diff --git "a/regional_perception_checkpoint-6379/trainer_state.json" "b/regional_perception_checkpoint-6379/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/regional_perception_checkpoint-6379/trainer_state.json"
@@ -0,0 +1,11550 @@
+{
+  "best_metric": 0.58554959,
+  "best_model_checkpoint": "/data1/wjx/model/swift/output/v1d+v3_prompt/output/internvl2-26b/v0-20240810-170945/checkpoint-6379",
+  "epoch": 4.999706084059959,
+  "eval_steps": 1,
+  "global_step": 6379,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "acc": 0.49760315,
+      "epoch": 0.0,
+      "learning_rate": 1.567398119122257e-07,
+      "loss": 1.98409796,
+      "memory(GiB)": 54.93,
+      "step": 1,
+      "train_speed(iter/s)": 0.023783
+    },
+    {
+      "acc": 0.52513188,
+      "epoch": 0.0,
+      "learning_rate": 7.836990595611285e-07,
+      "loss": 1.79844522,
+      "memory(GiB)": 54.93,
+      "step": 5,
+      "train_speed(iter/s)": 0.024289
+    },
+    {
+      "acc": 0.52413821,
+      "epoch": 0.01,
+      "learning_rate": 1.567398119122257e-06,
+      "loss": 1.82612877,
+      "memory(GiB)": 58.46,
+      "step": 10,
+      "train_speed(iter/s)": 0.023965
+    },
+    {
+      "acc": 0.51515636,
+      "epoch": 0.01,
+      "learning_rate": 2.3510971786833857e-06,
+      "loss": 1.82043953,
+      "memory(GiB)": 58.46,
+      "step": 15,
+      "train_speed(iter/s)": 0.024073
+    },
+    {
+      "acc": 0.51839466,
+      "epoch": 0.02,
+      "learning_rate": 3.134796238244514e-06,
+      "loss": 1.80918102,
+      "memory(GiB)": 58.46,
+      "step": 20,
+      "train_speed(iter/s)": 0.024066
+    },
+    {
+      "acc": 0.51010175,
+      "epoch": 0.02,
+      "learning_rate": 3.9184952978056436e-06,
+      "loss": 1.8428627,
+      "memory(GiB)": 58.46,
+      "step": 25,
+      "train_speed(iter/s)": 0.024068
+    },
+    {
+      "acc": 0.516012,
+      "epoch": 0.02,
+      "learning_rate": 4.7021943573667714e-06,
+      "loss": 1.81627235,
+      "memory(GiB)": 58.46,
+      "step": 30,
+      "train_speed(iter/s)": 0.024143
+    },
+    {
+      "acc": 0.50783205,
+      "epoch": 0.03,
+      "learning_rate": 5.4858934169279e-06,
+      "loss": 1.87089043,
+      "memory(GiB)": 58.46,
+      "step": 35,
+      "train_speed(iter/s)": 0.024184
+    },
+    {
+      "acc": 0.52221713,
+      "epoch": 0.03,
+      "learning_rate": 6.269592476489028e-06,
+      "loss": 1.78321209,
+      "memory(GiB)": 58.46,
+      "step": 40,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.53826532,
+      "epoch": 0.04,
+      "learning_rate": 7.053291536050157e-06,
+      "loss": 1.70393295,
+      "memory(GiB)": 58.46,
+      "step": 45,
+      "train_speed(iter/s)": 0.024212
+    },
+    {
+      "acc": 0.52223167,
+      "epoch": 0.04,
+      "learning_rate": 7.836990595611287e-06,
+      "loss": 1.74184074,
+      "memory(GiB)": 58.49,
+      "step": 50,
+      "train_speed(iter/s)": 0.024215
+    },
+    {
+      "acc": 0.55080438,
+      "epoch": 0.04,
+      "learning_rate": 8.620689655172414e-06,
+      "loss": 1.66886234,
+      "memory(GiB)": 58.49,
+      "step": 55,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.55040584,
+      "epoch": 0.05,
+      "learning_rate": 9.404388714733543e-06,
+      "loss": 1.62168198,
+      "memory(GiB)": 58.49,
+      "step": 60,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.56035452,
+      "epoch": 0.05,
+      "learning_rate": 1.0188087774294672e-05,
+      "loss": 1.61962128,
+      "memory(GiB)": 58.49,
+      "step": 65,
+      "train_speed(iter/s)": 0.024223
+    },
+    {
+      "acc": 0.5660913,
+      "epoch": 0.05,
+      "learning_rate": 1.09717868338558e-05,
+      "loss": 1.59430513,
+      "memory(GiB)": 58.49,
+      "step": 70,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.57812696,
+      "epoch": 0.06,
+      "learning_rate": 1.1755485893416929e-05,
+      "loss": 1.51112862,
+      "memory(GiB)": 58.49,
+      "step": 75,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.58022437,
+      "epoch": 0.06,
+      "learning_rate": 1.2539184952978056e-05,
+      "loss": 1.45745325,
+      "memory(GiB)": 58.49,
+      "step": 80,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.58660021,
+      "epoch": 0.07,
+      "learning_rate": 1.3322884012539186e-05,
+      "loss": 1.47154408,
+      "memory(GiB)": 58.49,
+      "step": 85,
+      "train_speed(iter/s)": 0.024254
+    },
+    {
+      "acc": 0.58845835,
+      "epoch": 0.07,
+      "learning_rate": 1.4106583072100313e-05,
+      "loss": 1.46421566,
+      "memory(GiB)": 60.96,
+      "step": 90,
+      "train_speed(iter/s)": 0.024265
+    },
+    {
+      "acc": 0.59027457,
+      "epoch": 0.07,
+      "learning_rate": 1.4890282131661442e-05,
+      "loss": 1.49178133,
+      "memory(GiB)": 60.96,
+      "step": 95,
+      "train_speed(iter/s)": 0.024276
+    },
+    {
+      "acc": 0.59797707,
+      "epoch": 0.08,
+      "learning_rate": 1.5673981191222574e-05,
+      "loss": 1.42499933,
+      "memory(GiB)": 60.96,
+      "step": 100,
+      "train_speed(iter/s)": 0.024276
+    },
+    {
+      "acc": 0.61143513,
+      "epoch": 0.08,
+      "learning_rate": 1.64576802507837e-05,
+      "loss": 1.35773869,
+      "memory(GiB)": 60.96,
+      "step": 105,
+      "train_speed(iter/s)": 0.024269
+    },
+    {
+      "acc": 0.60093813,
+      "epoch": 0.09,
+      "learning_rate": 1.7241379310344828e-05,
+      "loss": 1.42934065,
+      "memory(GiB)": 60.96,
+      "step": 110,
+      "train_speed(iter/s)": 0.024277
+    },
+    {
+      "acc": 0.60922484,
+      "epoch": 0.09,
+      "learning_rate": 1.8025078369905957e-05,
+      "loss": 1.38470116,
+      "memory(GiB)": 60.96,
+      "step": 115,
+      "train_speed(iter/s)": 0.024276
+    },
+    {
+      "acc": 0.59300284,
+      "epoch": 0.09,
+      "learning_rate": 1.8808777429467086e-05,
+      "loss": 1.40323668,
+      "memory(GiB)": 60.96,
+      "step": 120,
+      "train_speed(iter/s)": 0.02428
+    },
+    {
+      "acc": 0.59103971,
+      "epoch": 0.1,
+      "learning_rate": 1.9592476489028214e-05,
+      "loss": 1.42401762,
+      "memory(GiB)": 60.96,
+      "step": 125,
+      "train_speed(iter/s)": 0.024274
+    },
+    {
+      "acc": 0.59748178,
+      "epoch": 0.1,
+      "learning_rate": 2.0376175548589343e-05,
+      "loss": 1.43538437,
+      "memory(GiB)": 60.96,
+      "step": 130,
+      "train_speed(iter/s)": 0.024284
+    },
+    {
+      "acc": 0.59937601,
+      "epoch": 0.11,
+      "learning_rate": 2.1159874608150472e-05,
+      "loss": 1.37400723,
+      "memory(GiB)": 60.96,
+      "step": 135,
+      "train_speed(iter/s)": 0.024272
+    },
+    {
+      "acc": 0.60988913,
+      "epoch": 0.11,
+      "learning_rate": 2.19435736677116e-05,
+      "loss": 1.33568382,
+      "memory(GiB)": 60.96,
+      "step": 140,
+      "train_speed(iter/s)": 0.024269
+    },
+    {
+      "acc": 0.59887972,
+      "epoch": 0.11,
+      "learning_rate": 2.272727272727273e-05,
+      "loss": 1.37526703,
+      "memory(GiB)": 60.96,
+      "step": 145,
+      "train_speed(iter/s)": 0.024273
+    },
+    {
+      "acc": 0.60464563,
+      "epoch": 0.12,
+      "learning_rate": 2.3510971786833858e-05,
+      "loss": 1.38023567,
+      "memory(GiB)": 60.96,
+      "step": 150,
+      "train_speed(iter/s)": 0.02427
+    },
+    {
+      "acc": 0.62468381,
+      "epoch": 0.12,
+      "learning_rate": 2.4294670846394983e-05,
+      "loss": 1.31268759,
+      "memory(GiB)": 60.96,
+      "step": 155,
+      "train_speed(iter/s)": 0.024263
+    },
+    {
+      "acc": 0.615977,
+      "epoch": 0.13,
+      "learning_rate": 2.5078369905956112e-05,
+      "loss": 1.31516523,
+      "memory(GiB)": 60.96,
+      "step": 160,
+      "train_speed(iter/s)": 0.024272
+    },
+    {
+      "acc": 0.62613187,
+      "epoch": 0.13,
+      "learning_rate": 2.5862068965517244e-05,
+      "loss": 1.30936899,
+      "memory(GiB)": 60.96,
+      "step": 165,
+      "train_speed(iter/s)": 0.02427
+    },
+    {
+      "acc": 0.60663314,
+      "epoch": 0.13,
+      "learning_rate": 2.6645768025078373e-05,
+      "loss": 1.34656582,
+      "memory(GiB)": 60.96,
+      "step": 170,
+      "train_speed(iter/s)": 0.024273
+    },
+    {
+      "acc": 0.6247479,
+      "epoch": 0.14,
+      "learning_rate": 2.7429467084639498e-05,
+      "loss": 1.30112534,
+      "memory(GiB)": 60.96,
+      "step": 175,
+      "train_speed(iter/s)": 0.024272
+    },
+    {
+      "acc": 0.62302966,
+      "epoch": 0.14,
+      "learning_rate": 2.8213166144200627e-05,
+      "loss": 1.3161747,
+      "memory(GiB)": 63.69,
+      "step": 180,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.62723155,
+      "epoch": 0.14,
+      "learning_rate": 2.899686520376176e-05,
+      "loss": 1.28483057,
+      "memory(GiB)": 63.69,
+      "step": 185,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.61306233,
+      "epoch": 0.15,
+      "learning_rate": 2.9780564263322884e-05,
+      "loss": 1.3089962,
+      "memory(GiB)": 63.69,
+      "step": 190,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.60289783,
+      "epoch": 0.15,
+      "learning_rate": 3.056426332288401e-05,
+      "loss": 1.38051376,
+      "memory(GiB)": 63.69,
+      "step": 195,
+      "train_speed(iter/s)": 0.024228
+    },
+    {
+      "acc": 0.61080809,
+      "epoch": 0.16,
+      "learning_rate": 3.134796238244515e-05,
+      "loss": 1.34134455,
+      "memory(GiB)": 63.69,
+      "step": 200,
+      "train_speed(iter/s)": 0.024225
+    },
+    {
+      "acc": 0.62451038,
+      "epoch": 0.16,
+      "learning_rate": 3.213166144200627e-05,
+      "loss": 1.30470991,
+      "memory(GiB)": 63.69,
+      "step": 205,
+      "train_speed(iter/s)": 0.024222
+    },
+    {
+      "acc": 0.61880503,
+      "epoch": 0.16,
+      "learning_rate": 3.29153605015674e-05,
+      "loss": 1.31368427,
+      "memory(GiB)": 63.69,
+      "step": 210,
+      "train_speed(iter/s)": 0.02423
+    },
+    {
+      "acc": 0.62762785,
+      "epoch": 0.17,
+      "learning_rate": 3.369905956112853e-05,
+      "loss": 1.25870705,
+      "memory(GiB)": 63.69,
+      "step": 215,
+      "train_speed(iter/s)": 0.024226
+    },
+    {
+      "acc": 0.6347024,
+      "epoch": 0.17,
+      "learning_rate": 3.4482758620689657e-05,
+      "loss": 1.25454168,
+      "memory(GiB)": 63.69,
+      "step": 220,
+      "train_speed(iter/s)": 0.024227
+    },
+    {
+      "acc": 0.63634839,
+      "epoch": 0.18,
+      "learning_rate": 3.5266457680250785e-05,
+      "loss": 1.24715233,
+      "memory(GiB)": 63.69,
+      "step": 225,
+      "train_speed(iter/s)": 0.024228
+    },
+    {
+      "acc": 0.61799154,
+      "epoch": 0.18,
+      "learning_rate": 3.6050156739811914e-05,
+      "loss": 1.33669329,
+      "memory(GiB)": 63.69,
+      "step": 230,
+      "train_speed(iter/s)": 0.024225
+    },
+    {
+      "acc": 0.61765265,
+      "epoch": 0.18,
+      "learning_rate": 3.683385579937304e-05,
+      "loss": 1.29450636,
+      "memory(GiB)": 63.69,
+      "step": 235,
+      "train_speed(iter/s)": 0.024226
+    },
+    {
+      "acc": 0.62776251,
+      "epoch": 0.19,
+      "learning_rate": 3.761755485893417e-05,
+      "loss": 1.29685411,
+      "memory(GiB)": 63.69,
+      "step": 240,
+      "train_speed(iter/s)": 0.024226
+    },
+    {
+      "acc": 0.62579837,
+      "epoch": 0.19,
+      "learning_rate": 3.84012539184953e-05,
+      "loss": 1.29461927,
+      "memory(GiB)": 63.69,
+      "step": 245,
+      "train_speed(iter/s)": 0.024227
+    },
+    {
+      "acc": 0.62839761,
+      "epoch": 0.2,
+      "learning_rate": 3.918495297805643e-05,
+      "loss": 1.30260792,
+      "memory(GiB)": 63.69,
+      "step": 250,
+      "train_speed(iter/s)": 0.02423
+    },
+    {
+      "acc": 0.62507124,
+      "epoch": 0.2,
+      "learning_rate": 3.996865203761756e-05,
+      "loss": 1.28496952,
+      "memory(GiB)": 63.69,
+      "step": 255,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.62980595,
+      "epoch": 0.2,
+      "learning_rate": 4.0752351097178686e-05,
+      "loss": 1.29334612,
+      "memory(GiB)": 63.69,
+      "step": 260,
+      "train_speed(iter/s)": 0.024242
+    },
+    {
+      "acc": 0.63067842,
+      "epoch": 0.21,
+      "learning_rate": 4.1536050156739815e-05,
+      "loss": 1.30464935,
+      "memory(GiB)": 63.69,
+      "step": 265,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.637749,
+      "epoch": 0.21,
+      "learning_rate": 4.2319749216300944e-05,
+      "loss": 1.26530437,
+      "memory(GiB)": 63.69,
+      "step": 270,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.62997618,
+      "epoch": 0.22,
