diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,5442 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.3533881084901493,
+  "eval_steps": 100000,
+  "global_step": 3000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0001177960361633831,
+      "grad_norm": 18.875,
+      "learning_rate": 1e-05,
+      "loss": 0.42521548,
+      "memory(GiB)": 63.39,
+      "step": 1,
+      "train_speed(iter/s)": 0.060823
+    },
+    {
+      "epoch": 0.0005889801808169155,
+      "grad_norm": 5.6875,
+      "learning_rate": 9.999994520399261e-06,
+      "loss": 0.22830066,
+      "memory(GiB)": 74.41,
+      "step": 5,
+      "train_speed(iter/s)": 0.071086
+    },
+    {
+      "epoch": 0.001177960361633831,
+      "grad_norm": 5.40625,
+      "learning_rate": 9.999972259541841e-06,
+      "loss": 0.16523392,
+      "memory(GiB)": 74.41,
+      "step": 10,
+      "train_speed(iter/s)": 0.073351
+    },
+    {
+      "epoch": 0.0017669405424507465,
+      "grad_norm": 3.8125,
+      "learning_rate": 9.999932875028875e-06,
+      "loss": 0.12637961,
+      "memory(GiB)": 74.41,
+      "step": 15,
+      "train_speed(iter/s)": 0.073784
+    },
+    {
+      "epoch": 0.002355920723267662,
+      "grad_norm": 5.15625,
+      "learning_rate": 9.999876366995242e-06,
+      "loss": 0.14416599,
+      "memory(GiB)": 74.41,
+      "step": 20,
+      "train_speed(iter/s)": 0.074041
+    },
+    {
+      "epoch": 0.0029449009040845775,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.999802735634472e-06,
+      "loss": 0.12908924,
+      "memory(GiB)": 74.41,
+      "step": 25,
+      "train_speed(iter/s)": 0.074266
+    },
+    {
+      "epoch": 0.003533881084901493,
+      "grad_norm": 5.03125,
+      "learning_rate": 9.99971198119873e-06,
+      "loss": 0.15117922,
+      "memory(GiB)": 74.41,
+      "step": 30,
+      "train_speed(iter/s)": 0.074446
+    },
+    {
+      "epoch": 0.0041228612657184084,
+      "grad_norm": 5.53125,
+      "learning_rate": 9.99960410399883e-06,
+      "loss": 0.1358834,
+      "memory(GiB)": 74.41,
+      "step": 35,
+      "train_speed(iter/s)": 0.074609
+    },
+    {
+      "epoch": 0.004711841446535324,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.999479104404225e-06,
+      "loss": 0.13182677,
+      "memory(GiB)": 74.41,
+      "step": 40,
+      "train_speed(iter/s)": 0.074644
+    },
+    {
+      "epoch": 0.005300821627352239,
+      "grad_norm": 6.25,
+      "learning_rate": 9.999336982843003e-06,
+      "loss": 0.14286679,
+      "memory(GiB)": 74.41,
+      "step": 45,
+      "train_speed(iter/s)": 0.074647
+    },
+    {
+      "epoch": 0.005889801808169155,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.999177739801898e-06,
+      "loss": 0.12088619,
+      "memory(GiB)": 74.41,
+      "step": 50,
+      "train_speed(iter/s)": 0.074676
+    },
+    {
+      "epoch": 0.00647878198898607,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.999001375826279e-06,
+      "loss": 0.11520904,
+      "memory(GiB)": 74.41,
+      "step": 55,
+      "train_speed(iter/s)": 0.074701
+    },
+    {
+      "epoch": 0.007067762169802986,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.998807891520147e-06,
+      "loss": 0.11755368,
+      "memory(GiB)": 74.41,
+      "step": 60,
+      "train_speed(iter/s)": 0.07475
+    },
+    {
+      "epoch": 0.007656742350619901,
+      "grad_norm": 5.21875,
+      "learning_rate": 9.998597287546135e-06,
+      "loss": 0.1323318,
+      "memory(GiB)": 74.41,
+      "step": 65,
+      "train_speed(iter/s)": 0.074786
+    },
+    {
+      "epoch": 0.008245722531436817,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.998369564625513e-06,
+      "loss": 0.11528516,
+      "memory(GiB)": 74.41,
+      "step": 70,
+      "train_speed(iter/s)": 0.074767
+    },
+    {
+      "epoch": 0.008834702712253733,
+      "grad_norm": 4.625,
+      "learning_rate": 9.998124723538172e-06,
+      "loss": 0.1128188,
+      "memory(GiB)": 74.41,
+      "step": 75,
+      "train_speed(iter/s)": 0.074767
+    },
+    {
+      "epoch": 0.009423682893070648,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.997862765122634e-06,
+      "loss": 0.12012781,
+      "memory(GiB)": 74.41,
+      "step": 80,
+      "train_speed(iter/s)": 0.074792
+    },
+    {
+      "epoch": 0.010012663073887564,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.99758369027604e-06,
+      "loss": 0.13530364,
+      "memory(GiB)": 74.41,
+      "step": 85,
+      "train_speed(iter/s)": 0.074854
+    },
+    {
+      "epoch": 0.010601643254704479,
+      "grad_norm": 5.0,
+      "learning_rate": 9.997287499954151e-06,
+      "loss": 0.13009607,
+      "memory(GiB)": 74.41,
+      "step": 90,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.011190623435521395,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.996974195171348e-06,
+      "loss": 0.13199264,
+      "memory(GiB)": 74.41,
+      "step": 95,
+      "train_speed(iter/s)": 0.074953
+    },
+    {
+      "epoch": 0.01177960361633831,
+      "grad_norm": 5.6875,
+      "learning_rate": 9.996643777000617e-06,
+      "loss": 0.13707412,
+      "memory(GiB)": 74.41,
+      "step": 100,
+      "train_speed(iter/s)": 0.074985
+    },
+    {
+      "epoch": 0.012368583797155226,
+      "grad_norm": 4.75,
+      "learning_rate": 9.996296246573558e-06,
+      "loss": 0.10490649,
+      "memory(GiB)": 74.41,
+      "step": 105,
+      "train_speed(iter/s)": 0.074991
+    },
+    {
+      "epoch": 0.01295756397797214,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.99593160508038e-06,
+      "loss": 0.11323662,
+      "memory(GiB)": 74.41,
+      "step": 110,
+      "train_speed(iter/s)": 0.075013
+    },
+    {
+      "epoch": 0.013546544158789057,
+      "grad_norm": 4.75,
+      "learning_rate": 9.995549853769887e-06,
+      "loss": 0.13000023,
+      "memory(GiB)": 74.41,
+      "step": 115,
+      "train_speed(iter/s)": 0.075041
+    },
+    {
+      "epoch": 0.014135524339605972,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.995150993949479e-06,
+      "loss": 0.109146,
+      "memory(GiB)": 74.41,
+      "step": 120,
+      "train_speed(iter/s)": 0.075048
+    },
+    {
+      "epoch": 0.014724504520422888,
+      "grad_norm": 6.625,
+      "learning_rate": 9.994735026985157e-06,
+      "loss": 0.13355327,
+      "memory(GiB)": 74.41,
+      "step": 125,
+      "train_speed(iter/s)": 0.075064
+    },
+    {
+      "epoch": 0.015313484701239803,
+      "grad_norm": 4.5,
+      "learning_rate": 9.994301954301497e-06,
+      "loss": 0.10676175,
+      "memory(GiB)": 74.41,
+      "step": 130,
+      "train_speed(iter/s)": 0.075068
+    },
+    {
+      "epoch": 0.015902464882056717,
+      "grad_norm": 4.75,
+      "learning_rate": 9.99385177738167e-06,
+      "loss": 0.12428848,
+      "memory(GiB)": 74.41,
+      "step": 135,
+      "train_speed(iter/s)": 0.075069
+    },
+    {
+      "epoch": 0.016491445062873634,
+      "grad_norm": 3.875,
+      "learning_rate": 9.993384497767419e-06,
+      "loss": 0.11814754,
+      "memory(GiB)": 74.41,
+      "step": 140,
+      "train_speed(iter/s)": 0.075095
+    },
+    {
+      "epoch": 0.01708042524369055,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.992900117059056e-06,
+      "loss": 0.10705045,
+      "memory(GiB)": 74.41,
+      "step": 145,
+      "train_speed(iter/s)": 0.075056
+    },
+    {
+      "epoch": 0.017669405424507467,
+      "grad_norm": 4.375,
+      "learning_rate": 9.992398636915468e-06,
+      "loss": 0.11953868,
+      "memory(GiB)": 74.41,
+      "step": 150,
+      "train_speed(iter/s)": 0.07506
+    },
+    {
+      "epoch": 0.01825838560532438,
+      "grad_norm": 4.25,
+      "learning_rate": 9.991880059054097e-06,
+      "loss": 0.11556933,
+      "memory(GiB)": 74.41,
+      "step": 155,
+      "train_speed(iter/s)": 0.075058
+    },
+    {
+      "epoch": 0.018847365786141296,
+      "grad_norm": 4.625,
+      "learning_rate": 9.991344385250944e-06,
+      "loss": 0.11189389,
+      "memory(GiB)": 74.41,
+      "step": 160,
+      "train_speed(iter/s)": 0.075054
+    },
+    {
+      "epoch": 0.019436345966958212,
+      "grad_norm": 4.875,
+      "learning_rate": 9.990791617340558e-06,
+      "loss": 0.10822484,
+      "memory(GiB)": 74.41,
+      "step": 165,
+      "train_speed(iter/s)": 0.07506
+    },
+    {
+      "epoch": 0.02002532614777513,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.990221757216029e-06,
+      "loss": 0.11197646,
+      "memory(GiB)": 74.41,
+      "step": 170,
+      "train_speed(iter/s)": 0.075068
+    },
+    {
+      "epoch": 0.02061430632859204,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.989634806828987e-06,
+      "loss": 0.10931354,
+      "memory(GiB)": 74.41,
+      "step": 175,
+      "train_speed(iter/s)": 0.075063
+    },
+    {
+      "epoch": 0.021203286509408958,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.98903076818959e-06,
+      "loss": 0.11188631,
+      "memory(GiB)": 74.41,
+      "step": 180,
+      "train_speed(iter/s)": 0.075068
+    },
+    {
+      "epoch": 0.021792266690225874,
+      "grad_norm": 4.75,
+      "learning_rate": 9.988409643366518e-06,
+      "loss": 0.11905994,
+      "memory(GiB)": 74.41,
+      "step": 185,
+      "train_speed(iter/s)": 0.075071
+    },
+    {
+      "epoch": 0.02238124687104279,
+      "grad_norm": 6.09375,
+      "learning_rate": 9.987771434486972e-06,
+      "loss": 0.11119931,
+      "memory(GiB)": 74.41,
+      "step": 190,
+      "train_speed(iter/s)": 0.075058
+    },
+    {
+      "epoch": 0.022970227051859703,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.987116143736656e-06,
+      "loss": 0.07965935,
+      "memory(GiB)": 74.41,
+      "step": 195,
+      "train_speed(iter/s)": 0.075072
+    },
+    {
+      "epoch": 0.02355920723267662,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.986443773359776e-06,
+      "loss": 0.10195212,
+      "memory(GiB)": 74.41,
+      "step": 200,
+      "train_speed(iter/s)": 0.075089
+    },
+    {
+      "epoch": 0.024148187413493536,
+      "grad_norm": 5.15625,
+      "learning_rate": 9.985754325659033e-06,
+      "loss": 0.12634788,
+      "memory(GiB)": 74.41,
+      "step": 205,
+      "train_speed(iter/s)": 0.075075
+    },
+    {
+      "epoch": 0.024737167594310452,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.985047802995613e-06,
+      "loss": 0.11235939,
+      "memory(GiB)": 74.41,
+      "step": 210,
+      "train_speed(iter/s)": 0.075064
+    },
+    {
+      "epoch": 0.025326147775127365,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.98432420778918e-06,
+      "loss": 0.12128488,
+      "memory(GiB)": 74.41,
+      "step": 215,
+      "train_speed(iter/s)": 0.07508
+    },
+    {
+      "epoch": 0.02591512795594428,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.983583542517868e-06,
+      "loss": 0.09801877,
+      "memory(GiB)": 74.41,
+      "step": 220,
+      "train_speed(iter/s)": 0.075075
+    },
+    {
+      "epoch": 0.026504108136761198,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.982825809718268e-06,
+      "loss": 0.12451966,
+      "memory(GiB)": 74.41,
+      "step": 225,
+      "train_speed(iter/s)": 0.075076
+    },
+    {
+      "epoch": 0.027093088317578114,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.982051011985428e-06,
+      "loss": 0.10354398,
+      "memory(GiB)": 74.41,
+      "step": 230,
+      "train_speed(iter/s)": 0.075069
+    },
+    {
+      "epoch": 0.027682068498395027,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.981259151972835e-06,
+      "loss": 0.11414304,
+      "memory(GiB)": 74.41,
+      "step": 235,
+      "train_speed(iter/s)": 0.075059
+    },
+    {
+      "epoch": 0.028271048679211944,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.980450232392412e-06,
+      "loss": 0.11468437,
+      "memory(GiB)": 74.41,
+      "step": 240,
+      "train_speed(iter/s)": 0.075065
+    },
+    {
+      "epoch": 0.02886002886002886,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.979624256014505e-06,
+      "loss": 0.10513357,
+      "memory(GiB)": 74.41,
+      "step": 245,
+      "train_speed(iter/s)": 0.075058
+    },
+    {
+      "epoch": 0.029449009040845776,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.97878122566788e-06,
+      "loss": 0.11546416,
+      "memory(GiB)": 74.41,
+      "step": 250,
+      "train_speed(iter/s)": 0.075068
+    },
+    {
+      "epoch": 0.030037989221662693,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.977921144239701e-06,
+      "loss": 0.10937932,
+      "memory(GiB)": 74.41,
+      "step": 255,
+      "train_speed(iter/s)": 0.075063
+    },
+    {
+      "epoch": 0.030626969402479606,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.977044014675537e-06,
+      "loss": 0.11372919,
+      "memory(GiB)": 74.41,
+      "step": 260,
+      "train_speed(iter/s)": 0.07507
+    },
+    {
+      "epoch": 0.031215949583296522,
+      "grad_norm": 5.84375,
+      "learning_rate": 9.976149839979333e-06,
+      "loss": 0.13152323,
+      "memory(GiB)": 74.41,
+      "step": 265,
+      "train_speed(iter/s)": 0.075061
+    },
+    {
+      "epoch": 0.031804929764113435,
+      "grad_norm": 3.875,
+      "learning_rate": 9.975238623213417e-06,
+      "loss": 0.10404849,
+      "memory(GiB)": 74.41,
+      "step": 270,
+      "train_speed(iter/s)": 0.075064
+    },
+    {
+      "epoch": 0.032393909944930355,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.974310367498476e-06,
+      "loss": 0.09936564,
+      "memory(GiB)": 74.41,
+      "step": 275,
+      "train_speed(iter/s)": 0.075057
+    },
+    {
+      "epoch": 0.03298289012574727,
+      "grad_norm": 4.90625,
+      "learning_rate": 9.973365076013558e-06,
+      "loss": 0.11127584,
+      "memory(GiB)": 74.41,
+      "step": 280,
+      "train_speed(iter/s)": 0.07505
+    },
+    {
+      "epoch": 0.03357187030656419,
+      "grad_norm": 5.0625,
+      "learning_rate": 9.972402751996045e-06,
+      "loss": 0.1130929,
+      "memory(GiB)": 74.41,
+      "step": 285,
+      "train_speed(iter/s)": 0.075047
+    },
+    {
+      "epoch": 0.0341608504873811,
+      "grad_norm": 4.875,
+      "learning_rate": 9.97142339874166e-06,
+      "loss": 0.13001273,
+      "memory(GiB)": 74.41,
+      "step": 290,
+      "train_speed(iter/s)": 0.075049
+    },
+    {
+      "epoch": 0.03474983066819801,
+      "grad_norm": 4.75,
+      "learning_rate": 9.970427019604441e-06,
+      "loss": 0.11981708,
+      "memory(GiB)": 74.41,
+      "step": 295,
+      "train_speed(iter/s)": 0.075056
+    },
+    {
+      "epoch": 0.03533881084901493,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.96941361799674e-06,
+      "loss": 0.09800007,
+      "memory(GiB)": 74.41,
+      "step": 300,
+      "train_speed(iter/s)": 0.07505
+    },
+    {
+      "epoch": 0.035927791029831846,
+      "grad_norm": 5.5,
+      "learning_rate": 9.968383197389202e-06,
+      "loss": 0.10793682,
+      "memory(GiB)": 74.41,
+      "step": 305,
+      "train_speed(iter/s)": 0.075058
+    },
+    {
+      "epoch": 0.03651677121064876,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.967335761310761e-06,
+      "loss": 0.10909323,
+      "memory(GiB)": 74.41,
+      "step": 310,
+      "train_speed(iter/s)": 0.075054
+    },
+    {
+      "epoch": 0.03710575139146568,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.966271313348623e-06,
+      "loss": 0.10540984,
+      "memory(GiB)": 74.41,
+      "step": 315,
+      "train_speed(iter/s)": 0.075062
+    },
+    {
+      "epoch": 0.03769473157228259,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.96518985714826e-06,
+      "loss": 0.11170142,
+      "memory(GiB)": 74.41,
+      "step": 320,
+      "train_speed(iter/s)": 0.075084
+    },
+    {
+      "epoch": 0.03828371175309951,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.964091396413381e-06,
+      "loss": 0.1047295,
+      "memory(GiB)": 74.41,
+      "step": 325,
+      "train_speed(iter/s)": 0.075092
+    },
+    {
+      "epoch": 0.038872691933916424,
+      "grad_norm": 3.875,
+      "learning_rate": 9.962975934905947e-06,
+      "loss": 0.11229659,
+      "memory(GiB)": 74.41,
+      "step": 330,
+      "train_speed(iter/s)": 0.075096
+    },
+    {
+      "epoch": 0.03946167211473334,
+      "grad_norm": 5.75,
+      "learning_rate": 9.961843476446134e-06,
+      "loss": 0.12050457,
+      "memory(GiB)": 74.41,
+      "step": 335,
+      "train_speed(iter/s)": 0.075102
+    },
+    {
+      "epoch": 0.04005065229555026,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.960694024912328e-06,
+      "loss": 0.10951951,
+      "memory(GiB)": 74.41,
+      "step": 340,
+      "train_speed(iter/s)": 0.075106
+    },
+    {
+      "epoch": 0.04063963247636717,
+      "grad_norm": 3.21875,
+      "learning_rate": 9.959527584241112e-06,
+      "loss": 0.10334082,
+      "memory(GiB)": 74.41,
+      "step": 345,
+      "train_speed(iter/s)": 0.07511
+    },
+    {
+      "epoch": 0.04122861265718408,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.958344158427255e-06,
+      "loss": 0.1155634,
+      "memory(GiB)": 74.41,
+      "step": 350,
+      "train_speed(iter/s)": 0.075116
+    },
+    {
+      "epoch": 0.041817592838001,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.957143751523695e-06,
+      "loss": 0.12060295,
+      "memory(GiB)": 74.41,
+      "step": 355,
+      "train_speed(iter/s)": 0.075118
+    },
+    {
+      "epoch": 0.042406573018817915,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.955926367641524e-06,
+      "loss": 0.09431397,
+      "memory(GiB)": 74.41,
+      "step": 360,
+      "train_speed(iter/s)": 0.075117
+    },
+    {
+      "epoch": 0.042995553199634835,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.95469201094998e-06,
