05,
+      "loss": 0.46291208267211914,
+      "memory(GiB)": 76.61,
+      "step": 1455,
+      "token_acc": 0.8565744600227359,
+      "train_speed(iter/s)": 0.029581
+    },
+    {
+      "epoch": 0.4671813127474901,
+      "grad_norm": 0.05799269979733804,
+      "learning_rate": 1.1577165289105565e-05,
+      "loss": 0.4474311351776123,
+      "memory(GiB)": 76.61,
+      "step": 1460,
+      "token_acc": 0.8579789309403043,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.46878124875005,
+      "grad_norm": 0.057120675003187855,
+      "learning_rate": 1.1525867002700484e-05,
+      "loss": 0.46109714508056643,
+      "memory(GiB)": 76.61,
+      "step": 1465,
+      "token_acc": 0.8752182516587126,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.47038118475260987,
+      "grad_norm": 0.05696370798749074,
+      "learning_rate": 1.1474527572466847e-05,
+      "loss": 0.4501948833465576,
+      "memory(GiB)": 76.61,
+      "step": 1470,
+      "token_acc": 0.8529032258064516,
+      "train_speed(iter/s)": 0.029562
+    },
+    {
+      "epoch": 0.4719811207551698,
+      "grad_norm": 0.05518112754329221,
+      "learning_rate": 1.1423148382732854e-05,
+      "loss": 0.45941987037658694,
+      "memory(GiB)": 76.61,
+      "step": 1475,
+      "token_acc": 0.9009282399143164,
+      "train_speed(iter/s)": 0.029545
+    },
+    {
+      "epoch": 0.4735810567577297,
+      "grad_norm": 0.051496444525703684,
+      "learning_rate": 1.1371730818898785e-05,
+      "loss": 0.45296878814697267,
+      "memory(GiB)": 76.61,
+      "step": 1480,
+      "token_acc": 0.8814303638644918,
+      "train_speed(iter/s)": 0.029538
+    },
+    {
+      "epoch": 0.4751809927602896,
+      "grad_norm": 0.0677105428949175,
+      "learning_rate": 1.132027626739965e-05,
+      "loss": 0.45635080337524414,
+      "memory(GiB)": 76.61,
+      "step": 1485,
+      "token_acc": 0.880248833592535,
+      "train_speed(iter/s)": 0.029546
+    },
+    {
+      "epoch": 0.4767809287628495,
+      "grad_norm": 0.0673509631098402,
+      "learning_rate": 1.1268786115667798e-05,
+      "loss": 0.4614115715026855,
+      "memory(GiB)": 76.61,
+      "step": 1490,
+      "token_acc": 0.8609592251210748,
+      "train_speed(iter/s)": 0.029525
+    },
+    {
+      "epoch": 0.47838086476540936,
+      "grad_norm": 0.053337771378298794,
+      "learning_rate": 1.1217261752095518e-05,
+      "loss": 0.45500664710998534,
+      "memory(GiB)": 76.61,
+      "step": 1495,
+      "token_acc": 0.8794466403162056,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.4799808007679693,
+      "grad_norm": 0.05429302474155136,
+      "learning_rate": 1.1165704565997593e-05,
+      "loss": 0.44763407707214353,
+      "memory(GiB)": 76.61,
+      "step": 1500,
+      "token_acc": 0.8700440528634361,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.4799808007679693,
+      "eval_loss": 0.6668144464492798,
+      "eval_runtime": 124.2589,
+      "eval_samples_per_second": 161.662,
+      "eval_steps_per_second": 0.813,
+      "eval_token_acc": 0.8694583558206896,
+      "step": 1500
+    },
+    {
+      "epoch": 0.4815807367705292,
+      "grad_norm": 0.06146640524587408,
+      "learning_rate": 1.1114115947573834e-05,
+      "loss": 0.45711498260498046,
+      "memory(GiB)": 76.61,
+      "step": 1505,
+      "token_acc": 0.8695166967121641,
+      "train_speed(iter/s)": 0.029461
+    },
+    {
+      "epoch": 0.48318067277308907,
+      "grad_norm": 0.06183782693437151,
+      "learning_rate": 1.1062497287871606e-05,
+      "loss": 0.4499336242675781,
+      "memory(GiB)": 76.61,
+      "step": 1510,
+      "token_acc": 0.8487557381009906,
+      "train_speed(iter/s)": 0.029468
+    },
+    {
+      "epoch": 0.484780608775649,
+      "grad_norm": 0.056753269624682155,
+      "learning_rate": 1.1010849978748314e-05,
+      "loss": 0.4551094055175781,
+      "memory(GiB)": 76.61,
+      "step": 1515,
+      "token_acc": 0.8579035448045033,
+      "train_speed(iter/s)": 0.02948
+    },
+    {
+      "epoch": 0.48638054477820886,
+      "grad_norm": 0.05188962595699218,
+      "learning_rate": 1.0959175412833869e-05,
+      "loss": 0.4483503818511963,
+      "memory(GiB)": 76.61,
+      "step": 1520,
+      "token_acc": 0.8649127992905705,
+      "train_speed(iter/s)": 0.029477
+    },
+    {
+      "epoch": 0.4879804807807688,
+      "grad_norm": 0.0526414480661873,
+      "learning_rate": 1.0907474983493144e-05,
+      "loss": 0.45140752792358396,
+      "memory(GiB)": 76.61,
+      "step": 1525,
+      "token_acc": 0.8700204290091931,
+      "train_speed(iter/s)": 0.029464
+    },
+    {
+      "epoch": 0.48958041678332864,
+      "grad_norm": 0.06682159988119828,
+      "learning_rate": 1.08557500847884e-05,
+      "loss": 0.4480952262878418,
+      "memory(GiB)": 76.61,
+      "step": 1530,
+      "token_acc": 0.8906385187748745,
+      "train_speed(iter/s)": 0.029478
+    },
+    {
+      "epoch": 0.49118035278588856,
+      "grad_norm": 0.06117568492897364,
+      "learning_rate": 1.080400211144169e-05,
+      "loss": 0.453688907623291,
+      "memory(GiB)": 76.61,
+      "step": 1535,
+      "token_acc": 0.8130096719135217,
+      "train_speed(iter/s)": 0.029469
+    },
+    {
+      "epoch": 0.4927802887884485,
+      "grad_norm": 0.05143211947513191,
+      "learning_rate": 1.0752232458797262e-05,
+      "loss": 0.44568753242492676,
+      "memory(GiB)": 76.61,
+      "step": 1540,
+      "token_acc": 0.847257743677181,
+      "train_speed(iter/s)": 0.029464
+    },
+    {
+      "epoch": 0.49438022479100835,
+      "grad_norm": 0.05201971134010435,
+      "learning_rate": 1.070044252278393e-05,
+      "loss": 0.46500363349914553,
+      "memory(GiB)": 76.61,
+      "step": 1545,
+      "token_acc": 0.84,
+      "train_speed(iter/s)": 0.029474
+    },
+    {
+      "epoch": 0.49598016079356827,
+      "grad_norm": 0.05304880581645989,
+      "learning_rate": 1.064863369987743e-05,
+      "loss": 0.4501206398010254,
+      "memory(GiB)": 76.61,
+      "step": 1550,
+      "token_acc": 0.8888641920426762,
+      "train_speed(iter/s)": 0.029465
+    },
+    {
+      "epoch": 0.49758009679612814,
+      "grad_norm": 0.050584443072610216,
+      "learning_rate": 1.0596807387062772e-05,
+      "loss": 0.456621789932251,
+      "memory(GiB)": 76.61,
+      "step": 1555,
+      "token_acc": 0.8793768317137128,
+      "train_speed(iter/s)": 0.02948
+    },
+    {
+      "epoch": 0.49918003279868806,
+      "grad_norm": 0.05907676168100355,
+      "learning_rate": 1.0544964981796563e-05,
+      "loss": 0.4567122936248779,
+      "memory(GiB)": 76.61,
+      "step": 1560,
+      "token_acc": 0.8505747126436781,
+      "train_speed(iter/s)": 0.029477
+    },
+    {
+      "epoch": 0.500779968801248,
+      "grad_norm": 0.055037989511506104,
+      "learning_rate": 1.0493107881969335e-05,
+      "loss": 0.44720020294189455,
+      "memory(GiB)": 76.61,
+      "step": 1565,
+      "token_acc": 0.8853304383227032,
+      "train_speed(iter/s)": 0.029466
+    },
+    {
+      "epoch": 0.5023799048038079,
+      "grad_norm": 0.0597376748229471,
+      "learning_rate": 1.0441237485867845e-05,
+      "loss": 0.4492997169494629,
+      "memory(GiB)": 76.61,
+      "step": 1570,
+      "token_acc": 0.8809347181008902,
+      "train_speed(iter/s)": 0.029489
+    },
+    {
+      "epoch": 0.5039798408063677,
+      "grad_norm": 0.060265741182571844,
+      "learning_rate": 1.0389355192137379e-05,
+      "loss": 0.4525942325592041,
+      "memory(GiB)": 76.61,
+      "step": 1575,
+      "token_acc": 0.8839541547277937,
+      "train_speed(iter/s)": 0.029481
+    },
+    {
+      "epoch": 0.5055797768089276,
+      "grad_norm": 0.06015007204584338,
+      "learning_rate": 1.0337462399744025e-05,
+      "loss": 0.4606604099273682,
+      "memory(GiB)": 76.61,
+      "step": 1580,
+      "token_acc": 0.8439696373348328,
+      "train_speed(iter/s)": 0.029471
+    },
+    {
+      "epoch": 0.5071797128114875,
+      "grad_norm": 0.0539606724017438,
+      "learning_rate": 1.0285560507936962e-05,
+      "loss": 0.46471481323242186,
+      "memory(GiB)": 76.61,
+      "step": 1585,
+      "token_acc": 0.8212732305258995,
+      "train_speed(iter/s)": 0.029486
+    },
+    {
+      "epoch": 0.5087796488140475,
+      "grad_norm": 0.0588254805138369,
+      "learning_rate": 1.0233650916210736e-05,
+      "loss": 0.45154604911804197,
+      "memory(GiB)": 76.61,
+      "step": 1590,
+      "token_acc": 0.883357041251778,
+      "train_speed(iter/s)": 0.029474
+    },
+    {
+      "epoch": 0.5103795848166074,
+      "grad_norm": 0.06409304780777438,
+      "learning_rate": 1.0181735024267504e-05,
+      "loss": 0.45000271797180175,
+      "memory(GiB)": 76.61,
+      "step": 1595,
+      "token_acc": 0.8340197693574959,
+      "train_speed(iter/s)": 0.029485
+    },
+    {
+      "epoch": 0.5119795208191672,
+      "grad_norm": 0.058200459243944895,
+      "learning_rate": 1.012981423197931e-05,
+      "loss": 0.4608008861541748,
+      "memory(GiB)": 76.61,
+      "step": 1600,
+      "token_acc": 0.8793913904007917,
+      "train_speed(iter/s)": 0.029484
+    },
+    {
+      "epoch": 0.5135794568217271,
+      "grad_norm": 0.052541653818392466,
+      "learning_rate": 1.007788993935033e-05,
+      "loss": 0.45448942184448243,
+      "memory(GiB)": 76.61,
+      "step": 1605,
+      "token_acc": 0.8615229110512129,
+      "train_speed(iter/s)": 0.029472
+    },
+    {
+      "epoch": 0.515179392824287,
+      "grad_norm": 0.06526426191856917,
+      "learning_rate": 1.002596354647912e-05,
+      "loss": 0.45614914894104003,
+      "memory(GiB)": 76.61,
+      "step": 1610,
+      "token_acc": 0.871312462372065,
+      "train_speed(iter/s)": 0.029489
+    },
+    {
+      "epoch": 0.516779328826847,
+      "grad_norm": 0.05743481751682084,
+      "learning_rate": 9.974036453520881e-06,
+      "loss": 0.447450590133667,
+      "memory(GiB)": 76.61,
+      "step": 1615,
+      "token_acc": 0.8760352658295485,
+      "train_speed(iter/s)": 0.02948
+    },
+    {
+      "epoch": 0.5183792648294068,
+      "grad_norm": 0.06577378911981274,
+      "learning_rate": 9.922110060649672e-06,
+      "loss": 0.45809640884399416,
+      "memory(GiB)": 76.61,
+      "step": 1620,
+      "token_acc": 0.9048299514146899,
+      "train_speed(iter/s)": 0.029468
+    },
+    {
+      "epoch": 0.5199792008319667,
+      "grad_norm": 0.05180626448607971,
+      "learning_rate": 9.870185768020694e-06,
+      "loss": 0.4360641002655029,
+      "memory(GiB)": 76.61,
+      "step": 1625,
+      "token_acc": 0.890797148412184,
+      "train_speed(iter/s)": 0.029481
+    },
+    {
+      "epoch": 0.5215791368345266,
+      "grad_norm": 0.048795347699578454,
+      "learning_rate": 9.818264975732497e-06,
+      "loss": 0.4505919933319092,
+      "memory(GiB)": 76.61,
+      "step": 1630,
+      "token_acc": 0.8830155979202773,
+      "train_speed(iter/s)": 0.029462
+    },
+    {
+      "epoch": 0.5231790728370865,
+      "grad_norm": 0.054325754690003863,
+      "learning_rate": 9.766349083789266e-06,
+      "loss": 0.4518167495727539,
+      "memory(GiB)": 76.61,
+      "step": 1635,
+      "token_acc": 0.8740636704119851,
+      "train_speed(iter/s)": 0.029458
+    },
+    {
+      "epoch": 0.5247790088396465,
+      "grad_norm": 0.05473270131153146,
+      "learning_rate": 9.71443949206304e-06,
+      "loss": 0.4629377841949463,
+      "memory(GiB)": 76.61,
+      "step": 1640,
+      "token_acc": 0.8927648578811369,
+      "train_speed(iter/s)": 0.029469
+    },
+    {
+      "epoch": 0.5263789448422063,
+      "grad_norm": 0.05962553624327487,
+      "learning_rate": 9.662537600255979e-06,
+      "loss": 0.4535552501678467,
+      "memory(GiB)": 76.61,
+      "step": 1645,
+      "token_acc": 0.8980960623106881,