+      "learning_rate": 4.3103448275862066e-05,
+      "loss": 1.27895756,
+      "memory(GiB)": 63.69,
+      "step": 275,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.62871423,
+      "epoch": 0.22,
+      "learning_rate": 4.38871473354232e-05,
+      "loss": 1.24851456,
+      "memory(GiB)": 63.69,
+      "step": 280,
+      "train_speed(iter/s)": 0.024242
+    },
+    {
+      "acc": 0.62472744,
+      "epoch": 0.22,
+      "learning_rate": 4.467084639498433e-05,
+      "loss": 1.280616,
+      "memory(GiB)": 63.69,
+      "step": 285,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.6316843,
+      "epoch": 0.23,
+      "learning_rate": 4.545454545454546e-05,
+      "loss": 1.25226479,
+      "memory(GiB)": 66.18,
+      "step": 290,
+      "train_speed(iter/s)": 0.024242
+    },
+    {
+      "acc": 0.63331189,
+      "epoch": 0.23,
+      "learning_rate": 4.623824451410659e-05,
+      "loss": 1.24571609,
+      "memory(GiB)": 66.18,
+      "step": 295,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.6343452,
+      "epoch": 0.24,
+      "learning_rate": 4.7021943573667716e-05,
+      "loss": 1.28765945,
+      "memory(GiB)": 66.18,
+      "step": 300,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.61879272,
+      "epoch": 0.24,
+      "learning_rate": 4.7805642633228845e-05,
+      "loss": 1.32476501,
+      "memory(GiB)": 66.18,
+      "step": 305,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.64381948,
+      "epoch": 0.24,
+      "learning_rate": 4.858934169278997e-05,
+      "loss": 1.21850481,
+      "memory(GiB)": 66.18,
+      "step": 310,
+      "train_speed(iter/s)": 0.024237
+    },
+    {
+      "acc": 0.6378974,
+      "epoch": 0.25,
+      "learning_rate": 4.93730407523511e-05,
+      "loss": 1.23893595,
+      "memory(GiB)": 66.18,
+      "step": 315,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.64307256,
+      "epoch": 0.25,
+      "learning_rate": 5.0156739811912224e-05,
+      "loss": 1.21721163,
+      "memory(GiB)": 66.18,
+      "step": 320,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.63342023,
+      "epoch": 0.25,
+      "learning_rate": 5.094043887147336e-05,
+      "loss": 1.22513857,
+      "memory(GiB)": 66.18,
+      "step": 325,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.62956347,
+      "epoch": 0.26,
+      "learning_rate": 5.172413793103449e-05,
+      "loss": 1.273248,
+      "memory(GiB)": 66.18,
+      "step": 330,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.6396666,
+      "epoch": 0.26,
+      "learning_rate": 5.250783699059562e-05,
+      "loss": 1.2052968,
+      "memory(GiB)": 66.18,
+      "step": 335,
+      "train_speed(iter/s)": 0.024242
+    },
+    {
+      "acc": 0.63923178,
+      "epoch": 0.27,
+      "learning_rate": 5.3291536050156746e-05,
+      "loss": 1.19400892,
+      "memory(GiB)": 66.18,
+      "step": 340,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.64041128,
+      "epoch": 0.27,
+      "learning_rate": 5.407523510971787e-05,
+      "loss": 1.19951448,
+      "memory(GiB)": 66.18,
+      "step": 345,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.64085617,
+      "epoch": 0.27,
+      "learning_rate": 5.4858934169278996e-05,
+      "loss": 1.21989889,
+      "memory(GiB)": 66.18,
+      "step": 350,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.63642988,
+      "epoch": 0.28,
+      "learning_rate": 5.5642633228840125e-05,
+      "loss": 1.26623602,
+      "memory(GiB)": 66.18,
+      "step": 355,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.64299493,
+      "epoch": 0.28,
+      "learning_rate": 5.6426332288401254e-05,
+      "loss": 1.22799644,
+      "memory(GiB)": 66.18,
+      "step": 360,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.62601166,
+      "epoch": 0.29,
+      "learning_rate": 5.721003134796239e-05,
+      "loss": 1.27594986,
+      "memory(GiB)": 66.18,
+      "step": 365,
+      "train_speed(iter/s)": 0.024248
+    },
+    {
+      "acc": 0.63284698,
+      "epoch": 0.29,
+      "learning_rate": 5.799373040752352e-05,
+      "loss": 1.2637042,
+      "memory(GiB)": 66.18,
+      "step": 370,
+      "train_speed(iter/s)": 0.02425
+    },
+    {
+      "acc": 0.64749308,
+      "epoch": 0.29,
+      "learning_rate": 5.877742946708465e-05,
+      "loss": 1.17957096,
+      "memory(GiB)": 68.68,
+      "step": 375,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.63170052,
+      "epoch": 0.3,
+      "learning_rate": 5.956112852664577e-05,
+      "loss": 1.26501703,
+      "memory(GiB)": 68.68,
+      "step": 380,
+      "train_speed(iter/s)": 0.024248
+    },
+    {
+      "acc": 0.64609647,
+      "epoch": 0.3,
+      "learning_rate": 6.03448275862069e-05,
+      "loss": 1.1953764,
+      "memory(GiB)": 68.68,
+      "step": 385,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.63454189,
+      "epoch": 0.31,
+      "learning_rate": 6.112852664576803e-05,
+      "loss": 1.22050667,
+      "memory(GiB)": 68.68,
+      "step": 390,
+      "train_speed(iter/s)": 0.024249
+    },
+    {
+      "acc": 0.6379528,
+      "epoch": 0.31,
+      "learning_rate": 6.191222570532915e-05,
+      "loss": 1.21711063,
+      "memory(GiB)": 68.68,
+      "step": 395,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.64480844,
+      "epoch": 0.31,
+      "learning_rate": 6.26959247648903e-05,
+      "loss": 1.18095789,
+      "memory(GiB)": 68.68,
+      "step": 400,
+      "train_speed(iter/s)": 0.024249
+    },
+    {
+      "acc": 0.64001446,
+      "epoch": 0.32,
+      "learning_rate": 6.347962382445141e-05,
+      "loss": 1.20016241,
+      "memory(GiB)": 68.68,
+      "step": 405,
+      "train_speed(iter/s)": 0.02425
+    },
+    {
+      "acc": 0.64171805,
+      "epoch": 0.32,
+      "learning_rate": 6.426332288401254e-05,
+      "loss": 1.23874407,
+      "memory(GiB)": 68.68,
+      "step": 410,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.62769017,
+      "epoch": 0.33,
+      "learning_rate": 6.504702194357367e-05,
+      "loss": 1.28347731,
+      "memory(GiB)": 68.68,
+      "step": 415,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.6374104,
+      "epoch": 0.33,
+      "learning_rate": 6.58307210031348e-05,
+      "loss": 1.20551748,
+      "memory(GiB)": 68.68,
+      "step": 420,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.63077922,
+      "epoch": 0.33,
+      "learning_rate": 6.661442006269593e-05,
+      "loss": 1.23985853,
+      "memory(GiB)": 68.68,
+      "step": 425,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.63933249,
+      "epoch": 0.34,
+      "learning_rate": 6.739811912225706e-05,
+      "loss": 1.25593414,
+      "memory(GiB)": 68.68,
+      "step": 430,
+      "train_speed(iter/s)": 0.024248
+    },
+    {
+      "acc": 0.63769989,
+      "epoch": 0.34,
+      "learning_rate": 6.818181818181818e-05,
+      "loss": 1.22561283,
+      "memory(GiB)": 68.68,
+      "step": 435,
+      "train_speed(iter/s)": 0.024248
+    },
+    {
+      "acc": 0.62949352,
+      "epoch": 0.34,
+      "learning_rate": 6.896551724137931e-05,
+      "loss": 1.26371164,
+      "memory(GiB)": 68.68,
+      "step": 440,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.64247766,
+      "epoch": 0.35,
+      "learning_rate": 6.974921630094044e-05,
+      "loss": 1.20132828,
+      "memory(GiB)": 68.68,
+      "step": 445,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.63743186,
+      "epoch": 0.35,
+      "learning_rate": 7.053291536050157e-05,
+      "loss": 1.25252228,
+      "memory(GiB)": 68.68,
+      "step": 450,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.63471522,
+      "epoch": 0.36,
+      "learning_rate": 7.13166144200627e-05,
+      "loss": 1.21878319,
+      "memory(GiB)": 68.68,
+      "step": 455,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.64033532,
+      "epoch": 0.36,
+      "learning_rate": 7.210031347962383e-05,
+      "loss": 1.23047905,
+      "memory(GiB)": 68.68,
+      "step": 460,
+      "train_speed(iter/s)": 0.02425
+    },
+    {
+      "acc": 0.62159195,
+      "epoch": 0.36,
+      "learning_rate": 7.288401253918496e-05,
+      "loss": 1.26657543,
+      "memory(GiB)": 68.68,
+      "step": 465,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.63586564,
+      "epoch": 0.37,
+      "learning_rate": 7.366771159874609e-05,
+      "loss": 1.22125769,
+      "memory(GiB)": 68.68,
+      "step": 470,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.62739558,
+      "epoch": 0.37,
+      "learning_rate": 7.445141065830721e-05,
+      "loss": 1.24667158,
+      "memory(GiB)": 68.68,
+      "step": 475,
+      "train_speed(iter/s)": 0.02425
+    },
+    {
+      "acc": 0.64560175,
+      "epoch": 0.38,
+      "learning_rate": 7.523510971786834e-05,
+      "loss": 1.19969473,
+      "memory(GiB)": 68.68,
+      "step": 480,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64611869,
+      "epoch": 0.38,
+      "learning_rate": 7.601880877742947e-05,
+      "loss": 1.18089733,
+      "memory(GiB)": 68.68,
+      "step": 485,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.6526392,
+      "epoch": 0.38,
+      "learning_rate": 7.68025078369906e-05,
+      "loss": 1.18621178,
+      "memory(GiB)": 68.68,
+      "step": 490,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64487033,
+      "epoch": 0.39,
+      "learning_rate": 7.758620689655173e-05,
+      "loss": 1.22172861,
+      "memory(GiB)": 68.68,
+      "step": 495,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64801369,
+      "epoch": 0.39,
+      "learning_rate": 7.836990595611286e-05,
+      "loss": 1.1807312,
+      "memory(GiB)": 68.68,
+      "step": 500,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.65408354,
+      "epoch": 0.4,
+      "learning_rate": 7.915360501567399e-05,
+      "loss": 1.17569618,
+      "memory(GiB)": 68.68,
+      "step": 505,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.64917159,
+      "epoch": 0.4,
+      "learning_rate": 7.993730407523512e-05,
+      "loss": 1.18598318,
+      "memory(GiB)": 68.68,
+      "step": 510,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.6524426,
+      "epoch": 0.4,
+      "learning_rate": 8.072100313479624e-05,
+      "loss": 1.19354515,
+      "memory(GiB)": 68.68,
+      "step": 515,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.63392072,
+      "epoch": 0.41,
+      "learning_rate": 8.150470219435737e-05,
+      "loss": 1.25341883,
+      "memory(GiB)": 68.68,
+      "step": 520,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.64778457,
+      "epoch": 0.41,
+      "learning_rate": 8.22884012539185e-05,
+      "loss": 1.18014956,
+      "memory(GiB)": 68.68,
+      "step": 525,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.65192814,
+      "epoch": 0.42,
+      "learning_rate": 8.307210031347963e-05,
+      "loss": 1.14602833,
+      "memory(GiB)": 68.68,
+      "step": 530,
+      "train_speed(iter/s)": 0.024252
+    },
+    {
+      "acc": 0.6464828,
+      "epoch": 0.42,
+      "learning_rate": 8.385579937304076e-05,
+      "loss": 1.20210142,
+      "memory(GiB)": 68.68,
+      "step": 535,
+      "train_speed(iter/s)": 0.024249
+    },
+    {
+      "acc": 0.64848366,
+      "epoch": 0.42,
+      "learning_rate": 8.463949843260189e-05,
+      "loss": 1.16304598,
+      "memory(GiB)": 68.68,
+      "step": 540,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.65109825,
+      "epoch": 0.43,
+      "learning_rate": 8.542319749216302e-05,
+      "loss": 1.17073431,
+      "memory(GiB)": 68.68,
+      "step": 545,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.64547157,
+      "epoch": 0.43,
+      "learning_rate": 8.620689655172413e-05,
+      "loss": 1.21380081,
+      "memory(GiB)": 68.68,
+      "step": 550,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.63410115,
+      "epoch": 0.43,
+      "learning_rate": 8.699059561128527e-05,
+      "loss": 1.25124855,
+      "memory(GiB)": 68.68,
+      "step": 555,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.63803411,
+      "epoch": 0.44,
+      "learning_rate": 8.77742946708464e-05,
+      "loss": 1.2044651,
+      "memory(GiB)": 68.68,
+      "step": 560,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.64598598,
+      "epoch": 0.44,
+      "learning_rate": 8.855799373040753e-05,
+      "loss": 1.18607101,
+      "memory(GiB)": 68.68,
+      "step": 565,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.63589487,
+      "epoch": 0.45,
+      "learning_rate": 8.934169278996866e-05,
+      "loss": 1.21959095,
+      "memory(GiB)": 68.68,
+      "step": 570,
+      "train_speed(iter/s)": 0.024249
+    },
+    {