+      "loss": 0.10672688,
+      "memory(GiB)": 74.41,
+      "step": 365,
+      "train_speed(iter/s)": 0.075123
+    },
+    {
+      "epoch": 0.04358453338045175,
+      "grad_norm": 4.375,
+      "learning_rate": 9.953440685676425e-06,
+      "loss": 0.11097996,
+      "memory(GiB)": 74.41,
+      "step": 370,
+      "train_speed(iter/s)": 0.075131
+    },
+    {
+      "epoch": 0.04417351356126866,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.952172396106336e-06,
+      "loss": 0.11753557,
+      "memory(GiB)": 74.41,
+      "step": 375,
+      "train_speed(iter/s)": 0.07513
+    },
+    {
+      "epoch": 0.04476249374208558,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.950887146583288e-06,
+      "loss": 0.09597683,
+      "memory(GiB)": 74.41,
+      "step": 380,
+      "train_speed(iter/s)": 0.07513
+    },
+    {
+      "epoch": 0.045351473922902494,
+      "grad_norm": 4.5625,
+      "learning_rate": 9.949584941508937e-06,
+      "loss": 0.10678861,
+      "memory(GiB)": 74.41,
+      "step": 385,
+      "train_speed(iter/s)": 0.07512
+    },
+    {
+      "epoch": 0.04594045410371941,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.948265785343013e-06,
+      "loss": 0.10599846,
+      "memory(GiB)": 74.41,
+      "step": 390,
+      "train_speed(iter/s)": 0.07512
+    },
+    {
+      "epoch": 0.046529434284536327,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.946929682603295e-06,
+      "loss": 0.10928589,
+      "memory(GiB)": 74.41,
+      "step": 395,
+      "train_speed(iter/s)": 0.07511
+    },
+    {
+      "epoch": 0.04711841446535324,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.9455766378656e-06,
+      "loss": 0.10189447,
+      "memory(GiB)": 74.41,
+      "step": 400,
+      "train_speed(iter/s)": 0.075107
+    },
+    {
+      "epoch": 0.04770739464617016,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.94420665576377e-06,
+      "loss": 0.11017077,
+      "memory(GiB)": 74.41,
+      "step": 405,
+      "train_speed(iter/s)": 0.075097
+    },
+    {
+      "epoch": 0.04829637482698707,
+      "grad_norm": 3.171875,
+      "learning_rate": 9.942819740989651e-06,
+      "loss": 0.09610715,
+      "memory(GiB)": 74.41,
+      "step": 410,
+      "train_speed(iter/s)": 0.075098
+    },
+    {
+      "epoch": 0.048885355007803985,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.941415898293078e-06,
+      "loss": 0.09449079,
+      "memory(GiB)": 74.41,
+      "step": 415,
+      "train_speed(iter/s)": 0.075098
+    },
+    {
+      "epoch": 0.049474335188620905,
+      "grad_norm": 4.25,
+      "learning_rate": 9.939995132481863e-06,
+      "loss": 0.11824782,
+      "memory(GiB)": 74.41,
+      "step": 420,
+      "train_speed(iter/s)": 0.075098
+    },
+    {
+      "epoch": 0.05006331536943782,
+      "grad_norm": 3.953125,
+      "learning_rate": 9.938557448421772e-06,
+      "loss": 0.10327038,
+      "memory(GiB)": 74.41,
+      "step": 425,
+      "train_speed(iter/s)": 0.075095
+    },
+    {
+      "epoch": 0.05065229555025473,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.937102851036515e-06,
+      "loss": 0.101033,
+      "memory(GiB)": 74.41,
+      "step": 430,
+      "train_speed(iter/s)": 0.075096
+    },
+    {
+      "epoch": 0.05124127573107165,
+      "grad_norm": 5.625,
+      "learning_rate": 9.935631345307726e-06,
+      "loss": 0.12589303,
+      "memory(GiB)": 74.41,
+      "step": 435,
+      "train_speed(iter/s)": 0.075096
+    },
+    {
+      "epoch": 0.05183025591188856,
+      "grad_norm": 4.375,
+      "learning_rate": 9.934142936274941e-06,
+      "loss": 0.11252528,
+      "memory(GiB)": 74.41,
+      "step": 440,
+      "train_speed(iter/s)": 0.0751
+    },
+    {
+      "epoch": 0.05241923609270548,
+      "grad_norm": 5.03125,
+      "learning_rate": 9.932637629035594e-06,
+      "loss": 0.1076035,
+      "memory(GiB)": 74.41,
+      "step": 445,
+      "train_speed(iter/s)": 0.07511
+    },
+    {
+      "epoch": 0.053008216273522396,
+      "grad_norm": 4.8125,
+      "learning_rate": 9.931115428744983e-06,
+      "loss": 0.09129641,
+      "memory(GiB)": 74.41,
+      "step": 450,
+      "train_speed(iter/s)": 0.075106
+    },
+    {
+      "epoch": 0.05359719645433931,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.92957634061626e-06,
+      "loss": 0.10762049,
+      "memory(GiB)": 74.41,
+      "step": 455,
+      "train_speed(iter/s)": 0.075104
+    },
+    {
+      "epoch": 0.05418617663515623,
+      "grad_norm": 4.375,
+      "learning_rate": 9.928020369920422e-06,
+      "loss": 0.13035779,
+      "memory(GiB)": 74.41,
+      "step": 460,
+      "train_speed(iter/s)": 0.075115
+    },
+    {
+      "epoch": 0.05477515681597314,
+      "grad_norm": 4.90625,
+      "learning_rate": 9.92644752198628e-06,
+      "loss": 0.11500576,
+      "memory(GiB)": 74.41,
+      "step": 465,
+      "train_speed(iter/s)": 0.075106
+    },
+    {
+      "epoch": 0.055364136996790055,
+      "grad_norm": 5.59375,
+      "learning_rate": 9.924857802200443e-06,
+      "loss": 0.11347337,
+      "memory(GiB)": 74.41,
+      "step": 470,
+      "train_speed(iter/s)": 0.075116
+    },
+    {
+      "epoch": 0.055953117177606974,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.923251216007306e-06,
+      "loss": 0.11633272,
+      "memory(GiB)": 74.41,
+      "step": 475,
+      "train_speed(iter/s)": 0.075118
+    },
+    {
+      "epoch": 0.05654209735842389,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.921627768909022e-06,
+      "loss": 0.09800028,
+      "memory(GiB)": 74.41,
+      "step": 480,
+      "train_speed(iter/s)": 0.075124
+    },
+    {
+      "epoch": 0.05713107753924081,
+      "grad_norm": 8.4375,
+      "learning_rate": 9.919987466465495e-06,
+      "loss": 0.11815701,
+      "memory(GiB)": 74.41,
+      "step": 485,
+      "train_speed(iter/s)": 0.075127
+    },
+    {
+      "epoch": 0.05772005772005772,
+      "grad_norm": 4.9375,
+      "learning_rate": 9.91833031429435e-06,
+      "loss": 0.11462152,
+      "memory(GiB)": 74.41,
+      "step": 490,
+      "train_speed(iter/s)": 0.075127
+    },
+    {
+      "epoch": 0.05830903790087463,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.916656318070918e-06,
+      "loss": 0.09471865,
+      "memory(GiB)": 74.41,
+      "step": 495,
+      "train_speed(iter/s)": 0.075129
+    },
+    {
+      "epoch": 0.05889801808169155,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.91496548352822e-06,
+      "loss": 0.11097107,
+      "memory(GiB)": 74.41,
+      "step": 500,
+      "train_speed(iter/s)": 0.075131
+    },
+    {
+      "epoch": 0.059486998262508466,
+      "grad_norm": 6.75,
+      "learning_rate": 9.913257816456938e-06,
+      "loss": 0.12602438,
+      "memory(GiB)": 74.41,
+      "step": 505,
+      "train_speed(iter/s)": 0.074738
+    },
+    {
+      "epoch": 0.060075978443325385,
+      "grad_norm": 4.25,
+      "learning_rate": 9.91153332270541e-06,
+      "loss": 0.09413192,
+      "memory(GiB)": 74.41,
+      "step": 510,
+      "train_speed(iter/s)": 0.074752
+    },
+    {
+      "epoch": 0.0606649586241423,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.90979200817959e-06,
+      "loss": 0.10453418,
+      "memory(GiB)": 74.41,
+      "step": 515,
+      "train_speed(iter/s)": 0.074754
+    },
+    {
+      "epoch": 0.06125393880495921,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.90803387884305e-06,
+      "loss": 0.09584854,
+      "memory(GiB)": 74.41,
+      "step": 520,
+      "train_speed(iter/s)": 0.074762
+    },
+    {
+      "epoch": 0.06184291898577613,
+      "grad_norm": 3.625,
+      "learning_rate": 9.906258940716945e-06,
+      "loss": 0.11624388,
+      "memory(GiB)": 74.41,
+      "step": 525,
+      "train_speed(iter/s)": 0.074771
+    },
+    {
+      "epoch": 0.062431899166593044,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.904467199879988e-06,
+      "loss": 0.09951019,
+      "memory(GiB)": 74.41,
+      "step": 530,
+      "train_speed(iter/s)": 0.074776
+    },
+    {
+      "epoch": 0.06302087934740996,
+      "grad_norm": 4.75,
+      "learning_rate": 9.90265866246845e-06,
+      "loss": 0.11697209,
+      "memory(GiB)": 74.41,
+      "step": 535,
+      "train_speed(iter/s)": 0.074779
+    },
+    {
+      "epoch": 0.06360985952822687,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.900833334676116e-06,
+      "loss": 0.09333996,
+      "memory(GiB)": 74.41,
+      "step": 540,
+      "train_speed(iter/s)": 0.074775
+    },
+    {
+      "epoch": 0.06419883970904379,
+      "grad_norm": 4.125,
+      "learning_rate": 9.898991222754278e-06,
+      "loss": 0.09190307,
+      "memory(GiB)": 74.41,
+      "step": 545,
+      "train_speed(iter/s)": 0.07478
+    },
+    {
+      "epoch": 0.06478781988986071,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.89713233301171e-06,
+      "loss": 0.10609086,
+      "memory(GiB)": 74.41,
+      "step": 550,
+      "train_speed(iter/s)": 0.07479
+    },
+    {
+      "epoch": 0.06537680007067762,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.895256671814646e-06,
+      "loss": 0.11532583,
+      "memory(GiB)": 74.41,
+      "step": 555,
+      "train_speed(iter/s)": 0.07479
+    },
+    {
+      "epoch": 0.06596578025149454,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.893364245586755e-06,
+      "loss": 0.10446991,
+      "memory(GiB)": 74.41,
+      "step": 560,
+      "train_speed(iter/s)": 0.074794
+    },
+    {
+      "epoch": 0.06655476043231145,
+      "grad_norm": 4.125,
+      "learning_rate": 9.891455060809122e-06,
+      "loss": 0.08096789,
+      "memory(GiB)": 74.41,
+      "step": 565,
+      "train_speed(iter/s)": 0.074797
+    },
+    {
+      "epoch": 0.06714374061312837,
+      "grad_norm": 4.71875,
+      "learning_rate": 9.889529124020232e-06,
+      "loss": 0.10516653,
+      "memory(GiB)": 74.41,
+      "step": 570,
+      "train_speed(iter/s)": 0.074799
+    },
+    {
+      "epoch": 0.06773272079394528,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.887586441815934e-06,
+      "loss": 0.10007232,
+      "memory(GiB)": 74.41,
+      "step": 575,
+      "train_speed(iter/s)": 0.074804
+    },
+    {
+      "epoch": 0.0683217009747622,
+      "grad_norm": 3.5,
+      "learning_rate": 9.885627020849433e-06,
+      "loss": 0.09896673,
+      "memory(GiB)": 74.41,
+      "step": 580,
+      "train_speed(iter/s)": 0.074798
+    },
+    {
+      "epoch": 0.06891068115557912,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.883650867831252e-06,
+      "loss": 0.10226643,
+      "memory(GiB)": 74.41,
+      "step": 585,
+      "train_speed(iter/s)": 0.074805
+    },
+    {
+      "epoch": 0.06949966133639603,
+      "grad_norm": 4.84375,
+      "learning_rate": 9.881657989529222e-06,
+      "loss": 0.10619533,
+      "memory(GiB)": 74.41,
+      "step": 590,
+      "train_speed(iter/s)": 0.074807
+    },
+    {
+      "epoch": 0.07008864151721295,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.879648392768455e-06,
+      "loss": 0.1003757,
+      "memory(GiB)": 74.41,
+      "step": 595,
+      "train_speed(iter/s)": 0.074807
+    },
+    {
+      "epoch": 0.07067762169802987,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.877622084431316e-06,
+      "loss": 0.10820855,
+      "memory(GiB)": 74.41,
+      "step": 600,
+      "train_speed(iter/s)": 0.07481
+    },
+    {
+      "epoch": 0.07126660187884677,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.875579071457407e-06,
+      "loss": 0.09604223,
+      "memory(GiB)": 74.41,
+      "step": 605,
+      "train_speed(iter/s)": 0.074807
+    },
+    {
+      "epoch": 0.07185558205966369,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.873519360843535e-06,
+      "loss": 0.10592574,
+      "memory(GiB)": 74.41,
+      "step": 610,
+      "train_speed(iter/s)": 0.07481
+    },
+    {
+      "epoch": 0.07244456224048061,
+      "grad_norm": 4.375,
+      "learning_rate": 9.871442959643694e-06,
+      "loss": 0.10632453,
+      "memory(GiB)": 74.41,
+      "step": 615,
+      "train_speed(iter/s)": 0.074813
+    },
+    {
+      "epoch": 0.07303354242129752,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.869349874969039e-06,
+      "loss": 0.09836457,
+      "memory(GiB)": 74.41,
+      "step": 620,
+      "train_speed(iter/s)": 0.074819
+    },
+    {
+      "epoch": 0.07362252260211444,
+      "grad_norm": 4.46875,
+      "learning_rate": 9.867240113987863e-06,
+      "loss": 0.11561801,
+      "memory(GiB)": 74.41,
+      "step": 625,
+      "train_speed(iter/s)": 0.074823
+    },
+    {
+      "epoch": 0.07421150278293136,
+      "grad_norm": 3.109375,
+      "learning_rate": 9.865113683925571e-06,
+      "loss": 0.09328078,
+      "memory(GiB)": 74.41,
+      "step": 630,
+      "train_speed(iter/s)": 0.074825
+    },
+    {
+      "epoch": 0.07480048296374826,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.862970592064653e-06,
+      "loss": 0.10008674,
+      "memory(GiB)": 74.41,
+      "step": 635,
+      "train_speed(iter/s)": 0.074832
+    },
+    {
+      "epoch": 0.07538946314456518,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.860810845744664e-06,
+      "loss": 0.10500932,
+      "memory(GiB)": 74.41,
+      "step": 640,
+      "train_speed(iter/s)": 0.074834
+    },
+    {
+      "epoch": 0.0759784433253821,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.858634452362195e-06,
+      "loss": 0.10009133,
+      "memory(GiB)": 74.41,
+      "step": 645,
+      "train_speed(iter/s)": 0.074837
+    },
+    {
+      "epoch": 0.07656742350619902,
+      "grad_norm": 3.75,
+      "learning_rate": 9.856441419370849e-06,
+      "loss": 0.09874001,
+      "memory(GiB)": 74.41,
+      "step": 650,
+      "train_speed(iter/s)": 0.074838
+    },
+    {
+      "epoch": 0.07715640368701593,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.854231754281216e-06,
+      "loss": 0.10840924,
+      "memory(GiB)": 74.41,
+      "step": 655,
+      "train_speed(iter/s)": 0.074835
+    },
+    {
+      "epoch": 0.07774538386783285,
+      "grad_norm": 4.5,
+      "learning_rate": 9.852005464660847e-06,
+      "loss": 0.10447395,
+      "memory(GiB)": 74.41,
+      "step": 660,
+      "train_speed(iter/s)": 0.074838
+    },
+    {
+      "epoch": 0.07833436404864977,
+      "grad_norm": 3.515625,
+      "learning_rate": 9.84976255813423e-06,
+      "loss": 0.10944817,
+      "memory(GiB)": 74.41,
+      "step": 665,
+      "train_speed(iter/s)": 0.074837
+    },
+    {
+      "epoch": 0.07892334422946667,
+      "grad_norm": 4.375,
+      "learning_rate": 9.847503042382756e-06,
+      "loss": 0.11643915,
+      "memory(GiB)": 74.41,
+      "step": 670,
+      "train_speed(iter/s)": 0.074834
+    },
+    {
+      "epoch": 0.0795123244102836,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.845226925144702e-06,
+      "loss": 0.11636159,
+      "memory(GiB)": 74.41,
+      "step": 675,
+      "train_speed(iter/s)": 0.074838
+    },
+    {
+      "epoch": 0.08010130459110051,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.842934214215203e-06,
+      "loss": 0.11168916,
+      "memory(GiB)": 74.41,
+      "step": 680,
+      "train_speed(iter/s)": 0.074835
+    },
+    {
+      "epoch": 0.08069028477191742,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.84062491744622e-06,
+      "loss": 0.10953134,
+      "memory(GiB)": 74.41,
+      "step": 685,
+      "train_speed(iter/s)": 0.074839
+    },
+    {
+      "epoch": 0.08127926495273434,
+      "grad_norm": 4.78125,
+      "learning_rate": 9.838299042746518e-06,
+      "loss": 0.09663584,
+      "memory(GiB)": 74.41,
+      "step": 690,
+      "train_speed(iter/s)": 0.07484
+    },
+    {
+      "epoch": 0.08186824513355126,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.835956598081634e-06,
+      "loss": 0.10150684,
+      "memory(GiB)": 74.41,
+      "step": 695,
+      "train_speed(iter/s)": 0.074837
+    },
+    {
+      "epoch": 0.08245722531436817,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.833597591473858e-06,
+      "loss": 0.10742941,
+      "memory(GiB)": 74.41,
+      "step": 700,
+      "train_speed(iter/s)": 0.074837
+    },
+    {
+      "epoch": 0.08304620549518509,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.831222031002197e-06,
+      "loss": 0.10118611,
+      "memory(GiB)": 74.41,
+      "step": 705,
+      "train_speed(iter/s)": 0.074838
+    },
+    {
+      "epoch": 0.083635185676002,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.828829924802355e-06,
+      "loss": 0.0967037,
+      "memory(GiB)": 74.41,
+      "step": 710,
+      "train_speed(iter/s)": 0.074837
+    },
+    {
+      "epoch": 0.08422416585681892,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.826421281066693e-06,
+      "loss": 0.10264329,
+      "memory(GiB)": 74.41,
+      "step": 715,
+      "train_speed(iter/s)": 0.074838
+    },
+    {
+      "epoch": 0.08481314603763583,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.823996108044218e-06,
+      "loss": 0.10819044,
+      "memory(GiB)": 74.41,
+      "step": 720,
+      "train_speed(iter/s)": 0.07484
+    },
+    {
+      "epoch": 0.08540212621845275,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.82155441404054e-06,
+      "loss": 0.10065824,
+      "memory(GiB)": 74.41,
+      "step": 725,
+      "train_speed(iter/s)": 0.074845
+    },
+    {
+      "epoch": 0.08599110639926967,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.819096207417848e-06,
+      "loss": 0.08238539,
+      "memory(GiB)": 74.41,
+      "step": 730,
+      "train_speed(iter/s)": 0.074846
+    },
+    {
+      "epoch": 0.08658008658008658,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.816621496594891e-06,
+      "loss": 0.10270094,
+      "memory(GiB)": 74.41,
+      "step": 735,