+      "train_speed(iter/s)": 0.029457
+    },
+    {
+      "epoch": 0.5279788808447662,
+      "grad_norm": 0.06367541172972058,
+      "learning_rate": 9.610644807862625e-06,
+      "loss": 0.44418978691101074,
+      "memory(GiB)": 76.61,
+      "step": 1650,
+      "token_acc": 0.8769371011850501,
+      "train_speed(iter/s)": 0.029468
+    },
+    {
+      "epoch": 0.5295788168473261,
+      "grad_norm": 0.05288367644088033,
+      "learning_rate": 9.558762514132157e-06,
+      "loss": 0.4513704299926758,
+      "memory(GiB)": 76.61,
+      "step": 1655,
+      "token_acc": 0.8576478906434126,
+      "train_speed(iter/s)": 0.029464
+    },
+    {
+      "epoch": 0.531178752849886,
+      "grad_norm": 0.054919719691940275,
+      "learning_rate": 9.506892118030668e-06,
+      "loss": 0.4454075336456299,
+      "memory(GiB)": 76.61,
+      "step": 1660,
+      "token_acc": 0.8535078688042359,
+      "train_speed(iter/s)": 0.029456
+    },
+    {
+      "epoch": 0.532778688852446,
+      "grad_norm": 0.0561391497524031,
+      "learning_rate": 9.455035018203439e-06,
+      "loss": 0.4459484100341797,
+      "memory(GiB)": 76.61,
+      "step": 1665,
+      "token_acc": 0.8793124922157181,
+      "train_speed(iter/s)": 0.029471
+    },
+    {
+      "epoch": 0.5343786248550058,
+      "grad_norm": 0.051526152917333715,
+      "learning_rate": 9.40319261293723e-06,
+      "loss": 0.4593966484069824,
+      "memory(GiB)": 76.61,
+      "step": 1670,
+      "token_acc": 0.8957568638966378,
+      "train_speed(iter/s)": 0.029466
+    },
+    {
+      "epoch": 0.5359785608575657,
+      "grad_norm": 0.05336577516465571,
+      "learning_rate": 9.351366300122569e-06,
+      "loss": 0.45195541381835935,
+      "memory(GiB)": 76.61,
+      "step": 1675,
+      "token_acc": 0.8254120659305488,
+      "train_speed(iter/s)": 0.029459
+    },
+    {
+      "epoch": 0.5375784968601256,
+      "grad_norm": 0.05605931671991975,
+      "learning_rate": 9.299557477216073e-06,
+      "loss": 0.4473400115966797,
+      "memory(GiB)": 76.61,
+      "step": 1680,
+      "token_acc": 0.8684433164128595,
+      "train_speed(iter/s)": 0.029474
+    },
+    {
+      "epoch": 0.5391784328626855,
+      "grad_norm": 0.05284398220938548,
+      "learning_rate": 9.247767541202738e-06,
+      "loss": 0.4539934158325195,
+      "memory(GiB)": 76.61,
+      "step": 1685,
+      "token_acc": 0.8787515006002401,
+      "train_speed(iter/s)": 0.029458
+    },
+    {
+      "epoch": 0.5407783688652454,
+      "grad_norm": 0.06074778175058899,
+      "learning_rate": 9.195997888558312e-06,
+      "loss": 0.4540121078491211,
+      "memory(GiB)": 76.61,
+      "step": 1690,
+      "token_acc": 0.882076702321941,
+      "train_speed(iter/s)": 0.029458
+    },
+    {
+      "epoch": 0.5423783048678052,
+      "grad_norm": 0.06072504661929311,
+      "learning_rate": 9.144249915211605e-06,
+      "loss": 0.45176243782043457,
+      "memory(GiB)": 76.61,
+      "step": 1695,
+      "token_acc": 0.8652057386094908,
+      "train_speed(iter/s)": 0.029465
+    },
+    {
+      "epoch": 0.5439782408703652,
+      "grad_norm": 0.058552695609385315,
+      "learning_rate": 9.092525016506858e-06,
+      "loss": 0.4491862773895264,
+      "memory(GiB)": 76.61,
+      "step": 1700,
+      "token_acc": 0.8822588020118884,
+      "train_speed(iter/s)": 0.02945
+    },
+    {
+      "epoch": 0.5455781768729251,
+      "grad_norm": 0.056892490634495974,
+      "learning_rate": 9.040824587166136e-06,
+      "loss": 0.45043745040893557,
+      "memory(GiB)": 76.61,
+      "step": 1705,
+      "token_acc": 0.8825789923142613,
+      "train_speed(iter/s)": 0.029461
+    },
+    {
+      "epoch": 0.547178112875485,
+      "grad_norm": 0.05885692671807609,
+      "learning_rate": 8.98915002125169e-06,
+      "loss": 0.4475353240966797,
+      "memory(GiB)": 76.61,
+      "step": 1710,
+      "token_acc": 0.8721031538595574,
+      "train_speed(iter/s)": 0.029454
+    },
+    {
+      "epoch": 0.5487780488780449,
+      "grad_norm": 0.060276094585736115,
+      "learning_rate": 8.9375027121284e-06,
+      "loss": 0.4502556800842285,
+      "memory(GiB)": 76.61,
+      "step": 1715,
+      "token_acc": 0.8562842259917189,
+      "train_speed(iter/s)": 0.029449
+    },
+    {
+      "epoch": 0.5503779848806047,
+      "grad_norm": 0.06782068962590707,
+      "learning_rate": 8.885884052426168e-06,
+      "loss": 0.4532322883605957,
+      "memory(GiB)": 76.61,
+      "step": 1720,
+      "token_acc": 0.8593545573484518,
+      "train_speed(iter/s)": 0.029466
+    },
+    {
+      "epoch": 0.5519779208831647,
+      "grad_norm": 0.06070839045848377,
+      "learning_rate": 8.83429543400241e-06,
+      "loss": 0.45258092880249023,
+      "memory(GiB)": 76.61,
+      "step": 1725,
+      "token_acc": 0.8751242791807516,
+      "train_speed(iter/s)": 0.029452
+    },
+    {
+      "epoch": 0.5535778568857246,
+      "grad_norm": 0.049979952181739715,
+      "learning_rate": 8.78273824790448e-06,
+      "loss": 0.4340657234191895,
+      "memory(GiB)": 76.61,
+      "step": 1730,
+      "token_acc": 0.8650843222985634,
+      "train_speed(iter/s)": 0.029451
+    },
+    {
+      "epoch": 0.5551777928882845,
+      "grad_norm": 0.059124658124222323,
+      "learning_rate": 8.731213884332205e-06,
+      "loss": 0.43556828498840333,
+      "memory(GiB)": 76.61,
+      "step": 1735,
+      "token_acc": 0.8524390243902439,
+      "train_speed(iter/s)": 0.029459
+    },
+    {
+      "epoch": 0.5567777288908443,
+      "grad_norm": 0.05228309031135195,
+      "learning_rate": 8.679723732600355e-06,
+      "loss": 0.4483633041381836,
+      "memory(GiB)": 76.61,
+      "step": 1740,
+      "token_acc": 0.9039166284928997,
+      "train_speed(iter/s)": 0.029445
+    },
+    {
+      "epoch": 0.5583776648934042,
+      "grad_norm": 0.05659321921396489,
+      "learning_rate": 8.628269181101216e-06,
+      "loss": 0.45377864837646487,
+      "memory(GiB)": 76.61,
+      "step": 1745,
+      "token_acc": 0.8812897628687102,
+      "train_speed(iter/s)": 0.029449
+    },
+    {
+      "epoch": 0.5599776008959642,
+      "grad_norm": 0.0610469666222746,
+      "learning_rate": 8.576851617267151e-06,
+      "loss": 0.4495216369628906,
+      "memory(GiB)": 76.61,
+      "step": 1750,
+      "token_acc": 0.8734145104008117,
+      "train_speed(iter/s)": 0.029452
+    },
+    {
+      "epoch": 0.5599776008959642,
+      "eval_loss": 0.6640093922615051,
+      "eval_runtime": 114.9985,
+      "eval_samples_per_second": 174.681,
+      "eval_steps_per_second": 0.878,
+      "eval_token_acc": 0.8701128116616424,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5615775368985241,
+      "grad_norm": 0.061306770256929585,
+      "learning_rate": 8.525472427533156e-06,
+      "loss": 0.44908857345581055,
+      "memory(GiB)": 77.63,
+      "step": 1755,
+      "token_acc": 0.8715457946180765,
+      "train_speed(iter/s)": 0.029409
+    },
+    {
+      "epoch": 0.563177472901084,
+      "grad_norm": 0.05235639827008535,
+      "learning_rate": 8.474132997299521e-06,
+      "loss": 0.4579316139221191,
+      "memory(GiB)": 77.63,
+      "step": 1760,
+      "token_acc": 0.8922923256201098,
+      "train_speed(iter/s)": 0.029422
+    },
+    {
+      "epoch": 0.5647774089036438,
+      "grad_norm": 0.051281007426132216,
+      "learning_rate": 8.422834710894434e-06,
+      "loss": 0.45467004776000974,
+      "memory(GiB)": 77.63,
+      "step": 1765,
+      "token_acc": 0.903878366189924,
+      "train_speed(iter/s)": 0.029438
+    },
+    {
+      "epoch": 0.5663773449062037,
+      "grad_norm": 0.05049109520782513,
+      "learning_rate": 8.371578951536689e-06,
+      "loss": 0.45294957160949706,
+      "memory(GiB)": 77.63,
+      "step": 1770,
+      "token_acc": 0.8928110202324581,
+      "train_speed(iter/s)": 0.029439
+    },
+    {
+      "epoch": 0.5679772809087636,
+      "grad_norm": 0.04946427707777728,
+      "learning_rate": 8.320367101298351e-06,
+      "loss": 0.4473431587219238,
+      "memory(GiB)": 77.63,
+      "step": 1775,
+      "token_acc": 0.8723599632690542,
+      "train_speed(iter/s)": 0.029439
+    },
+    {
+      "epoch": 0.5695772169113236,
+      "grad_norm": 0.053606352244487274,
+      "learning_rate": 8.26920054106753e-06,
+      "loss": 0.4495864391326904,
+      "memory(GiB)": 77.63,
+      "step": 1780,
+      "token_acc": 0.8844315111203492,
+      "train_speed(iter/s)": 0.029459
+    },
+    {
+      "epoch": 0.5711771529138835,
+      "grad_norm": 0.05525614374940963,
+      "learning_rate": 8.218080650511107e-06,
+      "loss": 0.44890499114990234,
+      "memory(GiB)": 77.63,
+      "step": 1785,
+      "token_acc": 0.8749736453721273,
+      "train_speed(iter/s)": 0.02946
+    },
+    {
+      "epoch": 0.5727770889164433,
+      "grad_norm": 0.05882148265537131,
+      "learning_rate": 8.167008808037568e-06,
+      "loss": 0.44676194190979,
+      "memory(GiB)": 77.63,
+      "step": 1790,
+      "token_acc": 0.8807511737089202,
+      "train_speed(iter/s)": 0.029457
+    },
+    {
+      "epoch": 0.5743770249190032,
+      "grad_norm": 0.048821121641334515,
+      "learning_rate": 8.115986390759805e-06,
+      "loss": 0.4417415142059326,
+      "memory(GiB)": 77.63,
+      "step": 1795,
+      "token_acc": 0.8531673379714391,
+      "train_speed(iter/s)": 0.029469
+    },
+    {
+      "epoch": 0.5759769609215631,
+      "grad_norm": 0.054949264031140505,
+      "learning_rate": 8.065014774458004e-06,
+      "loss": 0.46439437866210936,
+      "memory(GiB)": 77.63,
+      "step": 1800,
+      "token_acc": 0.8333022213925705,
+      "train_speed(iter/s)": 0.029467
+    },
+    {
+      "epoch": 0.5775768969241231,
+      "grad_norm": 0.059507220518762304,
+      "learning_rate": 8.014095333542548e-06,
+      "loss": 0.4539642333984375,
+      "memory(GiB)": 77.63,
+      "step": 1805,
+      "token_acc": 0.8577178858942948,
+      "train_speed(iter/s)": 0.029483
+    },
+    {
+      "epoch": 0.579176832926683,
+      "grad_norm": 0.05302143027350534,
+      "learning_rate": 7.963229441016938e-06,
+      "loss": 0.4606470108032227,
+      "memory(GiB)": 77.63,
+      "step": 1810,
+      "token_acc": 0.8760775862068966,
+      "train_speed(iter/s)": 0.029484
+    },
+    {
+      "epoch": 0.5807767689292428,
+      "grad_norm": 0.0699581228289572,
+      "learning_rate": 7.912418468440794e-06,
+      "loss": 0.4488551139831543,
+      "memory(GiB)": 77.63,
+      "step": 1815,
+      "token_acc": 0.8892276422764228,
+      "train_speed(iter/s)": 0.029481
+    },
+    {
+      "epoch": 0.5823767049318027,
+      "grad_norm": 0.053456667148895104,
+      "learning_rate": 7.861663785892857e-06,
+      "loss": 0.45035881996154786,
+      "memory(GiB)": 77.63,
+      "step": 1820,
+      "token_acc": 0.8806643202815662,
+      "train_speed(iter/s)": 0.029498
+    },
+    {
+      "epoch": 0.5839766409343626,
+      "grad_norm": 0.05451209338463787,
+      "learning_rate": 7.810966761934053e-06,
+      "loss": 0.44800753593444825,
+      "memory(GiB)": 77.63,
+      "step": 1825,
+      "token_acc": 0.8771571298819255,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.5855765769369226,
+      "grad_norm": 0.05912934985203241,
+      "learning_rate": 7.760328763570589e-06,
+      "loss": 0.4499057769775391,
+      "memory(GiB)": 77.63,
+      "step": 1830,
+      "token_acc": 0.868710326675956,
+      "train_speed(iter/s)": 0.029487
+    },
+    {
+      "epoch": 0.5871765129394825,
+      "grad_norm": 0.052841905445767515,
+      "learning_rate": 7.709751156217088e-06,
+      "loss": 0.4497323989868164,
+      "memory(GiB)": 77.63,
+      "step": 1835,
+      "token_acc": 0.8117094325984822,
+      "train_speed(iter/s)": 0.029501
+    },
+    {