+      "acc": 0.66258736,
+      "epoch": 0.45,
+      "learning_rate": 9.012539184952979e-05,
+      "loss": 1.14576664,
+      "memory(GiB)": 68.68,
+      "step": 575,
+      "train_speed(iter/s)": 0.024249
+    },
+    {
+      "acc": 0.644449,
+      "epoch": 0.45,
+      "learning_rate": 9.090909090909092e-05,
+      "loss": 1.17768698,
+      "memory(GiB)": 68.68,
+      "step": 580,
+      "train_speed(iter/s)": 0.02425
+    },
+    {
+      "acc": 0.63850074,
+      "epoch": 0.46,
+      "learning_rate": 9.169278996865203e-05,
+      "loss": 1.21261606,
+      "memory(GiB)": 68.68,
+      "step": 585,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.66103854,
+      "epoch": 0.46,
+      "learning_rate": 9.247648902821317e-05,
+      "loss": 1.14370327,
+      "memory(GiB)": 68.68,
+      "step": 590,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.64929209,
+      "epoch": 0.47,
+      "learning_rate": 9.32601880877743e-05,
+      "loss": 1.17919035,
+      "memory(GiB)": 68.68,
+      "step": 595,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.6462472,
+      "epoch": 0.47,
+      "learning_rate": 9.404388714733543e-05,
+      "loss": 1.19705715,
+      "memory(GiB)": 68.68,
+      "step": 600,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.65177841,
+      "epoch": 0.47,
+      "learning_rate": 9.482758620689656e-05,
+      "loss": 1.17879667,
+      "memory(GiB)": 68.68,
+      "step": 605,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.63742218,
+      "epoch": 0.48,
+      "learning_rate": 9.561128526645769e-05,
+      "loss": 1.22292814,
+      "memory(GiB)": 68.68,
+      "step": 610,
+      "train_speed(iter/s)": 0.024257
+    },
+    {
+      "acc": 0.6366725,
+      "epoch": 0.48,
+      "learning_rate": 9.63949843260188e-05,
+      "loss": 1.21893911,
+      "memory(GiB)": 68.68,
+      "step": 615,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.6517189,
+      "epoch": 0.49,
+      "learning_rate": 9.717868338557993e-05,
+      "loss": 1.16371479,
+      "memory(GiB)": 68.68,
+      "step": 620,
+      "train_speed(iter/s)": 0.024257
+    },
+    {
+      "acc": 0.640413,
+      "epoch": 0.49,
+      "learning_rate": 9.796238244514106e-05,
+      "loss": 1.21149368,
+      "memory(GiB)": 68.68,
+      "step": 625,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.64837985,
+      "epoch": 0.49,
+      "learning_rate": 9.87460815047022e-05,
+      "loss": 1.18780231,
+      "memory(GiB)": 68.68,
+      "step": 630,
+      "train_speed(iter/s)": 0.024257
+    },
+    {
+      "acc": 0.64622922,
+      "epoch": 0.5,
+      "learning_rate": 9.952978056426333e-05,
+      "loss": 1.19707012,
+      "memory(GiB)": 68.68,
+      "step": 635,
+      "train_speed(iter/s)": 0.024259
+    },
+    {
+      "acc": 0.63969393,
+      "epoch": 0.5,
+      "learning_rate": 9.999999327227841e-05,
+      "loss": 1.19009142,
+      "memory(GiB)": 68.68,
+      "step": 640,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.66441765,
+      "epoch": 0.51,
+      "learning_rate": 9.999991758543125e-05,
+      "loss": 1.15019789,
+      "memory(GiB)": 68.68,
+      "step": 645,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.64071684,
+      "epoch": 0.51,
+      "learning_rate": 9.999975780221265e-05,
+      "loss": 1.21033449,
+      "memory(GiB)": 68.68,
+      "step": 650,
+      "train_speed(iter/s)": 0.024262
+    },
+    {
+      "acc": 0.65804129,
+      "epoch": 0.51,
+      "learning_rate": 9.999951392289139e-05,
+      "loss": 1.14957666,
+      "memory(GiB)": 68.68,
+      "step": 655,
+      "train_speed(iter/s)": 0.024264
+    },
+    {
+      "acc": 0.64765239,
+      "epoch": 0.52,
+      "learning_rate": 9.999918594787761e-05,
+      "loss": 1.19038534,
+      "memory(GiB)": 68.68,
+      "step": 660,
+      "train_speed(iter/s)": 0.024263
+    },
+    {
+      "acc": 0.64099531,
+      "epoch": 0.52,
+      "learning_rate": 9.999877387772296e-05,
+      "loss": 1.21312761,
+      "memory(GiB)": 68.68,
+      "step": 665,
+      "train_speed(iter/s)": 0.024263
+    },
+    {
+      "acc": 0.63610272,
+      "epoch": 0.53,
+      "learning_rate": 9.999827771312054e-05,
+      "loss": 1.20392857,
+      "memory(GiB)": 68.68,
+      "step": 670,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.65267296,
+      "epoch": 0.53,
+      "learning_rate": 9.999769745490481e-05,
+      "loss": 1.17815151,
+      "memory(GiB)": 68.68,
+      "step": 675,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.64978647,
+      "epoch": 0.53,
+      "learning_rate": 9.999703310405178e-05,
+      "loss": 1.17260685,
+      "memory(GiB)": 68.68,
+      "step": 680,
+      "train_speed(iter/s)": 0.024259
+    },
+    {
+      "acc": 0.64660273,
+      "epoch": 0.54,
+      "learning_rate": 9.999628466167881e-05,
+      "loss": 1.1884903,
+      "memory(GiB)": 68.68,
+      "step": 685,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.64088135,
+      "epoch": 0.54,
+      "learning_rate": 9.999545212904473e-05,
+      "loss": 1.21543169,
+      "memory(GiB)": 71.19,
+      "step": 690,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64940615,
+      "epoch": 0.54,
+      "learning_rate": 9.999453550754981e-05,
+      "loss": 1.16802692,
+      "memory(GiB)": 71.19,
+      "step": 695,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64227314,
+      "epoch": 0.55,
+      "learning_rate": 9.999353479873575e-05,
+      "loss": 1.19985647,
+      "memory(GiB)": 71.19,
+      "step": 700,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.65743675,
+      "epoch": 0.55,
+      "learning_rate": 9.999245000428563e-05,
+      "loss": 1.13964968,
+      "memory(GiB)": 71.19,
+      "step": 705,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.65150938,
+      "epoch": 0.56,
+      "learning_rate": 9.999128112602406e-05,
+      "loss": 1.13540983,
+      "memory(GiB)": 71.19,
+      "step": 710,
+      "train_speed(iter/s)": 0.024254
+    },
+    {
+      "acc": 0.65529537,
+      "epoch": 0.56,
+      "learning_rate": 9.999002816591696e-05,
+      "loss": 1.148839,
+      "memory(GiB)": 71.19,
+      "step": 715,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.65492792,
+      "epoch": 0.56,
+      "learning_rate": 9.998869112607177e-05,
+      "loss": 1.16663084,
+      "memory(GiB)": 71.19,
+      "step": 720,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.64847426,
+      "epoch": 0.57,
+      "learning_rate": 9.998727000873724e-05,
+      "loss": 1.19009991,
+      "memory(GiB)": 71.19,
+      "step": 725,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.65983605,
+      "epoch": 0.57,
+      "learning_rate": 9.998576481630364e-05,
+      "loss": 1.17517796,
+      "memory(GiB)": 71.19,
+      "step": 730,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.66514916,
+      "epoch": 0.58,
+      "learning_rate": 9.998417555130258e-05,
+      "loss": 1.11018133,
+      "memory(GiB)": 71.19,
+      "step": 735,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.65130382,
+      "epoch": 0.58,
+      "learning_rate": 9.998250221640708e-05,
+      "loss": 1.18467026,
+      "memory(GiB)": 71.19,
+      "step": 740,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.64631677,
+      "epoch": 0.58,
+      "learning_rate": 9.99807448144316e-05,
+      "loss": 1.16014452,
+      "memory(GiB)": 71.19,
+      "step": 745,
+      "train_speed(iter/s)": 0.024258
+    },
+    {
+      "acc": 0.64330449,
+      "epoch": 0.59,
+      "learning_rate": 9.997890334833194e-05,
+      "loss": 1.16912346,
+      "memory(GiB)": 71.19,
+      "step": 750,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.64438591,
+      "epoch": 0.59,
+      "learning_rate": 9.997697782120535e-05,
+      "loss": 1.18001528,
+      "memory(GiB)": 71.19,
+      "step": 755,
+      "train_speed(iter/s)": 0.024262
+    },
+    {
+      "acc": 0.64493594,
+      "epoch": 0.6,
+      "learning_rate": 9.997496823629038e-05,
+      "loss": 1.18717165,
+      "memory(GiB)": 71.19,
+      "step": 760,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.64458404,
+      "epoch": 0.6,
+      "learning_rate": 9.997287459696707e-05,
+      "loss": 1.2015686,
+      "memory(GiB)": 71.19,
+      "step": 765,
+      "train_speed(iter/s)": 0.024262
+    },
+    {
+      "acc": 0.65334735,
+      "epoch": 0.6,
+      "learning_rate": 9.997069690675673e-05,
+      "loss": 1.13985786,
+      "memory(GiB)": 71.19,
+      "step": 770,
+      "train_speed(iter/s)": 0.024263
+    },
+    {
+      "acc": 0.65646448,
+      "epoch": 0.61,
+      "learning_rate": 9.996843516932212e-05,
+      "loss": 1.14643917,
+      "memory(GiB)": 71.19,
+      "step": 775,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.63831444,
+      "epoch": 0.61,
+      "learning_rate": 9.99660893884673e-05,
+      "loss": 1.20351677,
+      "memory(GiB)": 71.19,
+      "step": 780,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.64026852,
+      "epoch": 0.62,
+      "learning_rate": 9.996365956813771e-05,
+      "loss": 1.21742287,
+      "memory(GiB)": 71.19,
+      "step": 785,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.65602589,
+      "epoch": 0.62,
+      "learning_rate": 9.996114571242014e-05,
+      "loss": 1.1601553,
+      "memory(GiB)": 71.19,
+      "step": 790,
+      "train_speed(iter/s)": 0.024259
+    },
+    {
+      "acc": 0.64420905,
+      "epoch": 0.62,
+      "learning_rate": 9.995854782554275e-05,
+      "loss": 1.19600077,
+      "memory(GiB)": 71.19,
+      "step": 795,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.64605112,
+      "epoch": 0.63,
+      "learning_rate": 9.995586591187496e-05,
+      "loss": 1.18536654,
+      "memory(GiB)": 71.19,
+      "step": 800,
+      "train_speed(iter/s)": 0.02426
+    },
+    {
+      "acc": 0.63376851,
+      "epoch": 0.63,
+      "learning_rate": 9.995309997592757e-05,
+      "loss": 1.2327877,
+      "memory(GiB)": 71.19,
+      "step": 805,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.63126373,
+      "epoch": 0.63,
+      "learning_rate": 9.995025002235272e-05,
+      "loss": 1.24970627,
+      "memory(GiB)": 71.19,
+      "step": 810,
+      "train_speed(iter/s)": 0.024261
+    },
+    {
+      "acc": 0.64044433,
+      "epoch": 0.64,
+      "learning_rate": 9.994731605594381e-05,
+      "loss": 1.19621964,
+      "memory(GiB)": 71.19,
+      "step": 815,
+      "train_speed(iter/s)": 0.024259
+    },
+    {
+      "acc": 0.64977989,
+      "epoch": 0.64,
+      "learning_rate": 9.994429808163556e-05,
+      "loss": 1.17283697,
+      "memory(GiB)": 71.19,
+      "step": 820,
+      "train_speed(iter/s)": 0.024256
+    },
+    {
+      "acc": 0.64557714,
+      "epoch": 0.65,
+      "learning_rate": 9.994119610450401e-05,
+      "loss": 1.18671465,
+      "memory(GiB)": 71.19,
+      "step": 825,
+      "train_speed(iter/s)": 0.024257
+    },
+    {
+      "acc": 0.66244955,
+      "epoch": 0.65,
+      "learning_rate": 9.993801012976647e-05,
+      "loss": 1.1419055,
+      "memory(GiB)": 71.19,
+      "step": 830,
+      "train_speed(iter/s)": 0.024255
+    },
+    {
+      "acc": 0.66743221,
+      "epoch": 0.65,
+      "learning_rate": 9.99347401627815e-05,
+      "loss": 1.10382252,
+      "memory(GiB)": 71.19,
+      "step": 835,
+      "train_speed(iter/s)": 0.024253
+    },
+    {
+      "acc": 0.65264039,
+      "epoch": 0.66,
+      "learning_rate": 9.993138620904901e-05,
+      "loss": 1.14855709,
+      "memory(GiB)": 71.19,
+      "step": 840,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.65064726,
+      "epoch": 0.66,
+      "learning_rate": 9.992794827421004e-05,
+      "loss": 1.15438948,
+      "memory(GiB)": 71.19,
+      "step": 845,
+      "train_speed(iter/s)": 0.024251
+    },
+    {
+      "acc": 0.64738302,
+      "epoch": 0.67,
+      "learning_rate": 9.992442636404701e-05,
+      "loss": 1.18792486,
+      "memory(GiB)": 71.19,
+      "step": 850,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.63004446,
+      "epoch": 0.67,
+      "learning_rate": 9.992082048448353e-05,
+      "loss": 1.24233475,
+      "memory(GiB)": 71.19,
+      "step": 855,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65914078,
+      "epoch": 0.67,
+      "learning_rate": 9.991713064158442e-05,
+      "loss": 1.13226604,
+      "memory(GiB)": 71.19,
+      "step": 860,
+      "train_speed(iter/s)": 0.024242
+    },
+    {
+      "acc": 0.65967417,
+      "epoch": 0.68,
+      "learning_rate": 9.991335684155574e-05,
+      "loss": 1.1242136,
+      "memory(GiB)": 71.19,
+      "step": 865,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.64683952,
+      "epoch": 0.68,