+      "train_speed(iter/s)": 0.074848
+    },
+    {
+      "epoch": 0.0871690667609035,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.814130290046933e-06,
+      "loss": 0.11216452,
+      "memory(GiB)": 74.41,
+      "step": 740,
+      "train_speed(iter/s)": 0.074848
+    },
+    {
+      "epoch": 0.08775804694172042,
+      "grad_norm": 4.375,
+      "learning_rate": 9.811622596305733e-06,
+      "loss": 0.08987691,
+      "memory(GiB)": 74.41,
+      "step": 745,
+      "train_speed(iter/s)": 0.074847
+    },
+    {
+      "epoch": 0.08834702712253732,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.809098423959513e-06,
+      "loss": 0.10679505,
+      "memory(GiB)": 74.41,
+      "step": 750,
+      "train_speed(iter/s)": 0.074847
+    },
+    {
+      "epoch": 0.08893600730335424,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.806557781652938e-06,
+      "loss": 0.09105108,
+      "memory(GiB)": 74.41,
+      "step": 755,
+      "train_speed(iter/s)": 0.074852
+    },
+    {
+      "epoch": 0.08952498748417116,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.804000678087069e-06,
+      "loss": 0.0892078,
+      "memory(GiB)": 74.41,
+      "step": 760,
+      "train_speed(iter/s)": 0.074853
+    },
+    {
+      "epoch": 0.09011396766498807,
+      "grad_norm": 4.625,
+      "learning_rate": 9.80142712201935e-06,
+      "loss": 0.10482442,
+      "memory(GiB)": 74.41,
+      "step": 765,
+      "train_speed(iter/s)": 0.07486
+    },
+    {
+      "epoch": 0.09070294784580499,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.798837122263563e-06,
+      "loss": 0.10331622,
+      "memory(GiB)": 74.41,
+      "step": 770,
+      "train_speed(iter/s)": 0.074863
+    },
+    {
+      "epoch": 0.09129192802662191,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.796230687689812e-06,
+      "loss": 0.11459937,
+      "memory(GiB)": 74.41,
+      "step": 775,
+      "train_speed(iter/s)": 0.074864
+    },
+    {
+      "epoch": 0.09188090820743881,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.793607827224485e-06,
+      "loss": 0.09125889,
+      "memory(GiB)": 74.41,
+      "step": 780,
+      "train_speed(iter/s)": 0.074869
+    },
+    {
+      "epoch": 0.09246988838825573,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.790968549850224e-06,
+      "loss": 0.11352601,
+      "memory(GiB)": 74.41,
+      "step": 785,
+      "train_speed(iter/s)": 0.074873
+    },
+    {
+      "epoch": 0.09305886856907265,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.788312864605893e-06,
+      "loss": 0.10174282,
+      "memory(GiB)": 74.41,
+      "step": 790,
+      "train_speed(iter/s)": 0.074874
+    },
+    {
+      "epoch": 0.09364784874988957,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.785640780586551e-06,
+      "loss": 0.09074538,
+      "memory(GiB)": 74.41,
+      "step": 795,
+      "train_speed(iter/s)": 0.074875
+    },
+    {
+      "epoch": 0.09423682893070648,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.782952306943418e-06,
+      "loss": 0.08487504,
+      "memory(GiB)": 74.41,
+      "step": 800,
+      "train_speed(iter/s)": 0.074881
+    },
+    {
+      "epoch": 0.0948258091115234,
+      "grad_norm": 4.625,
+      "learning_rate": 9.780247452883845e-06,
+      "loss": 0.10566314,
+      "memory(GiB)": 74.41,
+      "step": 805,
+      "train_speed(iter/s)": 0.074884
+    },
+    {
+      "epoch": 0.09541478929234032,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.77752622767128e-06,
+      "loss": 0.11225069,
+      "memory(GiB)": 74.41,
+      "step": 810,
+      "train_speed(iter/s)": 0.074886
+    },
+    {
+      "epoch": 0.09600376947315722,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.774788640625242e-06,
+      "loss": 0.08767406,
+      "memory(GiB)": 74.41,
+      "step": 815,
+      "train_speed(iter/s)": 0.074889
+    },
+    {
+      "epoch": 0.09659274965397414,
+      "grad_norm": 4.125,
+      "learning_rate": 9.772034701121278e-06,
+      "loss": 0.10022966,
+      "memory(GiB)": 74.41,
+      "step": 820,
+      "train_speed(iter/s)": 0.074892
+    },
+    {
+      "epoch": 0.09718172983479106,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.769264418590944e-06,
+      "loss": 0.11698074,
+      "memory(GiB)": 74.41,
+      "step": 825,
+      "train_speed(iter/s)": 0.074893
+    },
+    {
+      "epoch": 0.09777071001560797,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.76647780252177e-06,
+      "loss": 0.09307386,
+      "memory(GiB)": 74.41,
+      "step": 830,
+      "train_speed(iter/s)": 0.074895
+    },
+    {
+      "epoch": 0.09835969019642489,
+      "grad_norm": 4.75,
+      "learning_rate": 9.763674862457214e-06,
+      "loss": 0.09483533,
+      "memory(GiB)": 74.41,
+      "step": 835,
+      "train_speed(iter/s)": 0.074896
+    },
+    {
+      "epoch": 0.09894867037724181,
+      "grad_norm": 4.9375,
+      "learning_rate": 9.760855607996648e-06,
+      "loss": 0.10245972,
+      "memory(GiB)": 74.41,
+      "step": 840,
+      "train_speed(iter/s)": 0.074893
+    },
+    {
+      "epoch": 0.09953765055805872,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.758020048795313e-06,
+      "loss": 0.0943541,
+      "memory(GiB)": 74.41,
+      "step": 845,
+      "train_speed(iter/s)": 0.074898
+    },
+    {
+      "epoch": 0.10012663073887564,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.755168194564291e-06,
+      "loss": 0.09881738,
+      "memory(GiB)": 74.41,
+      "step": 850,
+      "train_speed(iter/s)": 0.074898
+    },
+    {
+      "epoch": 0.10071561091969256,
+      "grad_norm": 5.625,
+      "learning_rate": 9.752300055070472e-06,
+      "loss": 0.10225842,
+      "memory(GiB)": 74.41,
+      "step": 855,
+      "train_speed(iter/s)": 0.074904
+    },
+    {
+      "epoch": 0.10130459110050946,
+      "grad_norm": 3.75,
+      "learning_rate": 9.749415640136515e-06,
+      "loss": 0.09013618,
+      "memory(GiB)": 74.41,
+      "step": 860,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.10189357128132638,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.74651495964082e-06,
+      "loss": 0.09407109,
+      "memory(GiB)": 74.41,
+      "step": 865,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.1024825514621433,
+      "grad_norm": 4.5,
+      "learning_rate": 9.743598023517493e-06,
+      "loss": 0.10868713,
+      "memory(GiB)": 74.41,
+      "step": 870,
+      "train_speed(iter/s)": 0.074912
+    },
+    {
+      "epoch": 0.10307153164296022,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.740664841756313e-06,
+      "loss": 0.11703372,
+      "memory(GiB)": 74.41,
+      "step": 875,
+      "train_speed(iter/s)": 0.074918
+    },
+    {
+      "epoch": 0.10366051182377713,
+      "grad_norm": 3.484375,
+      "learning_rate": 9.737715424402691e-06,
+      "loss": 0.10799139,
+      "memory(GiB)": 74.41,
+      "step": 880,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.10424949200459405,
+      "grad_norm": 5.09375,
+      "learning_rate": 9.73474978155765e-06,
+      "loss": 0.10723754,
+      "memory(GiB)": 74.41,
+      "step": 885,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.10483847218541097,
+      "grad_norm": 3.75,
+      "learning_rate": 9.73176792337777e-06,
+      "loss": 0.11358409,
+      "memory(GiB)": 74.41,
+      "step": 890,
+      "train_speed(iter/s)": 0.074927
+    },
+    {
+      "epoch": 0.10542745236622787,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.728769860075175e-06,
+      "loss": 0.09777128,
+      "memory(GiB)": 74.41,
+      "step": 895,
+      "train_speed(iter/s)": 0.074932
+    },
+    {
+      "epoch": 0.10601643254704479,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.725755601917478e-06,
+      "loss": 0.10329788,
+      "memory(GiB)": 74.41,
+      "step": 900,
+      "train_speed(iter/s)": 0.074938
+    },
+    {
+      "epoch": 0.10660541272786171,
+      "grad_norm": 5.84375,
+      "learning_rate": 9.722725159227765e-06,
+      "loss": 0.10509249,
+      "memory(GiB)": 74.41,
+      "step": 905,
+      "train_speed(iter/s)": 0.074943
+    },
+    {
+      "epoch": 0.10719439290867862,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.71967854238454e-06,
+      "loss": 0.1047749,
+      "memory(GiB)": 74.41,
+      "step": 910,
+      "train_speed(iter/s)": 0.074947
+    },
+    {
+      "epoch": 0.10778337308949554,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.716615761821708e-06,
+      "loss": 0.10192933,
+      "memory(GiB)": 74.41,
+      "step": 915,
+      "train_speed(iter/s)": 0.074949
+    },
+    {
+      "epoch": 0.10837235327031246,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.713536828028524e-06,
+      "loss": 0.11062961,
+      "memory(GiB)": 74.41,
+      "step": 920,
+      "train_speed(iter/s)": 0.074953
+    },
+    {
+      "epoch": 0.10896133345112936,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.710441751549569e-06,
+      "loss": 0.09096059,
+      "memory(GiB)": 74.41,
+      "step": 925,
+      "train_speed(iter/s)": 0.074956
+    },
+    {
+      "epoch": 0.10955031363194628,
+      "grad_norm": 4.625,
+      "learning_rate": 9.707330542984706e-06,
+      "loss": 0.09158493,
+      "memory(GiB)": 74.41,
+      "step": 930,
+      "train_speed(iter/s)": 0.07496
+    },
+    {
+      "epoch": 0.1101392938127632,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.704203212989048e-06,
+      "loss": 0.10105819,
+      "memory(GiB)": 74.41,
+      "step": 935,
+      "train_speed(iter/s)": 0.07496
+    },
+    {
+      "epoch": 0.11072827399358011,
+      "grad_norm": 5.34375,
+      "learning_rate": 9.701059772272915e-06,
+      "loss": 0.09538058,
+      "memory(GiB)": 74.41,
+      "step": 940,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.11131725417439703,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.69790023160181e-06,
+      "loss": 0.10084579,
+      "memory(GiB)": 74.41,
+      "step": 945,
+      "train_speed(iter/s)": 0.074966
+    },
+    {
+      "epoch": 0.11190623435521395,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.69472460179637e-06,
+      "loss": 0.10691296,
+      "memory(GiB)": 74.41,
+      "step": 950,
+      "train_speed(iter/s)": 0.074968
+    },
+    {
+      "epoch": 0.11249521453603087,
+      "grad_norm": 4.25,
+      "learning_rate": 9.691532893732331e-06,
+      "loss": 0.096218,
+      "memory(GiB)": 74.41,
+      "step": 955,
+      "train_speed(iter/s)": 0.074971
+    },
+    {
+      "epoch": 0.11308419471684777,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.688325118340496e-06,
+      "loss": 0.10754682,
+      "memory(GiB)": 74.41,
+      "step": 960,
+      "train_speed(iter/s)": 0.074974
+    },
+    {
+      "epoch": 0.1136731748976647,
+      "grad_norm": 3.875,
+      "learning_rate": 9.685101286606695e-06,
+      "loss": 0.10555474,
+      "memory(GiB)": 74.41,
+      "step": 965,
+      "train_speed(iter/s)": 0.074978
+    },
+    {
+      "epoch": 0.11426215507848161,
+      "grad_norm": 3.5625,
+      "learning_rate": 9.68186140957175e-06,
+      "loss": 0.08726187,
+      "memory(GiB)": 74.41,
+      "step": 970,
+      "train_speed(iter/s)": 0.074983
+    },
+    {
+      "epoch": 0.11485113525929852,
+      "grad_norm": 3.84375,
+      "learning_rate": 9.678605498331422e-06,
+      "loss": 0.08722601,
+      "memory(GiB)": 74.41,
+      "step": 975,
+      "train_speed(iter/s)": 0.074988
+    },
+    {
+      "epoch": 0.11544011544011544,
+      "grad_norm": 4.78125,
+      "learning_rate": 9.675333564036398e-06,
+      "loss": 0.09676975,
+      "memory(GiB)": 74.41,
+      "step": 980,
+      "train_speed(iter/s)": 0.07499
+    },
+    {
+      "epoch": 0.11602909562093236,
+      "grad_norm": 4.875,
+      "learning_rate": 9.672045617892235e-06,
+      "loss": 0.0948953,
+      "memory(GiB)": 74.41,
+      "step": 985,
+      "train_speed(iter/s)": 0.074991
+    },
+    {
+      "epoch": 0.11661807580174927,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.668741671159329e-06,
+      "loss": 0.10143999,
+      "memory(GiB)": 74.41,
+      "step": 990,
+      "train_speed(iter/s)": 0.074992
+    },
+    {
+      "epoch": 0.11720705598256619,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.66542173515287e-06,
+      "loss": 0.08537531,
+      "memory(GiB)": 74.41,
+      "step": 995,
+      "train_speed(iter/s)": 0.074997
+    },
+    {
+      "epoch": 0.1177960361633831,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.662085821242807e-06,
+      "loss": 0.0954556,
+      "memory(GiB)": 74.41,
+      "step": 1000,
+      "train_speed(iter/s)": 0.074998
+    },
+    {
+      "epoch": 0.11838501634420001,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.658733940853815e-06,
+      "loss": 0.09058597,
+      "memory(GiB)": 74.41,
+      "step": 1005,
+      "train_speed(iter/s)": 0.074817
+    },
+    {
+      "epoch": 0.11897399652501693,
+      "grad_norm": 4.53125,
+      "learning_rate": 9.655366105465244e-06,
+      "loss": 0.09935067,
+      "memory(GiB)": 74.41,
+      "step": 1010,
+      "train_speed(iter/s)": 0.07482
+    },
+    {
+      "epoch": 0.11956297670583385,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.65198232661109e-06,
+      "loss": 0.08947166,
+      "memory(GiB)": 74.41,
+      "step": 1015,
+      "train_speed(iter/s)": 0.074825
+    },
+    {
+      "epoch": 0.12015195688665077,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.648582615879952e-06,
+      "loss": 0.10199349,
+      "memory(GiB)": 74.41,
+      "step": 1020,
+      "train_speed(iter/s)": 0.074824
+    },
+    {
+      "epoch": 0.12074093706746768,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.645166984914985e-06,
+      "loss": 0.10013585,
+      "memory(GiB)": 74.41,
+      "step": 1025,
+      "train_speed(iter/s)": 0.074828
+    },
+    {
+      "epoch": 0.1213299172482846,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.641735445413875e-06,
+      "loss": 0.11086836,
+      "memory(GiB)": 74.41,
+      "step": 1030,
+      "train_speed(iter/s)": 0.074831
+    },
+    {
+      "epoch": 0.12191889742910152,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.638288009128789e-06,
+      "loss": 0.09356855,
+      "memory(GiB)": 74.41,
+      "step": 1035,
+      "train_speed(iter/s)": 0.074832
+    },
+    {
+      "epoch": 0.12250787760991842,
+      "grad_norm": 3.5,
+      "learning_rate": 9.634824687866329e-06,
+      "loss": 0.09086785,
+      "memory(GiB)": 74.41,
+      "step": 1040,
+      "train_speed(iter/s)": 0.074832
+    },
+    {
+      "epoch": 0.12309685779073534,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.631345493487509e-06,
+      "loss": 0.1021161,
+      "memory(GiB)": 74.41,
+      "step": 1045,
+      "train_speed(iter/s)": 0.074834
+    },
+    {
+      "epoch": 0.12368583797155226,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.6278504379077e-06,
+      "loss": 0.10499916,
+      "memory(GiB)": 74.41,
+      "step": 1050,
+      "train_speed(iter/s)": 0.074835
+    },
+    {
+      "epoch": 0.12427481815236917,
+      "grad_norm": 3.5,
+      "learning_rate": 9.624339533096594e-06,
+      "loss": 0.08981788,
+      "memory(GiB)": 74.41,
+      "step": 1055,
+      "train_speed(iter/s)": 0.07484
+    },
+    {
+      "epoch": 0.12486379833318609,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.620812791078161e-06,
+      "loss": 0.08886168,
+      "memory(GiB)": 74.41,
+      "step": 1060,
+      "train_speed(iter/s)": 0.074844
+    },
+    {
+      "epoch": 0.125452778514003,
+      "grad_norm": 3.828125,
+      "learning_rate": 9.617270223930613e-06,
+      "loss": 0.08076782,
+      "memory(GiB)": 74.41,
+      "step": 1065,
+      "train_speed(iter/s)": 0.074847
+    },
+    {
+      "epoch": 0.12604175869481993,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.613711843786356e-06,
+      "loss": 0.08811791,
+      "memory(GiB)": 74.41,
+      "step": 1070,
+      "train_speed(iter/s)": 0.07485
+    },
+    {
+      "epoch": 0.12663073887563683,
+      "grad_norm": 5.71875,
+      "learning_rate": 9.610137662831953e-06,
+      "loss": 0.10126686,
+      "memory(GiB)": 74.41,
+      "step": 1075,
+      "train_speed(iter/s)": 0.074854
+    },
+    {
+      "epoch": 0.12721971905645374,
+      "grad_norm": 4.375,
+      "learning_rate": 9.60654769330808e-06,
+      "loss": 0.09718223,
+      "memory(GiB)": 74.41,
+      "step": 1080,
+      "train_speed(iter/s)": 0.074858
+    },
+    {
+      "epoch": 0.12780869923727067,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.60294194750949e-06,
+      "loss": 0.09517021,
+      "memory(GiB)": 74.41,
+      "step": 1085,
+      "train_speed(iter/s)": 0.074863
+    },
+    {
+      "epoch": 0.12839767941808758,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.599320437784953e-06,
+      "loss": 0.09423549,
+      "memory(GiB)": 74.41,
+      "step": 1090,
+      "train_speed(iter/s)": 0.074865
+    },
+    {
+      "epoch": 0.12898665959890449,
+      "grad_norm": 4.90625,
+      "learning_rate": 9.595683176537242e-06,
+      "loss": 0.10256985,
+      "memory(GiB)": 74.41,
+      "step": 1095,
+      "train_speed(iter/s)": 0.074868
+    },
+    {
+      "epoch": 0.12957563977972142,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.592030176223066e-06,
+      "loss": 0.08859265,
+      "memory(GiB)": 74.41,
+      "step": 1100,
+      "train_speed(iter/s)": 0.074869
+    },
+    {
+      "epoch": 0.13016461996053832,
+      "grad_norm": 4.40625,
+      "learning_rate": 9.588361449353036e-06,
+      "loss": 0.08601279,
+      "memory(GiB)": 74.41,
+      "step": 1105,
+      "train_speed(iter/s)": 0.074871
+    },
+    {
+      "epoch": 0.13075360014135523,
+      "grad_norm": 4.625,
+      "learning_rate": 9.584677008491628e-06,