+      "epoch": 0.5887764489420423,
+      "grad_norm": 0.060076953422732254,
+      "learning_rate": 7.659235303659784e-06,
+      "loss": 0.4582187652587891,
+      "memory(GiB)": 77.63,
+      "step": 1840,
+      "token_acc": 0.8795674258561363,
+      "train_speed(iter/s)": 0.029492
+    },
+    {
+      "epoch": 0.5903763849446022,
+      "grad_norm": 0.06307528499562465,
+      "learning_rate": 7.608782568019729e-06,
+      "loss": 0.4430552005767822,
+      "memory(GiB)": 77.63,
+      "step": 1845,
+      "token_acc": 0.8452444922084901,
+      "train_speed(iter/s)": 0.029498
+    },
+    {
+      "epoch": 0.5919763209471621,
+      "grad_norm": 0.05378691938628143,
+      "learning_rate": 7.558394309716088e-06,
+      "loss": 0.459810209274292,
+      "memory(GiB)": 77.63,
+      "step": 1850,
+      "token_acc": 0.8506092736192435,
+      "train_speed(iter/s)": 0.029503
+    },
+    {
+      "epoch": 0.593576256949722,
+      "grad_norm": 0.0586506530143339,
+      "learning_rate": 7.508071887429433e-06,
+      "loss": 0.46239190101623534,
+      "memory(GiB)": 77.63,
+      "step": 1855,
+      "token_acc": 0.9115304709141274,
+      "train_speed(iter/s)": 0.029495
+    },
+    {
+      "epoch": 0.5951761929522819,
+      "grad_norm": 0.053290473896441634,
+      "learning_rate": 7.4578166580651335e-06,
+      "loss": 0.4524221897125244,
+      "memory(GiB)": 77.63,
+      "step": 1860,
+      "token_acc": 0.8817879571481345,
+      "train_speed(iter/s)": 0.029508
+    },
+    {
+      "epoch": 0.5967761289548418,
+      "grad_norm": 0.051901913358510056,
+      "learning_rate": 7.4076299767167325e-06,
+      "loss": 0.4579325675964355,
+      "memory(GiB)": 77.63,
+      "step": 1865,
+      "token_acc": 0.8617401668653158,
+      "train_speed(iter/s)": 0.029506
+    },
+    {
+      "epoch": 0.5983760649574017,
+      "grad_norm": 0.05256077511072294,
+      "learning_rate": 7.35751319662945e-06,
+      "loss": 0.45406513214111327,
+      "memory(GiB)": 77.63,
+      "step": 1870,
+      "token_acc": 0.8924402944873406,
+      "train_speed(iter/s)": 0.029497
+    },
+    {
+      "epoch": 0.5999760009599616,
+      "grad_norm": 0.056121622843709036,
+      "learning_rate": 7.307467669163655e-06,
+      "loss": 0.450104284286499,
+      "memory(GiB)": 77.63,
+      "step": 1875,
+      "token_acc": 0.8646184340931615,
+      "train_speed(iter/s)": 0.02951
+    },
+    {
+      "epoch": 0.6015759369625215,
+      "grad_norm": 0.051068951060234354,
+      "learning_rate": 7.25749474375846e-06,
+      "loss": 0.45695791244506834,
+      "memory(GiB)": 77.63,
+      "step": 1880,
+      "token_acc": 0.9112655568126717,
+      "train_speed(iter/s)": 0.029503
+    },
+    {
+      "epoch": 0.6031758729650813,
+      "grad_norm": 0.05120698584703106,
+      "learning_rate": 7.207595767895303e-06,
+      "loss": 0.4460740089416504,
+      "memory(GiB)": 77.63,
+      "step": 1885,
+      "token_acc": 0.8637192342752963,
+      "train_speed(iter/s)": 0.029499
+    },
+    {
+      "epoch": 0.6047758089676413,
+      "grad_norm": 0.05826366701259215,
+      "learning_rate": 7.157772087061645e-06,
+      "loss": 0.4498391628265381,
+      "memory(GiB)": 77.63,
+      "step": 1890,
+      "token_acc": 0.8602477214302408,
+      "train_speed(iter/s)": 0.029509
+    },
+    {
+      "epoch": 0.6063757449702012,
+      "grad_norm": 0.05454678875604061,
+      "learning_rate": 7.108025044714661e-06,
+      "loss": 0.44768247604370115,
+      "memory(GiB)": 77.63,
+      "step": 1895,
+      "token_acc": 0.8998014357721094,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.6079756809727611,
+      "grad_norm": 0.04862560763785379,
+      "learning_rate": 7.058355982245038e-06,
+      "loss": 0.44283151626586914,
+      "memory(GiB)": 77.63,
+      "step": 1900,
+      "token_acc": 0.8749580958766343,
+      "train_speed(iter/s)": 0.029501
+    },
+    {
+      "epoch": 0.609575616975321,
+      "grad_norm": 0.05390239428952395,
+      "learning_rate": 7.00876623894079e-06,
+      "loss": 0.4445077419281006,
+      "memory(GiB)": 77.63,
+      "step": 1905,
+      "token_acc": 0.8588156123822341,
+      "train_speed(iter/s)": 0.029504
+    },
+    {
+      "epoch": 0.6111755529778808,
+      "grad_norm": 0.052917745372876655,
+      "learning_rate": 6.959257151951153e-06,
+      "loss": 0.45001955032348634,
+      "memory(GiB)": 77.63,
+      "step": 1910,
+      "token_acc": 0.8768155911013054,
+      "train_speed(iter/s)": 0.029494
+    },
+    {
+      "epoch": 0.6127754889804408,
+      "grad_norm": 0.05432256049056495,
+      "learning_rate": 6.909830056250527e-06,
+      "loss": 0.44944238662719727,
+      "memory(GiB)": 77.63,
+      "step": 1915,
+      "token_acc": 0.8941244909831297,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.6143754249830007,
+      "grad_norm": 0.05852297407331436,
+      "learning_rate": 6.860486284602479e-06,
+      "loss": 0.4477729797363281,
+      "memory(GiB)": 77.63,
+      "step": 1920,
+      "token_acc": 0.8854845719252499,
+      "train_speed(iter/s)": 0.029501
+    },
+    {
+      "epoch": 0.6159753609855606,
+      "grad_norm": 0.05474007805899836,
+      "learning_rate": 6.8112271675238154e-06,
+      "loss": 0.4501204013824463,
+      "memory(GiB)": 77.63,
+      "step": 1925,
+      "token_acc": 0.8803290949887809,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.6175752969881205,
+      "grad_norm": 0.05545012433641634,
+      "learning_rate": 6.762054033248681e-06,
+      "loss": 0.44565958976745607,
+      "memory(GiB)": 77.63,
+      "step": 1930,
+      "token_acc": 0.8480542195015304,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.6191752329906803,
+      "grad_norm": 0.05495247298953925,
+      "learning_rate": 6.712968207692778e-06,
+      "loss": 0.44170804023742677,
+      "memory(GiB)": 77.63,
+      "step": 1935,
+      "token_acc": 0.8709073900841908,
+      "train_speed(iter/s)": 0.029498
+    },
+    {
+      "epoch": 0.6207751689932403,
+      "grad_norm": 0.05792014047889592,
+      "learning_rate": 6.663971014417585e-06,
+      "loss": 0.4454016208648682,
+      "memory(GiB)": 77.63,
+      "step": 1940,
+      "token_acc": 0.8606651376146789,
+      "train_speed(iter/s)": 0.0295
+    },
+    {
+      "epoch": 0.6223751049958002,
+      "grad_norm": 0.04853659131630362,
+      "learning_rate": 6.615063774594677e-06,
+      "loss": 0.4387532711029053,
+      "memory(GiB)": 77.63,
+      "step": 1945,
+      "token_acc": 0.8920454545454546,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.6239750409983601,
+      "grad_norm": 0.05266495974136303,
+      "learning_rate": 6.566247806970119e-06,
+      "loss": 0.4472493171691895,
+      "memory(GiB)": 77.63,
+      "step": 1950,
+      "token_acc": 0.848505251817937,
+      "train_speed(iter/s)": 0.029497
+    },
+    {
+      "epoch": 0.62557497700092,
+      "grad_norm": 0.054994759694813,
+      "learning_rate": 6.5175244278288705e-06,
+      "loss": 0.44487895965576174,
+      "memory(GiB)": 77.63,
+      "step": 1955,
+      "token_acc": 0.8689320388349514,
+      "train_speed(iter/s)": 0.029506
+    },
+    {
+      "epoch": 0.6271749130034798,
+      "grad_norm": 0.057067387083368365,
+      "learning_rate": 6.468894950959336e-06,
+      "loss": 0.4466127395629883,
+      "memory(GiB)": 77.63,
+      "step": 1960,
+      "token_acc": 0.846737755286463,
+      "train_speed(iter/s)": 0.0295
+    },
+    {
+      "epoch": 0.6287748490060397,
+      "grad_norm": 0.05417940634204734,
+      "learning_rate": 6.420360687617897e-06,
+      "loss": 0.44883151054382325,
+      "memory(GiB)": 77.63,
+      "step": 1965,
+      "token_acc": 0.8795967892477132,
+      "train_speed(iter/s)": 0.02949
+    },
+    {
+      "epoch": 0.6303747850085997,
+      "grad_norm": 0.05176488752108695,
+      "learning_rate": 6.3719229464935915e-06,
+      "loss": 0.4542849063873291,
+      "memory(GiB)": 77.63,
+      "step": 1970,
+      "token_acc": 0.8568893191352049,
+      "train_speed(iter/s)": 0.029503
+    },
+    {
+      "epoch": 0.6319747210111596,
+      "grad_norm": 0.04867135924369273,
+      "learning_rate": 6.323583033672799e-06,
+      "loss": 0.44331774711608884,
+      "memory(GiB)": 77.63,
+      "step": 1975,
+      "token_acc": 0.8647865559204172,
+      "train_speed(iter/s)": 0.029491
+    },
+    {
+      "epoch": 0.6335746570137194,
+      "grad_norm": 0.06076783884358601,
+      "learning_rate": 6.275342252604044e-06,
+      "loss": 0.44751858711242676,
+      "memory(GiB)": 77.63,
+      "step": 1980,
+      "token_acc": 0.871765773944621,
+      "train_speed(iter/s)": 0.029486
+    },
+    {
+      "epoch": 0.6351745930162793,
+      "grad_norm": 0.0520886449098567,
+      "learning_rate": 6.22720190406283e-06,
+      "loss": 0.46150927543640136,
+      "memory(GiB)": 77.63,
+      "step": 1985,
+      "token_acc": 0.8921661480178595,
+      "train_speed(iter/s)": 0.029497
+    },
+    {
+      "epoch": 0.6367745290188392,
+      "grad_norm": 0.058090405193780774,
+      "learning_rate": 6.179163286116581e-06,
+      "loss": 0.44019436836242676,
+      "memory(GiB)": 77.63,
+      "step": 1990,
+      "token_acc": 0.9157033805888768,
+      "train_speed(iter/s)": 0.029488
+    },
+    {
+      "epoch": 0.6383744650213992,
+      "grad_norm": 0.057472120727550105,
+      "learning_rate": 6.13122769408963e-06,
+      "loss": 0.4466409683227539,
+      "memory(GiB)": 77.63,
+      "step": 1995,
+      "token_acc": 0.8608313968499871,
+      "train_speed(iter/s)": 0.029492
+    },
+    {
+      "epoch": 0.6399744010239591,
+      "grad_norm": 0.05665485079826101,
+      "learning_rate": 6.083396420528298e-06,
+      "loss": 0.451153039932251,
+      "memory(GiB)": 77.63,
+      "step": 2000,
+      "token_acc": 0.8910367046369808,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.6399744010239591,
+      "eval_loss": 0.6622327566146851,
+      "eval_runtime": 115.9166,
+      "eval_samples_per_second": 173.297,
+      "eval_steps_per_second": 0.871,
+      "eval_token_acc": 0.8709580958089251,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6415743370265189,
+      "grad_norm": 0.05247711719910443,
+      "learning_rate": 6.0356707551660434e-06,
+      "loss": 0.45055346488952636,
+      "memory(GiB)": 77.63,
+      "step": 2005,
+      "token_acc": 0.8961562482257424,
+      "train_speed(iter/s)": 0.029453
+    },
+    {
+      "epoch": 0.6431742730290788,
+      "grad_norm": 0.0533769803289562,
+      "learning_rate": 5.988051984888668e-06,
+      "loss": 0.4436792373657227,
+      "memory(GiB)": 77.63,
+      "step": 2010,
+      "token_acc": 0.8894836272040302,
+      "train_speed(iter/s)": 0.029463
+    },
+    {
+      "epoch": 0.6447742090316387,
+      "grad_norm": 0.05428299581856707,
+      "learning_rate": 5.940541393699646e-06,
+      "loss": 0.44562363624572754,
+      "memory(GiB)": 77.63,
+      "step": 2015,
+      "token_acc": 0.8804031789106416,
+      "train_speed(iter/s)": 0.029477
+    },
+    {
+      "epoch": 0.6463741450341987,
+      "grad_norm": 0.055930584071511934,
+      "learning_rate": 5.893140262685469e-06,
+      "loss": 0.4412201404571533,
+      "memory(GiB)": 77.63,
+      "step": 2020,
+      "token_acc": 0.8791348600508906,
+      "train_speed(iter/s)": 0.029482
+    },
+    {
+      "epoch": 0.6479740810367586,
+      "grad_norm": 0.06077731970466293,
+      "learning_rate": 5.845849869981137e-06,
+      "loss": 0.44964237213134767,
+      "memory(GiB)": 77.63,
+      "step": 2025,
+      "token_acc": 0.8710450018908358,
+      "train_speed(iter/s)": 0.029484
+    },
+    {
+      "epoch": 0.6495740170393184,
+      "grad_norm": 0.05824848510516177,
+      "learning_rate": 5.7986714907356614e-06,
+      "loss": 0.4586543083190918,
+      "memory(GiB)": 77.63,