+      "learning_rate": 9.990949909074476e-05,
+      "loss": 1.14824209,
+      "memory(GiB)": 71.19,
+      "step": 870,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.64689598,
+      "epoch": 0.69,
+      "learning_rate": 9.990555739563994e-05,
+      "loss": 1.18441563,
+      "memory(GiB)": 71.19,
+      "step": 875,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.64474206,
+      "epoch": 0.69,
+      "learning_rate": 9.990153176287094e-05,
+      "loss": 1.17308598,
+      "memory(GiB)": 71.19,
+      "step": 880,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.64102311,
+      "epoch": 0.69,
+      "learning_rate": 9.989742219920861e-05,
+      "loss": 1.1535264,
+      "memory(GiB)": 71.19,
+      "step": 885,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.63249907,
+      "epoch": 0.7,
+      "learning_rate": 9.989322871156492e-05,
+      "loss": 1.24110155,
+      "memory(GiB)": 71.19,
+      "step": 890,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.65820117,
+      "epoch": 0.7,
+      "learning_rate": 9.988895130699305e-05,
+      "loss": 1.14757719,
+      "memory(GiB)": 71.19,
+      "step": 895,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.6568809,
+      "epoch": 0.71,
+      "learning_rate": 9.988458999268728e-05,
+      "loss": 1.1805316,
+      "memory(GiB)": 71.19,
+      "step": 900,
+      "train_speed(iter/s)": 0.024247
+    },
+    {
+      "acc": 0.65315137,
+      "epoch": 0.71,
+      "learning_rate": 9.988014477598304e-05,
+      "loss": 1.18291893,
+      "memory(GiB)": 71.19,
+      "step": 905,
+      "train_speed(iter/s)": 0.024248
+    },
+    {
+      "acc": 0.65974207,
+      "epoch": 0.71,
+      "learning_rate": 9.987561566435689e-05,
+      "loss": 1.11334858,
+      "memory(GiB)": 71.19,
+      "step": 910,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.66105995,
+      "epoch": 0.72,
+      "learning_rate": 9.987100266542644e-05,
+      "loss": 1.10593357,
+      "memory(GiB)": 71.19,
+      "step": 915,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65526924,
+      "epoch": 0.72,
+      "learning_rate": 9.986630578695047e-05,
+      "loss": 1.139046,
+      "memory(GiB)": 71.19,
+      "step": 920,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.64982562,
+      "epoch": 0.72,
+      "learning_rate": 9.986152503682879e-05,
+      "loss": 1.15638342,
+      "memory(GiB)": 71.19,
+      "step": 925,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.65965147,
+      "epoch": 0.73,
+      "learning_rate": 9.985666042310229e-05,
+      "loss": 1.10007477,
+      "memory(GiB)": 71.19,
+      "step": 930,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.64175363,
+      "epoch": 0.73,
+      "learning_rate": 9.98517119539529e-05,
+      "loss": 1.2035121,
+      "memory(GiB)": 71.19,
+      "step": 935,
+      "train_speed(iter/s)": 0.024246
+    },
+    {
+      "acc": 0.67316985,
+      "epoch": 0.74,
+      "learning_rate": 9.984667963770361e-05,
+      "loss": 1.0980998,
+      "memory(GiB)": 71.19,
+      "step": 940,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.65934963,
+      "epoch": 0.74,
+      "learning_rate": 9.984156348281842e-05,
+      "loss": 1.15233898,
+      "memory(GiB)": 71.19,
+      "step": 945,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65463395,
+      "epoch": 0.74,
+      "learning_rate": 9.983636349790235e-05,
+      "loss": 1.17552748,
+      "memory(GiB)": 71.19,
+      "step": 950,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.64568629,
+      "epoch": 0.75,
+      "learning_rate": 9.983107969170141e-05,
+      "loss": 1.18645792,
+      "memory(GiB)": 71.19,
+      "step": 955,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.64675536,
+      "epoch": 0.75,
+      "learning_rate": 9.982571207310259e-05,
+      "loss": 1.16107082,
+      "memory(GiB)": 71.19,
+      "step": 960,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.66028018,
+      "epoch": 0.76,
+      "learning_rate": 9.982026065113386e-05,
+      "loss": 1.14361153,
+      "memory(GiB)": 71.19,
+      "step": 965,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65468411,
+      "epoch": 0.76,
+      "learning_rate": 9.981472543496412e-05,
+      "loss": 1.14505243,
+      "memory(GiB)": 71.19,
+      "step": 970,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65591521,
+      "epoch": 0.76,
+      "learning_rate": 9.980910643390321e-05,
+      "loss": 1.12250948,
+      "memory(GiB)": 71.19,
+      "step": 975,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.66190786,
+      "epoch": 0.77,
+      "learning_rate": 9.980340365740193e-05,
+      "loss": 1.12542,
+      "memory(GiB)": 71.19,
+      "step": 980,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65410662,
+      "epoch": 0.77,
+      "learning_rate": 9.979761711505191e-05,
+      "loss": 1.14729557,
+      "memory(GiB)": 71.19,
+      "step": 985,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.66674676,
+      "epoch": 0.78,
+      "learning_rate": 9.979174681658574e-05,
+      "loss": 1.10079012,
+      "memory(GiB)": 71.19,
+      "step": 990,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.65360141,
+      "epoch": 0.78,
+      "learning_rate": 9.978579277187684e-05,
+      "loss": 1.16522446,
+      "memory(GiB)": 71.19,
+      "step": 995,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.66134276,
+      "epoch": 0.78,
+      "learning_rate": 9.97797549909395e-05,
+      "loss": 1.11977797,
+      "memory(GiB)": 71.19,
+      "step": 1000,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65579562,
+      "epoch": 0.79,
+      "learning_rate": 9.977363348392886e-05,
+      "loss": 1.15444326,
+      "memory(GiB)": 71.19,
+      "step": 1005,
+      "train_speed(iter/s)": 0.024244
+    },
+    {
+      "acc": 0.66116667,
+      "epoch": 0.79,
+      "learning_rate": 9.976742826114083e-05,
+      "loss": 1.12424622,
+      "memory(GiB)": 71.19,
+      "step": 1010,
+      "train_speed(iter/s)": 0.024245
+    },
+    {
+      "acc": 0.6683176,
+      "epoch": 0.8,
+      "learning_rate": 9.97611393330122e-05,
+      "loss": 1.11872005,
+      "memory(GiB)": 71.19,
+      "step": 1015,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65369892,
+      "epoch": 0.8,
+      "learning_rate": 9.975476671012049e-05,
+      "loss": 1.1429491,
+      "memory(GiB)": 71.19,
+      "step": 1020,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.64377575,
+      "epoch": 0.8,
+      "learning_rate": 9.974831040318403e-05,
+      "loss": 1.19284143,
+      "memory(GiB)": 71.19,
+      "step": 1025,
+      "train_speed(iter/s)": 0.024243
+    },
+    {
+      "acc": 0.65243697,
+      "epoch": 0.81,
+      "learning_rate": 9.974177042306183e-05,
+      "loss": 1.157862,
+      "memory(GiB)": 71.19,
+      "step": 1030,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.64958396,
+      "epoch": 0.81,
+      "learning_rate": 9.973514678075372e-05,
+      "loss": 1.18240442,
+      "memory(GiB)": 71.19,
+      "step": 1035,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.6682889,
+      "epoch": 0.82,
+      "learning_rate": 9.972843948740019e-05,
+      "loss": 1.11338902,
+      "memory(GiB)": 71.19,
+      "step": 1040,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.65257483,
+      "epoch": 0.82,
+      "learning_rate": 9.972164855428244e-05,
+      "loss": 1.13974934,
+      "memory(GiB)": 71.19,
+      "step": 1045,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.6569396,
+      "epoch": 0.82,
+      "learning_rate": 9.971477399282236e-05,
+      "loss": 1.13563824,
+      "memory(GiB)": 71.19,
+      "step": 1050,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.66365395,
+      "epoch": 0.83,
+      "learning_rate": 9.970781581458246e-05,
+      "loss": 1.13148432,
+      "memory(GiB)": 71.19,
+      "step": 1055,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.66655474,
+      "epoch": 0.83,
+      "learning_rate": 9.970077403126592e-05,
+      "loss": 1.12719841,
+      "memory(GiB)": 71.19,
+      "step": 1060,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.65451961,
+      "epoch": 0.83,
+      "learning_rate": 9.969364865471654e-05,
+      "loss": 1.17055464,
+      "memory(GiB)": 71.19,
+      "step": 1065,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.66282625,
+      "epoch": 0.84,
+      "learning_rate": 9.968643969691868e-05,
+      "loss": 1.11539078,
+      "memory(GiB)": 71.19,
+      "step": 1070,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.66324406,
+      "epoch": 0.84,
+      "learning_rate": 9.96791471699973e-05,
+      "loss": 1.12122641,
+      "memory(GiB)": 71.19,
+      "step": 1075,
+      "train_speed(iter/s)": 0.024237
+    },
+    {
+      "acc": 0.66220465,
+      "epoch": 0.85,
+      "learning_rate": 9.967177108621798e-05,
+      "loss": 1.1304287,
+      "memory(GiB)": 71.19,
+      "step": 1080,
+      "train_speed(iter/s)": 0.024237
+    },
+    {
+      "acc": 0.65334153,
+      "epoch": 0.85,
+      "learning_rate": 9.966431145798672e-05,
+      "loss": 1.15781803,
+      "memory(GiB)": 71.19,
+      "step": 1085,
+      "train_speed(iter/s)": 0.024235
+    },
+    {
+      "acc": 0.6635685,
+      "epoch": 0.85,
+      "learning_rate": 9.965676829785011e-05,
+      "loss": 1.10384178,
+      "memory(GiB)": 71.19,
+      "step": 1090,
+      "train_speed(iter/s)": 0.024235
+    },
+    {
+      "acc": 0.66880417,
+      "epoch": 0.86,
+      "learning_rate": 9.964914161849522e-05,
+      "loss": 1.1105998,
+      "memory(GiB)": 71.19,
+      "step": 1095,
+      "train_speed(iter/s)": 0.024235
+    },
+    {
+      "acc": 0.65008655,
+      "epoch": 0.86,
+      "learning_rate": 9.96414314327496e-05,
+      "loss": 1.15002518,
+      "memory(GiB)": 71.19,
+      "step": 1100,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.660391,
+      "epoch": 0.87,
+      "learning_rate": 9.963363775358123e-05,
+      "loss": 1.13650742,
+      "memory(GiB)": 71.19,
+      "step": 1105,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.65939445,
+      "epoch": 0.87,
+      "learning_rate": 9.962576059409854e-05,
+      "loss": 1.14001417,
+      "memory(GiB)": 71.19,
+      "step": 1110,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.65705018,
+      "epoch": 0.87,
+      "learning_rate": 9.961779996755036e-05,
+      "loss": 1.13998594,
+      "memory(GiB)": 71.19,
+      "step": 1115,
+      "train_speed(iter/s)": 0.024237
+    },
+    {
+      "acc": 0.65304127,
+      "epoch": 0.88,
+      "learning_rate": 9.96097558873259e-05,
+      "loss": 1.13927307,
+      "memory(GiB)": 71.19,
+      "step": 1120,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.67214541,
+      "epoch": 0.88,
+      "learning_rate": 9.960162836695478e-05,
+      "loss": 1.10992355,
+      "memory(GiB)": 71.19,
+      "step": 1125,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.64525504,
+      "epoch": 0.89,
+      "learning_rate": 9.959341742010688e-05,
+      "loss": 1.18042841,
+      "memory(GiB)": 71.19,
+      "step": 1130,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.6544466,
+      "epoch": 0.89,
+      "learning_rate": 9.958512306059247e-05,
+      "loss": 1.14733744,
+      "memory(GiB)": 71.19,
+      "step": 1135,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.65201364,
+      "epoch": 0.89,
+      "learning_rate": 9.957674530236205e-05,
+      "loss": 1.13061304,
+      "memory(GiB)": 71.19,
+      "step": 1140,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.66318994,
+      "epoch": 0.9,
+      "learning_rate": 9.956828415950645e-05,
+      "loss": 1.1062932,
+      "memory(GiB)": 71.19,
+      "step": 1145,
+      "train_speed(iter/s)": 0.02424
+    },
+    {
+      "acc": 0.65637937,
+      "epoch": 0.9,
+      "learning_rate": 9.955973964625672e-05,
+      "loss": 1.14308052,
+      "memory(GiB)": 71.19,
+      "step": 1150,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.65574675,
+      "epoch": 0.91,
+      "learning_rate": 9.955111177698412e-05,
+      "loss": 1.14501066,
+      "memory(GiB)": 71.19,
+      "step": 1155,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.6562243,
+      "epoch": 0.91,
+      "learning_rate": 9.954240056620014e-05,
+      "loss": 1.16590223,
+      "memory(GiB)": 71.19,
+      "step": 1160,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.66819267,
+      "epoch": 0.91,
+      "learning_rate": 9.953360602855641e-05,
+      "loss": 1.10895786,
+      "memory(GiB)": 71.19,
+      "step": 1165,
+      "train_speed(iter/s)": 0.024241
+    },
+    {
+      "acc": 0.66646504,
+      "epoch": 0.92,