+      "loss": 0.12187407,
+      "memory(GiB)": 74.41,
+      "step": 1110,
+      "train_speed(iter/s)": 0.074876
+    },
+    {
+      "epoch": 0.13134258032217216,
+      "grad_norm": 3.34375,
+      "learning_rate": 9.580976866257129e-06,
+      "loss": 0.09723064,
+      "memory(GiB)": 74.41,
+      "step": 1115,
+      "train_speed(iter/s)": 0.074877
+    },
+    {
+      "epoch": 0.13193156050298907,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.577261035321601e-06,
+      "loss": 0.09539672,
+      "memory(GiB)": 74.41,
+      "step": 1120,
+      "train_speed(iter/s)": 0.074878
+    },
+    {
+      "epoch": 0.13252054068380598,
+      "grad_norm": 4.25,
+      "learning_rate": 9.573529528410838e-06,
+      "loss": 0.09372305,
+      "memory(GiB)": 74.41,
+      "step": 1125,
+      "train_speed(iter/s)": 0.074879
+    },
+    {
+      "epoch": 0.1331095208646229,
+      "grad_norm": 3.296875,
+      "learning_rate": 9.569782358304319e-06,
+      "loss": 0.07629693,
+      "memory(GiB)": 74.41,
+      "step": 1130,
+      "train_speed(iter/s)": 0.074883
+    },
+    {
+      "epoch": 0.13369850104543982,
+      "grad_norm": 5.03125,
+      "learning_rate": 9.566019537835166e-06,
+      "loss": 0.0996025,
+      "memory(GiB)": 74.41,
+      "step": 1135,
+      "train_speed(iter/s)": 0.074886
+    },
+    {
+      "epoch": 0.13428748122625675,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.562241079890093e-06,
+      "loss": 0.09327987,
+      "memory(GiB)": 74.41,
+      "step": 1140,
+      "train_speed(iter/s)": 0.074887
+    },
+    {
+      "epoch": 0.13487646140707366,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.558446997409382e-06,
+      "loss": 0.09146149,
+      "memory(GiB)": 74.41,
+      "step": 1145,
+      "train_speed(iter/s)": 0.074889
+    },
+    {
+      "epoch": 0.13546544158789056,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.554637303386813e-06,
+      "loss": 0.08869413,
+      "memory(GiB)": 74.41,
+      "step": 1150,
+      "train_speed(iter/s)": 0.074893
+    },
+    {
+      "epoch": 0.1360544217687075,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.550812010869636e-06,
+      "loss": 0.08775733,
+      "memory(GiB)": 74.41,
+      "step": 1155,
+      "train_speed(iter/s)": 0.074892
+    },
+    {
+      "epoch": 0.1366434019495244,
+      "grad_norm": 5.0,
+      "learning_rate": 9.546971132958524e-06,
+      "loss": 0.09848154,
+      "memory(GiB)": 74.41,
+      "step": 1160,
+      "train_speed(iter/s)": 0.074895
+    },
+    {
+      "epoch": 0.1372323821303413,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.543114682807525e-06,
+      "loss": 0.09634271,
+      "memory(GiB)": 74.41,
+      "step": 1165,
+      "train_speed(iter/s)": 0.074894
+    },
+    {
+      "epoch": 0.13782136231115824,
+      "grad_norm": 4.375,
+      "learning_rate": 9.539242673624012e-06,
+      "loss": 0.10609367,
+      "memory(GiB)": 74.41,
+      "step": 1170,
+      "train_speed(iter/s)": 0.074897
+    },
+    {
+      "epoch": 0.13841034249197515,
+      "grad_norm": 3.796875,
+      "learning_rate": 9.535355118668652e-06,
+      "loss": 0.10536243,
+      "memory(GiB)": 74.41,
+      "step": 1175,
+      "train_speed(iter/s)": 0.074901
+    },
+    {
+      "epoch": 0.13899932267279205,
+      "grad_norm": 3.578125,
+      "learning_rate": 9.53145203125535e-06,
+      "loss": 0.08514765,
+      "memory(GiB)": 74.41,
+      "step": 1180,
+      "train_speed(iter/s)": 0.074904
+    },
+    {
+      "epoch": 0.139588302853609,
+      "grad_norm": 3.765625,
+      "learning_rate": 9.527533424751206e-06,
+      "loss": 0.09435316,
+      "memory(GiB)": 74.41,
+      "step": 1185,
+      "train_speed(iter/s)": 0.074905
+    },
+    {
+      "epoch": 0.1401772830344259,
+      "grad_norm": 3.03125,
+      "learning_rate": 9.523599312576464e-06,
+      "loss": 0.0877773,
+      "memory(GiB)": 74.41,
+      "step": 1190,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.1407662632152428,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.519649708204478e-06,
+      "loss": 0.08990437,
+      "memory(GiB)": 74.41,
+      "step": 1195,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.14135524339605973,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.515684625161656e-06,
+      "loss": 0.11489943,
+      "memory(GiB)": 74.41,
+      "step": 1200,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.14194422357687664,
+      "grad_norm": 7.59375,
+      "learning_rate": 9.511704077027414e-06,
+      "loss": 0.0885065,
+      "memory(GiB)": 74.41,
+      "step": 1205,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.14253320375769354,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.507708077434138e-06,
+      "loss": 0.08773028,
+      "memory(GiB)": 74.41,
+      "step": 1210,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.14312218393851048,
+      "grad_norm": 2.84375,
+      "learning_rate": 9.503696640067129e-06,
+      "loss": 0.09397745,
+      "memory(GiB)": 74.41,
+      "step": 1215,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.14371116411932738,
+      "grad_norm": 5.1875,
+      "learning_rate": 9.499669778664553e-06,
+      "loss": 0.10290816,
+      "memory(GiB)": 74.41,
+      "step": 1220,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.1443001443001443,
+      "grad_norm": 2.59375,
+      "learning_rate": 9.495627507017407e-06,
+      "loss": 0.09706104,
+      "memory(GiB)": 74.41,
+      "step": 1225,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.14488912448096122,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.491569838969459e-06,
+      "loss": 0.08967851,
+      "memory(GiB)": 74.41,
+      "step": 1230,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.14547810466177813,
+      "grad_norm": 3.875,
+      "learning_rate": 9.487496788417211e-06,
+      "loss": 0.07626108,
+      "memory(GiB)": 74.41,
+      "step": 1235,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.14606708484259504,
+      "grad_norm": 5.125,
+      "learning_rate": 9.483408369309843e-06,
+      "loss": 0.11157024,
+      "memory(GiB)": 74.41,
+      "step": 1240,
+      "train_speed(iter/s)": 0.074906
+    },
+    {
+      "epoch": 0.14665606502341197,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.479304595649166e-06,
+      "loss": 0.09179769,
+      "memory(GiB)": 74.41,
+      "step": 1245,
+      "train_speed(iter/s)": 0.074905
+    },
+    {
+      "epoch": 0.14724504520422887,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.475185481489581e-06,
+      "loss": 0.10004801,
+      "memory(GiB)": 74.41,
+      "step": 1250,
+      "train_speed(iter/s)": 0.074906
+    },
+    {
+      "epoch": 0.14783402538504578,
+      "grad_norm": 3.265625,
+      "learning_rate": 9.471051040938026e-06,
+      "loss": 0.09620751,
+      "memory(GiB)": 74.41,
+      "step": 1255,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.14842300556586271,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.466901288153927e-06,
+      "loss": 0.08952096,
+      "memory(GiB)": 74.41,
+      "step": 1260,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.14901198574667962,
+      "grad_norm": 3.59375,
+      "learning_rate": 9.462736237349145e-06,
+      "loss": 0.08996918,
+      "memory(GiB)": 74.41,
+      "step": 1265,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.14960096592749653,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.458555902787945e-06,
+      "loss": 0.09044787,
+      "memory(GiB)": 74.41,
+      "step": 1270,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.15018994610831346,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.454360298786924e-06,
+      "loss": 0.08726504,
+      "memory(GiB)": 74.41,
+      "step": 1275,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.15077892628913037,
+      "grad_norm": 4.96875,
+      "learning_rate": 9.45014943971498e-06,
+      "loss": 0.0934383,
+      "memory(GiB)": 74.41,
+      "step": 1280,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.1513679064699473,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.44592333999325e-06,
+      "loss": 0.08363155,
+      "memory(GiB)": 74.41,
+      "step": 1285,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.1519568866507642,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.441682014095071e-06,
+      "loss": 0.08424637,
+      "memory(GiB)": 74.41,
+      "step": 1290,
+      "train_speed(iter/s)": 0.074918
+    },
+    {
+      "epoch": 0.1525458668315811,
+      "grad_norm": 3.734375,
+      "learning_rate": 9.437425476545926e-06,
+      "loss": 0.09210879,
+      "memory(GiB)": 74.41,
+      "step": 1295,
+      "train_speed(iter/s)": 0.074919
+    },
+    {
+      "epoch": 0.15313484701239805,
+      "grad_norm": 3.234375,
+      "learning_rate": 9.43315374192339e-06,
+      "loss": 0.09213892,
+      "memory(GiB)": 74.41,
+      "step": 1300,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.15372382719321495,
+      "grad_norm": 4.5,
+      "learning_rate": 9.428866824857087e-06,
+      "loss": 0.08642803,
+      "memory(GiB)": 74.41,
+      "step": 1305,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.15431280737403186,
+      "grad_norm": 4.0,
+      "learning_rate": 9.424564740028638e-06,
+      "loss": 0.08265491,
+      "memory(GiB)": 74.41,
+      "step": 1310,
+      "train_speed(iter/s)": 0.074928
+    },
+    {
+      "epoch": 0.1549017875548488,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.420247502171606e-06,
+      "loss": 0.08584201,
+      "memory(GiB)": 74.41,
+      "step": 1315,
+      "train_speed(iter/s)": 0.07493
+    },
+    {
+      "epoch": 0.1554907677356657,
+      "grad_norm": 5.21875,
+      "learning_rate": 9.415915126071456e-06,
+      "loss": 0.09583381,
+      "memory(GiB)": 74.41,
+      "step": 1320,
+      "train_speed(iter/s)": 0.074932
+    },
+    {
+      "epoch": 0.1560797479164826,
+      "grad_norm": 5.8125,
+      "learning_rate": 9.41156762656549e-06,
+      "loss": 0.07853642,
+      "memory(GiB)": 74.41,
+      "step": 1325,
+      "train_speed(iter/s)": 0.074936
+    },
+    {
+      "epoch": 0.15666872809729954,
+      "grad_norm": 4.4375,
+      "learning_rate": 9.407205018542807e-06,
+      "loss": 0.11025757,
+      "memory(GiB)": 74.41,
+      "step": 1330,
+      "train_speed(iter/s)": 0.07494
+    },
+    {
+      "epoch": 0.15725770827811644,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.402827316944253e-06,
+      "loss": 0.10197206,
+      "memory(GiB)": 74.41,
+      "step": 1335,
+      "train_speed(iter/s)": 0.074941
+    },
+    {
+      "epoch": 0.15784668845893335,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.398434536762359e-06,
+      "loss": 0.09212865,
+      "memory(GiB)": 74.41,
+      "step": 1340,
+      "train_speed(iter/s)": 0.074943
+    },
+    {
+      "epoch": 0.15843566863975028,
+      "grad_norm": 3.703125,
+      "learning_rate": 9.3940266930413e-06,
+      "loss": 0.08865182,
+      "memory(GiB)": 74.41,
+      "step": 1345,
+      "train_speed(iter/s)": 0.074945
+    },
+    {
+      "epoch": 0.1590246488205672,
+      "grad_norm": 2.890625,
+      "learning_rate": 9.389603800876838e-06,
+      "loss": 0.08562068,
+      "memory(GiB)": 74.41,
+      "step": 1350,
+      "train_speed(iter/s)": 0.074948
+    },
+    {
+      "epoch": 0.1596136290013841,
+      "grad_norm": 4.6875,
+      "learning_rate": 9.385165875416278e-06,
+      "loss": 0.1068722,
+      "memory(GiB)": 74.41,
+      "step": 1355,
+      "train_speed(iter/s)": 0.074947
+    },
+    {
+      "epoch": 0.16020260918220103,
+      "grad_norm": 3.5,
+      "learning_rate": 9.3807129318584e-06,
+      "loss": 0.09370728,
+      "memory(GiB)": 74.41,
+      "step": 1360,
+      "train_speed(iter/s)": 0.074948
+    },
+    {
+      "epoch": 0.16079158936301793,
+      "grad_norm": 5.0625,
+      "learning_rate": 9.376244985453427e-06,
+      "loss": 0.08163351,
+      "memory(GiB)": 74.41,
+      "step": 1365,
+      "train_speed(iter/s)": 0.07495
+    },
+    {
+      "epoch": 0.16138056954383484,
+      "grad_norm": 5.0,
+      "learning_rate": 9.371762051502957e-06,
+      "loss": 0.10225484,
+      "memory(GiB)": 74.41,
+      "step": 1370,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.16196954972465177,
+      "grad_norm": 3.5,
+      "learning_rate": 9.36726414535992e-06,
+      "loss": 0.10129716,
+      "memory(GiB)": 74.41,
+      "step": 1375,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.16255852990546868,
+      "grad_norm": 3.640625,
+      "learning_rate": 9.362751282428519e-06,
+      "loss": 0.08826185,
+      "memory(GiB)": 74.41,
+      "step": 1380,
+      "train_speed(iter/s)": 0.074953
+    },
+    {
+      "epoch": 0.16314751008628559,
+      "grad_norm": 3.890625,
+      "learning_rate": 9.35822347816418e-06,
+      "loss": 0.10849725,
+      "memory(GiB)": 74.41,
+      "step": 1385,
+      "train_speed(iter/s)": 0.074955
+    },
+    {
+      "epoch": 0.16373649026710252,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.353680748073509e-06,
+      "loss": 0.10136243,
+      "memory(GiB)": 74.41,
+      "step": 1390,
+      "train_speed(iter/s)": 0.074957
+    },
+    {
+      "epoch": 0.16432547044791943,
+      "grad_norm": 3.75,
+      "learning_rate": 9.349123107714211e-06,
+      "loss": 0.08352109,
+      "memory(GiB)": 74.41,
+      "step": 1395,
+      "train_speed(iter/s)": 0.07496
+    },
+    {
+      "epoch": 0.16491445062873633,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.344550572695073e-06,
+      "loss": 0.1029326,
+      "memory(GiB)": 74.41,
+      "step": 1400,
+      "train_speed(iter/s)": 0.074961
+    },
+    {
+      "epoch": 0.16550343080955326,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.33996315867588e-06,
+      "loss": 0.10244679,
+      "memory(GiB)": 74.41,
+      "step": 1405,
+      "train_speed(iter/s)": 0.074964
+    },
+    {
+      "epoch": 0.16609241099037017,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.335360881367381e-06,
+      "loss": 0.09310257,
+      "memory(GiB)": 74.41,
+      "step": 1410,
+      "train_speed(iter/s)": 0.074966
+    },
+    {
+      "epoch": 0.16668139117118708,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.330743756531227e-06,
+      "loss": 0.09983196,
+      "memory(GiB)": 74.41,
+      "step": 1415,
+      "train_speed(iter/s)": 0.074968
+    },
+    {
+      "epoch": 0.167270371352004,
+      "grad_norm": 3.453125,
+      "learning_rate": 9.326111799979915e-06,
+      "loss": 0.09484429,
+      "memory(GiB)": 74.41,
+      "step": 1420,
+      "train_speed(iter/s)": 0.074967
+    },
+    {
+      "epoch": 0.16785935153282092,
+      "grad_norm": 4.25,
+      "learning_rate": 9.321465027576738e-06,
+      "loss": 0.08307941,
+      "memory(GiB)": 74.41,
+      "step": 1425,
+      "train_speed(iter/s)": 0.074967
+    },
+    {
+      "epoch": 0.16844833171363785,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.316803455235732e-06,
+      "loss": 0.09872603,
+      "memory(GiB)": 74.41,
+      "step": 1430,
+      "train_speed(iter/s)": 0.074966
+    },
+    {
+      "epoch": 0.16903731189445476,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.312127098921614e-06,
+      "loss": 0.09692515,
+      "memory(GiB)": 74.41,
+      "step": 1435,
+      "train_speed(iter/s)": 0.074965
+    },
+    {
+      "epoch": 0.16962629207527166,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.307435974649744e-06,
+      "loss": 0.09557692,
+      "memory(GiB)": 74.41,
+      "step": 1440,
+      "train_speed(iter/s)": 0.074968
+    },
+    {
+      "epoch": 0.1702152722560886,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.30273009848604e-06,
+      "loss": 0.10825093,
+      "memory(GiB)": 74.41,
+      "step": 1445,
+      "train_speed(iter/s)": 0.074971
+    },
+    {
+      "epoch": 0.1708042524369055,
+      "grad_norm": 4.03125,
+      "learning_rate": 9.29800948654696e-06,
+      "loss": 0.09957713,
+      "memory(GiB)": 74.41,
+      "step": 1450,
+      "train_speed(iter/s)": 0.074971
+    },
+    {
+      "epoch": 0.1713932326177224,
+      "grad_norm": 3.71875,
+      "learning_rate": 9.293274154999415e-06,
+      "loss": 0.08825269,
+      "memory(GiB)": 74.41,
+      "step": 1455,
+      "train_speed(iter/s)": 0.074974
+    },
+    {
+      "epoch": 0.17198221279853934,
+      "grad_norm": 3.6875,
+      "learning_rate": 9.288524120060733e-06,
+      "loss": 0.1051899,
+      "memory(GiB)": 74.41,
+      "step": 1460,
+      "train_speed(iter/s)": 0.074977
+    },
+    {
+      "epoch": 0.17257119297935625,
+      "grad_norm": 3.859375,
+      "learning_rate": 9.2837593979986e-06,
+      "loss": 0.09597876,
+      "memory(GiB)": 74.41,
+      "step": 1465,
+      "train_speed(iter/s)": 0.07498
+    },
+    {
+      "epoch": 0.17316017316017315,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.278980005130996e-06,
+      "loss": 0.10092977,
+      "memory(GiB)": 74.41,
+      "step": 1470,
+      "train_speed(iter/s)": 0.074982
+    },
+    {
+      "epoch": 0.1737491533409901,
+      "grad_norm": 4.0,
+      "learning_rate": 9.274185957826147e-06,
+      "loss": 0.10035901,
+      "memory(GiB)": 74.41,
+      "step": 1475,
+      "train_speed(iter/s)": 0.074984
+    },
+    {
+      "epoch": 0.174338133521807,
+      "grad_norm": 4.125,
+      "learning_rate": 9.26937727250247e-06,
+      "loss": 0.08626782,
+      "memory(GiB)": 74.41,
+      "step": 1480,
+      "train_speed(iter/s)": 0.074985
+    },
+    {
+      "epoch": 0.1749271137026239,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.264553965628506e-06,
+      "loss": 0.087044,
+      "memory(GiB)": 74.41,
+      "step": 1485,
+      "train_speed(iter/s)": 0.074984
+    },
+    {