+      "step": 2030,
+      "token_acc": 0.8852591792656588,
+      "train_speed(iter/s)": 0.029498
+    },
+    {
+      "epoch": 0.6511739530418783,
+      "grad_norm": 0.06066761562869553,
+      "learning_rate": 5.751606397077703e-06,
+      "loss": 0.44632205963134763,
+      "memory(GiB)": 77.63,
+      "step": 2035,
+      "token_acc": 0.8871352785145888,
+      "train_speed(iter/s)": 0.029494
+    },
+    {
+      "epoch": 0.6527738890444382,
+      "grad_norm": 0.055201144436432543,
+      "learning_rate": 5.704655858081268e-06,
+      "loss": 0.43164916038513185,
+      "memory(GiB)": 77.63,
+      "step": 2040,
+      "token_acc": 0.8937977909940527,
+      "train_speed(iter/s)": 0.029496
+    },
+    {
+      "epoch": 0.6543738250469981,
+      "grad_norm": 0.05987076771844116,
+      "learning_rate": 5.6578211397314765e-06,
+      "loss": 0.4560856819152832,
+      "memory(GiB)": 77.63,
+      "step": 2045,
+      "token_acc": 0.8462914545204349,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.6559737610495581,
+      "grad_norm": 0.05549604363093839,
+      "learning_rate": 5.611103504890444e-06,
+      "loss": 0.44809746742248535,
+      "memory(GiB)": 77.63,
+      "step": 2050,
+      "token_acc": 0.8783254279232832,
+      "train_speed(iter/s)": 0.029503
+    },
+    {
+      "epoch": 0.6575736970521179,
+      "grad_norm": 0.060605872447174955,
+      "learning_rate": 5.564504213263205e-06,
+      "loss": 0.43492536544799804,
+      "memory(GiB)": 77.63,
+      "step": 2055,
+      "token_acc": 0.8383036405886909,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.6591736330546778,
+      "grad_norm": 0.05003885513998493,
+      "learning_rate": 5.5180245213637785e-06,
+      "loss": 0.44741315841674806,
+      "memory(GiB)": 77.63,
+      "step": 2060,
+      "token_acc": 0.8784857874174862,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.6607735690572377,
+      "grad_norm": 0.053248832099036005,
+      "learning_rate": 5.4716656824812505e-06,
+      "loss": 0.4469279766082764,
+      "memory(GiB)": 77.63,
+      "step": 2065,
+      "token_acc": 0.8853107344632768,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.6623735050597976,
+      "grad_norm": 0.050937098304756526,
+      "learning_rate": 5.425428946646016e-06,
+      "loss": 0.44948582649230956,
+      "memory(GiB)": 77.63,
+      "step": 2070,
+      "token_acc": 0.8934210526315789,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.6639734410623576,
+      "grad_norm": 0.050018563981163396,
+      "learning_rate": 5.379315560596038e-06,
+      "loss": 0.4475410461425781,
+      "memory(GiB)": 77.63,
+      "step": 2075,
+      "token_acc": 0.8478816857555876,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.6655733770649174,
+      "grad_norm": 0.05745892430696422,
+      "learning_rate": 5.333326767743263e-06,
+      "loss": 0.45008225440979005,
+      "memory(GiB)": 77.63,
+      "step": 2080,
+      "token_acc": 0.8264099454214675,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.6671733130674773,
+      "grad_norm": 0.05536507134607956,
+      "learning_rate": 5.287463808140069e-06,
+      "loss": 0.4393789291381836,
+      "memory(GiB)": 77.63,
+      "step": 2085,
+      "token_acc": 0.8450008816787162,
+      "train_speed(iter/s)": 0.029527
+    },
+    {
+      "epoch": 0.6687732490700372,
+      "grad_norm": 0.06142641017026178,
+      "learning_rate": 5.241727918445836e-06,
+      "loss": 0.4437687873840332,
+      "memory(GiB)": 77.63,
+      "step": 2090,
+      "token_acc": 0.8837277242185217,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.6703731850725971,
+      "grad_norm": 0.05426196603270913,
+      "learning_rate": 5.1961203318936116e-06,
+      "loss": 0.4427367687225342,
+      "memory(GiB)": 77.63,
+      "step": 2095,
+      "token_acc": 0.856048805815161,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.6719731210751569,
+      "grad_norm": 0.054169398190345976,
+      "learning_rate": 5.1506422782568345e-06,
+      "loss": 0.4520686626434326,
+      "memory(GiB)": 77.63,
+      "step": 2100,
+      "token_acc": 0.8747133027522935,
+      "train_speed(iter/s)": 0.029529
+    },
+    {
+      "epoch": 0.6735730570777169,
+      "grad_norm": 0.054436537230257924,
+      "learning_rate": 5.105294983816203e-06,
+      "loss": 0.44482645988464353,
+      "memory(GiB)": 77.63,
+      "step": 2105,
+      "token_acc": 0.8637377049180328,
+      "train_speed(iter/s)": 0.029517
+    },
+    {
+      "epoch": 0.6751729930802768,
+      "grad_norm": 0.05860088154390529,
+      "learning_rate": 5.060079671326577e-06,
+      "loss": 0.44719686508178713,
+      "memory(GiB)": 77.63,
+      "step": 2110,
+      "token_acc": 0.8593150866058442,
+      "train_speed(iter/s)": 0.029529
+    },
+    {
+      "epoch": 0.6767729290828367,
+      "grad_norm": 0.05264024149284518,
+      "learning_rate": 5.014997559984045e-06,
+      "loss": 0.43972039222717285,
+      "memory(GiB)": 77.63,
+      "step": 2115,
+      "token_acc": 0.8533221194280909,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.6783728650853966,
+      "grad_norm": 0.0534652970629265,
+      "learning_rate": 4.970049865393009e-06,
+      "loss": 0.4468375205993652,
+      "memory(GiB)": 77.63,
+      "step": 2120,
+      "token_acc": 0.8628782287822878,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.6799728010879564,
+      "grad_norm": 0.05246927821047006,
+      "learning_rate": 4.925237799533445e-06,
+      "loss": 0.4498266696929932,
+      "memory(GiB)": 77.63,
+      "step": 2125,
+      "token_acc": 0.9048205760049284,
+      "train_speed(iter/s)": 0.029533
+    },
+    {
+      "epoch": 0.6815727370905164,
+      "grad_norm": 0.05104237083350841,
+      "learning_rate": 4.880562570728188e-06,
+      "loss": 0.4389338970184326,
+      "memory(GiB)": 77.63,
+      "step": 2130,
+      "token_acc": 0.8844444444444445,
+      "train_speed(iter/s)": 0.029525
+    },
+    {
+      "epoch": 0.6831726730930763,
+      "grad_norm": 0.05297787940328326,
+      "learning_rate": 4.836025383610382e-06,
+      "loss": 0.4495584487915039,
+      "memory(GiB)": 77.63,
+      "step": 2135,
+      "token_acc": 0.8647426233038984,
+      "train_speed(iter/s)": 0.029524
+    },
+    {
+      "epoch": 0.6847726090956362,
+      "grad_norm": 0.05092547333787791,
+      "learning_rate": 4.791627439090975e-06,
+      "loss": 0.4421692848205566,
+      "memory(GiB)": 77.63,
+      "step": 2140,
+      "token_acc": 0.8828041384231181,
+      "train_speed(iter/s)": 0.029534
+    },
+    {
+      "epoch": 0.6863725450981961,
+      "grad_norm": 0.053418572817851825,
+      "learning_rate": 4.74736993432634e-06,
+      "loss": 0.44208922386169436,
+      "memory(GiB)": 77.63,
+      "step": 2145,
+      "token_acc": 0.8888520238885202,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.6879724811007559,
+      "grad_norm": 0.053760421496406786,
+      "learning_rate": 4.703254062686017e-06,
+      "loss": 0.4469425201416016,
+      "memory(GiB)": 77.63,
+      "step": 2150,
+      "token_acc": 0.8764145324597975,
+      "train_speed(iter/s)": 0.02953
+    },
+    {
+      "epoch": 0.6895724171033158,
+      "grad_norm": 0.05280613332203591,
+      "learning_rate": 4.6592810137205e-06,
+      "loss": 0.45023741722106936,
+      "memory(GiB)": 77.63,
+      "step": 2155,
+      "token_acc": 0.8968010517090271,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.6911723531058758,
+      "grad_norm": 0.05438679052798784,
+      "learning_rate": 4.615451973129196e-06,
+      "loss": 0.4470167636871338,
+      "memory(GiB)": 77.63,
+      "step": 2160,
+      "token_acc": 0.8761696818465378,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.6927722891084357,
+      "grad_norm": 0.056426544102266905,
+      "learning_rate": 4.571768122728421e-06,
+      "loss": 0.4486443042755127,
+      "memory(GiB)": 77.63,
+      "step": 2165,
+      "token_acc": 0.8781996587030717,
+      "train_speed(iter/s)": 0.029533
+    },
+    {
+      "epoch": 0.6943722251109956,
+      "grad_norm": 0.05337656902490804,
+      "learning_rate": 4.528230640419562e-06,
+      "loss": 0.4497722625732422,
+      "memory(GiB)": 77.63,
+      "step": 2170,
+      "token_acc": 0.8571428571428571,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.6959721611135554,
+      "grad_norm": 0.054129658100736014,
+      "learning_rate": 4.4848407001572945e-06,
+      "loss": 0.44121665954589845,
+      "memory(GiB)": 77.63,
+      "step": 2175,
+      "token_acc": 0.8674278464954012,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.6975720971161153,
+      "grad_norm": 0.05332136363084243,
+      "learning_rate": 4.441599471917946e-06,
+      "loss": 0.43872866630554197,
+      "memory(GiB)": 77.63,
+      "step": 2180,
+      "token_acc": 0.8575067664384652,
+      "train_speed(iter/s)": 0.029529
+    },
+    {
+      "epoch": 0.6991720331186753,
+      "grad_norm": 0.06093731322456081,
+      "learning_rate": 4.398508121667925e-06,
+      "loss": 0.42902402877807616,
+      "memory(GiB)": 77.63,
+      "step": 2185,
+      "token_acc": 0.8526187576126675,
+      "train_speed(iter/s)": 0.029519
+    },
+    {
+      "epoch": 0.7007719691212352,
+      "grad_norm": 0.05959591977220614,
+      "learning_rate": 4.355567811332311e-06,
+      "loss": 0.44504075050354003,
+      "memory(GiB)": 77.63,
+      "step": 2190,
+      "token_acc": 0.8604511878618487,
+      "train_speed(iter/s)": 0.029519
+    },
+    {
+      "epoch": 0.702371905123795,
+      "grad_norm": 0.052994813328955795,
+      "learning_rate": 4.312779698763493e-06,
+      "loss": 0.4408130168914795,
+      "memory(GiB)": 77.63,
+      "step": 2195,
+      "token_acc": 0.8787728847105394,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.7039718411263549,
+      "grad_norm": 0.05402320661185779,
+      "learning_rate": 4.270144937709981e-06,
+      "loss": 0.4396658897399902,
+      "memory(GiB)": 77.63,
+      "step": 2200,
+      "token_acc": 0.8677248677248677,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.7055717771289148,
+      "grad_norm": 0.053270864469091045,
+      "learning_rate": 4.227664677785264e-06,
+      "loss": 0.4493250846862793,
+      "memory(GiB)": 77.63,
+      "step": 2205,
+      "token_acc": 0.8801270542742715,
+      "train_speed(iter/s)": 0.029523
+    },
+    {
+      "epoch": 0.7071717131314748,
+      "grad_norm": 0.06248819511930574,
+      "learning_rate": 4.1853400644368395e-06,
+      "loss": 0.44740095138549807,
+      "memory(GiB)": 77.63,
+      "step": 2210,
+      "token_acc": 0.8988747408942849,
+      "train_speed(iter/s)": 0.029514
+    },
+    {
+      "epoch": 0.7087716491340347,
+      "grad_norm": 0.05807018681919018,
+      "learning_rate": 4.143172238915302e-06,
+      "loss": 0.4508991241455078,
+      "memory(GiB)": 77.63,
+      "step": 2215,
+      "token_acc": 0.8739803562510404,
+      "train_speed(iter/s)": 0.029506
+    },
+    {
+      "epoch": 0.7103715851365945,
+      "grad_norm": 0.05376168037907147,
+      "learning_rate": 4.101162338243595e-06,
+      "loss": 0.4486696243286133,
+      "memory(GiB)": 77.63,
+      "step": 2220,
+      "token_acc": 0.8750778169744761,
+      "train_speed(iter/s)": 0.029514
+    },
+    {
+      "epoch": 0.7119715211391544,
+      "grad_norm": 0.05460944158847028,
+      "learning_rate": 4.059311495186338e-06,
+      "loss": 0.4484865188598633,
+      "memory(GiB)": 77.63,
+      "step": 2225,
+      "token_acc": 0.8524350054924936,
+      "train_speed(iter/s)": 0.029506
+    },
+    {
+      "epoch": 0.7135714571417143,
+      "grad_norm": 0.05716955035585288,
+      "learning_rate": 4.017620838219276e-06,
+      "loss": 0.44258599281311034,
+      "memory(GiB)": 77.63,
+      "step": 2230,
+      "token_acc": 0.8597191629955947,