+      "learning_rate": 9.952472817884476e-05,
+      "loss": 1.13673782,
+      "memory(GiB)": 71.19,
+      "step": 1170,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.67009692,
+      "epoch": 0.92,
+      "learning_rate": 9.951576703199708e-05,
+      "loss": 1.0828022,
+      "memory(GiB)": 71.19,
+      "step": 1175,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.64940991,
+      "epoch": 0.92,
+      "learning_rate": 9.95067226030854e-05,
+      "loss": 1.18824711,
+      "memory(GiB)": 71.19,
+      "step": 1180,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.67086515,
+      "epoch": 0.93,
+      "learning_rate": 9.949759490732185e-05,
+      "loss": 1.11339579,
+      "memory(GiB)": 71.19,
+      "step": 1185,
+      "train_speed(iter/s)": 0.024238
+    },
+    {
+      "acc": 0.65760446,
+      "epoch": 0.93,
+      "learning_rate": 9.948838396005854e-05,
+      "loss": 1.1283742,
+      "memory(GiB)": 71.19,
+      "step": 1190,
+      "train_speed(iter/s)": 0.024239
+    },
+    {
+      "acc": 0.66921711,
+      "epoch": 0.94,
+      "learning_rate": 9.947908977678766e-05,
+      "loss": 1.09572687,
+      "memory(GiB)": 71.19,
+      "step": 1195,
+      "train_speed(iter/s)": 0.024237
+    },
+    {
+      "acc": 0.65912962,
+      "epoch": 0.94,
+      "learning_rate": 9.946971237314136e-05,
+      "loss": 1.14600134,
+      "memory(GiB)": 71.19,
+      "step": 1200,
+      "train_speed(iter/s)": 0.024236
+    },
+    {
+      "acc": 0.66213255,
+      "epoch": 0.94,
+      "learning_rate": 9.94602517648918e-05,
+      "loss": 1.12797499,
+      "memory(GiB)": 71.19,
+      "step": 1205,
+      "train_speed(iter/s)": 0.024234
+    },
+    {
+      "acc": 0.64736218,
+      "epoch": 0.95,
+      "learning_rate": 9.945070796795105e-05,
+      "loss": 1.15010433,
+      "memory(GiB)": 71.19,
+      "step": 1210,
+      "train_speed(iter/s)": 0.024234
+    },
+    {
+      "acc": 0.65648313,
+      "epoch": 0.95,
+      "learning_rate": 9.94410809983711e-05,
+      "loss": 1.13810177,
+      "memory(GiB)": 71.19,
+      "step": 1215,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.65412345,
+      "epoch": 0.96,
+      "learning_rate": 9.943137087234385e-05,
+      "loss": 1.14729261,
+      "memory(GiB)": 71.19,
+      "step": 1220,
+      "train_speed(iter/s)": 0.024231
+    },
+    {
+      "acc": 0.65371003,
+      "epoch": 0.96,
+      "learning_rate": 9.942157760620108e-05,
+      "loss": 1.15231876,
+      "memory(GiB)": 71.19,
+      "step": 1225,
+      "train_speed(iter/s)": 0.024231
+    },
+    {
+      "acc": 0.6497436,
+      "epoch": 0.96,
+      "learning_rate": 9.941170121641435e-05,
+      "loss": 1.15340414,
+      "memory(GiB)": 71.19,
+      "step": 1230,
+      "train_speed(iter/s)": 0.024231
+    },
+    {
+      "acc": 0.66141253,
+      "epoch": 0.97,
+      "learning_rate": 9.940174171959504e-05,
+      "loss": 1.12856808,
+      "memory(GiB)": 71.19,
+      "step": 1235,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.66497359,
+      "epoch": 0.97,
+      "learning_rate": 9.939169913249438e-05,
+      "loss": 1.12229996,
+      "memory(GiB)": 71.19,
+      "step": 1240,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.67611599,
+      "epoch": 0.98,
+      "learning_rate": 9.938157347200327e-05,
+      "loss": 1.0430521,
+      "memory(GiB)": 71.19,
+      "step": 1245,
+      "train_speed(iter/s)": 0.024233
+    },
+    {
+      "acc": 0.65242562,
+      "epoch": 0.98,
+      "learning_rate": 9.937136475515237e-05,
+      "loss": 1.16849995,
+      "memory(GiB)": 71.19,
+      "step": 1250,
+      "train_speed(iter/s)": 0.02423
+    },
+    {
+      "acc": 0.65185585,
+      "epoch": 0.98,
+      "learning_rate": 9.936107299911203e-05,
+      "loss": 1.17074966,
+      "memory(GiB)": 71.19,
+      "step": 1255,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.6592226,
+      "epoch": 0.99,
+      "learning_rate": 9.935069822119226e-05,
+      "loss": 1.11288481,
+      "memory(GiB)": 71.19,
+      "step": 1260,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.66193151,
+      "epoch": 0.99,
+      "learning_rate": 9.934024043884271e-05,
+      "loss": 1.16120825,
+      "memory(GiB)": 71.19,
+      "step": 1265,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.66481395,
+      "epoch": 1.0,
+      "learning_rate": 9.932969966965267e-05,
+      "loss": 1.09738159,
+      "memory(GiB)": 71.19,
+      "step": 1270,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "acc": 0.65508566,
+      "epoch": 1.0,
+      "learning_rate": 9.931907593135093e-05,
+      "loss": 1.14331837,
+      "memory(GiB)": 71.19,
+      "step": 1275,
+      "train_speed(iter/s)": 0.024232
+    },
+    {
+      "epoch": 1.0,
+      "eval_acc": 0.6856999749561733,
+      "eval_loss": 1.018600344657898,
+      "eval_runtime": 107.9063,
+      "eval_samples_per_second": 0.862,
+      "eval_steps_per_second": 0.862,
+      "step": 1275
+    },
+    {
+      "acc": 0.68564477,
+      "epoch": 1.0,
+      "learning_rate": 9.93083692418059e-05,
+      "loss": 1.04102726,
+      "memory(GiB)": 71.19,
+      "step": 1280,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.66535759,
+      "epoch": 1.01,
+      "learning_rate": 9.929757961902549e-05,
+      "loss": 1.10843706,
+      "memory(GiB)": 71.19,
+      "step": 1285,
+      "train_speed(iter/s)": 0.02418
+    },
+    {
+      "acc": 0.67624335,
+      "epoch": 1.01,
+      "learning_rate": 9.928670708115708e-05,
+      "loss": 1.03700457,
+      "memory(GiB)": 71.19,
+      "step": 1290,
+      "train_speed(iter/s)": 0.02418
+    },
+    {
+      "acc": 0.6848104,
+      "epoch": 1.01,
+      "learning_rate": 9.927575164648754e-05,
+      "loss": 1.01924419,
+      "memory(GiB)": 71.19,
+      "step": 1295,
+      "train_speed(iter/s)": 0.024179
+    },
+    {
+      "acc": 0.65978332,
+      "epoch": 1.02,
+      "learning_rate": 9.926471333344311e-05,
+      "loss": 1.10564556,
+      "memory(GiB)": 71.19,
+      "step": 1300,
+      "train_speed(iter/s)": 0.024181
+    },
+    {
+      "acc": 0.66694851,
+      "epoch": 1.02,
+      "learning_rate": 9.925359216058952e-05,
+      "loss": 1.06940336,
+      "memory(GiB)": 71.19,
+      "step": 1305,
+      "train_speed(iter/s)": 0.024181
+    },
+    {
+      "acc": 0.67258253,
+      "epoch": 1.03,
+      "learning_rate": 9.924238814663173e-05,
+      "loss": 1.05483885,
+      "memory(GiB)": 71.19,
+      "step": 1310,
+      "train_speed(iter/s)": 0.024181
+    },
+    {
+      "acc": 0.67171688,
+      "epoch": 1.03,
+      "learning_rate": 9.923110131041419e-05,
+      "loss": 1.0581152,
+      "memory(GiB)": 71.19,
+      "step": 1315,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.66035485,
+      "epoch": 1.03,
+      "learning_rate": 9.92197316709205e-05,
+      "loss": 1.10017042,
+      "memory(GiB)": 71.19,
+      "step": 1320,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.67783785,
+      "epoch": 1.04,
+      "learning_rate": 9.920827924727366e-05,
+      "loss": 1.05424404,
+      "memory(GiB)": 71.19,
+      "step": 1325,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.67227278,
+      "epoch": 1.04,
+      "learning_rate": 9.91967440587358e-05,
+      "loss": 1.07807531,
+      "memory(GiB)": 71.19,
+      "step": 1330,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.65477858,
+      "epoch": 1.05,
+      "learning_rate": 9.918512612470834e-05,
+      "loss": 1.10319042,
+      "memory(GiB)": 71.19,
+      "step": 1335,
+      "train_speed(iter/s)": 0.024182
+    },
+    {
+      "acc": 0.67027154,
+      "epoch": 1.05,
+      "learning_rate": 9.917342546473181e-05,
+      "loss": 1.07305937,
+      "memory(GiB)": 71.19,
+      "step": 1340,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66486812,
+      "epoch": 1.05,
+      "learning_rate": 9.916164209848588e-05,
+      "loss": 1.11207981,
+      "memory(GiB)": 71.19,
+      "step": 1345,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66329069,
+      "epoch": 1.06,
+      "learning_rate": 9.914977604578941e-05,
+      "loss": 1.10326147,
+      "memory(GiB)": 71.19,
+      "step": 1350,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.68474355,
+      "epoch": 1.06,
+      "learning_rate": 9.913782732660024e-05,
+      "loss": 1.02233744,
+      "memory(GiB)": 71.19,
+      "step": 1355,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.67932301,
+      "epoch": 1.07,
+      "learning_rate": 9.912579596101525e-05,
+      "loss": 1.06197271,
+      "memory(GiB)": 71.19,
+      "step": 1360,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66768179,
+      "epoch": 1.07,
+      "learning_rate": 9.911368196927043e-05,
+      "loss": 1.07399158,
+      "memory(GiB)": 71.19,
+      "step": 1365,
+      "train_speed(iter/s)": 0.024184
+    },
+    {
+      "acc": 0.67560496,
+      "epoch": 1.07,
+      "learning_rate": 9.91014853717406e-05,
+      "loss": 1.05643578,
+      "memory(GiB)": 71.19,
+      "step": 1370,
+      "train_speed(iter/s)": 0.024184
+    },
+    {
+      "acc": 0.67212873,
+      "epoch": 1.08,
+      "learning_rate": 9.908920618893962e-05,
+      "loss": 1.0583807,
+      "memory(GiB)": 71.19,
+      "step": 1375,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66627192,
+      "epoch": 1.08,
+      "learning_rate": 9.90768444415202e-05,
+      "loss": 1.07695656,
+      "memory(GiB)": 71.19,
+      "step": 1380,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66721344,
+      "epoch": 1.09,
+      "learning_rate": 9.906440015027399e-05,
+      "loss": 1.04808693,
+      "memory(GiB)": 71.19,
+      "step": 1385,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.67704191,
+      "epoch": 1.09,
+      "learning_rate": 9.905187333613134e-05,
+      "loss": 1.06024771,
+      "memory(GiB)": 71.19,
+      "step": 1390,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66961083,
+      "epoch": 1.09,
+      "learning_rate": 9.903926402016153e-05,
+      "loss": 1.07331381,
+      "memory(GiB)": 71.19,
+      "step": 1395,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.67959518,
+      "epoch": 1.1,
+      "learning_rate": 9.902657222357252e-05,
+      "loss": 1.04364729,
+      "memory(GiB)": 71.19,
+      "step": 1400,
+      "train_speed(iter/s)": 0.024183
+    },
+    {
+      "acc": 0.66278548,
+      "epoch": 1.1,
+      "learning_rate": 9.901379796771107e-05,
+      "loss": 1.08651028,
+      "memory(GiB)": 71.19,
+      "step": 1405,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.65881038,
+      "epoch": 1.11,
+      "learning_rate": 9.900094127406253e-05,
+      "loss": 1.09732819,
+      "memory(GiB)": 71.19,
+      "step": 1410,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.66367035,
+      "epoch": 1.11,
+      "learning_rate": 9.898800216425099e-05,
+      "loss": 1.083144,
+      "memory(GiB)": 71.19,
+      "step": 1415,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.65925193,
+      "epoch": 1.11,
+      "learning_rate": 9.897498066003913e-05,
+      "loss": 1.12018261,
+      "memory(GiB)": 71.19,
+      "step": 1420,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.67458215,
+      "epoch": 1.12,
+      "learning_rate": 9.89618767833282e-05,
+      "loss": 1.06256332,
+      "memory(GiB)": 71.19,
+      "step": 1425,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.66883864,
+      "epoch": 1.12,
+      "learning_rate": 9.894869055615803e-05,
+      "loss": 1.1081831,
+      "memory(GiB)": 71.19,
+      "step": 1430,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.66981688,
+      "epoch": 1.12,
+      "learning_rate": 9.893542200070691e-05,
+      "loss": 1.07718334,
+      "memory(GiB)": 71.19,
+      "step": 1435,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.67857156,
+      "epoch": 1.13,
+      "learning_rate": 9.892207113929164e-05,
+      "loss": 1.04210625,
+      "memory(GiB)": 71.19,
+      "step": 1440,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.66746936,
+      "epoch": 1.13,
+      "learning_rate": 9.890863799436743e-05,
+      "loss": 1.08562384,
+      "memory(GiB)": 71.19,
+      "step": 1445,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.65901647,
+      "epoch": 1.14,
+      "learning_rate": 9.889512258852789e-05,
+      "loss": 1.09942312,
+      "memory(GiB)": 71.19,
+      "step": 1450,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.67867947,
+      "epoch": 1.14,
+      "learning_rate": 9.888152494450498e-05,
+      "loss": 1.06383791,
+      "memory(GiB)": 71.19,
+      "step": 1455,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.66191411,