+      "epoch": 0.17551609388344083,
+      "grad_norm": 4.09375,
+      "learning_rate": 9.259716053722878e-06,
+      "loss": 0.10160702,
+      "memory(GiB)": 74.41,
+      "step": 1490,
+      "train_speed(iter/s)": 0.074985
+    },
+    {
+      "epoch": 0.17610507406425774,
+      "grad_norm": 3.875,
+      "learning_rate": 9.254863553354224e-06,
+      "loss": 0.07840201,
+      "memory(GiB)": 74.41,
+      "step": 1495,
+      "train_speed(iter/s)": 0.074988
+    },
+    {
+      "epoch": 0.17669405424507464,
+      "grad_norm": 3.25,
+      "learning_rate": 9.249996481141149e-06,
+      "loss": 0.08488679,
+      "memory(GiB)": 74.41,
+      "step": 1500,
+      "train_speed(iter/s)": 0.074991
+    },
+    {
+      "epoch": 0.17728303442589158,
+      "grad_norm": 3.96875,
+      "learning_rate": 9.245114853752157e-06,
+      "loss": 0.10635732,
+      "memory(GiB)": 74.41,
+      "step": 1505,
+      "train_speed(iter/s)": 0.074857
+    },
+    {
+      "epoch": 0.17787201460670848,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.240218687905598e-06,
+      "loss": 0.08825208,
+      "memory(GiB)": 74.41,
+      "step": 1510,
+      "train_speed(iter/s)": 0.074857
+    },
+    {
+      "epoch": 0.1784609947875254,
+      "grad_norm": 4.28125,
+      "learning_rate": 9.235308000369623e-06,
+      "loss": 0.09694942,
+      "memory(GiB)": 74.41,
+      "step": 1515,
+      "train_speed(iter/s)": 0.074857
+    },
+    {
+      "epoch": 0.17904997496834232,
+      "grad_norm": 4.21875,
+      "learning_rate": 9.230382807962106e-06,
+      "loss": 0.10525234,
+      "memory(GiB)": 74.41,
+      "step": 1520,
+      "train_speed(iter/s)": 0.074859
+    },
+    {
+      "epoch": 0.17963895514915923,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.225443127550601e-06,
+      "loss": 0.10098197,
+      "memory(GiB)": 74.41,
+      "step": 1525,
+      "train_speed(iter/s)": 0.07486
+    },
+    {
+      "epoch": 0.18022793532997614,
+      "grad_norm": 4.9375,
+      "learning_rate": 9.220488976052279e-06,
+      "loss": 0.09873599,
+      "memory(GiB)": 74.41,
+      "step": 1530,
+      "train_speed(iter/s)": 0.074861
+    },
+    {
+      "epoch": 0.18081691551079307,
+      "grad_norm": 3.546875,
+      "learning_rate": 9.21552037043387e-06,
+      "loss": 0.08211206,
+      "memory(GiB)": 74.41,
+      "step": 1535,
+      "train_speed(iter/s)": 0.074863
+    },
+    {
+      "epoch": 0.18140589569160998,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.210537327711608e-06,
+      "loss": 0.09001173,
+      "memory(GiB)": 74.41,
+      "step": 1540,
+      "train_speed(iter/s)": 0.074863
+    },
+    {
+      "epoch": 0.18199487587242688,
+      "grad_norm": 5.53125,
+      "learning_rate": 9.205539864951169e-06,
+      "loss": 0.1033685,
+      "memory(GiB)": 74.41,
+      "step": 1545,
+      "train_speed(iter/s)": 0.074866
+    },
+    {
+      "epoch": 0.18258385605324381,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.200527999267614e-06,
+      "loss": 0.09862096,
+      "memory(GiB)": 74.41,
+      "step": 1550,
+      "train_speed(iter/s)": 0.074867
+    },
+    {
+      "epoch": 0.18317283623406072,
+      "grad_norm": 3.609375,
+      "learning_rate": 9.19550174782533e-06,
+      "loss": 0.07522287,
+      "memory(GiB)": 74.41,
+      "step": 1555,
+      "train_speed(iter/s)": 0.074869
+    },
+    {
+      "epoch": 0.18376181641487763,
+      "grad_norm": 4.15625,
+      "learning_rate": 9.190461127837971e-06,
+      "loss": 0.07987702,
+      "memory(GiB)": 74.41,
+      "step": 1560,
+      "train_speed(iter/s)": 0.074872
+    },
+    {
+      "epoch": 0.18435079659569456,
+      "grad_norm": 3.1875,
+      "learning_rate": 9.185406156568404e-06,
+      "loss": 0.08131886,
+      "memory(GiB)": 74.41,
+      "step": 1565,
+      "train_speed(iter/s)": 0.074875
+    },
+    {
+      "epoch": 0.18493977677651147,
+      "grad_norm": 3.15625,
+      "learning_rate": 9.18033685132864e-06,
+      "loss": 0.09074624,
+      "memory(GiB)": 74.41,
+      "step": 1570,
+      "train_speed(iter/s)": 0.074877
+    },
+    {
+      "epoch": 0.18552875695732837,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.175253229479784e-06,
+      "loss": 0.08689606,
+      "memory(GiB)": 74.41,
+      "step": 1575,
+      "train_speed(iter/s)": 0.074878
+    },
+    {
+      "epoch": 0.1861177371381453,
+      "grad_norm": 2.953125,
+      "learning_rate": 9.17015530843197e-06,
+      "loss": 0.09036893,
+      "memory(GiB)": 74.41,
+      "step": 1580,
+      "train_speed(iter/s)": 0.07488
+    },
+    {
+      "epoch": 0.1867067173189622,
+      "grad_norm": 3.921875,
+      "learning_rate": 9.165043105644303e-06,
+      "loss": 0.08858083,
+      "memory(GiB)": 74.41,
+      "step": 1585,
+      "train_speed(iter/s)": 0.074883
+    },
+    {
+      "epoch": 0.18729569749977915,
+      "grad_norm": 3.65625,
+      "learning_rate": 9.159916638624801e-06,
+      "loss": 0.08610859,
+      "memory(GiB)": 74.41,
+      "step": 1590,
+      "train_speed(iter/s)": 0.074887
+    },
+    {
+      "epoch": 0.18788467768059605,
+      "grad_norm": 3.4375,
+      "learning_rate": 9.154775924930333e-06,
+      "loss": 0.09216586,
+      "memory(GiB)": 74.41,
+      "step": 1595,
+      "train_speed(iter/s)": 0.074888
+    },
+    {
+      "epoch": 0.18847365786141296,
+      "grad_norm": 4.59375,
+      "learning_rate": 9.149620982166557e-06,
+      "loss": 0.09575672,
+      "memory(GiB)": 74.41,
+      "step": 1600,
+      "train_speed(iter/s)": 0.074892
+    },
+    {
+      "epoch": 0.1890626380422299,
+      "grad_norm": 3.5,
+      "learning_rate": 9.144451827987865e-06,
+      "loss": 0.10303175,
+      "memory(GiB)": 74.41,
+      "step": 1605,
+      "train_speed(iter/s)": 0.074892
+    },
+    {
+      "epoch": 0.1896516182230468,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.139268480097317e-06,
+      "loss": 0.10116973,
+      "memory(GiB)": 74.41,
+      "step": 1610,
+      "train_speed(iter/s)": 0.074894
+    },
+    {
+      "epoch": 0.1902405984038637,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.134070956246586e-06,
+      "loss": 0.08978627,
+      "memory(GiB)": 74.41,
+      "step": 1615,
+      "train_speed(iter/s)": 0.074895
+    },
+    {
+      "epoch": 0.19082957858468064,
+      "grad_norm": 3.46875,
+      "learning_rate": 9.128859274235893e-06,
+      "loss": 0.09291114,
+      "memory(GiB)": 74.41,
+      "step": 1620,
+      "train_speed(iter/s)": 0.074897
+    },
+    {
+      "epoch": 0.19141855876549754,
+      "grad_norm": 4.1875,
+      "learning_rate": 9.123633451913945e-06,
+      "loss": 0.0884718,
+      "memory(GiB)": 74.41,
+      "step": 1625,
+      "train_speed(iter/s)": 0.074899
+    },
+    {
+      "epoch": 0.19200753894631445,
+      "grad_norm": 3.78125,
+      "learning_rate": 9.118393507177877e-06,
+      "loss": 0.08941454,
+      "memory(GiB)": 74.41,
+      "step": 1630,
+      "train_speed(iter/s)": 0.074901
+    },
+    {
+      "epoch": 0.19259651912713138,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.11313945797319e-06,
+      "loss": 0.08903505,
+      "memory(GiB)": 74.41,
+      "step": 1635,
+      "train_speed(iter/s)": 0.074903
+    },
+    {
+      "epoch": 0.1931854993079483,
+      "grad_norm": 3.390625,
+      "learning_rate": 9.107871322293694e-06,
+      "loss": 0.06867069,
+      "memory(GiB)": 74.41,
+      "step": 1640,
+      "train_speed(iter/s)": 0.074906
+    },
+    {
+      "epoch": 0.1937744794887652,
+      "grad_norm": 4.25,
+      "learning_rate": 9.102589118181432e-06,
+      "loss": 0.09329748,
+      "memory(GiB)": 74.41,
+      "step": 1645,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.19436345966958213,
+      "grad_norm": 6.375,
+      "learning_rate": 9.097292863726635e-06,
+      "loss": 0.10539714,
+      "memory(GiB)": 74.41,
+      "step": 1650,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.19495243985039903,
+      "grad_norm": 4.0625,
+      "learning_rate": 9.091982577067651e-06,
+      "loss": 0.09411023,
+      "memory(GiB)": 74.41,
+      "step": 1655,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.19554142003121594,
+      "grad_norm": 3.25,
+      "learning_rate": 9.086658276390883e-06,
+      "loss": 0.08853958,
+      "memory(GiB)": 74.41,
+      "step": 1660,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.19613040021203287,
+      "grad_norm": 4.34375,
+      "learning_rate": 9.081319979930735e-06,
+      "loss": 0.08764254,
+      "memory(GiB)": 74.41,
+      "step": 1665,
+      "train_speed(iter/s)": 0.074914
+    },
+    {
+      "epoch": 0.19671938039284978,
+      "grad_norm": 3.078125,
+      "learning_rate": 9.075967705969536e-06,
+      "loss": 0.09640335,
+      "memory(GiB)": 74.41,
+      "step": 1670,
+      "train_speed(iter/s)": 0.074915
+    },
+    {
+      "epoch": 0.19730836057366669,
+      "grad_norm": 3.875,
+      "learning_rate": 9.070601472837486e-06,
+      "loss": 0.08481975,
+      "memory(GiB)": 74.41,
+      "step": 1675,
+      "train_speed(iter/s)": 0.074917
+    },
+    {
+      "epoch": 0.19789734075448362,
+      "grad_norm": 3.28125,
+      "learning_rate": 9.065221298912592e-06,
+      "loss": 0.09183933,
+      "memory(GiB)": 74.41,
+      "step": 1680,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.19848632093530053,
+      "grad_norm": 5.0,
+      "learning_rate": 9.059827202620608e-06,
+      "loss": 0.08611407,
+      "memory(GiB)": 74.41,
+      "step": 1685,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.19907530111611743,
+      "grad_norm": 3.9375,
+      "learning_rate": 9.054419202434963e-06,
+      "loss": 0.08965172,
+      "memory(GiB)": 74.41,
+      "step": 1690,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.19966428129693436,
+      "grad_norm": 3.984375,
+      "learning_rate": 9.048997316876707e-06,
+      "loss": 0.09860962,
+      "memory(GiB)": 74.41,
+      "step": 1695,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.20025326147775127,
+      "grad_norm": 3.671875,
+      "learning_rate": 9.043561564514445e-06,
+      "loss": 0.09100879,
+      "memory(GiB)": 74.41,
+      "step": 1700,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.20084224165856818,
+      "grad_norm": 4.375,
+      "learning_rate": 9.038111963964266e-06,
+      "loss": 0.09089627,
+      "memory(GiB)": 74.41,
+      "step": 1705,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.2014312218393851,
+      "grad_norm": 2.734375,
+      "learning_rate": 9.032648533889697e-06,
+      "loss": 0.08130147,
+      "memory(GiB)": 74.41,
+      "step": 1710,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.20202020202020202,
+      "grad_norm": 3.140625,
+      "learning_rate": 9.027171293001615e-06,
+      "loss": 0.09625549,
+      "memory(GiB)": 74.41,
+      "step": 1715,
+      "train_speed(iter/s)": 0.074928
+    },
+    {
+      "epoch": 0.20260918220101892,
+      "grad_norm": 4.65625,
+      "learning_rate": 9.021680260058203e-06,
+      "loss": 0.09061633,
+      "memory(GiB)": 74.41,
+      "step": 1720,
+      "train_speed(iter/s)": 0.07493
+    },
+    {
+      "epoch": 0.20319816238183586,
+      "grad_norm": 4.3125,
+      "learning_rate": 9.016175453864877e-06,
+      "loss": 0.09252499,
+      "memory(GiB)": 74.41,
+      "step": 1725,
+      "train_speed(iter/s)": 0.074931
+    },
+    {
+      "epoch": 0.20378714256265276,
+      "grad_norm": 3.40625,
+      "learning_rate": 9.010656893274222e-06,
+      "loss": 0.0898882,
+      "memory(GiB)": 74.41,
+      "step": 1730,
+      "train_speed(iter/s)": 0.074933
+    },
+    {
+      "epoch": 0.2043761227434697,
+      "grad_norm": 3.90625,
+      "learning_rate": 9.00512459718593e-06,
+      "loss": 0.08716853,
+      "memory(GiB)": 74.41,
+      "step": 1735,
+      "train_speed(iter/s)": 0.074935
+    },
+    {
+      "epoch": 0.2049651029242866,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.999578584546733e-06,
+      "loss": 0.08696725,
+      "memory(GiB)": 74.41,
+      "step": 1740,
+      "train_speed(iter/s)": 0.074937
+    },
+    {
+      "epoch": 0.2055540831051035,
+      "grad_norm": 4.25,
+      "learning_rate": 8.994018874350337e-06,
+      "loss": 0.09265312,
+      "memory(GiB)": 74.41,
+      "step": 1745,
+      "train_speed(iter/s)": 0.074939
+    },
+    {
+      "epoch": 0.20614306328592044,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.98844548563736e-06,
+      "loss": 0.09432794,
+      "memory(GiB)": 74.41,
+      "step": 1750,
+      "train_speed(iter/s)": 0.07494
+    },
+    {
+      "epoch": 0.20673204346673735,
+      "grad_norm": 3.171875,
+      "learning_rate": 8.982858437495267e-06,
+      "loss": 0.08902328,
+      "memory(GiB)": 74.41,
+      "step": 1755,
+      "train_speed(iter/s)": 0.07494
+    },
+    {
+      "epoch": 0.20732102364755425,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.977257749058302e-06,
+      "loss": 0.09804455,
+      "memory(GiB)": 74.41,
+      "step": 1760,
+      "train_speed(iter/s)": 0.074942
+    },
+    {
+      "epoch": 0.2079100038283712,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.97164343950742e-06,
+      "loss": 0.08721474,
+      "memory(GiB)": 74.41,
+      "step": 1765,
+      "train_speed(iter/s)": 0.074944
+    },
+    {
+      "epoch": 0.2084989840091881,
+      "grad_norm": 4.59375,
+      "learning_rate": 8.966015528070232e-06,
+      "loss": 0.09626728,
+      "memory(GiB)": 74.41,
+      "step": 1770,
+      "train_speed(iter/s)": 0.074946
+    },
+    {
+      "epoch": 0.209087964190005,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.960374034020926e-06,
+      "loss": 0.0885976,
+      "memory(GiB)": 74.41,
+      "step": 1775,
+      "train_speed(iter/s)": 0.074948
+    },
+    {
+      "epoch": 0.20967694437082193,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.954718976680212e-06,
+      "loss": 0.09599138,
+      "memory(GiB)": 74.41,
+      "step": 1780,
+      "train_speed(iter/s)": 0.074949
+    },
+    {
+      "epoch": 0.21026592455163884,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.949050375415245e-06,
+      "loss": 0.09163067,
+      "memory(GiB)": 74.41,
+      "step": 1785,
+      "train_speed(iter/s)": 0.07495
+    },
+    {
+      "epoch": 0.21085490473245574,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.94336824963957e-06,
+      "loss": 0.09342519,
+      "memory(GiB)": 74.41,
+      "step": 1790,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.21144388491327268,
+      "grad_norm": 3.375,
+      "learning_rate": 8.937672618813046e-06,
+      "loss": 0.08690483,
+      "memory(GiB)": 74.41,
+      "step": 1795,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.21203286509408958,
+      "grad_norm": 3.25,
+      "learning_rate": 8.931963502441788e-06,
+      "loss": 0.0816748,
+      "memory(GiB)": 74.41,
+      "step": 1800,
+      "train_speed(iter/s)": 0.07495
+    },
+    {
+      "epoch": 0.2126218452749065,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.926240920078091e-06,
+      "loss": 0.08076277,
+      "memory(GiB)": 74.41,
+      "step": 1805,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.21321082545572342,
+      "grad_norm": 2.90625,
+      "learning_rate": 8.92050489132037e-06,
+      "loss": 0.08003956,
+      "memory(GiB)": 74.41,
+      "step": 1810,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.21379980563654033,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.91475543581309e-06,
+      "loss": 0.09223965,
+      "memory(GiB)": 74.41,
+      "step": 1815,
+      "train_speed(iter/s)": 0.074954
+    },
+    {
+      "epoch": 0.21438878581735724,
+      "grad_norm": 6.0625,
+      "learning_rate": 8.9089925732467e-06,
+      "loss": 0.09530929,
+      "memory(GiB)": 74.41,
+      "step": 1820,
+      "train_speed(iter/s)": 0.074956
+    },
+    {
+      "epoch": 0.21497776599817417,
+      "grad_norm": 3.765625,
+      "learning_rate": 8.903216323357562e-06,
+      "loss": 0.09788524,
+      "memory(GiB)": 74.41,
+      "step": 1825,
+      "train_speed(iter/s)": 0.074956
+    },
+    {
+      "epoch": 0.21556674617899108,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.897426705927892e-06,
+      "loss": 0.07674338,
+      "memory(GiB)": 74.41,
+      "step": 1830,
+      "train_speed(iter/s)": 0.074958
+    },
+    {
+      "epoch": 0.21615572635980798,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.89162374078568e-06,
+      "loss": 0.09034337,
+      "memory(GiB)": 74.41,
+      "step": 1835,
+      "train_speed(iter/s)": 0.074957
+    },
+    {
+      "epoch": 0.21674470654062492,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.885807447804633e-06,
+      "loss": 0.08671494,
+      "memory(GiB)": 74.41,
+      "step": 1840,
+      "train_speed(iter/s)": 0.074959
+    },
+    {
+      "epoch": 0.21733368672144182,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.8799778469041e-06,
+      "loss": 0.08138096,
+      "memory(GiB)": 74.41,
+      "step": 1845,
+      "train_speed(iter/s)": 0.074961
+    },
+    {
+      "epoch": 0.21792266690225873,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.87413495804901e-06,
+      "loss": 0.08538883,
+      "memory(GiB)": 74.41,
+      "step": 1850,
+      "train_speed(iter/s)": 0.074961
+    },
+    {
+      "epoch": 0.21851164708307566,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.868278801249794e-06,
+      "loss": 0.09670607,
+      "memory(GiB)": 74.41,
+      "step": 1855,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.21910062726389257,
+      "grad_norm": 3.375,
+      "learning_rate": 8.862409396562326e-06,
+      "loss": 0.09224599,
+      "memory(GiB)": 74.41,
+      "step": 1860,
+      "train_speed(iter/s)": 0.074963
+    },
+    {
+      "epoch": 0.21968960744470947,