+      "train_speed(iter/s)": 0.029504
+    },
+    {
+      "epoch": 0.7151713931442742,
+      "grad_norm": 0.05984913995816041,
+      "learning_rate": 3.9760914914988716e-06,
+      "loss": 0.4547589778900146,
+      "memory(GiB)": 77.63,
+      "step": 2235,
+      "token_acc": 0.8679617117117117,
+      "train_speed(iter/s)": 0.029511
+    },
+    {
+      "epoch": 0.7167713291468342,
+      "grad_norm": 0.05686589162715874,
+      "learning_rate": 3.93472457483197e-06,
+      "loss": 0.4416301727294922,
+      "memory(GiB)": 77.63,
+      "step": 2240,
+      "token_acc": 0.826577064816822,
+      "train_speed(iter/s)": 0.029498
+    },
+    {
+      "epoch": 0.718371265149394,
+      "grad_norm": 0.05780707586931182,
+      "learning_rate": 3.893521203645618e-06,
+      "loss": 0.45052361488342285,
+      "memory(GiB)": 77.63,
+      "step": 2245,
+      "token_acc": 0.8836182062608028,
+      "train_speed(iter/s)": 0.0295
+    },
+    {
+      "epoch": 0.7199712011519539,
+      "grad_norm": 0.049110615928360885,
+      "learning_rate": 3.852482488956992e-06,
+      "loss": 0.4427218437194824,
+      "memory(GiB)": 77.63,
+      "step": 2250,
+      "token_acc": 0.8621255642183012,
+      "train_speed(iter/s)": 0.029501
+    },
+    {
+      "epoch": 0.7199712011519539,
+      "eval_loss": 0.659950852394104,
+      "eval_runtime": 108.3142,
+      "eval_samples_per_second": 185.461,
+      "eval_steps_per_second": 0.932,
+      "eval_token_acc": 0.8716289458342705,
+      "step": 2250
+    },
+    {
+      "epoch": 0.7215711371545138,
+      "grad_norm": 0.04899882607919235,
+      "learning_rate": 3.8116095373434204e-06,
+      "loss": 0.4487879753112793,
+      "memory(GiB)": 77.63,
+      "step": 2255,
+      "token_acc": 0.8912671818368324,
+      "train_speed(iter/s)": 0.029473
+    },
+    {
+      "epoch": 0.7231710731570737,
+      "grad_norm": 0.05080548488435112,
+      "learning_rate": 3.7709034509125706e-06,
+      "loss": 0.44452829360961915,
+      "memory(GiB)": 77.63,
+      "step": 2260,
+      "token_acc": 0.8442477876106195,
+      "train_speed(iter/s)": 0.029482
+    },
+    {
+      "epoch": 0.7247710091596337,
+      "grad_norm": 0.048986009146357284,
+      "learning_rate": 3.7303653272727057e-06,
+      "loss": 0.4472095012664795,
+      "memory(GiB)": 77.63,
+      "step": 2265,
+      "token_acc": 0.870567815521944,
+      "train_speed(iter/s)": 0.029495
+    },
+    {
+      "epoch": 0.7263709451621935,
+      "grad_norm": 0.05152412916361422,
+      "learning_rate": 3.689996259503116e-06,
+      "loss": 0.440493106842041,
+      "memory(GiB)": 77.63,
+      "step": 2270,
+      "token_acc": 0.8795436455293181,
+      "train_speed(iter/s)": 0.029499
+    },
+    {
+      "epoch": 0.7279708811647534,
+      "grad_norm": 0.055480142184644934,
+      "learning_rate": 3.6497973361246153e-06,
+      "loss": 0.4417555809020996,
+      "memory(GiB)": 77.63,
+      "step": 2275,
+      "token_acc": 0.8660460713158725,
+      "train_speed(iter/s)": 0.029499
+    },
+    {
+      "epoch": 0.7295708171673133,
+      "grad_norm": 0.05625540509082736,
+      "learning_rate": 3.609769641070221e-06,
+      "loss": 0.4407214164733887,
+      "memory(GiB)": 77.63,
+      "step": 2280,
+      "token_acc": 0.8890608875128999,
+      "train_speed(iter/s)": 0.029513
+    },
+    {
+      "epoch": 0.7311707531698732,
+      "grad_norm": 0.05002588428206622,
+      "learning_rate": 3.569914253655896e-06,
+      "loss": 0.4413386344909668,
+      "memory(GiB)": 77.63,
+      "step": 2285,
+      "token_acc": 0.8921049390319005,
+      "train_speed(iter/s)": 0.029513
+    },
+    {
+      "epoch": 0.7327706891724332,
+      "grad_norm": 0.05088814815973685,
+      "learning_rate": 3.530232248551466e-06,
+      "loss": 0.4507819652557373,
+      "memory(GiB)": 77.63,
+      "step": 2290,
+      "token_acc": 0.8278411830895355,
+      "train_speed(iter/s)": 0.029513
+    },
+    {
+      "epoch": 0.734370625174993,
+      "grad_norm": 0.05399937134620822,
+      "learning_rate": 3.4907246957516416e-06,
+      "loss": 0.4447961330413818,
+      "memory(GiB)": 77.63,
+      "step": 2295,
+      "token_acc": 0.8888263967004124,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.7359705611775529,
+      "grad_norm": 0.06200035405309708,
+      "learning_rate": 3.4513926605471504e-06,
+      "loss": 0.45868444442749023,
+      "memory(GiB)": 77.63,
+      "step": 2300,
+      "token_acc": 0.8513141426783479,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.7375704971801128,
+      "grad_norm": 0.055876255912378235,
+      "learning_rate": 3.412237203496036e-06,
+      "loss": 0.4431456089019775,
+      "memory(GiB)": 77.63,
+      "step": 2305,
+      "token_acc": 0.8651997041420119,
+      "train_speed(iter/s)": 0.029532
+    },
+    {
+      "epoch": 0.7391704331826727,
+      "grad_norm": 0.06032844036632358,
+      "learning_rate": 3.3732593803950354e-06,
+      "loss": 0.4452229976654053,
+      "memory(GiB)": 77.63,
+      "step": 2310,
+      "token_acc": 0.8915232899706252,
+      "train_speed(iter/s)": 0.02953
+    },
+    {
+      "epoch": 0.7407703691852325,
+      "grad_norm": 0.05255216039270682,
+      "learning_rate": 3.3344602422511343e-06,
+      "loss": 0.4414207458496094,
+      "memory(GiB)": 77.63,
+      "step": 2315,
+      "token_acc": 0.8901802257032171,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.7423703051877925,
+      "grad_norm": 0.05535966142690852,
+      "learning_rate": 3.2958408352532055e-06,
+      "loss": 0.43938393592834474,
+      "memory(GiB)": 77.63,
+      "step": 2320,
+      "token_acc": 0.8354077253218885,
+      "train_speed(iter/s)": 0.029536
+    },
+    {
+      "epoch": 0.7439702411903524,
+      "grad_norm": 0.0505418855319798,
+      "learning_rate": 3.257402200743821e-06,
+      "loss": 0.44445362091064455,
+      "memory(GiB)": 77.63,
+      "step": 2325,
+      "token_acc": 0.864262790258637,
+      "train_speed(iter/s)": 0.02953
+    },
+    {
+      "epoch": 0.7455701771929123,
+      "grad_norm": 0.0563107101835597,
+      "learning_rate": 3.2191453751911505e-06,
+      "loss": 0.45569453239440916,
+      "memory(GiB)": 77.63,
+      "step": 2330,
+      "token_acc": 0.8784313725490196,
+      "train_speed(iter/s)": 0.029523
+    },
+    {
+      "epoch": 0.7471701131954722,
+      "grad_norm": 0.05000152007266613,
+      "learning_rate": 3.1810713901610367e-06,
+      "loss": 0.4395348072052002,
+      "memory(GiB)": 77.63,
+      "step": 2335,
+      "token_acc": 0.8867111781175964,
+      "train_speed(iter/s)": 0.029536
+    },
+    {
+      "epoch": 0.748770049198032,
+      "grad_norm": 0.057169590375126145,
+      "learning_rate": 3.1431812722891598e-06,
+      "loss": 0.4397278785705566,
+      "memory(GiB)": 77.63,
+      "step": 2340,
+      "token_acc": 0.8577532891037895,
+      "train_speed(iter/s)": 0.029529
+    },
+    {
+      "epoch": 0.750369985200592,
+      "grad_norm": 0.05730745865195846,
+      "learning_rate": 3.1054760432533626e-06,
+      "loss": 0.45998029708862304,
+      "memory(GiB)": 77.63,
+      "step": 2345,
+      "token_acc": 0.8845755097339016,
+      "train_speed(iter/s)": 0.029534
+    },
+    {
+      "epoch": 0.7519699212031519,
+      "grad_norm": 0.05180470840824953,
+      "learning_rate": 3.0679567197461135e-06,
+      "loss": 0.45008273124694825,
+      "memory(GiB)": 77.63,
+      "step": 2350,
+      "token_acc": 0.8394425931535898,
+      "train_speed(iter/s)": 0.029537
+    },
+    {
+      "epoch": 0.7535698572057118,
+      "grad_norm": 0.06025883780673481,
+      "learning_rate": 3.0306243134470668e-06,
+      "loss": 0.4444745540618896,
+      "memory(GiB)": 77.63,
+      "step": 2355,
+      "token_acc": 0.889631386074585,
+      "train_speed(iter/s)": 0.02953
+    },
+    {
+      "epoch": 0.7551697932082717,
+      "grad_norm": 0.05199872680450009,
+      "learning_rate": 2.993479830995815e-06,
+      "loss": 0.451768159866333,
+      "memory(GiB)": 77.63,
+      "step": 2360,
+      "token_acc": 0.8736520199581522,
+      "train_speed(iter/s)": 0.029541
+    },
+    {
+      "epoch": 0.7567697292108315,
+      "grad_norm": 0.05489027404588469,
+      "learning_rate": 2.9565242739647115e-06,
+      "loss": 0.4442115306854248,
+      "memory(GiB)": 77.63,
+      "step": 2365,
+      "token_acc": 0.8865552903739061,
+      "train_speed(iter/s)": 0.029538
+    },
+    {
+      "epoch": 0.7583696652133914,
+      "grad_norm": 0.06334021131539457,
+      "learning_rate": 2.919758638831893e-06,
+      "loss": 0.4570741653442383,
+      "memory(GiB)": 77.63,
+      "step": 2370,
+      "token_acc": 0.8652606912712361,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.7599696012159514,
+      "grad_norm": 0.053831666314624105,
+      "learning_rate": 2.8831839169543998e-06,
+      "loss": 0.44495415687561035,
+      "memory(GiB)": 77.63,
+      "step": 2375,
+      "token_acc": 0.8756476683937824,
+      "train_speed(iter/s)": 0.029541
+    },
+    {
+      "epoch": 0.7615695372185113,
+      "grad_norm": 0.0527583973457582,
+      "learning_rate": 2.84680109454143e-06,
+      "loss": 0.4472104549407959,
+      "memory(GiB)": 77.63,
+      "step": 2380,
+      "token_acc": 0.8725328947368421,
+      "train_speed(iter/s)": 0.029536
+    },
+    {
+      "epoch": 0.7631694732210712,
+      "grad_norm": 0.058941021305098804,
+      "learning_rate": 2.810611152627777e-06,
+      "loss": 0.4499720573425293,
+      "memory(GiB)": 77.63,
+      "step": 2385,
+      "token_acc": 0.8632213889794588,
+      "train_speed(iter/s)": 0.029535
+    },
+    {
+      "epoch": 0.764769409223631,
+      "grad_norm": 0.05393008170855123,
+      "learning_rate": 2.774615067047346e-06,
+      "loss": 0.43872222900390623,
+      "memory(GiB)": 77.63,
+      "step": 2390,
+      "token_acc": 0.8742202234150588,
+      "train_speed(iter/s)": 0.02954
+    },
+    {
+      "epoch": 0.7663693452261909,
+      "grad_norm": 0.0556417645809335,
+      "learning_rate": 2.738813808406866e-06,
+      "loss": 0.4399220943450928,
+      "memory(GiB)": 77.63,
+      "step": 2395,
+      "token_acc": 0.8997599039615847,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.7679692812287509,
+      "grad_norm": 0.05379432234404155,
+      "learning_rate": 2.7032083420597e-06,
+      "loss": 0.4382453441619873,
+      "memory(GiB)": 77.63,
+      "step": 2400,
+      "token_acc": 0.8875784668061633,
+      "train_speed(iter/s)": 0.029541
+    },
+    {
+      "epoch": 0.7695692172313108,
+      "grad_norm": 0.05806842163630925,
+      "learning_rate": 2.667799628079829e-06,
+      "loss": 0.44454326629638674,
+      "memory(GiB)": 77.63,
+      "step": 2405,
+      "token_acc": 0.8880662020905923,
+      "train_speed(iter/s)": 0.029535
+    },
+    {
+      "epoch": 0.7711691532338707,
+      "grad_norm": 0.06148704112133217,
+      "learning_rate": 2.6325886212359496e-06,
+      "loss": 0.43945813179016113,
+      "memory(GiB)": 77.63,
+      "step": 2410,
+      "token_acc": 0.8767772511848341,
+      "train_speed(iter/s)": 0.029527
+    },
+    {
+      "epoch": 0.7727690892364305,
+      "grad_norm": 0.056530065759685846,
+      "learning_rate": 2.5975762709657506e-06,
+      "loss": 0.4438450813293457,
+      "memory(GiB)": 77.63,
+      "step": 2415,
+      "token_acc": 0.8570975416336241,
+      "train_speed(iter/s)": 0.029538
+    },
+    {
+      "epoch": 0.7743690252389904,
+      "grad_norm": 0.056107845444701834,
+      "learning_rate": 2.5627635213502832e-06,
+      "loss": 0.43836054801940916,
+      "memory(GiB)": 77.63,
+      "step": 2420,
+      "token_acc": 0.8966822253059165,
+      "train_speed(iter/s)": 0.029532
+    },
+    {
+      "epoch": 0.7759689612415503,
+      "grad_norm": 0.05796065696017405,