+      "epoch": 1.14,
+      "learning_rate": 9.886784508516901e-05,
+      "loss": 1.09565401,
+      "memory(GiB)": 71.19,
+      "step": 1460,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.67523494,
+      "epoch": 1.15,
+      "learning_rate": 9.885408303352854e-05,
+      "loss": 1.07071381,
+      "memory(GiB)": 71.19,
+      "step": 1465,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.68043942,
+      "epoch": 1.15,
+      "learning_rate": 9.884023881273037e-05,
+      "loss": 1.05898991,
+      "memory(GiB)": 71.19,
+      "step": 1470,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.66896992,
+      "epoch": 1.16,
+      "learning_rate": 9.882631244605952e-05,
+      "loss": 1.08150454,
+      "memory(GiB)": 71.19,
+      "step": 1475,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.66858001,
+      "epoch": 1.16,
+      "learning_rate": 9.881230395693917e-05,
+      "loss": 1.07318954,
+      "memory(GiB)": 71.19,
+      "step": 1480,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.67361059,
+      "epoch": 1.16,
+      "learning_rate": 9.879821336893062e-05,
+      "loss": 1.0633007,
+      "memory(GiB)": 71.19,
+      "step": 1485,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.66314216,
+      "epoch": 1.17,
+      "learning_rate": 9.878404070573327e-05,
+      "loss": 1.10530052,
+      "memory(GiB)": 71.19,
+      "step": 1490,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.66712093,
+      "epoch": 1.17,
+      "learning_rate": 9.876978599118452e-05,
+      "loss": 1.074893,
+      "memory(GiB)": 71.19,
+      "step": 1495,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.67996116,
+      "epoch": 1.18,
+      "learning_rate": 9.875544924925981e-05,
+      "loss": 1.03155136,
+      "memory(GiB)": 71.19,
+      "step": 1500,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.66647635,
+      "epoch": 1.18,
+      "learning_rate": 9.874103050407256e-05,
+      "loss": 1.07530098,
+      "memory(GiB)": 71.19,
+      "step": 1505,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.66543493,
+      "epoch": 1.18,
+      "learning_rate": 9.872652977987408e-05,
+      "loss": 1.0699563,
+      "memory(GiB)": 71.19,
+      "step": 1510,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.66304975,
+      "epoch": 1.19,
+      "learning_rate": 9.871194710105357e-05,
+      "loss": 1.10289001,
+      "memory(GiB)": 71.19,
+      "step": 1515,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.67143154,
+      "epoch": 1.19,
+      "learning_rate": 9.86972824921381e-05,
+      "loss": 1.07342091,
+      "memory(GiB)": 71.19,
+      "step": 1520,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.66971889,
+      "epoch": 1.2,
+      "learning_rate": 9.868253597779249e-05,
+      "loss": 1.08200026,
+      "memory(GiB)": 71.19,
+      "step": 1525,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.66585174,
+      "epoch": 1.2,
+      "learning_rate": 9.866770758281937e-05,
+      "loss": 1.07775135,
+      "memory(GiB)": 71.19,
+      "step": 1530,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.67415438,
+      "epoch": 1.2,
+      "learning_rate": 9.865279733215905e-05,
+      "loss": 1.03115015,
+      "memory(GiB)": 71.19,
+      "step": 1535,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.6658217,
+      "epoch": 1.21,
+      "learning_rate": 9.863780525088955e-05,
+      "loss": 1.11420527,
+      "memory(GiB)": 71.19,
+      "step": 1540,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.68223243,
+      "epoch": 1.21,
+      "learning_rate": 9.862273136422648e-05,
+      "loss": 1.03579702,
+      "memory(GiB)": 71.19,
+      "step": 1545,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.68120694,
+      "epoch": 1.21,
+      "learning_rate": 9.860757569752309e-05,
+      "loss": 1.06588163,
+      "memory(GiB)": 71.19,
+      "step": 1550,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.65559621,
+      "epoch": 1.22,
+      "learning_rate": 9.859233827627013e-05,
+      "loss": 1.12858868,
+      "memory(GiB)": 71.19,
+      "step": 1555,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.67324719,
+      "epoch": 1.22,
+      "learning_rate": 9.857701912609589e-05,
+      "loss": 1.10106726,
+      "memory(GiB)": 71.19,
+      "step": 1560,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.68056149,
+      "epoch": 1.23,
+      "learning_rate": 9.856161827276613e-05,
+      "loss": 1.03890066,
+      "memory(GiB)": 71.19,
+      "step": 1565,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.66202497,
+      "epoch": 1.23,
+      "learning_rate": 9.854613574218396e-05,
+      "loss": 1.10919619,
+      "memory(GiB)": 71.19,
+      "step": 1570,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.66488538,
+      "epoch": 1.23,
+      "learning_rate": 9.853057156038998e-05,
+      "loss": 1.07370605,
+      "memory(GiB)": 71.19,
+      "step": 1575,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.66696358,
+      "epoch": 1.24,
+      "learning_rate": 9.851492575356201e-05,
+      "loss": 1.0641902,
+      "memory(GiB)": 71.19,
+      "step": 1580,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68181887,
+      "epoch": 1.24,
+      "learning_rate": 9.849919834801522e-05,
+      "loss": 1.04376001,
+      "memory(GiB)": 71.19,
+      "step": 1585,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67546906,
+      "epoch": 1.25,
+      "learning_rate": 9.8483389370202e-05,
+      "loss": 1.03430681,
+      "memory(GiB)": 71.19,
+      "step": 1590,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.67184181,
+      "epoch": 1.25,
+      "learning_rate": 9.846749884671198e-05,
+      "loss": 1.09638357,
+      "memory(GiB)": 71.19,
+      "step": 1595,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.66743326,
+      "epoch": 1.25,
+      "learning_rate": 9.845152680427186e-05,
+      "loss": 1.08324871,
+      "memory(GiB)": 71.19,
+      "step": 1600,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67404108,
+      "epoch": 1.26,
+      "learning_rate": 9.843547326974555e-05,
+      "loss": 1.05259237,
+      "memory(GiB)": 71.19,
+      "step": 1605,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.67859106,
+      "epoch": 1.26,
+      "learning_rate": 9.841933827013394e-05,
+      "loss": 1.02058706,
+      "memory(GiB)": 71.19,
+      "step": 1610,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67385159,
+      "epoch": 1.27,
+      "learning_rate": 9.840312183257498e-05,
+      "loss": 1.0660202,
+      "memory(GiB)": 71.19,
+      "step": 1615,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68775396,
+      "epoch": 1.27,
+      "learning_rate": 9.83868239843436e-05,
+      "loss": 1.01889,
+      "memory(GiB)": 71.19,
+      "step": 1620,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.67444701,
+      "epoch": 1.27,
+      "learning_rate": 9.837044475285165e-05,
+      "loss": 1.0787632,
+      "memory(GiB)": 71.19,
+      "step": 1625,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.67301526,
+      "epoch": 1.28,
+      "learning_rate": 9.83539841656478e-05,
+      "loss": 1.07971487,
+      "memory(GiB)": 71.19,
+      "step": 1630,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.6624773,
+      "epoch": 1.28,
+      "learning_rate": 9.833744225041767e-05,
+      "loss": 1.07760115,
+      "memory(GiB)": 71.19,
+      "step": 1635,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67510743,
+      "epoch": 1.29,
+      "learning_rate": 9.832081903498359e-05,
+      "loss": 1.06419029,
+      "memory(GiB)": 71.19,
+      "step": 1640,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.68025455,
+      "epoch": 1.29,
+      "learning_rate": 9.830411454730464e-05,
+      "loss": 1.03052721,
+      "memory(GiB)": 71.19,
+      "step": 1645,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.66855297,
+      "epoch": 1.29,
+      "learning_rate": 9.82873288154766e-05,
+      "loss": 1.06540508,
+      "memory(GiB)": 71.19,
+      "step": 1650,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.68221936,
+      "epoch": 1.3,
+      "learning_rate": 9.82704618677319e-05,
+      "loss": 1.02048054,
+      "memory(GiB)": 71.19,
+      "step": 1655,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67271628,
+      "epoch": 1.3,
+      "learning_rate": 9.825351373243957e-05,
+      "loss": 1.05244284,
+      "memory(GiB)": 71.19,
+      "step": 1660,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.67585492,
+      "epoch": 1.3,
+      "learning_rate": 9.82364844381052e-05,
+      "loss": 1.03952274,
+      "memory(GiB)": 71.19,
+      "step": 1665,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67327933,
+      "epoch": 1.31,
+      "learning_rate": 9.821937401337087e-05,
+      "loss": 1.06375408,
+      "memory(GiB)": 71.19,
+      "step": 1670,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67473063,
+      "epoch": 1.31,
+      "learning_rate": 9.820218248701512e-05,
+      "loss": 1.03807678,
+      "memory(GiB)": 71.19,
+      "step": 1675,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.66647797,
+      "epoch": 1.32,
+      "learning_rate": 9.818490988795289e-05,
+      "loss": 1.08794069,
+      "memory(GiB)": 71.19,
+      "step": 1680,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.6725491,
+      "epoch": 1.32,
+      "learning_rate": 9.816755624523548e-05,
+      "loss": 1.04948406,
+      "memory(GiB)": 71.19,
+      "step": 1685,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67765465,
+      "epoch": 1.32,
+      "learning_rate": 9.815012158805054e-05,
+      "loss": 1.07902899,
+      "memory(GiB)": 71.19,
+      "step": 1690,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68021631,
+      "epoch": 1.33,
+      "learning_rate": 9.813260594572192e-05,
+      "loss": 1.03020191,
+      "memory(GiB)": 71.19,
+      "step": 1695,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.66895852,
+      "epoch": 1.33,
+      "learning_rate": 9.811500934770969e-05,
+      "loss": 1.10171871,
+      "memory(GiB)": 71.19,
+      "step": 1700,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.65738878,
+      "epoch": 1.34,
+      "learning_rate": 9.809733182361014e-05,
+      "loss": 1.12308722,
+      "memory(GiB)": 71.19,
+      "step": 1705,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.68673587,
+      "epoch": 1.34,
+      "learning_rate": 9.80795734031556e-05,
+      "loss": 1.04205341,
+      "memory(GiB)": 71.19,
+      "step": 1710,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.66318197,
+      "epoch": 1.34,
+      "learning_rate": 9.806173411621451e-05,
+      "loss": 1.07137575,
+      "memory(GiB)": 71.19,
+      "step": 1715,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67586851,
+      "epoch": 1.35,
+      "learning_rate": 9.80438139927913e-05,
+      "loss": 1.04353619,
+      "memory(GiB)": 71.19,
+      "step": 1720,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67039175,
+      "epoch": 1.35,
+      "learning_rate": 9.802581306302638e-05,
+      "loss": 1.0834878,
+      "memory(GiB)": 71.19,
+      "step": 1725,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67330465,
+      "epoch": 1.36,
+      "learning_rate": 9.800773135719604e-05,
+      "loss": 1.04026947,
+      "memory(GiB)": 71.19,
+      "step": 1730,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.65766926,
+      "epoch": 1.36,
+      "learning_rate": 9.798956890571244e-05,
+      "loss": 1.11574898,
+      "memory(GiB)": 71.19,
+      "step": 1735,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.66705904,
+      "epoch": 1.36,
+      "learning_rate": 9.79713257391236e-05,
+      "loss": 1.09899406,
+      "memory(GiB)": 71.19,
+      "step": 1740,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.66609573,
+      "epoch": 1.37,
+      "learning_rate": 9.795300188811319e-05,
+      "loss": 1.10992451,
+      "memory(GiB)": 71.19,
+      "step": 1745,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.6764576,
+      "epoch": 1.37,
+      "learning_rate": 9.793459738350069e-05,
+      "loss": 1.05841427,
+      "memory(GiB)": 71.19,
+      "step": 1750,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.65810442,
+      "epoch": 1.38,
+      "learning_rate": 9.791611225624118e-05,
+      "loss": 1.12776289,
+      "memory(GiB)": 71.19,
+      "step": 1755,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.6705864,
+      "epoch": 1.38,
+      "learning_rate": 9.789754653742537e-05,
+      "loss": 1.07466335,
+      "memory(GiB)": 71.19,
+      "step": 1760,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.67177191,
+      "epoch": 1.38,