+      "grad_norm": 3.625,
+      "learning_rate": 8.856526764087852e-06,
+      "loss": 0.10004935,
+      "memory(GiB)": 74.41,
+      "step": 1865,
+      "train_speed(iter/s)": 0.074963
+    },
+    {
+      "epoch": 0.2202785876255264,
+      "grad_norm": 2.484375,
+      "learning_rate": 8.850630923972919e-06,
+      "loss": 0.08341603,
+      "memory(GiB)": 74.41,
+      "step": 1870,
+      "train_speed(iter/s)": 0.074965
+    },
+    {
+      "epoch": 0.2208675678063433,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.844721896409302e-06,
+      "loss": 0.08883305,
+      "memory(GiB)": 74.41,
+      "step": 1875,
+      "train_speed(iter/s)": 0.074967
+    },
+    {
+      "epoch": 0.22145654798716022,
+      "grad_norm": 4.8125,
+      "learning_rate": 8.838799701633946e-06,
+      "loss": 0.09026896,
+      "memory(GiB)": 74.41,
+      "step": 1880,
+      "train_speed(iter/s)": 0.07497
+    },
+    {
+      "epoch": 0.22204552816797715,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.832864359928892e-06,
+      "loss": 0.07970605,
+      "memory(GiB)": 74.41,
+      "step": 1885,
+      "train_speed(iter/s)": 0.07497
+    },
+    {
+      "epoch": 0.22263450834879406,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.826915891621196e-06,
+      "loss": 0.09585873,
+      "memory(GiB)": 74.41,
+      "step": 1890,
+      "train_speed(iter/s)": 0.074971
+    },
+    {
+      "epoch": 0.223223488529611,
+      "grad_norm": 4.375,
+      "learning_rate": 8.820954317082879e-06,
+      "loss": 0.08822346,
+      "memory(GiB)": 74.41,
+      "step": 1895,
+      "train_speed(iter/s)": 0.074973
+    },
+    {
+      "epoch": 0.2238124687104279,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.814979656730841e-06,
+      "loss": 0.08403029,
+      "memory(GiB)": 74.41,
+      "step": 1900,
+      "train_speed(iter/s)": 0.074974
+    },
+    {
+      "epoch": 0.2244014488912448,
+      "grad_norm": 3.21875,
+      "learning_rate": 8.808991931026806e-06,
+      "loss": 0.08884512,
+      "memory(GiB)": 74.41,
+      "step": 1905,
+      "train_speed(iter/s)": 0.074976
+    },
+    {
+      "epoch": 0.22499042907206174,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.802991160477235e-06,
+      "loss": 0.07620925,
+      "memory(GiB)": 74.41,
+      "step": 1910,
+      "train_speed(iter/s)": 0.074977
+    },
+    {
+      "epoch": 0.22557940925287864,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.796977365633265e-06,
+      "loss": 0.08231162,
+      "memory(GiB)": 74.41,
+      "step": 1915,
+      "train_speed(iter/s)": 0.074979
+    },
+    {
+      "epoch": 0.22616838943369555,
+      "grad_norm": 3.328125,
+      "learning_rate": 8.790950567090646e-06,
+      "loss": 0.08745655,
+      "memory(GiB)": 74.41,
+      "step": 1920,
+      "train_speed(iter/s)": 0.07498
+    },
+    {
+      "epoch": 0.22675736961451248,
+      "grad_norm": 3.984375,
+      "learning_rate": 8.784910785489653e-06,
+      "loss": 0.09518582,
+      "memory(GiB)": 74.41,
+      "step": 1925,
+      "train_speed(iter/s)": 0.074983
+    },
+    {
+      "epoch": 0.2273463497953294,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.77885804151503e-06,
+      "loss": 0.08628204,
+      "memory(GiB)": 74.41,
+      "step": 1930,
+      "train_speed(iter/s)": 0.074984
+    },
+    {
+      "epoch": 0.2279353299761463,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.772792355895914e-06,
+      "loss": 0.11409738,
+      "memory(GiB)": 74.41,
+      "step": 1935,
+      "train_speed(iter/s)": 0.074984
+    },
+    {
+      "epoch": 0.22852431015696323,
+      "grad_norm": 2.984375,
+      "learning_rate": 8.76671374940576e-06,
+      "loss": 0.08102646,
+      "memory(GiB)": 74.41,
+      "step": 1940,
+      "train_speed(iter/s)": 0.074986
+    },
+    {
+      "epoch": 0.22911329033778013,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.760622242862278e-06,
+      "loss": 0.08864189,
+      "memory(GiB)": 74.41,
+      "step": 1945,
+      "train_speed(iter/s)": 0.074989
+    },
+    {
+      "epoch": 0.22970227051859704,
+      "grad_norm": 3.671875,
+      "learning_rate": 8.754517857127354e-06,
+      "loss": 0.07894121,
+      "memory(GiB)": 74.41,
+      "step": 1950,
+      "train_speed(iter/s)": 0.074991
+    },
+    {
+      "epoch": 0.23029125069941397,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.748400613106987e-06,
+      "loss": 0.08113719,
+      "memory(GiB)": 74.41,
+      "step": 1955,
+      "train_speed(iter/s)": 0.074993
+    },
+    {
+      "epoch": 0.23088023088023088,
+      "grad_norm": 3.359375,
+      "learning_rate": 8.742270531751205e-06,
+      "loss": 0.07365735,
+      "memory(GiB)": 74.41,
+      "step": 1960,
+      "train_speed(iter/s)": 0.074995
+    },
+    {
+      "epoch": 0.23146921106104779,
+      "grad_norm": 2.765625,
+      "learning_rate": 8.736127634054005e-06,
+      "loss": 0.09221531,
+      "memory(GiB)": 74.41,
+      "step": 1965,
+      "train_speed(iter/s)": 0.074996
+    },
+    {
+      "epoch": 0.23205819124186472,
+      "grad_norm": 3.71875,
+      "learning_rate": 8.729971941053281e-06,
+      "loss": 0.07759452,
+      "memory(GiB)": 74.41,
+      "step": 1970,
+      "train_speed(iter/s)": 0.074998
+    },
+    {
+      "epoch": 0.23264717142268163,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.723803473830742e-06,
+      "loss": 0.07012342,
+      "memory(GiB)": 74.41,
+      "step": 1975,
+      "train_speed(iter/s)": 0.074999
+    },
+    {
+      "epoch": 0.23323615160349853,
+      "grad_norm": 4.25,
+      "learning_rate": 8.717622253511841e-06,
+      "loss": 0.09737064,
+      "memory(GiB)": 74.41,
+      "step": 1980,
+      "train_speed(iter/s)": 0.075
+    },
+    {
+      "epoch": 0.23382513178431547,
+      "grad_norm": 3.640625,
+      "learning_rate": 8.71142830126572e-06,
+      "loss": 0.0820079,
+      "memory(GiB)": 74.41,
+      "step": 1985,
+      "train_speed(iter/s)": 0.075002
+    },
+    {
+      "epoch": 0.23441411196513237,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.705221638305115e-06,
+      "loss": 0.11287713,
+      "memory(GiB)": 74.41,
+      "step": 1990,
+      "train_speed(iter/s)": 0.075002
+    },
+    {
+      "epoch": 0.23500309214594928,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.699002285886296e-06,
+      "loss": 0.08315295,
+      "memory(GiB)": 74.41,
+      "step": 1995,
+      "train_speed(iter/s)": 0.075001
+    },
+    {
+      "epoch": 0.2355920723267662,
+      "grad_norm": 3.203125,
+      "learning_rate": 8.69277026530899e-06,
+      "loss": 0.08922192,
+      "memory(GiB)": 74.41,
+      "step": 2000,
+      "train_speed(iter/s)": 0.075003
+    },
+    {
+      "epoch": 0.23618105250758312,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.68652559791631e-06,
+      "loss": 0.07833961,
+      "memory(GiB)": 74.41,
+      "step": 2005,
+      "train_speed(iter/s)": 0.074902
+    },
+    {
+      "epoch": 0.23677003268840002,
+      "grad_norm": 4.71875,
+      "learning_rate": 8.680268305094682e-06,
+      "loss": 0.07653344,
+      "memory(GiB)": 74.41,
+      "step": 2010,
+      "train_speed(iter/s)": 0.074902
+    },
+    {
+      "epoch": 0.23735901286921696,
+      "grad_norm": 2.9375,
+      "learning_rate": 8.673998408273772e-06,
+      "loss": 0.08819388,
+      "memory(GiB)": 74.41,
+      "step": 2015,
+      "train_speed(iter/s)": 0.074904
+    },
+    {
+      "epoch": 0.23794799305003386,
+      "grad_norm": 3.90625,
+      "learning_rate": 8.667715928926406e-06,
+      "loss": 0.08014967,
+      "memory(GiB)": 74.41,
+      "step": 2020,
+      "train_speed(iter/s)": 0.074905
+    },
+    {
+      "epoch": 0.23853697323085077,
+      "grad_norm": 2.609375,
+      "learning_rate": 8.661420888568508e-06,
+      "loss": 0.07681896,
+      "memory(GiB)": 74.41,
+      "step": 2025,
+      "train_speed(iter/s)": 0.074908
+    },
+    {
+      "epoch": 0.2391259534116677,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.65511330875902e-06,
+      "loss": 0.08645278,
+      "memory(GiB)": 74.41,
+      "step": 2030,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.2397149335924846,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.648793211099823e-06,
+      "loss": 0.0930752,
+      "memory(GiB)": 74.41,
+      "step": 2035,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.24030391377330154,
+      "grad_norm": 4.375,
+      "learning_rate": 8.642460617235675e-06,
+      "loss": 0.08526615,
+      "memory(GiB)": 74.41,
+      "step": 2040,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.24089289395411845,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.636115548854125e-06,
+      "loss": 0.0950511,
+      "memory(GiB)": 74.41,
+      "step": 2045,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.24148187413493535,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.629758027685452e-06,
+      "loss": 0.07512583,
+      "memory(GiB)": 74.41,
+      "step": 2050,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.2420708543157523,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.623388075502569e-06,
+      "loss": 0.09064118,
+      "memory(GiB)": 74.41,
+      "step": 2055,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.2426598344965692,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.617005714120977e-06,
+      "loss": 0.08695101,
+      "memory(GiB)": 74.41,
+      "step": 2060,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.2432488146773861,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.610610965398665e-06,
+      "loss": 0.09054219,
+      "memory(GiB)": 74.41,
+      "step": 2065,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.24383779485820303,
+      "grad_norm": 4.6875,
+      "learning_rate": 8.604203851236053e-06,
+      "loss": 0.08845918,
+      "memory(GiB)": 74.41,
+      "step": 2070,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.24442677503901994,
+      "grad_norm": 4.40625,
+      "learning_rate": 8.597784393575902e-06,
+      "loss": 0.08291144,
+      "memory(GiB)": 74.41,
+      "step": 2075,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.24501575521983684,
+      "grad_norm": 3.3125,
+      "learning_rate": 8.591352614403255e-06,
+      "loss": 0.07554498,
+      "memory(GiB)": 74.41,
+      "step": 2080,
+      "train_speed(iter/s)": 0.074912
+    },
+    {
+      "epoch": 0.24560473540065378,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.584908535745344e-06,
+      "loss": 0.07999554,
+      "memory(GiB)": 74.41,
+      "step": 2085,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.24619371558147068,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.57845217967153e-06,
+      "loss": 0.09628522,
+      "memory(GiB)": 74.41,
+      "step": 2090,
+      "train_speed(iter/s)": 0.074914
+    },
+    {
+      "epoch": 0.2467826957622876,
+      "grad_norm": 2.3125,
+      "learning_rate": 8.57198356829322e-06,
+      "loss": 0.07814129,
+      "memory(GiB)": 74.41,
+      "step": 2095,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.24737167594310452,
+      "grad_norm": 3.5,
+      "learning_rate": 8.56550272376379e-06,
+      "loss": 0.09396058,
+      "memory(GiB)": 74.41,
+      "step": 2100,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.24796065612392143,
+      "grad_norm": 3.59375,
+      "learning_rate": 8.559009668278516e-06,
+      "loss": 0.08171555,
+      "memory(GiB)": 74.41,
+      "step": 2105,
+      "train_speed(iter/s)": 0.074915
+    },
+    {
+      "epoch": 0.24854963630473834,
+      "grad_norm": 4.25,
+      "learning_rate": 8.552504424074484e-06,
+      "loss": 0.08422858,
+      "memory(GiB)": 74.41,
+      "step": 2110,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.24913861648555527,
+      "grad_norm": 3.625,
+      "learning_rate": 8.545987013430536e-06,
+      "loss": 0.08350295,
+      "memory(GiB)": 74.41,
+      "step": 2115,
+      "train_speed(iter/s)": 0.074915
+    },
+    {
+      "epoch": 0.24972759666637218,
+      "grad_norm": 4.0,
+      "learning_rate": 8.539457458667173e-06,
+      "loss": 0.09641573,
+      "memory(GiB)": 74.41,
+      "step": 2120,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.2503165768471891,
+      "grad_norm": 4.375,
+      "learning_rate": 8.532915782146488e-06,
+      "loss": 0.09828712,
+      "memory(GiB)": 74.41,
+      "step": 2125,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.250905557028006,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.526362006272087e-06,
+      "loss": 0.08949637,
+      "memory(GiB)": 74.41,
+      "step": 2130,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.2514945372088229,
+      "grad_norm": 4.0,
+      "learning_rate": 8.519796153489018e-06,
+      "loss": 0.08683264,
+      "memory(GiB)": 74.41,
+      "step": 2135,
+      "train_speed(iter/s)": 0.074915
+    },
+    {
+      "epoch": 0.25208351738963986,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.513218246283684e-06,
+      "loss": 0.08473853,
+      "memory(GiB)": 74.41,
+      "step": 2140,
+      "train_speed(iter/s)": 0.074915
+    },
+    {
+      "epoch": 0.25267249757045673,
+      "grad_norm": 3.546875,
+      "learning_rate": 8.506628307183773e-06,
+      "loss": 0.08321443,
+      "memory(GiB)": 74.41,
+      "step": 2145,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.25326147775127367,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.50002635875818e-06,
+      "loss": 0.09309025,
+      "memory(GiB)": 74.41,
+      "step": 2150,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.2538504579320906,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.493412423616929e-06,
+      "loss": 0.08694515,
+      "memory(GiB)": 74.41,
+      "step": 2155,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.2544394381129075,
+      "grad_norm": 4.625,
+      "learning_rate": 8.486786524411096e-06,
+      "loss": 0.08189753,
+      "memory(GiB)": 74.41,
+      "step": 2160,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.2550284182937244,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.480148683832727e-06,
+      "loss": 0.09474859,
+      "memory(GiB)": 74.41,
+      "step": 2165,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.25561739847454135,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.47349892461477e-06,
+      "loss": 0.08173348,
+      "memory(GiB)": 74.41,
+      "step": 2170,
+      "train_speed(iter/s)": 0.074918
+    },
+    {
+      "epoch": 0.2562063786553582,
+      "grad_norm": 4.09375,
+      "learning_rate": 8.466837269530991e-06,
+      "loss": 0.08846778,
+      "memory(GiB)": 74.41,
+      "step": 2175,
+      "train_speed(iter/s)": 0.074918
+    },
+    {
+      "epoch": 0.25679535883617516,
+      "grad_norm": 3.078125,
+      "learning_rate": 8.460163741395888e-06,
+      "loss": 0.08622532,
+      "memory(GiB)": 74.41,
+      "step": 2180,
+      "train_speed(iter/s)": 0.074919
+    },
+    {
+      "epoch": 0.2573843390169921,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.45347836306463e-06,
+      "loss": 0.10147046,
+      "memory(GiB)": 74.41,
+      "step": 2185,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.25797331919780897,
+      "grad_norm": 4.125,
+      "learning_rate": 8.446781157432969e-06,
+      "loss": 0.07713024,
+      "memory(GiB)": 74.41,
+      "step": 2190,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.2585622993786259,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.440072147437158e-06,
+      "loss": 0.08459938,
+      "memory(GiB)": 74.41,
+      "step": 2195,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.25915127955944284,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.433351356053878e-06,
+      "loss": 0.08433875,
+      "memory(GiB)": 74.41,
+      "step": 2200,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.2597402597402597,
+      "grad_norm": 3.828125,
+      "learning_rate": 8.426618806300167e-06,
+      "loss": 0.08870918,
+      "memory(GiB)": 74.41,
+      "step": 2205,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.26032923992107665,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.41987452123332e-06,
+      "loss": 0.07198619,
+      "memory(GiB)": 74.41,
+      "step": 2210,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.2609182201018936,
+      "grad_norm": 3.921875,
+      "learning_rate": 8.413118523950827e-06,
+      "loss": 0.07893941,
+      "memory(GiB)": 74.41,
+      "step": 2215,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.26150720028271046,
+      "grad_norm": 3.84375,
+      "learning_rate": 8.406350837590296e-06,
+      "loss": 0.09590834,
+      "memory(GiB)": 74.41,
+      "step": 2220,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.2620961804635274,
+      "grad_norm": 4.53125,
+      "learning_rate": 8.399571485329356e-06,
+      "loss": 0.07775668,
+      "memory(GiB)": 74.41,
+      "step": 2225,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.26268516064434433,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.392780490385599e-06,
+      "loss": 0.08939707,
+      "memory(GiB)": 74.41,
+      "step": 2230,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.2632741408251612,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.385977876016481e-06,
+      "loss": 0.0896252,
+      "memory(GiB)": 74.41,
+      "step": 2235,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.26386312100597814,
+      "grad_norm": 2.78125,
+      "learning_rate": 8.37916366551926e-06,
+      "loss": 0.09092049,
+      "memory(GiB)": 74.41,
+      "step": 2240,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.2644521011867951,
+      "grad_norm": 2.6875,
+      "learning_rate": 8.372337882230904e-06,
+      "loss": 0.07872928,
+      "memory(GiB)": 74.41,