+      "learning_rate": 2.528151311088537e-06,
+      "loss": 0.4400279998779297,
+      "memory(GiB)": 77.63,
+      "step": 2425,
+      "token_acc": 0.8552805280528053,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.7775688972441103,
+      "grad_norm": 0.05418546630146028,
+      "learning_rate": 2.4937405734720964e-06,
+      "loss": 0.44541444778442385,
+      "memory(GiB)": 77.63,
+      "step": 2430,
+      "token_acc": 0.8620764552562988,
+      "train_speed(iter/s)": 0.029532
+    },
+    {
+      "epoch": 0.7791688332466701,
+      "grad_norm": 0.05654250831277805,
+      "learning_rate": 2.459532236360007e-06,
+      "loss": 0.43491110801696775,
+      "memory(GiB)": 77.63,
+      "step": 2435,
+      "token_acc": 0.8570184983677911,
+      "train_speed(iter/s)": 0.029522
+    },
+    {
+      "epoch": 0.78076876924923,
+      "grad_norm": 0.05526372242621089,
+      "learning_rate": 2.4255272221537295e-06,
+      "loss": 0.4378859043121338,
+      "memory(GiB)": 77.63,
+      "step": 2440,
+      "token_acc": 0.8631236857197476,
+      "train_speed(iter/s)": 0.029526
+    },
+    {
+      "epoch": 0.7823687052517899,
+      "grad_norm": 0.05404315424969483,
+      "learning_rate": 2.391726447772279e-06,
+      "loss": 0.45857391357421873,
+      "memory(GiB)": 77.63,
+      "step": 2445,
+      "token_acc": 0.8634816932081122,
+      "train_speed(iter/s)": 0.029527
+    },
+    {
+      "epoch": 0.7839686412543498,
+      "grad_norm": 0.05765113554061621,
+      "learning_rate": 2.3581308246275103e-06,
+      "loss": 0.4473139762878418,
+      "memory(GiB)": 77.63,
+      "step": 2450,
+      "token_acc": 0.8979846898922044,
+      "train_speed(iter/s)": 0.029518
+    },
+    {
+      "epoch": 0.7855685772569098,
+      "grad_norm": 0.058353221842389495,
+      "learning_rate": 2.324741258599521e-06,
+      "loss": 0.44444866180419923,
+      "memory(GiB)": 77.63,
+      "step": 2455,
+      "token_acc": 0.8648913576213038,
+      "train_speed(iter/s)": 0.029527
+    },
+    {
+      "epoch": 0.7871685132594696,
+      "grad_norm": 0.05309289512529438,
+      "learning_rate": 2.29155865001225e-06,
+      "loss": 0.43857607841491697,
+      "memory(GiB)": 77.63,
+      "step": 2460,
+      "token_acc": 0.894580549368968,
+      "train_speed(iter/s)": 0.029521
+    },
+    {
+      "epoch": 0.7887684492620295,
+      "grad_norm": 0.0540412312893473,
+      "learning_rate": 2.2585838936091753e-06,
+      "loss": 0.43953213691711424,
+      "memory(GiB)": 77.63,
+      "step": 2465,
+      "token_acc": 0.8868672731513879,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.7903683852645894,
+      "grad_norm": 0.05880520800742855,
+      "learning_rate": 2.225817878529214e-06,
+      "loss": 0.4457580089569092,
+      "memory(GiB)": 77.63,
+      "step": 2470,
+      "token_acc": 0.8630282437884901,
+      "train_speed(iter/s)": 0.029525
+    },
+    {
+      "epoch": 0.7919683212671493,
+      "grad_norm": 0.058328885138964066,
+      "learning_rate": 2.1932614882827196e-06,
+      "loss": 0.4424918174743652,
+      "memory(GiB)": 77.63,
+      "step": 2475,
+      "token_acc": 0.8814697747925722,
+      "train_speed(iter/s)": 0.029517
+    },
+    {
+      "epoch": 0.7935682572697093,
+      "grad_norm": 0.05685263085809571,
+      "learning_rate": 2.160915600727688e-06,
+      "loss": 0.43921732902526855,
+      "memory(GiB)": 77.63,
+      "step": 2480,
+      "token_acc": 0.913681738109219,
+      "train_speed(iter/s)": 0.029516
+    },
+    {
+      "epoch": 0.7951681932722691,
+      "grad_norm": 0.056639845561812056,
+      "learning_rate": 2.1287810880460636e-06,
+      "loss": 0.44060502052307127,
+      "memory(GiB)": 77.63,
+      "step": 2485,
+      "token_acc": 0.8829075425790754,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.796768129274829,
+      "grad_norm": 0.05230960490349676,
+      "learning_rate": 2.0968588167202265e-06,
+      "loss": 0.43935480117797854,
+      "memory(GiB)": 77.63,
+      "step": 2490,
+      "token_acc": 0.8856997455470738,
+      "train_speed(iter/s)": 0.029511
+    },
+    {
+      "epoch": 0.7983680652773889,
+      "grad_norm": 0.05305183045142263,
+      "learning_rate": 2.0651496475096455e-06,
+      "loss": 0.4360368728637695,
+      "memory(GiB)": 77.63,
+      "step": 2495,
+      "token_acc": 0.8394655704008221,
+      "train_speed(iter/s)": 0.029517
+    },
+    {
+      "epoch": 0.7999680012799488,
+      "grad_norm": 0.05620012484228566,
+      "learning_rate": 2.03365443542764e-06,
+      "loss": 0.44507203102111814,
+      "memory(GiB)": 77.63,
+      "step": 2500,
+      "token_acc": 0.8857098429482195,
+      "train_speed(iter/s)": 0.029515
+    },
+    {
+      "epoch": 0.7999680012799488,
+      "eval_loss": 0.6586322784423828,
+      "eval_runtime": 105.1966,
+      "eval_samples_per_second": 190.957,
+      "eval_steps_per_second": 0.96,
+      "eval_token_acc": 0.8721292963419328,
+      "step": 2500
+    },
+    {
+      "epoch": 0.8015679372825087,
+      "grad_norm": 0.05541548293022976,
+      "learning_rate": 2.0023740297183536e-06,
+      "loss": 0.44654192924499514,
+      "memory(GiB)": 77.63,
+      "step": 2505,
+      "token_acc": 0.8819252077562327,
+      "train_speed(iter/s)": 0.029489
+    },
+    {
+      "epoch": 0.8031678732850686,
+      "grad_norm": 0.057374199419424524,
+      "learning_rate": 1.971309273833828e-06,
+      "loss": 0.44596128463745116,
+      "memory(GiB)": 77.63,
+      "step": 2510,
+      "token_acc": 0.855553561815898,
+      "train_speed(iter/s)": 0.029494
+    },
+    {
+      "epoch": 0.8047678092876285,
+      "grad_norm": 0.05297595418967434,
+      "learning_rate": 1.940461005411288e-06,
+      "loss": 0.45099148750305174,
+      "memory(GiB)": 77.63,
+      "step": 2515,
+      "token_acc": 0.8958361962347121,
+      "train_speed(iter/s)": 0.029503
+    },
+    {
+      "epoch": 0.8063677452901884,
+      "grad_norm": 0.056714681257095015,
+      "learning_rate": 1.9098300562505266e-06,
+      "loss": 0.4423669338226318,
+      "memory(GiB)": 77.63,
+      "step": 2520,
+      "token_acc": 0.879045996592845,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.8079676812927483,
+      "grad_norm": 0.05868047347236466,
+      "learning_rate": 1.8794172522915022e-06,
+      "loss": 0.4462554931640625,
+      "memory(GiB)": 77.63,
+      "step": 2525,
+      "token_acc": 0.8811320754716981,
+      "train_speed(iter/s)": 0.029507
+    },
+    {
+      "epoch": 0.8095676172953082,
+      "grad_norm": 0.04860202133053876,
+      "learning_rate": 1.849223413592046e-06,
+      "loss": 0.4488513946533203,
+      "memory(GiB)": 77.63,
+      "step": 2530,
+      "token_acc": 0.8654490616621984,
+      "train_speed(iter/s)": 0.029517
+    },
+    {
+      "epoch": 0.811167553297868,
+      "grad_norm": 0.055005201314914515,
+      "learning_rate": 1.8192493543057676e-06,
+      "loss": 0.45094904899597166,
+      "memory(GiB)": 77.63,
+      "step": 2535,
+      "token_acc": 0.8800874078120732,
+      "train_speed(iter/s)": 0.029516
+    },
+    {
+      "epoch": 0.812767489300428,
+      "grad_norm": 0.058350813115584405,
+      "learning_rate": 1.7894958826600884e-06,
+      "loss": 0.4489152908325195,
+      "memory(GiB)": 77.63,
+      "step": 2540,
+      "token_acc": 0.8784655623365301,
+      "train_speed(iter/s)": 0.02952
+    },
+    {
+      "epoch": 0.8143674253029879,
+      "grad_norm": 0.05243037926696882,
+      "learning_rate": 1.7599638009344566e-06,
+      "loss": 0.4506648063659668,
+      "memory(GiB)": 77.63,
+      "step": 2545,
+      "token_acc": 0.8711162255466053,
+      "train_speed(iter/s)": 0.029531
+    },
+    {
+      "epoch": 0.8159673613055478,
+      "grad_norm": 0.055874525296942985,
+      "learning_rate": 1.730653905438714e-06,
+      "loss": 0.451121187210083,
+      "memory(GiB)": 77.63,
+      "step": 2550,
+      "token_acc": 0.875531914893617,
+      "train_speed(iter/s)": 0.029527
+    },
+    {
+      "epoch": 0.8175672973081076,
+      "grad_norm": 0.052382476873803714,
+      "learning_rate": 1.701566986491614e-06,
+      "loss": 0.43659415245056155,
+      "memory(GiB)": 77.63,
+      "step": 2555,
+      "token_acc": 0.8824301518844928,
+      "train_speed(iter/s)": 0.029539
+    },
+    {
+      "epoch": 0.8191672333106675,
+      "grad_norm": 0.05679998544269888,
+      "learning_rate": 1.672703828399529e-06,
+      "loss": 0.44143290519714357,
+      "memory(GiB)": 77.63,
+      "step": 2560,
+      "token_acc": 0.9194786645241921,
+      "train_speed(iter/s)": 0.029538
+    },
+    {
+      "epoch": 0.8207671693132275,
+      "grad_norm": 0.05177151194489443,
+      "learning_rate": 1.6440652094352838e-06,
+      "loss": 0.44036478996276857,
+      "memory(GiB)": 77.63,
+      "step": 2565,
+      "token_acc": 0.86801315171442,
+      "train_speed(iter/s)": 0.029534
+    },
+    {
+      "epoch": 0.8223671053157874,
+      "grad_norm": 0.047985617439266506,
+      "learning_rate": 1.6156519018171856e-06,
+      "loss": 0.44090909957885743,
+      "memory(GiB)": 77.63,
+      "step": 2570,
+      "token_acc": 0.8926761055759482,
+      "train_speed(iter/s)": 0.029546
+    },
+    {
+      "epoch": 0.8239670413183473,
+      "grad_norm": 0.06022869670658613,
+      "learning_rate": 1.587464671688187e-06,
+      "loss": 0.4480876922607422,
+      "memory(GiB)": 77.63,
+      "step": 2575,
+      "token_acc": 0.8749468913751594,
+      "train_speed(iter/s)": 0.029542
+    },
+    {
+      "epoch": 0.8255669773209071,
+      "grad_norm": 0.05591525813204934,
+      "learning_rate": 1.5595042790952442e-06,
+      "loss": 0.4516183853149414,
+      "memory(GiB)": 77.63,
+      "step": 2580,
+      "token_acc": 0.8408594319009468,
+      "train_speed(iter/s)": 0.029539
+    },
+    {
+      "epoch": 0.827166913323467,
+      "grad_norm": 0.05315795994218538,
+      "learning_rate": 1.5317714779688076e-06,
+      "loss": 0.44116387367248533,
+      "memory(GiB)": 77.63,
+      "step": 2585,
+      "token_acc": 0.8697747394374089,
+      "train_speed(iter/s)": 0.029549
+    },
+    {
+      "epoch": 0.828766849326027,
+      "grad_norm": 0.054322305095737905,
+      "learning_rate": 1.5042670161024975e-06,
+      "loss": 0.4457075119018555,
+      "memory(GiB)": 77.63,
+      "step": 2590,
+      "token_acc": 0.8946940985381701,
+      "train_speed(iter/s)": 0.029543
+    },
+    {
+      "epoch": 0.8303667853285869,
+      "grad_norm": 0.05625328099370699,
+      "learning_rate": 1.4769916351329495e-06,
+      "loss": 0.4413478851318359,
+      "memory(GiB)": 77.63,
+      "step": 2595,
+      "token_acc": 0.8992218637312583,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.8319667213311468,
+      "grad_norm": 0.055362561635933255,
+      "learning_rate": 1.4499460705198e-06,
+      "loss": 0.4511932373046875,
+      "memory(GiB)": 77.63,
+      "step": 2600,
+      "token_acc": 0.8438552188552189,
+      "train_speed(iter/s)": 0.029552
+    },
+    {
+      "epoch": 0.8335666573337066,
+      "grad_norm": 0.052438473367666195,
+      "learning_rate": 1.4231310515258745e-06,
+      "loss": 0.441973352432251,
+      "memory(GiB)": 77.63,
+      "step": 2605,
+      "token_acc": 0.8753952017853822,
+      "train_speed(iter/s)": 0.029546
+    },
+    {
+      "epoch": 0.8351665933362665,
+      "grad_norm": 0.05274974730016364,
+      "learning_rate": 1.396547301197504e-06,
+      "loss": 0.4393311977386475,
+      "memory(GiB)": 77.63,
+      "step": 2610,
+      "token_acc": 0.8518848700967906,
+      "train_speed(iter/s)": 0.029557
+    },
+    {
+      "epoch": 0.8367665293388264,
+      "grad_norm": 0.05853294004614818,
+      "learning_rate": 1.3701955363450447e-06,
+      "loss": 0.4380232810974121,
+      "memory(GiB)": 77.63,
+      "step": 2615,