+      "learning_rate": 9.787890025827948e-05,
+      "loss": 1.08449554,
+      "memory(GiB)": 71.19,
+      "step": 1765,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67626877,
+      "epoch": 1.39,
+      "learning_rate": 9.786017345016524e-05,
+      "loss": 1.06005878,
+      "memory(GiB)": 71.19,
+      "step": 1770,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.68222189,
+      "epoch": 1.39,
+      "learning_rate": 9.784136614457988e-05,
+      "loss": 1.06014185,
+      "memory(GiB)": 71.19,
+      "step": 1775,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.66160769,
+      "epoch": 1.4,
+      "learning_rate": 9.782247837315595e-05,
+      "loss": 1.11386976,
+      "memory(GiB)": 71.19,
+      "step": 1780,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.68455191,
+      "epoch": 1.4,
+      "learning_rate": 9.780351016766136e-05,
+      "loss": 1.04914351,
+      "memory(GiB)": 71.19,
+      "step": 1785,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.66824789,
+      "epoch": 1.4,
+      "learning_rate": 9.77844615599993e-05,
+      "loss": 1.05908461,
+      "memory(GiB)": 71.19,
+      "step": 1790,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.66619282,
+      "epoch": 1.41,
+      "learning_rate": 9.776533258220819e-05,
+      "loss": 1.09789734,
+      "memory(GiB)": 71.19,
+      "step": 1795,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67257457,
+      "epoch": 1.41,
+      "learning_rate": 9.774612326646169e-05,
+      "loss": 1.05887022,
+      "memory(GiB)": 71.19,
+      "step": 1800,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67153988,
+      "epoch": 1.41,
+      "learning_rate": 9.772683364506847e-05,
+      "loss": 1.07383604,
+      "memory(GiB)": 71.19,
+      "step": 1805,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.67205477,
+      "epoch": 1.42,
+      "learning_rate": 9.770746375047235e-05,
+      "loss": 1.06401491,
+      "memory(GiB)": 71.19,
+      "step": 1810,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.66954546,
+      "epoch": 1.42,
+      "learning_rate": 9.768801361525215e-05,
+      "loss": 1.08306198,
+      "memory(GiB)": 71.19,
+      "step": 1815,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68025694,
+      "epoch": 1.43,
+      "learning_rate": 9.766848327212161e-05,
+      "loss": 1.02885761,
+      "memory(GiB)": 71.19,
+      "step": 1820,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68164105,
+      "epoch": 1.43,
+      "learning_rate": 9.764887275392943e-05,
+      "loss": 1.05446472,
+      "memory(GiB)": 71.19,
+      "step": 1825,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.66762023,
+      "epoch": 1.43,
+      "learning_rate": 9.762918209365913e-05,
+      "loss": 1.10211878,
+      "memory(GiB)": 71.19,
+      "step": 1830,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67623882,
+      "epoch": 1.44,
+      "learning_rate": 9.760941132442902e-05,
+      "loss": 1.06018152,
+      "memory(GiB)": 71.19,
+      "step": 1835,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.6666759,
+      "epoch": 1.44,
+      "learning_rate": 9.758956047949215e-05,
+      "loss": 1.11666231,
+      "memory(GiB)": 71.19,
+      "step": 1840,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67013221,
+      "epoch": 1.45,
+      "learning_rate": 9.756962959223628e-05,
+      "loss": 1.06344328,
+      "memory(GiB)": 71.19,
+      "step": 1845,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66876597,
+      "epoch": 1.45,
+      "learning_rate": 9.754961869618373e-05,
+      "loss": 1.08094902,
+      "memory(GiB)": 71.19,
+      "step": 1850,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.66471753,
+      "epoch": 1.45,
+      "learning_rate": 9.752952782499147e-05,
+      "loss": 1.0886961,
+      "memory(GiB)": 71.19,
+      "step": 1855,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.67831697,
+      "epoch": 1.46,
+      "learning_rate": 9.750935701245092e-05,
+      "loss": 1.03748837,
+      "memory(GiB)": 71.19,
+      "step": 1860,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.6669693,
+      "epoch": 1.46,
+      "learning_rate": 9.748910629248798e-05,
+      "loss": 1.08023806,
+      "memory(GiB)": 71.19,
+      "step": 1865,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.67126279,
+      "epoch": 1.47,
+      "learning_rate": 9.746877569916297e-05,
+      "loss": 1.07919216,
+      "memory(GiB)": 71.19,
+      "step": 1870,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67684112,
+      "epoch": 1.47,
+      "learning_rate": 9.74483652666705e-05,
+      "loss": 1.04165659,
+      "memory(GiB)": 71.19,
+      "step": 1875,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.68273034,
+      "epoch": 1.47,
+      "learning_rate": 9.74278750293395e-05,
+      "loss": 1.04425545,
+      "memory(GiB)": 71.19,
+      "step": 1880,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66332841,
+      "epoch": 1.48,
+      "learning_rate": 9.740730502163311e-05,
+      "loss": 1.07659769,
+      "memory(GiB)": 71.19,
+      "step": 1885,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68010182,
+      "epoch": 1.48,
+      "learning_rate": 9.738665527814867e-05,
+      "loss": 1.0395319,
+      "memory(GiB)": 71.19,
+      "step": 1890,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68821034,
+      "epoch": 1.49,
+      "learning_rate": 9.736592583361762e-05,
+      "loss": 1.01133175,
+      "memory(GiB)": 71.19,
+      "step": 1895,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66931405,
+      "epoch": 1.49,
+      "learning_rate": 9.73451167229054e-05,
+      "loss": 1.06255102,
+      "memory(GiB)": 71.19,
+      "step": 1900,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67967229,
+      "epoch": 1.49,
+      "learning_rate": 9.732422798101149e-05,
+      "loss": 1.02841129,
+      "memory(GiB)": 71.19,
+      "step": 1905,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67584767,
+      "epoch": 1.5,
+      "learning_rate": 9.73032596430693e-05,
+      "loss": 1.05436392,
+      "memory(GiB)": 71.19,
+      "step": 1910,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66209011,
+      "epoch": 1.5,
+      "learning_rate": 9.728221174434613e-05,
+      "loss": 1.10967607,
+      "memory(GiB)": 71.19,
+      "step": 1915,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68050909,
+      "epoch": 1.5,
+      "learning_rate": 9.726108432024306e-05,
+      "loss": 1.02168255,
+      "memory(GiB)": 71.19,
+      "step": 1920,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68102293,
+      "epoch": 1.51,
+      "learning_rate": 9.723987740629494e-05,
+      "loss": 1.05781345,
+      "memory(GiB)": 71.19,
+      "step": 1925,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.6717145,
+      "epoch": 1.51,
+      "learning_rate": 9.721859103817033e-05,
+      "loss": 1.06613684,
+      "memory(GiB)": 71.19,
+      "step": 1930,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.66321344,
+      "epoch": 1.52,
+      "learning_rate": 9.719722525167141e-05,
+      "loss": 1.09566288,
+      "memory(GiB)": 71.19,
+      "step": 1935,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.6747292,
+      "epoch": 1.52,
+      "learning_rate": 9.717578008273393e-05,
+      "loss": 1.04526234,
+      "memory(GiB)": 71.19,
+      "step": 1940,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68077927,
+      "epoch": 1.52,
+      "learning_rate": 9.715425556742716e-05,
+      "loss": 1.02439919,
+      "memory(GiB)": 71.19,
+      "step": 1945,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.67475858,
+      "epoch": 1.53,
+      "learning_rate": 9.713265174195388e-05,
+      "loss": 1.05720959,
+      "memory(GiB)": 71.19,
+      "step": 1950,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.66691504,
+      "epoch": 1.53,
+      "learning_rate": 9.711096864265016e-05,
+      "loss": 1.08331633,
+      "memory(GiB)": 71.19,
+      "step": 1955,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.66824608,
+      "epoch": 1.54,
+      "learning_rate": 9.708920630598551e-05,
+      "loss": 1.08959455,
+      "memory(GiB)": 71.19,
+      "step": 1960,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.66691604,
+      "epoch": 1.54,
+      "learning_rate": 9.706736476856264e-05,
+      "loss": 1.05755281,
+      "memory(GiB)": 71.19,
+      "step": 1965,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66844997,
+      "epoch": 1.54,
+      "learning_rate": 9.704544406711746e-05,
+      "loss": 1.07190256,
+      "memory(GiB)": 71.19,
+      "step": 1970,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.6775053,
+      "epoch": 1.55,
+      "learning_rate": 9.702344423851911e-05,
+      "loss": 1.04533443,
+      "memory(GiB)": 71.19,
+      "step": 1975,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.66825333,
+      "epoch": 1.55,
+      "learning_rate": 9.700136531976974e-05,
+      "loss": 1.1034709,
+      "memory(GiB)": 71.19,
+      "step": 1980,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.67021503,
+      "epoch": 1.56,
+      "learning_rate": 9.697920734800456e-05,
+      "loss": 1.08550053,
+      "memory(GiB)": 71.19,
+      "step": 1985,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.66171827,
+      "epoch": 1.56,
+      "learning_rate": 9.695697036049172e-05,
+      "loss": 1.09154787,
+      "memory(GiB)": 71.19,
+      "step": 1990,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68564267,
+      "epoch": 1.56,
+      "learning_rate": 9.693465439463228e-05,
+      "loss": 1.02461205,
+      "memory(GiB)": 71.19,
+      "step": 1995,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.67194061,
+      "epoch": 1.57,
+      "learning_rate": 9.691225948796014e-05,
+      "loss": 1.05746717,
+      "memory(GiB)": 71.19,
+      "step": 2000,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.67832704,
+      "epoch": 1.57,
+      "learning_rate": 9.6889785678142e-05,
+      "loss": 1.05458746,
+      "memory(GiB)": 71.19,
+      "step": 2005,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.67701178,
+      "epoch": 1.58,
+      "learning_rate": 9.68672330029772e-05,
+      "loss": 1.05478268,
+      "memory(GiB)": 71.19,
+      "step": 2010,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.66319699,
+      "epoch": 1.58,
+      "learning_rate": 9.684460150039779e-05,
+      "loss": 1.1452549,
+      "memory(GiB)": 71.19,
+      "step": 2015,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.6763217,
+      "epoch": 1.58,
+      "learning_rate": 9.682189120846834e-05,
+      "loss": 1.04298344,
+      "memory(GiB)": 71.19,
+      "step": 2020,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.68313031,
+      "epoch": 1.59,
+      "learning_rate": 9.679910216538601e-05,
+      "loss": 1.02893553,
+      "memory(GiB)": 71.19,
+      "step": 2025,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67663183,
+      "epoch": 1.59,
+      "learning_rate": 9.677623440948038e-05,
+      "loss": 1.04474058,
+      "memory(GiB)": 71.19,
+      "step": 2030,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.66433477,
+      "epoch": 1.59,
+      "learning_rate": 9.675328797921342e-05,
+      "loss": 1.07190542,
+      "memory(GiB)": 71.19,
+      "step": 2035,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.68374462,
+      "epoch": 1.6,
+      "learning_rate": 9.67302629131794e-05,
+      "loss": 1.07545433,
+      "memory(GiB)": 71.19,
+      "step": 2040,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67807226,
+      "epoch": 1.6,
+      "learning_rate": 9.670715925010489e-05,
+      "loss": 1.05842772,
+      "memory(GiB)": 71.19,
+      "step": 2045,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67401872,
+      "epoch": 1.61,
+      "learning_rate": 9.668397702884866e-05,
+      "loss": 1.07067871,
+      "memory(GiB)": 71.19,
+      "step": 2050,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.67526712,
+      "epoch": 1.61,
+      "learning_rate": 9.666071628840154e-05,
+      "loss": 1.04234457,
+      "memory(GiB)": 71.19,
+      "step": 2055,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67841988,
+      "epoch": 1.61,
+      "learning_rate": 9.663737706788652e-05,
+      "loss": 1.05335855,
+      "memory(GiB)": 71.19,
+      "step": 2060,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67320924,
+      "epoch": 1.62,
+      "learning_rate": 9.66139594065585e-05,
+      "loss": 1.07291307,
+      "memory(GiB)": 71.19,
+      "step": 2065,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67182488,
+      "epoch": 1.62,