+      "step": 2245,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.26504108136761195,
+      "grad_norm": 5.28125,
+      "learning_rate": 8.365500549528012e-06,
+      "loss": 0.08535323,
+      "memory(GiB)": 74.41,
+      "step": 2250,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.2656300615484289,
+      "grad_norm": 2.953125,
+      "learning_rate": 8.358651690826742e-06,
+      "loss": 0.09457763,
+      "memory(GiB)": 74.41,
+      "step": 2255,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.2662190417292458,
+      "grad_norm": 4.9375,
+      "learning_rate": 8.351791329582724e-06,
+      "loss": 0.10520123,
+      "memory(GiB)": 74.41,
+      "step": 2260,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.26680802191006275,
+      "grad_norm": 3.34375,
+      "learning_rate": 8.34491948929098e-06,
+      "loss": 0.08100507,
+      "memory(GiB)": 74.41,
+      "step": 2265,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.26739700209087963,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.338036193485846e-06,
+      "loss": 0.09872182,
+      "memory(GiB)": 74.41,
+      "step": 2270,
+      "train_speed(iter/s)": 0.074926
+    },
+    {
+      "epoch": 0.26798598227169657,
+      "grad_norm": 5.46875,
+      "learning_rate": 8.331141465740893e-06,
+      "loss": 0.08186264,
+      "memory(GiB)": 74.41,
+      "step": 2275,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.2685749624525135,
+      "grad_norm": 4.25,
+      "learning_rate": 8.324235329668835e-06,
+      "loss": 0.09497812,
+      "memory(GiB)": 74.41,
+      "step": 2280,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.2691639426333304,
+      "grad_norm": 5.1875,
+      "learning_rate": 8.317317808921466e-06,
+      "loss": 0.09269063,
+      "memory(GiB)": 74.41,
+      "step": 2285,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.2697529228141473,
+      "grad_norm": 4.5625,
+      "learning_rate": 8.310388927189564e-06,
+      "loss": 0.080677,
+      "memory(GiB)": 74.41,
+      "step": 2290,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.27034190299496424,
+      "grad_norm": 2.59375,
+      "learning_rate": 8.30344870820282e-06,
+      "loss": 0.07890679,
+      "memory(GiB)": 74.41,
+      "step": 2295,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.2709308831757811,
+      "grad_norm": 3.4375,
+      "learning_rate": 8.296497175729747e-06,
+      "loss": 0.08448692,
+      "memory(GiB)": 74.41,
+      "step": 2300,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.27151986335659806,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.289534353577609e-06,
+      "loss": 0.07329035,
+      "memory(GiB)": 74.41,
+      "step": 2305,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.272108843537415,
+      "grad_norm": 7.875,
+      "learning_rate": 8.28256026559233e-06,
+      "loss": 0.07775943,
+      "memory(GiB)": 74.41,
+      "step": 2310,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.27269782371823187,
+      "grad_norm": 4.78125,
+      "learning_rate": 8.275574935658421e-06,
+      "loss": 0.09190539,
+      "memory(GiB)": 74.41,
+      "step": 2315,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.2732868038990488,
+      "grad_norm": 3.96875,
+      "learning_rate": 8.26857838769889e-06,
+      "loss": 0.08333533,
+      "memory(GiB)": 74.41,
+      "step": 2320,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.27387578407986574,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.261570645675164e-06,
+      "loss": 0.10267622,
+      "memory(GiB)": 74.41,
+      "step": 2325,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.2744647642606826,
+      "grad_norm": 3.46875,
+      "learning_rate": 8.254551733587009e-06,
+      "loss": 0.07961779,
+      "memory(GiB)": 74.41,
+      "step": 2330,
+      "train_speed(iter/s)": 0.074927
+    },
+    {
+      "epoch": 0.27505374444149955,
+      "grad_norm": 4.46875,
+      "learning_rate": 8.247521675472447e-06,
+      "loss": 0.08598344,
+      "memory(GiB)": 74.41,
+      "step": 2335,
+      "train_speed(iter/s)": 0.074928
+    },
+    {
+      "epoch": 0.2756427246223165,
+      "grad_norm": 4.1875,
+      "learning_rate": 8.240480495407667e-06,
+      "loss": 0.09493701,
+      "memory(GiB)": 74.41,
+      "step": 2340,
+      "train_speed(iter/s)": 0.07493
+    },
+    {
+      "epoch": 0.27623170480313336,
+      "grad_norm": 4.3125,
+      "learning_rate": 8.233428217506954e-06,
+      "loss": 0.09267281,
+      "memory(GiB)": 74.41,
+      "step": 2345,
+      "train_speed(iter/s)": 0.074932
+    },
+    {
+      "epoch": 0.2768206849839503,
+      "grad_norm": 2.734375,
+      "learning_rate": 8.226364865922598e-06,
+      "loss": 0.09006658,
+      "memory(GiB)": 74.41,
+      "step": 2350,
+      "train_speed(iter/s)": 0.074933
+    },
+    {
+      "epoch": 0.2774096651647672,
+      "grad_norm": 3.125,
+      "learning_rate": 8.21929046484481e-06,
+      "loss": 0.07196817,
+      "memory(GiB)": 74.41,
+      "step": 2355,
+      "train_speed(iter/s)": 0.074934
+    },
+    {
+      "epoch": 0.2779986453455841,
+      "grad_norm": 3.453125,
+      "learning_rate": 8.212205038501648e-06,
+      "loss": 0.08442661,
+      "memory(GiB)": 74.41,
+      "step": 2360,
+      "train_speed(iter/s)": 0.074936
+    },
+    {
+      "epoch": 0.27858762552640104,
+      "grad_norm": 4.96875,
+      "learning_rate": 8.205108611158926e-06,
+      "loss": 0.09003518,
+      "memory(GiB)": 74.41,
+      "step": 2365,
+      "train_speed(iter/s)": 0.074937
+    },
+    {
+      "epoch": 0.279176605707218,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.198001207120135e-06,
+      "loss": 0.09039379,
+      "memory(GiB)": 74.41,
+      "step": 2370,
+      "train_speed(iter/s)": 0.074938
+    },
+    {
+      "epoch": 0.27976558588803485,
+      "grad_norm": 4.0,
+      "learning_rate": 8.190882850726357e-06,
+      "loss": 0.08618115,
+      "memory(GiB)": 74.41,
+      "step": 2375,
+      "train_speed(iter/s)": 0.07494
+    },
+    {
+      "epoch": 0.2803545660688518,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.183753566356182e-06,
+      "loss": 0.08356151,
+      "memory(GiB)": 74.41,
+      "step": 2380,
+      "train_speed(iter/s)": 0.074941
+    },
+    {
+      "epoch": 0.2809435462496687,
+      "grad_norm": 3.890625,
+      "learning_rate": 8.176613378425633e-06,
+      "loss": 0.07245021,
+      "memory(GiB)": 74.41,
+      "step": 2385,
+      "train_speed(iter/s)": 0.074942
+    },
+    {
+      "epoch": 0.2815325264304856,
+      "grad_norm": 4.25,
+      "learning_rate": 8.169462311388066e-06,
+      "loss": 0.07743685,
+      "memory(GiB)": 74.41,
+      "step": 2390,
+      "train_speed(iter/s)": 0.074945
+    },
+    {
+      "epoch": 0.28212150661130253,
+      "grad_norm": 3.953125,
+      "learning_rate": 8.162300389734094e-06,
+      "loss": 0.08839593,
+      "memory(GiB)": 74.41,
+      "step": 2395,
+      "train_speed(iter/s)": 0.074947
+    },
+    {
+      "epoch": 0.28271048679211946,
+      "grad_norm": 3.78125,
+      "learning_rate": 8.155127637991518e-06,
+      "loss": 0.08605137,
+      "memory(GiB)": 74.41,
+      "step": 2400,
+      "train_speed(iter/s)": 0.074948
+    },
+    {
+      "epoch": 0.28329946697293634,
+      "grad_norm": 3.609375,
+      "learning_rate": 8.147944080725214e-06,
+      "loss": 0.0692309,
+      "memory(GiB)": 74.41,
+      "step": 2405,
+      "train_speed(iter/s)": 0.07495
+    },
+    {
+      "epoch": 0.2838884471537533,
+      "grad_norm": 3.859375,
+      "learning_rate": 8.140749742537073e-06,
+      "loss": 0.08173347,
+      "memory(GiB)": 74.41,
+      "step": 2410,
+      "train_speed(iter/s)": 0.07495
+    },
+    {
+      "epoch": 0.2844774273345702,
+      "grad_norm": 3.390625,
+      "learning_rate": 8.133544648065905e-06,
+      "loss": 0.07887983,
+      "memory(GiB)": 74.41,
+      "step": 2415,
+      "train_speed(iter/s)": 0.074951
+    },
+    {
+      "epoch": 0.2850664075153871,
+      "grad_norm": 4.40625,
+      "learning_rate": 8.126328821987357e-06,
+      "loss": 0.09297481,
+      "memory(GiB)": 74.41,
+      "step": 2420,
+      "train_speed(iter/s)": 0.074952
+    },
+    {
+      "epoch": 0.285655387696204,
+      "grad_norm": 3.421875,
+      "learning_rate": 8.119102289013832e-06,
+      "loss": 0.0871433,
+      "memory(GiB)": 74.41,
+      "step": 2425,
+      "train_speed(iter/s)": 0.074955
+    },
+    {
+      "epoch": 0.28624436787702096,
+      "grad_norm": 4.28125,
+      "learning_rate": 8.1118650738944e-06,
+      "loss": 0.08925153,
+      "memory(GiB)": 74.41,
+      "step": 2430,
+      "train_speed(iter/s)": 0.074956
+    },
+    {
+      "epoch": 0.28683334805783783,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.104617201414711e-06,
+      "loss": 0.08430476,
+      "memory(GiB)": 74.41,
+      "step": 2435,
+      "train_speed(iter/s)": 0.074957
+    },
+    {
+      "epoch": 0.28742232823865477,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.09735869639692e-06,
+      "loss": 0.1002124,
+      "memory(GiB)": 74.41,
+      "step": 2440,
+      "train_speed(iter/s)": 0.074959
+    },
+    {
+      "epoch": 0.2880113084194717,
+      "grad_norm": 4.0625,
+      "learning_rate": 8.090089583699591e-06,
+      "loss": 0.07936279,
+      "memory(GiB)": 74.41,
+      "step": 2445,
+      "train_speed(iter/s)": 0.07496
+    },
+    {
+      "epoch": 0.2886002886002886,
+      "grad_norm": 4.65625,
+      "learning_rate": 8.082809888217622e-06,
+      "loss": 0.08244927,
+      "memory(GiB)": 74.41,
+      "step": 2450,
+      "train_speed(iter/s)": 0.074961
+    },
+    {
+      "epoch": 0.2891892687811055,
+      "grad_norm": 3.5,
+      "learning_rate": 8.075519634882146e-06,
+      "loss": 0.09104651,
+      "memory(GiB)": 74.41,
+      "step": 2455,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.28977824896192245,
+      "grad_norm": 3.65625,
+      "learning_rate": 8.068218848660461e-06,
+      "loss": 0.07435999,
+      "memory(GiB)": 74.41,
+      "step": 2460,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.2903672291427393,
+      "grad_norm": 3.75,
+      "learning_rate": 8.060907554555937e-06,
+      "loss": 0.07937492,
+      "memory(GiB)": 74.41,
+      "step": 2465,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.29095620932355626,
+      "grad_norm": 3.53125,
+      "learning_rate": 8.053585777607929e-06,
+      "loss": 0.08450289,
+      "memory(GiB)": 74.41,
+      "step": 2470,
+      "train_speed(iter/s)": 0.074963
+    },
+    {
+      "epoch": 0.2915451895043732,
+      "grad_norm": 3.734375,
+      "learning_rate": 8.046253542891689e-06,
+      "loss": 0.06627862,
+      "memory(GiB)": 74.41,
+      "step": 2475,
+      "train_speed(iter/s)": 0.074962
+    },
+    {
+      "epoch": 0.29213416968519007,
+      "grad_norm": 4.03125,
+      "learning_rate": 8.038910875518295e-06,
+      "loss": 0.09286552,
+      "memory(GiB)": 74.41,
+      "step": 2480,
+      "train_speed(iter/s)": 0.074963
+    },
+    {
+      "epoch": 0.292723149866007,
+      "grad_norm": 4.34375,
+      "learning_rate": 8.031557800634544e-06,
+      "loss": 0.07346339,
+      "memory(GiB)": 74.41,
+      "step": 2485,
+      "train_speed(iter/s)": 0.074964
+    },
+    {
+      "epoch": 0.29331213004682394,
+      "grad_norm": 3.703125,
+      "learning_rate": 8.024194343422882e-06,
+      "loss": 0.08958888,
+      "memory(GiB)": 74.41,
+      "step": 2490,
+      "train_speed(iter/s)": 0.074963
+    },
+    {
+      "epoch": 0.2939011102276408,
+      "grad_norm": 3.1875,
+      "learning_rate": 8.016820529101308e-06,
+      "loss": 0.08174566,
+      "memory(GiB)": 74.41,
+      "step": 2495,
+      "train_speed(iter/s)": 0.074964
+    },
+    {
+      "epoch": 0.29449009040845775,
+      "grad_norm": 3.375,
+      "learning_rate": 8.009436382923294e-06,
+      "loss": 0.09179025,
+      "memory(GiB)": 74.41,
+      "step": 2500,
+      "train_speed(iter/s)": 0.074965
+    },
+    {
+      "epoch": 0.2950790705892747,
+      "grad_norm": 3.515625,
+      "learning_rate": 8.002041930177696e-06,
+      "loss": 0.08729964,
+      "memory(GiB)": 74.41,
+      "step": 2505,
+      "train_speed(iter/s)": 0.074862
+    },
+    {
+      "epoch": 0.29566805077009156,
+      "grad_norm": 4.375,
+      "learning_rate": 7.994637196188666e-06,
+      "loss": 0.09242439,
+      "memory(GiB)": 74.41,
+      "step": 2510,
+      "train_speed(iter/s)": 0.074861
+    },
+    {
+      "epoch": 0.2962570309509085,
+      "grad_norm": 4.125,
+      "learning_rate": 7.987222206315568e-06,
+      "loss": 0.08498147,
+      "memory(GiB)": 74.41,
+      "step": 2515,
+      "train_speed(iter/s)": 0.074862
+    },
+    {
+      "epoch": 0.29684601113172543,
+      "grad_norm": 3.359375,
+      "learning_rate": 7.979796985952892e-06,
+      "loss": 0.07728409,
+      "memory(GiB)": 74.41,
+      "step": 2520,
+      "train_speed(iter/s)": 0.074864
+    },
+    {
+      "epoch": 0.2974349913125423,
+      "grad_norm": 3.953125,
+      "learning_rate": 7.972361560530157e-06,
+      "loss": 0.08274835,
+      "memory(GiB)": 74.41,
+      "step": 2525,
+      "train_speed(iter/s)": 0.074864
+    },
+    {
+      "epoch": 0.29802397149335924,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.96491595551184e-06,
+      "loss": 0.09903065,
+      "memory(GiB)": 74.41,
+      "step": 2530,
+      "train_speed(iter/s)": 0.074865
+    },
+    {
+      "epoch": 0.2986129516741762,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.95746019639728e-06,
+      "loss": 0.07592887,
+      "memory(GiB)": 74.41,
+      "step": 2535,
+      "train_speed(iter/s)": 0.074865
+    },
+    {
+      "epoch": 0.29920193185499305,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.949994308720584e-06,
+      "loss": 0.10274631,
+      "memory(GiB)": 74.41,
+      "step": 2540,
+      "train_speed(iter/s)": 0.074867
+    },
+    {
+      "epoch": 0.29979091203581,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.942518318050555e-06,
+      "loss": 0.08352513,
+      "memory(GiB)": 74.41,
+      "step": 2545,
+      "train_speed(iter/s)": 0.074868
+    },
+    {
+      "epoch": 0.3003798922166269,
+      "grad_norm": 3.125,
+      "learning_rate": 7.93503224999059e-06,
+      "loss": 0.08402941,
+      "memory(GiB)": 74.41,
+      "step": 2550,
+      "train_speed(iter/s)": 0.074869
+    },
+    {
+      "epoch": 0.3009688723974438,
+      "grad_norm": 3.0625,
+      "learning_rate": 7.927536130178605e-06,
+      "loss": 0.07813259,
+      "memory(GiB)": 74.41,
+      "step": 2555,
+      "train_speed(iter/s)": 0.074869
+    },
+    {
+      "epoch": 0.30155785257826073,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.920029984286935e-06,
+      "loss": 0.07638797,
+      "memory(GiB)": 74.41,
+      "step": 2560,
+      "train_speed(iter/s)": 0.07487
+    },
+    {
+      "epoch": 0.30214683275907767,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.912513838022253e-06,
+      "loss": 0.08150139,
+      "memory(GiB)": 74.41,
+      "step": 2565,
+      "train_speed(iter/s)": 0.074872
+    },
+    {
+      "epoch": 0.3027358129398946,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.904987717125485e-06,
+      "loss": 0.07865617,
+      "memory(GiB)": 74.41,
+      "step": 2570,
+      "train_speed(iter/s)": 0.074873
+    },
+    {
+      "epoch": 0.3033247931207115,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.89745164737171e-06,
+      "loss": 0.10058107,
+      "memory(GiB)": 74.41,
+      "step": 2575,
+      "train_speed(iter/s)": 0.074875
+    },
+    {
+      "epoch": 0.3039137733015284,
+      "grad_norm": 4.625,
+      "learning_rate": 7.88990565457009e-06,
+      "loss": 0.08737856,
+      "memory(GiB)": 74.41,
+      "step": 2580,
+      "train_speed(iter/s)": 0.074876
+    },
+    {
+      "epoch": 0.30450275348234535,
+      "grad_norm": 3.796875,
+      "learning_rate": 7.882349764563758e-06,
+      "loss": 0.08744036,
+      "memory(GiB)": 74.41,
+      "step": 2585,
+      "train_speed(iter/s)": 0.074879
+    },
+    {
+      "epoch": 0.3050917336631622,
+      "grad_norm": 4.34375,
+      "learning_rate": 7.874784003229751e-06,
+      "loss": 0.11328788,
+      "memory(GiB)": 74.41,
+      "step": 2590,
+      "train_speed(iter/s)": 0.07488
+    },
+    {
+      "epoch": 0.30568071384397916,
+      "grad_norm": 4.84375,
+      "learning_rate": 7.867208396478914e-06,
+      "loss": 0.08754985,
+      "memory(GiB)": 74.41,
+      "step": 2595,
+      "train_speed(iter/s)": 0.074881
+    },
+    {
+      "epoch": 0.3062696940247961,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.859622970255804e-06,
+      "loss": 0.07025304,
+      "memory(GiB)": 74.41,
+      "step": 2600,
+      "train_speed(iter/s)": 0.074882
+    },
+    {
+      "epoch": 0.30685867420561297,
+      "grad_norm": 3.75,
+      "learning_rate": 7.852027750538613e-06,
+      "loss": 0.09671483,
+      "memory(GiB)": 74.41,
+      "step": 2605,
+      "train_speed(iter/s)": 0.074883
+    },
+    {
+      "epoch": 0.3074476543864299,
+      "grad_norm": 3.484375,
+      "learning_rate": 7.844422763339066e-06,
+      "loss": 0.08538601,
+      "memory(GiB)": 74.41,
+      "step": 2610,
+      "train_speed(iter/s)": 0.074883
+    },
+    {
+      "epoch": 0.30803663456724684,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.836808034702348e-06,
+      "loss": 0.0819549,
+      "memory(GiB)": 74.41,
+      "step": 2615,
+      "train_speed(iter/s)": 0.074885
+    },
+    {
+      "epoch": 0.3086256147480637,
+      "grad_norm": 3.125,
+      "learning_rate": 7.829183590707001e-06,
+      "loss": 0.06693002,
+      "memory(GiB)": 74.41,
+      "step": 2620,
+      "train_speed(iter/s)": 0.074886
+    },
+    {
+      "epoch": 0.30921459492888065,