+      "token_acc": 0.8570597362296354,
+      "train_speed(iter/s)": 0.029554
+    },
+    {
+      "epoch": 0.8383664653413864,
+      "grad_norm": 0.05410978790127522,
+      "learning_rate": 1.3440764675235384e-06,
+      "loss": 0.4373164653778076,
+      "memory(GiB)": 77.63,
+      "step": 2620,
+      "token_acc": 0.8798353909465021,
+      "train_speed(iter/s)": 0.029552
+    },
+    {
+      "epoch": 0.8399664013439463,
+      "grad_norm": 0.048967955063799855,
+      "learning_rate": 1.3181907990135624e-06,
+      "loss": 0.4333020210266113,
+      "memory(GiB)": 77.63,
+      "step": 2625,
+      "token_acc": 0.8836341008089608,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.8415663373465061,
+      "grad_norm": 0.048274089580157754,
+      "learning_rate": 1.2925392288022299e-06,
+      "loss": 0.4414947509765625,
+      "memory(GiB)": 77.63,
+      "step": 2630,
+      "token_acc": 0.8760546404178385,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.843166273349066,
+      "grad_norm": 0.053684823491607934,
+      "learning_rate": 1.267122448564374e-06,
+      "loss": 0.44922800064086915,
+      "memory(GiB)": 77.63,
+      "step": 2635,
+      "token_acc": 0.8554064052425748,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.8447662093516259,
+      "grad_norm": 0.05262528572569429,
+      "learning_rate": 1.2419411436439021e-06,
+      "loss": 0.4328805923461914,
+      "memory(GiB)": 77.63,
+      "step": 2640,
+      "token_acc": 0.8400081317340923,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.8463661453541859,
+      "grad_norm": 0.05549273276976771,
+      "learning_rate": 1.2169959930353049e-06,
+      "loss": 0.4460554599761963,
+      "memory(GiB)": 77.63,
+      "step": 2645,
+      "token_acc": 0.8804424157303371,
+      "train_speed(iter/s)": 0.029559
+    },
+    {
+      "epoch": 0.8479660813567458,
+      "grad_norm": 0.05180537064402683,
+      "learning_rate": 1.1922876693653584e-06,
+      "loss": 0.4503427505493164,
+      "memory(GiB)": 77.63,
+      "step": 2650,
+      "token_acc": 0.8934362934362934,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.8495660173593056,
+      "grad_norm": 0.05011013702559624,
+      "learning_rate": 1.1678168388749788e-06,
+      "loss": 0.4415099620819092,
+      "memory(GiB)": 77.63,
+      "step": 2655,
+      "token_acc": 0.8995949690897463,
+      "train_speed(iter/s)": 0.029566
+    },
+    {
+      "epoch": 0.8511659533618655,
+      "grad_norm": 0.057170743989864214,
+      "learning_rate": 1.1435841614012666e-06,
+      "loss": 0.44884433746337893,
+      "memory(GiB)": 77.63,
+      "step": 2660,
+      "token_acc": 0.854816112084063,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.8527658893644254,
+      "grad_norm": 0.054388272142607975,
+      "learning_rate": 1.1195902903597023e-06,
+      "loss": 0.439667797088623,
+      "memory(GiB)": 77.63,
+      "step": 2665,
+      "token_acc": 0.8846260387811634,
+      "train_speed(iter/s)": 0.02957
+    },
+    {
+      "epoch": 0.8543658253669854,
+      "grad_norm": 0.05039207218424233,
+      "learning_rate": 1.0958358727265438e-06,
+      "loss": 0.4384475231170654,
+      "memory(GiB)": 77.63,
+      "step": 2670,
+      "token_acc": 0.8525793222533995,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.8559657613695452,
+      "grad_norm": 0.0543645938545219,
+      "learning_rate": 1.0723215490213635e-06,
+      "loss": 0.4338691711425781,
+      "memory(GiB)": 77.63,
+      "step": 2675,
+      "token_acc": 0.853824495541999,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.8575656973721051,
+      "grad_norm": 0.0579168704227633,
+      "learning_rate": 1.0490479532897946e-06,
+      "loss": 0.458463716506958,
+      "memory(GiB)": 77.63,
+      "step": 2680,
+      "token_acc": 0.867092866756393,
+      "train_speed(iter/s)": 0.029566
+    },
+    {
+      "epoch": 0.859165633374665,
+      "grad_norm": 0.04996147776053655,
+      "learning_rate": 1.0260157130864178e-06,
+      "loss": 0.43754091262817385,
+      "memory(GiB)": 77.63,
+      "step": 2685,
+      "token_acc": 0.8611873713109128,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.8607655693772249,
+      "grad_norm": 0.053537353272037014,
+      "learning_rate": 1.0032254494578519e-06,
+      "loss": 0.44204487800598147,
+      "memory(GiB)": 77.63,
+      "step": 2690,
+      "token_acc": 0.8575780654988576,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.8623655053797848,
+      "grad_norm": 0.055119127073411836,
+      "learning_rate": 9.806777769260034e-07,
+      "loss": 0.4500781536102295,
+      "memory(GiB)": 77.63,
+      "step": 2695,
+      "token_acc": 0.8872294372294373,
+      "train_speed(iter/s)": 0.029562
+    },
+    {
+      "epoch": 0.8639654413823447,
+      "grad_norm": 0.0558713442289911,
+      "learning_rate": 9.583733034714982e-07,
+      "loss": 0.43947248458862304,
+      "memory(GiB)": 77.63,
+      "step": 2700,
+      "token_acc": 0.8926744522729466,
+      "train_speed(iter/s)": 0.029555
+    },
+    {
+      "epoch": 0.8655653773849046,
+      "grad_norm": 0.0552196144062876,
+      "learning_rate": 9.363126305172831e-07,
+      "loss": 0.4443229675292969,
+      "memory(GiB)": 77.63,
+      "step": 2705,
+      "token_acc": 0.9038031319910514,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.8671653133874645,
+      "grad_norm": 0.055723745057826034,
+      "learning_rate": 9.144963529124163e-07,
+      "loss": 0.42942004203796386,
+      "memory(GiB)": 77.63,
+      "step": 2710,
+      "token_acc": 0.8680161943319838,
+      "train_speed(iter/s)": 0.029557
+    },
+    {
+      "epoch": 0.8687652493900244,
+      "grad_norm": 0.0589746686821641,
+      "learning_rate": 8.929250589160166e-07,
+      "loss": 0.4397599220275879,
+      "memory(GiB)": 77.63,
+      "step": 2715,
+      "token_acc": 0.8713878713878714,
+      "train_speed(iter/s)": 0.029552
+    },
+    {
+      "epoch": 0.8703651853925843,
+      "grad_norm": 0.04909314017257213,
+      "learning_rate": 8.715993301814174e-07,
+      "loss": 0.44155421257019045,
+      "memory(GiB)": 77.63,
+      "step": 2720,
+      "token_acc": 0.8710053650571495,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.8719651213951441,
+      "grad_norm": 0.05047518453544575,
+      "learning_rate": 8.505197417404687e-07,
+      "loss": 0.43677616119384766,
+      "memory(GiB)": 77.63,
+      "step": 2725,
+      "token_acc": 0.8809886575249704,
+      "train_speed(iter/s)": 0.029556
+    },
+    {
+      "epoch": 0.8735650573977041,
+      "grad_norm": 0.05102151204327215,
+      "learning_rate": 8.296868619880372e-07,
+      "loss": 0.44188566207885743,
+      "memory(GiB)": 77.63,
+      "step": 2730,
+      "token_acc": 0.8547172833573602,
+      "train_speed(iter/s)": 0.029553
+    },
+    {
+      "epoch": 0.875164993400264,
+      "grad_norm": 0.04729834705444575,
+      "learning_rate": 8.091012526666797e-07,
+      "loss": 0.4441237926483154,
+      "memory(GiB)": 77.63,
+      "step": 2735,
+      "token_acc": 0.8537975972307066,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.8767649294028239,
+      "grad_norm": 0.047668539210598965,
+      "learning_rate": 7.887634688515e-07,
+      "loss": 0.4462736129760742,
+      "memory(GiB)": 77.63,
+      "step": 2740,
+      "token_acc": 0.903437815975733,
+      "train_speed(iter/s)": 0.029554
+    },
+    {
+      "epoch": 0.8783648654053838,
+      "grad_norm": 0.052216823887528664,
+      "learning_rate": 7.686740589351704e-07,
+      "loss": 0.44857120513916016,
+      "memory(GiB)": 77.63,
+      "step": 2745,
+      "token_acc": 0.8033573141486811,
+      "train_speed(iter/s)": 0.029556
+    },
+    {
+      "epoch": 0.8799648014079436,
+      "grad_norm": 0.055862979558343906,
+      "learning_rate": 7.488335646131628e-07,
+      "loss": 0.44959425926208496,
+      "memory(GiB)": 77.63,
+      "step": 2750,
+      "token_acc": 0.8605054151624548,
+      "train_speed(iter/s)": 0.029558
+    },
+    {
+      "epoch": 0.8799648014079436,
+      "eval_loss": 0.6577034592628479,
+      "eval_runtime": 106.4875,
+      "eval_samples_per_second": 188.642,
+      "eval_steps_per_second": 0.948,
+      "eval_token_acc": 0.8722958612553617,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8815647374105036,
+      "grad_norm": 0.051528350081992934,
+      "learning_rate": 7.292425208691212e-07,
+      "loss": 0.43878631591796874,
+      "memory(GiB)": 77.63,
+      "step": 2755,
+      "token_acc": 0.8812832745626772,
+      "train_speed(iter/s)": 0.029532
+    },
+    {
+      "epoch": 0.8831646734130635,
+      "grad_norm": 0.05310175482611414,
+      "learning_rate": 7.099014559604556e-07,
+      "loss": 0.45635418891906737,
+      "memory(GiB)": 77.63,
+      "step": 2760,
+      "token_acc": 0.8894999360532038,
+      "train_speed(iter/s)": 0.029537
+    },
+    {
+      "epoch": 0.8847646094156234,
+      "grad_norm": 0.04975738369955541,
+      "learning_rate": 6.908108914040823e-07,
+      "loss": 0.4421397686004639,
+      "memory(GiB)": 77.63,
+      "step": 2765,
+      "token_acc": 0.9070493575117089,
+      "train_speed(iter/s)": 0.029548
+    },
+    {
+      "epoch": 0.8863645454181833,
+      "grad_norm": 0.053564472486824076,
+      "learning_rate": 6.71971341962373e-07,
+      "loss": 0.4513510227203369,
+      "memory(GiB)": 77.63,
+      "step": 2770,
+      "token_acc": 0.8660589060308556,
+      "train_speed(iter/s)": 0.02955
+    },
+    {
+      "epoch": 0.8879644814207431,
+      "grad_norm": 0.06332925713320489,
+      "learning_rate": 6.53383315629268e-07,
+      "loss": 0.4404273509979248,
+      "memory(GiB)": 77.63,
+      "step": 2775,
+      "token_acc": 0.8507806501151779,
+      "train_speed(iter/s)": 0.029546
+    },
+    {
+      "epoch": 0.889564417423303,
+      "grad_norm": 0.063294227744794,
+      "learning_rate": 6.350473136165836e-07,
+      "loss": 0.4379493236541748,
+      "memory(GiB)": 77.63,
+      "step": 2780,
+      "token_acc": 0.8879898461050294,
+      "train_speed(iter/s)": 0.029561
+    },
+    {
+      "epoch": 0.891164353425863,
+      "grad_norm": 0.05151642870451994,
+      "learning_rate": 6.169638303404912e-07,
+      "loss": 0.4380655765533447,
+      "memory(GiB)": 77.63,
+      "step": 2785,
+      "token_acc": 0.8904059040590406,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.8927642894284229,
+      "grad_norm": 0.05406168921762394,
+      "learning_rate": 5.991333534081878e-07,
+      "loss": 0.4479250907897949,
+      "memory(GiB)": 77.63,
+      "step": 2790,
+      "token_acc": 0.8831118813787792,
+      "train_speed(iter/s)": 0.02956
+    },
+    {
+      "epoch": 0.8943642254309827,
+      "grad_norm": 0.054911478200183335,
+      "learning_rate": 5.815563636047539e-07,
+      "loss": 0.43634886741638185,
+      "memory(GiB)": 77.63,
+      "step": 2795,
+      "token_acc": 0.8714865708931917,
+      "train_speed(iter/s)": 0.02957
+    },
+    {
+      "epoch": 0.8959641614335426,
+      "grad_norm": 0.05257509941727236,
+      "learning_rate": 5.64233334880181e-07,
+      "loss": 0.44048466682434084,
+      "memory(GiB)": 77.63,
+      "step": 2800,
+      "token_acc": 0.891296869625043,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.8975640974361025,
+      "grad_norm": 0.05532206374459385,
+      "learning_rate": 5.471647343365982e-07,
+      "loss": 0.44726853370666503,
+      "memory(GiB)": 77.63,
+      "step": 2805,
+      "token_acc": 0.8828892005610098,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.8991640334386625,
+      "grad_norm": 0.05107208389411162,
+      "learning_rate": 5.303510222156716e-07,
+      "loss": 0.4470540523529053,
+      "memory(GiB)": 77.63,
+      "step": 2810,
+      "token_acc": 0.8489765812281025,