+      "grad_norm": 3.75,
+      "learning_rate": 7.82154945746484e-06,
+      "loss": 0.07030878,
+      "memory(GiB)": 74.41,
+      "step": 2625,
+      "train_speed(iter/s)": 0.074887
+    },
+    {
+      "epoch": 0.3098035751096976,
+      "grad_norm": 3.234375,
+      "learning_rate": 7.813905661120862e-06,
+      "loss": 0.08062052,
+      "memory(GiB)": 74.41,
+      "step": 2630,
+      "train_speed(iter/s)": 0.074888
+    },
+    {
+      "epoch": 0.31039255529051446,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.806252227853161e-06,
+      "loss": 0.09003597,
+      "memory(GiB)": 74.41,
+      "step": 2635,
+      "train_speed(iter/s)": 0.074889
+    },
+    {
+      "epoch": 0.3109815354713314,
+      "grad_norm": 3.09375,
+      "learning_rate": 7.798589183872833e-06,
+      "loss": 0.08022001,
+      "memory(GiB)": 74.41,
+      "step": 2640,
+      "train_speed(iter/s)": 0.074889
+    },
+    {
+      "epoch": 0.3115705156521483,
+      "grad_norm": 4.59375,
+      "learning_rate": 7.79091655542389e-06,
+      "loss": 0.09506249,
+      "memory(GiB)": 74.41,
+      "step": 2645,
+      "train_speed(iter/s)": 0.07489
+    },
+    {
+      "epoch": 0.3121594958329652,
+      "grad_norm": 4.09375,
+      "learning_rate": 7.783234368783166e-06,
+      "loss": 0.07678704,
+      "memory(GiB)": 74.41,
+      "step": 2650,
+      "train_speed(iter/s)": 0.07489
+    },
+    {
+      "epoch": 0.31274847601378214,
+      "grad_norm": 3.21875,
+      "learning_rate": 7.775542650260231e-06,
+      "loss": 0.08948871,
+      "memory(GiB)": 74.41,
+      "step": 2655,
+      "train_speed(iter/s)": 0.074892
+    },
+    {
+      "epoch": 0.3133374561945991,
+      "grad_norm": 4.78125,
+      "learning_rate": 7.767841426197297e-06,
+      "loss": 0.09192417,
+      "memory(GiB)": 74.41,
+      "step": 2660,
+      "train_speed(iter/s)": 0.074893
+    },
+    {
+      "epoch": 0.31392643637541595,
+      "grad_norm": 3.5,
+      "learning_rate": 7.760130722969135e-06,
+      "loss": 0.07221487,
+      "memory(GiB)": 74.41,
+      "step": 2665,
+      "train_speed(iter/s)": 0.074894
+    },
+    {
+      "epoch": 0.3145154165562329,
+      "grad_norm": 4.6875,
+      "learning_rate": 7.752410566982973e-06,
+      "loss": 0.07583697,
+      "memory(GiB)": 74.41,
+      "step": 2670,
+      "train_speed(iter/s)": 0.074897
+    },
+    {
+      "epoch": 0.3151043967370498,
+      "grad_norm": 3.203125,
+      "learning_rate": 7.74468098467842e-06,
+      "loss": 0.07944269,
+      "memory(GiB)": 74.41,
+      "step": 2675,
+      "train_speed(iter/s)": 0.074896
+    },
+    {
+      "epoch": 0.3156933769178667,
+      "grad_norm": 3.078125,
+      "learning_rate": 7.736942002527362e-06,
+      "loss": 0.0761774,
+      "memory(GiB)": 74.41,
+      "step": 2680,
+      "train_speed(iter/s)": 0.074897
+    },
+    {
+      "epoch": 0.31628235709868363,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.729193647033879e-06,
+      "loss": 0.07489694,
+      "memory(GiB)": 74.41,
+      "step": 2685,
+      "train_speed(iter/s)": 0.074899
+    },
+    {
+      "epoch": 0.31687133727950056,
+      "grad_norm": 3.9375,
+      "learning_rate": 7.72143594473415e-06,
+      "loss": 0.08410535,
+      "memory(GiB)": 74.41,
+      "step": 2690,
+      "train_speed(iter/s)": 0.074901
+    },
+    {
+      "epoch": 0.31746031746031744,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.713668922196372e-06,
+      "loss": 0.08352896,
+      "memory(GiB)": 74.41,
+      "step": 2695,
+      "train_speed(iter/s)": 0.074903
+    },
+    {
+      "epoch": 0.3180492976411344,
+      "grad_norm": 2.765625,
+      "learning_rate": 7.705892606020652e-06,
+      "loss": 0.0845727,
+      "memory(GiB)": 74.41,
+      "step": 2700,
+      "train_speed(iter/s)": 0.074904
+    },
+    {
+      "epoch": 0.3186382778219513,
+      "grad_norm": 3.671875,
+      "learning_rate": 7.698107022838931e-06,
+      "loss": 0.09101894,
+      "memory(GiB)": 74.41,
+      "step": 2705,
+      "train_speed(iter/s)": 0.074905
+    },
+    {
+      "epoch": 0.3192272580027682,
+      "grad_norm": 3.890625,
+      "learning_rate": 7.690312199314887e-06,
+      "loss": 0.08560072,
+      "memory(GiB)": 74.41,
+      "step": 2710,
+      "train_speed(iter/s)": 0.074906
+    },
+    {
+      "epoch": 0.3198162381835851,
+      "grad_norm": 3.375,
+      "learning_rate": 7.682508162143843e-06,
+      "loss": 0.09032207,
+      "memory(GiB)": 74.41,
+      "step": 2715,
+      "train_speed(iter/s)": 0.074907
+    },
+    {
+      "epoch": 0.32040521836440206,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.674694938052673e-06,
+      "loss": 0.08928435,
+      "memory(GiB)": 74.41,
+      "step": 2720,
+      "train_speed(iter/s)": 0.074909
+    },
+    {
+      "epoch": 0.32099419854521893,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.666872553799719e-06,
+      "loss": 0.06355714,
+      "memory(GiB)": 74.41,
+      "step": 2725,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.32158317872603587,
+      "grad_norm": 3.609375,
+      "learning_rate": 7.659041036174693e-06,
+      "loss": 0.09267324,
+      "memory(GiB)": 74.41,
+      "step": 2730,
+      "train_speed(iter/s)": 0.07491
+    },
+    {
+      "epoch": 0.3221721589068528,
+      "grad_norm": 4.5,
+      "learning_rate": 7.651200411998585e-06,
+      "loss": 0.08836827,
+      "memory(GiB)": 74.41,
+      "step": 2735,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.3227611390876697,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.643350708123574e-06,
+      "loss": 0.09088976,
+      "memory(GiB)": 74.41,
+      "step": 2740,
+      "train_speed(iter/s)": 0.074912
+    },
+    {
+      "epoch": 0.3233501192684866,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.635491951432931e-06,
+      "loss": 0.08111639,
+      "memory(GiB)": 74.41,
+      "step": 2745,
+      "train_speed(iter/s)": 0.074913
+    },
+    {
+      "epoch": 0.32393909944930355,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.62762416884094e-06,
+      "loss": 0.07648246,
+      "memory(GiB)": 74.41,
+      "step": 2750,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.3245280796301204,
+      "grad_norm": 4.1875,
+      "learning_rate": 7.619747387292786e-06,
+      "loss": 0.08834222,
+      "memory(GiB)": 74.41,
+      "step": 2755,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.32511705981093736,
+      "grad_norm": 3.53125,
+      "learning_rate": 7.611861633764479e-06,
+      "loss": 0.08350364,
+      "memory(GiB)": 74.41,
+      "step": 2760,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.3257060399917543,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.603966935262752e-06,
+      "loss": 0.08396969,
+      "memory(GiB)": 74.41,
+      "step": 2765,
+      "train_speed(iter/s)": 0.074911
+    },
+    {
+      "epoch": 0.32629502017257117,
+      "grad_norm": 2.9375,
+      "learning_rate": 7.596063318824978e-06,
+      "loss": 0.06955547,
+      "memory(GiB)": 74.41,
+      "step": 2770,
+      "train_speed(iter/s)": 0.074912
+    },
+    {
+      "epoch": 0.3268840003533881,
+      "grad_norm": 2.875,
+      "learning_rate": 7.588150811519066e-06,
+      "loss": 0.08324391,
+      "memory(GiB)": 74.41,
+      "step": 2775,
+      "train_speed(iter/s)": 0.074914
+    },
+    {
+      "epoch": 0.32747298053420504,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.580229440443378e-06,
+      "loss": 0.08206114,
+      "memory(GiB)": 74.41,
+      "step": 2780,
+      "train_speed(iter/s)": 0.074916
+    },
+    {
+      "epoch": 0.3280619607150219,
+      "grad_norm": 3.703125,
+      "learning_rate": 7.572299232726627e-06,
+      "loss": 0.08156523,
+      "memory(GiB)": 74.41,
+      "step": 2785,
+      "train_speed(iter/s)": 0.074917
+    },
+    {
+      "epoch": 0.32865094089583885,
+      "grad_norm": 3.5625,
+      "learning_rate": 7.564360215527797e-06,
+      "loss": 0.07584106,
+      "memory(GiB)": 74.41,
+      "step": 2790,
+      "train_speed(iter/s)": 0.074919
+    },
+    {
+      "epoch": 0.3292399210766558,
+      "grad_norm": 4.28125,
+      "learning_rate": 7.556412416036036e-06,
+      "loss": 0.08185716,
+      "memory(GiB)": 74.41,
+      "step": 2795,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.32982890125747266,
+      "grad_norm": 3.625,
+      "learning_rate": 7.5484558614705705e-06,
+      "loss": 0.0776104,
+      "memory(GiB)": 74.41,
+      "step": 2800,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.3304178814382896,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.540490579080613e-06,
+      "loss": 0.07412539,
+      "memory(GiB)": 74.41,
+      "step": 2805,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.33100686161910653,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.532516596145263e-06,
+      "loss": 0.06569082,
+      "memory(GiB)": 74.41,
+      "step": 2810,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.3315958417999234,
+      "grad_norm": 3.890625,
+      "learning_rate": 7.524533939973424e-06,
+      "loss": 0.08522917,
+      "memory(GiB)": 74.41,
+      "step": 2815,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.33218482198074034,
+      "grad_norm": 4.03125,
+      "learning_rate": 7.516542637903692e-06,
+      "loss": 0.10329256,
+      "memory(GiB)": 74.41,
+      "step": 2820,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.3327738021615573,
+      "grad_norm": 2.640625,
+      "learning_rate": 7.508542717304286e-06,
+      "loss": 0.06652896,
+      "memory(GiB)": 74.41,
+      "step": 2825,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.33336278234237415,
+      "grad_norm": 3.03125,
+      "learning_rate": 7.500534205572932e-06,
+      "loss": 0.07229707,
+      "memory(GiB)": 74.41,
+      "step": 2830,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.3339517625231911,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.492517130136781e-06,
+      "loss": 0.07162905,
+      "memory(GiB)": 74.41,
+      "step": 2835,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.334540742704008,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.484491518452315e-06,
+      "loss": 0.06838317,
+      "memory(GiB)": 74.41,
+      "step": 2840,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.3351297228848249,
+      "grad_norm": 3.75,
+      "learning_rate": 7.476457398005246e-06,
+      "loss": 0.09434291,
+      "memory(GiB)": 74.41,
+      "step": 2845,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.33571870306564183,
+      "grad_norm": 3.9375,
+      "learning_rate": 7.46841479631043e-06,
+      "loss": 0.08622286,
+      "memory(GiB)": 74.41,
+      "step": 2850,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.33630768324645877,
+      "grad_norm": 3.171875,
+      "learning_rate": 7.460363740911769e-06,
+      "loss": 0.0905494,
+      "memory(GiB)": 74.41,
+      "step": 2855,
+      "train_speed(iter/s)": 0.074926
+    },
+    {
+      "epoch": 0.3368966634272757,
+      "grad_norm": 3.71875,
+      "learning_rate": 7.452304259382115e-06,
+      "loss": 0.08356917,
+      "memory(GiB)": 74.41,
+      "step": 2860,
+      "train_speed(iter/s)": 0.074927
+    },
+    {
+      "epoch": 0.3374856436080926,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.4442363793231775e-06,
+      "loss": 0.09314897,
+      "memory(GiB)": 74.41,
+      "step": 2865,
+      "train_speed(iter/s)": 0.074926
+    },
+    {
+      "epoch": 0.3380746237889095,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.436160128365431e-06,
+      "loss": 0.0789938,
+      "memory(GiB)": 74.41,
+      "step": 2870,
+      "train_speed(iter/s)": 0.074925
+    },
+    {
+      "epoch": 0.33866360396972645,
+      "grad_norm": 3.34375,
+      "learning_rate": 7.428075534168015e-06,
+      "loss": 0.08869631,
+      "memory(GiB)": 74.41,
+      "step": 2875,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.3392525841505433,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.419982624418647e-06,
+      "loss": 0.07138891,
+      "memory(GiB)": 74.41,
+      "step": 2880,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.33984156433136026,
+      "grad_norm": 4.40625,
+      "learning_rate": 7.4118814268335205e-06,
+      "loss": 0.09039434,
+      "memory(GiB)": 74.41,
+      "step": 2885,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.3404305445121772,
+      "grad_norm": 3.640625,
+      "learning_rate": 7.403771969157209e-06,
+      "loss": 0.08224541,
+      "memory(GiB)": 74.41,
+      "step": 2890,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.34101952469299407,
+      "grad_norm": 3.25,
+      "learning_rate": 7.39565427916258e-06,
+      "loss": 0.08286152,
+      "memory(GiB)": 74.41,
+      "step": 2895,
+      "train_speed(iter/s)": 0.074924
+    },
+    {
+      "epoch": 0.341608504873811,
+      "grad_norm": 3.140625,
+      "learning_rate": 7.387528384650699e-06,
+      "loss": 0.06858167,
+      "memory(GiB)": 74.41,
+      "step": 2900,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.34219748505462794,
+      "grad_norm": 3.40625,
+      "learning_rate": 7.379394313450714e-06,
+      "loss": 0.07839249,
+      "memory(GiB)": 74.41,
+      "step": 2905,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.3427864652354448,
+      "grad_norm": 3.734375,
+      "learning_rate": 7.371252093419795e-06,
+      "loss": 0.06531736,
+      "memory(GiB)": 74.41,
+      "step": 2910,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.34337544541626175,
+      "grad_norm": 2.953125,
+      "learning_rate": 7.363101752443007e-06,
+      "loss": 0.07913511,
+      "memory(GiB)": 74.41,
+      "step": 2915,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.3439644255970787,
+      "grad_norm": 3.015625,
+      "learning_rate": 7.3549433184332306e-06,
+      "loss": 0.08093199,
+      "memory(GiB)": 74.41,
+      "step": 2920,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.34455340577789556,
+      "grad_norm": 4.84375,
+      "learning_rate": 7.346776819331065e-06,
+      "loss": 0.06076463,
+      "memory(GiB)": 74.41,
+      "step": 2925,
+      "train_speed(iter/s)": 0.074923
+    },
+    {
+      "epoch": 0.3451423859587125,
+      "grad_norm": 4.15625,
+      "learning_rate": 7.338602283104731e-06,
+      "loss": 0.08175699,
+      "memory(GiB)": 74.41,
+      "step": 2930,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.34573136613952943,
+      "grad_norm": 3.4375,
+      "learning_rate": 7.330419737749967e-06,
+      "loss": 0.06969818,
+      "memory(GiB)": 74.41,
+      "step": 2935,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.3463203463203463,
+      "grad_norm": 3.546875,
+      "learning_rate": 7.3222292112899505e-06,
+      "loss": 0.08905634,
+      "memory(GiB)": 74.41,
+      "step": 2940,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.34690932650116324,
+      "grad_norm": 3.390625,
+      "learning_rate": 7.314030731775187e-06,
+      "loss": 0.07148146,
+      "memory(GiB)": 74.41,
+      "step": 2945,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.3474983066819802,
+      "grad_norm": 3.375,
+      "learning_rate": 7.305824327283419e-06,
+      "loss": 0.08839546,
+      "memory(GiB)": 74.41,
+      "step": 2950,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.34808728686279705,
+      "grad_norm": 4.5625,
+      "learning_rate": 7.297610025919533e-06,
+      "loss": 0.079582,
+      "memory(GiB)": 74.41,
+      "step": 2955,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.348676267043614,
+      "grad_norm": 5.0,
+      "learning_rate": 7.289387855815455e-06,
+      "loss": 0.07907141,
+      "memory(GiB)": 74.41,
+      "step": 2960,
+      "train_speed(iter/s)": 0.07492
+    },
+    {
+      "epoch": 0.3492652472244309,
+      "grad_norm": 3.46875,
+      "learning_rate": 7.281157845130067e-06,
+      "loss": 0.08737714,
+      "memory(GiB)": 74.41,
+      "step": 2965,
+      "train_speed(iter/s)": 0.074919
+    },
+    {
+      "epoch": 0.3498542274052478,
+      "grad_norm": 3.5,
+      "learning_rate": 7.272920022049097e-06,
+      "loss": 0.07906392,
+      "memory(GiB)": 74.41,
+      "step": 2970,
+      "train_speed(iter/s)": 0.074919
+    },
+    {
+      "epoch": 0.35044320758606473,
+      "grad_norm": 3.453125,
+      "learning_rate": 7.264674414785031e-06,
+      "loss": 0.08779007,
+      "memory(GiB)": 74.41,
+      "step": 2975,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.35103218776688166,
+      "grad_norm": 3.265625,
+      "learning_rate": 7.256421051577013e-06,
+      "loss": 0.06584383,
+      "memory(GiB)": 74.41,
+      "step": 2980,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.35162116794769854,
+      "grad_norm": 4.21875,
+      "learning_rate": 7.24815996069075e-06,
+      "loss": 0.08898224,
+      "memory(GiB)": 74.41,
+      "step": 2985,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.3522101481285155,
+      "grad_norm": 3.6875,
+      "learning_rate": 7.239891170418414e-06,
+      "loss": 0.08532063,
+      "memory(GiB)": 74.41,
+      "step": 2990,
+      "train_speed(iter/s)": 0.074921
+    },
+    {
+      "epoch": 0.3527991283093324,
+      "grad_norm": 3.28125,
+      "learning_rate": 7.231614709078545e-06,
+      "loss": 0.08233502,
+      "memory(GiB)": 74.41,
+      "step": 2995,
+      "train_speed(iter/s)": 0.074922
+    },
+    {
+      "epoch": 0.3533881084901493,
+      "grad_norm": 3.3125,
+      "learning_rate": 7.223330605015955e-06,
+      "loss": 0.09144391,
+      "memory(GiB)": 74.41,
+      "step": 3000,
+      "train_speed(iter/s)": 0.074923
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 8489,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.858899770166477e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}