+      "train_speed(iter/s)": 0.029574
+    },
+    {
+      "epoch": 0.9007639694412224,
+      "grad_norm": 0.055452171728558604,
+      "learning_rate": 5.137926518862013e-07,
+      "loss": 0.4417248249053955,
+      "memory(GiB)": 77.63,
+      "step": 2815,
+      "token_acc": 0.8739084132055378,
+      "train_speed(iter/s)": 0.029569
+    },
+    {
+      "epoch": 0.9023639054437822,
+      "grad_norm": 0.04866974543895629,
+      "learning_rate": 4.974900698318885e-07,
+      "loss": 0.4414045810699463,
+      "memory(GiB)": 77.63,
+      "step": 2820,
+      "token_acc": 0.8659420289855072,
+      "train_speed(iter/s)": 0.029581
+    },
+    {
+      "epoch": 0.9039638414463421,
+      "grad_norm": 0.05275403785935388,
+      "learning_rate": 4.814437156393048e-07,
+      "loss": 0.4543337821960449,
+      "memory(GiB)": 77.63,
+      "step": 2825,
+      "token_acc": 0.8396122896854425,
+      "train_speed(iter/s)": 0.029574
+    },
+    {
+      "epoch": 0.905563777448902,
+      "grad_norm": 0.05372217346495556,
+      "learning_rate": 4.656540219860317e-07,
+      "loss": 0.45271754264831543,
+      "memory(GiB)": 77.63,
+      "step": 2830,
+      "token_acc": 0.8707617789520036,
+      "train_speed(iter/s)": 0.029571
+    },
+    {
+      "epoch": 0.907163713451462,
+      "grad_norm": 0.05715427837146615,
+      "learning_rate": 4.501214146289956e-07,
+      "loss": 0.4418344497680664,
+      "memory(GiB)": 77.63,
+      "step": 2835,
+      "token_acc": 0.8788416882939489,
+      "train_speed(iter/s)": 0.02958
+    },
+    {
+      "epoch": 0.9087636494540219,
+      "grad_norm": 0.04986985198768239,
+      "learning_rate": 4.3484631239299356e-07,
+      "loss": 0.4437891960144043,
+      "memory(GiB)": 77.63,
+      "step": 2840,
+      "token_acc": 0.8431597023468803,
+      "train_speed(iter/s)": 0.029572
+    },
+    {
+      "epoch": 0.9103635854565817,
+      "grad_norm": 0.04999552116510165,
+      "learning_rate": 4.198291271593924e-07,
+      "loss": 0.44283204078674315,
+      "memory(GiB)": 77.63,
+      "step": 2845,
+      "token_acc": 0.8843727072633896,
+      "train_speed(iter/s)": 0.029575
+    },
+    {
+      "epoch": 0.9119635214591416,
+      "grad_norm": 0.047844581858855956,
+      "learning_rate": 4.0507026385502747e-07,
+      "loss": 0.4449836254119873,
+      "memory(GiB)": 77.63,
+      "step": 2850,
+      "token_acc": 0.9034812490661885,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.9135634574617015,
+      "grad_norm": 0.05675006848739315,
+      "learning_rate": 3.9057012044127817e-07,
+      "loss": 0.44204154014587405,
+      "memory(GiB)": 77.63,
+      "step": 2855,
+      "token_acc": 0.8655569782330346,
+      "train_speed(iter/s)": 0.029569
+    },
+    {
+      "epoch": 0.9151633934642615,
+      "grad_norm": 0.054077703664588216,
+      "learning_rate": 3.7632908790334656e-07,
+      "loss": 0.4383398532867432,
+      "memory(GiB)": 77.63,
+      "step": 2860,
+      "token_acc": 0.8899396929824561,
+      "train_speed(iter/s)": 0.029578
+    },
+    {
+      "epoch": 0.9167633294668214,
+      "grad_norm": 0.055142279260510525,
+      "learning_rate": 3.6234755023970447e-07,
+      "loss": 0.4388674259185791,
+      "memory(GiB)": 77.63,
+      "step": 2865,
+      "token_acc": 0.8406133828996283,
+      "train_speed(iter/s)": 0.029574
+    },
+    {
+      "epoch": 0.9183632654693812,
+      "grad_norm": 0.05242909357272202,
+      "learning_rate": 3.4862588445174985e-07,
+      "loss": 0.44350008964538573,
+      "memory(GiB)": 77.63,
+      "step": 2870,
+      "token_acc": 0.8773854961832062,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.9199632014719411,
+      "grad_norm": 0.051431062939426,
+      "learning_rate": 3.3516446053363015e-07,
+      "loss": 0.43948516845703123,
+      "memory(GiB)": 77.63,
+      "step": 2875,
+      "token_acc": 0.8812238692512353,
+      "train_speed(iter/s)": 0.029578
+    },
+    {
+      "epoch": 0.921563137474501,
+      "grad_norm": 0.04842108410960974,
+      "learning_rate": 3.219636414622751e-07,
+      "loss": 0.44395694732666013,
+      "memory(GiB)": 77.63,
+      "step": 2880,
+      "token_acc": 0.872349158571624,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.923163073477061,
+      "grad_norm": 0.050854557927068264,
+      "learning_rate": 3.090237831876053e-07,
+      "loss": 0.4437469482421875,
+      "memory(GiB)": 77.63,
+      "step": 2885,
+      "token_acc": 0.8500566251415629,
+      "train_speed(iter/s)": 0.029572
+    },
+    {
+      "epoch": 0.9247630094796209,
+      "grad_norm": 0.05289158935619814,
+      "learning_rate": 2.9634523462293005e-07,
+      "loss": 0.439394474029541,
+      "memory(GiB)": 77.63,
+      "step": 2890,
+      "token_acc": 0.9039268013724743,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.9263629454821807,
+      "grad_norm": 0.0532633695315608,
+      "learning_rate": 2.839283376355506e-07,
+      "loss": 0.4414195537567139,
+      "memory(GiB)": 77.63,
+      "step": 2895,
+      "token_acc": 0.8979206049149339,
+      "train_speed(iter/s)": 0.029567
+    },
+    {
+      "epoch": 0.9279628814847406,
+      "grad_norm": 0.05197494947423765,
+      "learning_rate": 2.717734270375272e-07,
+      "loss": 0.4303572177886963,
+      "memory(GiB)": 77.63,
+      "step": 2900,
+      "token_acc": 0.8932318992654774,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.9295628174873005,
+      "grad_norm": 0.05715199054523632,
+      "learning_rate": 2.5988083057666534e-07,
+      "loss": 0.4488718032836914,
+      "memory(GiB)": 77.63,
+      "step": 2905,
+      "token_acc": 0.8734599589322382,
+      "train_speed(iter/s)": 0.029572
+    },
+    {
+      "epoch": 0.9311627534898604,
+      "grad_norm": 0.052180972248620894,
+      "learning_rate": 2.4825086892766745e-07,
+      "loss": 0.44499683380126953,
+      "memory(GiB)": 77.63,
+      "step": 2910,
+      "token_acc": 0.8798773215198501,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.9327626894924202,
+      "grad_norm": 0.05195036988843101,
+      "learning_rate": 2.3688385568349515e-07,
+      "loss": 0.4348268508911133,
+      "memory(GiB)": 77.63,
+      "step": 2915,
+      "token_acc": 0.8380835380835381,
+      "train_speed(iter/s)": 0.029578
+    },
+    {
+      "epoch": 0.9343626254949802,
+      "grad_norm": 0.05340475761749026,
+      "learning_rate": 2.2578009734690264e-07,
+      "loss": 0.4533662796020508,
+      "memory(GiB)": 77.63,
+      "step": 2920,
+      "token_acc": 0.8854700854700854,
+      "train_speed(iter/s)": 0.029569
+    },
+    {
+      "epoch": 0.9359625614975401,
+      "grad_norm": 0.047434143218971526,
+      "learning_rate": 2.1493989332218468e-07,
+      "loss": 0.4382183074951172,
+      "memory(GiB)": 77.63,
+      "step": 2925,
+      "token_acc": 0.880750496121234,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.9375624975001,
+      "grad_norm": 0.052509681961836426,
+      "learning_rate": 2.043635359070928e-07,
+      "loss": 0.44708704948425293,
+      "memory(GiB)": 77.63,
+      "step": 2930,
+      "token_acc": 0.8942012598933936,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.9391624335026599,
+      "grad_norm": 0.0531539858574761,
+      "learning_rate": 1.9405131028495838e-07,
+      "loss": 0.45058341026306153,
+      "memory(GiB)": 77.63,
+      "step": 2935,
+      "token_acc": 0.8614694335389792,
+      "train_speed(iter/s)": 0.029569
+    },
+    {
+      "epoch": 0.9407623695052197,
+      "grad_norm": 0.05931701393554437,
+      "learning_rate": 1.8400349451700438e-07,
+      "loss": 0.44367156028747556,
+      "memory(GiB)": 77.63,
+      "step": 2940,
+      "token_acc": 0.881156184096266,
+      "train_speed(iter/s)": 0.029573
+    },
+    {
+      "epoch": 0.9423623055077797,
+      "grad_norm": 0.05380572562718686,
+      "learning_rate": 1.742203595348435e-07,
+      "loss": 0.4424111843109131,
+      "memory(GiB)": 77.63,
+      "step": 2945,
+      "token_acc": 0.8769617074701821,
+      "train_speed(iter/s)": 0.029576
+    },
+    {
+      "epoch": 0.9439622415103396,
+      "grad_norm": 0.06156259021000458,
+      "learning_rate": 1.6470216913317628e-07,
+      "loss": 0.4509577751159668,
+      "memory(GiB)": 77.63,
+      "step": 2950,
+      "token_acc": 0.8544532947139754,
+      "train_speed(iter/s)": 0.029567
+    },
+    {
+      "epoch": 0.9455621775128995,
+      "grad_norm": 0.050697077683688974,
+      "learning_rate": 1.5544917996267562e-07,
+      "loss": 0.44117283821105957,
+      "memory(GiB)": 77.63,
+      "step": 2955,
+      "token_acc": 0.8515226026101759,
+      "train_speed(iter/s)": 0.029575
+    },
+    {
+      "epoch": 0.9471621135154594,
+      "grad_norm": 0.05205296295346333,
+      "learning_rate": 1.464616415230702e-07,
+      "loss": 0.4488182067871094,
+      "memory(GiB)": 77.63,
+      "step": 2960,
+      "token_acc": 0.874439461883408,
+      "train_speed(iter/s)": 0.029569
+    },
+    {
+      "epoch": 0.9487620495180192,
+      "grad_norm": 0.047632029637579856,
+      "learning_rate": 1.3773979615640976e-07,
+      "loss": 0.4415272235870361,
+      "memory(GiB)": 77.63,
+      "step": 2965,
+      "token_acc": 0.8889883616830797,
+      "train_speed(iter/s)": 0.029564
+    },
+    {
+      "epoch": 0.9503619855205792,
+      "grad_norm": 0.051226023380966074,
+      "learning_rate": 1.292838790405393e-07,
+      "loss": 0.4453396797180176,
+      "memory(GiB)": 77.63,
+      "step": 2970,
+      "token_acc": 0.8701866977829639,
+      "train_speed(iter/s)": 0.029572
+    },
+    {
+      "epoch": 0.9519619215231391,
+      "grad_norm": 0.05296626405711913,
+      "learning_rate": 1.2109411818274851e-07,
+      "loss": 0.44417614936828614,
+      "memory(GiB)": 77.63,
+      "step": 2975,
+      "token_acc": 0.8997547959036493,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.953561857525699,
+      "grad_norm": 0.0535756814263424,
+      "learning_rate": 1.1317073441363458e-07,
+      "loss": 0.444796085357666,
+      "memory(GiB)": 77.63,
+      "step": 2980,
+      "token_acc": 0.9101887677336147,
+      "train_speed(iter/s)": 0.029563
+    },
+    {
+      "epoch": 0.9551617935282589,
+      "grad_norm": 0.05005027675979017,
+      "learning_rate": 1.055139413811379e-07,
+      "loss": 0.45203323364257814,
+      "memory(GiB)": 77.63,
+      "step": 2985,
+      "token_acc": 0.8569892473118279,
+      "train_speed(iter/s)": 0.029568
+    },
+    {
+      "epoch": 0.9567617295308187,
+      "grad_norm": 0.04934807615166247,
+      "learning_rate": 9.812394554478355e-08,
+      "loss": 0.43912034034729003,
+      "memory(GiB)": 77.63,
+      "step": 2990,
+      "token_acc": 0.8557346268189642,
+      "train_speed(iter/s)": 0.029559
+    },
+    {
+      "epoch": 0.9583616655333786,
+      "grad_norm": 0.051618825055470385,
+      "learning_rate": 9.10009461701189e-08,
+      "loss": 0.4506105899810791,
+      "memory(GiB)": 77.63,
+      "step": 2995,
+      "token_acc": 0.7809948032665182,
+      "train_speed(iter/s)": 0.029565
+    },
+    {
+      "epoch": 0.9599616015359386,
+      "grad_norm": 0.054833054855342726,
+      "learning_rate": 8.41451353233369e-08,
+      "loss": 0.442844820022583,
+      "memory(GiB)": 77.63,
+      "step": 3000,
+      "token_acc": 0.8733862959285005,
+      "train_speed(iter/s)": 0.029563
+    },
+    {
+      "epoch": 0.9599616015359386,
+      "eval_loss": 0.6573547720909119,
+      "eval_runtime": 106.0877,
+      "eval_samples_per_second": 189.353,
+      "eval_steps_per_second": 0.952,
+      "eval_token_acc": 0.8724322608695082,
+      "step": 3000
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 3125,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 3.944159044825703e+20,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}