+      "learning_rate": 9.659046334380439e-05,
+      "loss": 1.06681614,
+      "memory(GiB)": 71.19,
+      "step": 2070,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.68178086,
+      "epoch": 1.63,
+      "learning_rate": 9.65668889191429e-05,
+      "loss": 1.02115288,
+      "memory(GiB)": 71.19,
+      "step": 2075,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.67527838,
+      "epoch": 1.63,
+      "learning_rate": 9.654323617222456e-05,
+      "loss": 1.02676697,
+      "memory(GiB)": 71.19,
+      "step": 2080,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.68160815,
+      "epoch": 1.63,
+      "learning_rate": 9.651950514283166e-05,
+      "loss": 1.0352355,
+      "memory(GiB)": 71.19,
+      "step": 2085,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.67300172,
+      "epoch": 1.64,
+      "learning_rate": 9.649569587087814e-05,
+      "loss": 1.06639357,
+      "memory(GiB)": 71.19,
+      "step": 2090,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67614121,
+      "epoch": 1.64,
+      "learning_rate": 9.647180839640951e-05,
+      "loss": 1.05420818,
+      "memory(GiB)": 71.19,
+      "step": 2095,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.67984071,
+      "epoch": 1.65,
+      "learning_rate": 9.644784275960286e-05,
+      "loss": 1.04375839,
+      "memory(GiB)": 71.19,
+      "step": 2100,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.67311459,
+      "epoch": 1.65,
+      "learning_rate": 9.64237990007667e-05,
+      "loss": 1.05994186,
+      "memory(GiB)": 71.19,
+      "step": 2105,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.66600237,
+      "epoch": 1.65,
+      "learning_rate": 9.639967716034095e-05,
+      "loss": 1.08946924,
+      "memory(GiB)": 71.19,
+      "step": 2110,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.68244076,
+      "epoch": 1.66,
+      "learning_rate": 9.637547727889688e-05,
+      "loss": 0.99340763,
+      "memory(GiB)": 71.19,
+      "step": 2115,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.6899344,
+      "epoch": 1.66,
+      "learning_rate": 9.6351199397137e-05,
+      "loss": 1.00486383,
+      "memory(GiB)": 71.19,
+      "step": 2120,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.68560667,
+      "epoch": 1.67,
+      "learning_rate": 9.632684355589499e-05,
+      "loss": 1.02728119,
+      "memory(GiB)": 71.19,
+      "step": 2125,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.67438531,
+      "epoch": 1.67,
+      "learning_rate": 9.630240979613569e-05,
+      "loss": 1.08428488,
+      "memory(GiB)": 71.19,
+      "step": 2130,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.67004066,
+      "epoch": 1.67,
+      "learning_rate": 9.627789815895498e-05,
+      "loss": 1.06729488,
+      "memory(GiB)": 71.19,
+      "step": 2135,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.66677608,
+      "epoch": 1.68,
+      "learning_rate": 9.625330868557973e-05,
+      "loss": 1.08347321,
+      "memory(GiB)": 71.19,
+      "step": 2140,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.68743095,
+      "epoch": 1.68,
+      "learning_rate": 9.622864141736772e-05,
+      "loss": 1.02547398,
+      "memory(GiB)": 71.19,
+      "step": 2145,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.66881833,
+      "epoch": 1.69,
+      "learning_rate": 9.620389639580753e-05,
+      "loss": 1.10209799,
+      "memory(GiB)": 71.19,
+      "step": 2150,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.68243523,
+      "epoch": 1.69,
+      "learning_rate": 9.617907366251862e-05,
+      "loss": 1.02223969,
+      "memory(GiB)": 71.19,
+      "step": 2155,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.67003131,
+      "epoch": 1.69,
+      "learning_rate": 9.615417325925106e-05,
+      "loss": 1.05287113,
+      "memory(GiB)": 71.19,
+      "step": 2160,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.67350674,
+      "epoch": 1.7,
+      "learning_rate": 9.612919522788559e-05,
+      "loss": 1.05360003,
+      "memory(GiB)": 71.19,
+      "step": 2165,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.67382431,
+      "epoch": 1.7,
+      "learning_rate": 9.610413961043354e-05,
+      "loss": 1.06978779,
+      "memory(GiB)": 71.19,
+      "step": 2170,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.69041929,
+      "epoch": 1.7,
+      "learning_rate": 9.607900644903667e-05,
+      "loss": 0.99977674,
+      "memory(GiB)": 71.19,
+      "step": 2175,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.66950564,
+      "epoch": 1.71,
+      "learning_rate": 9.605379578596724e-05,
+      "loss": 1.07388229,
+      "memory(GiB)": 71.19,
+      "step": 2180,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.67316561,
+      "epoch": 1.71,
+      "learning_rate": 9.60285076636278e-05,
+      "loss": 1.06690502,
+      "memory(GiB)": 71.19,
+      "step": 2185,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.68053436,
+      "epoch": 1.72,
+      "learning_rate": 9.600314212455123e-05,
+      "loss": 1.02924366,
+      "memory(GiB)": 71.19,
+      "step": 2190,
+      "train_speed(iter/s)": 0.024205
+    },
+    {
+      "acc": 0.68438997,
+      "epoch": 1.72,
+      "learning_rate": 9.597769921140059e-05,
+      "loss": 1.04086161,
+      "memory(GiB)": 71.19,
+      "step": 2195,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.68536568,
+      "epoch": 1.72,
+      "learning_rate": 9.595217896696906e-05,
+      "loss": 1.02934484,
+      "memory(GiB)": 71.19,
+      "step": 2200,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.66752796,
+      "epoch": 1.73,
+      "learning_rate": 9.592658143417993e-05,
+      "loss": 1.07700911,
+      "memory(GiB)": 71.19,
+      "step": 2205,
+      "train_speed(iter/s)": 0.024206
+    },
+    {
+      "acc": 0.6667316,
+      "epoch": 1.73,
+      "learning_rate": 9.590090665608647e-05,
+      "loss": 1.10646029,
+      "memory(GiB)": 71.19,
+      "step": 2210,
+      "train_speed(iter/s)": 0.024207
+    },
+    {
+      "acc": 0.67603645,
+      "epoch": 1.74,
+      "learning_rate": 9.587515467587184e-05,
+      "loss": 1.01645374,
+      "memory(GiB)": 71.19,
+      "step": 2215,
+      "train_speed(iter/s)": 0.024207
+    },
+    {
+      "acc": 0.67964444,
+      "epoch": 1.74,
+      "learning_rate": 9.584932553684912e-05,
+      "loss": 1.02113829,
+      "memory(GiB)": 71.19,
+      "step": 2220,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67175584,
+      "epoch": 1.74,
+      "learning_rate": 9.582341928246105e-05,
+      "loss": 1.07317562,
+      "memory(GiB)": 71.19,
+      "step": 2225,
+      "train_speed(iter/s)": 0.024207
+    },
+    {
+      "acc": 0.67587581,
+      "epoch": 1.75,
+      "learning_rate": 9.57974359562802e-05,
+      "loss": 1.05869913,
+      "memory(GiB)": 71.19,
+      "step": 2230,
+      "train_speed(iter/s)": 0.024207
+    },
+    {
+      "acc": 0.67292519,
+      "epoch": 1.75,
+      "learning_rate": 9.577137560200868e-05,
+      "loss": 1.05732136,
+      "memory(GiB)": 71.19,
+      "step": 2235,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67593732,
+      "epoch": 1.76,
+      "learning_rate": 9.574523826347821e-05,
+      "loss": 1.05148783,
+      "memory(GiB)": 71.19,
+      "step": 2240,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66758423,
+      "epoch": 1.76,
+      "learning_rate": 9.571902398464996e-05,
+      "loss": 1.08099422,
+      "memory(GiB)": 71.19,
+      "step": 2245,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67744985,
+      "epoch": 1.76,
+      "learning_rate": 9.569273280961451e-05,
+      "loss": 1.0564353,
+      "memory(GiB)": 71.19,
+      "step": 2250,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.68401761,
+      "epoch": 1.77,
+      "learning_rate": 9.566636478259178e-05,
+      "loss": 1.03796558,
+      "memory(GiB)": 71.19,
+      "step": 2255,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66767297,
+      "epoch": 1.77,
+      "learning_rate": 9.563991994793095e-05,
+      "loss": 1.07227659,
+      "memory(GiB)": 71.19,
+      "step": 2260,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.66152043,
+      "epoch": 1.78,
+      "learning_rate": 9.56133983501104e-05,
+      "loss": 1.12642975,
+      "memory(GiB)": 71.19,
+      "step": 2265,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.68994551,
+      "epoch": 1.78,
+      "learning_rate": 9.558680003373761e-05,
+      "loss": 1.02917881,
+      "memory(GiB)": 71.19,
+      "step": 2270,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66738191,
+      "epoch": 1.78,
+      "learning_rate": 9.556012504354907e-05,
+      "loss": 1.0579689,
+      "memory(GiB)": 71.19,
+      "step": 2275,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.68090143,
+      "epoch": 1.79,
+      "learning_rate": 9.553337342441026e-05,
+      "loss": 1.02114449,
+      "memory(GiB)": 71.19,
+      "step": 2280,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.68244548,
+      "epoch": 1.79,
+      "learning_rate": 9.550654522131554e-05,
+      "loss": 1.05710659,
+      "memory(GiB)": 71.19,
+      "step": 2285,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.67936034,
+      "epoch": 1.79,
+      "learning_rate": 9.547964047938807e-05,
+      "loss": 1.06786947,
+      "memory(GiB)": 71.19,
+      "step": 2290,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.66966805,
+      "epoch": 1.8,
+      "learning_rate": 9.545265924387976e-05,
+      "loss": 1.07189226,
+      "memory(GiB)": 71.19,
+      "step": 2295,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66073766,
+      "epoch": 1.8,
+      "learning_rate": 9.542560156017114e-05,
+      "loss": 1.10265789,
+      "memory(GiB)": 71.19,
+      "step": 2300,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.6612524,
+      "epoch": 1.81,
+      "learning_rate": 9.539846747377137e-05,
+      "loss": 1.10375538,
+      "memory(GiB)": 71.19,
+      "step": 2305,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.6832036,
+      "epoch": 1.81,
+      "learning_rate": 9.537125703031808e-05,
+      "loss": 1.01659298,
+      "memory(GiB)": 71.19,
+      "step": 2310,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67649841,
+      "epoch": 1.81,
+      "learning_rate": 9.534397027557734e-05,
+      "loss": 1.05449762,
+      "memory(GiB)": 71.19,
+      "step": 2315,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.6723896,
+      "epoch": 1.82,
+      "learning_rate": 9.531660725544357e-05,
+      "loss": 1.08471155,
+      "memory(GiB)": 71.19,
+      "step": 2320,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67170587,
+      "epoch": 1.82,
+      "learning_rate": 9.528916801593945e-05,
+      "loss": 1.06638412,
+      "memory(GiB)": 71.19,
+      "step": 2325,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66721401,
+      "epoch": 1.83,
+      "learning_rate": 9.526165260321587e-05,
+      "loss": 1.07490387,
+      "memory(GiB)": 71.19,
+      "step": 2330,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.66922231,
+      "epoch": 1.83,
+      "learning_rate": 9.523406106355181e-05,
+      "loss": 1.07525682,
+      "memory(GiB)": 71.19,
+      "step": 2335,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.68418026,
+      "epoch": 1.83,
+      "learning_rate": 9.520639344335436e-05,
+      "loss": 1.02238379,
+      "memory(GiB)": 71.19,
+      "step": 2340,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66266646,
+      "epoch": 1.84,
+      "learning_rate": 9.517864978915852e-05,
+      "loss": 1.107903,
+      "memory(GiB)": 71.19,
+      "step": 2345,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66747322,
+      "epoch": 1.84,
+      "learning_rate": 9.515083014762714e-05,
+      "loss": 1.05063229,
+      "memory(GiB)": 71.19,
+      "step": 2350,
+      "train_speed(iter/s)": 0.024207
+    },
+    {
+      "acc": 0.68329477,
+      "epoch": 1.85,
+      "learning_rate": 9.512293456555094e-05,
+      "loss": 1.04089613,
+      "memory(GiB)": 71.19,
+      "step": 2355,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67805209,
+      "epoch": 1.85,
+      "learning_rate": 9.509496308984834e-05,
+      "loss": 1.05391035,
+      "memory(GiB)": 71.19,
+      "step": 2360,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.68161922,
+      "epoch": 1.85,
+      "learning_rate": 9.506691576756542e-05,
+      "loss": 1.0376071,
+      "memory(GiB)": 71.19,
+      "step": 2365,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.68472705,
+      "epoch": 1.86,
+      "learning_rate": 9.50387926458758e-05,
+      "loss": 1.03176441,
+      "memory(GiB)": 71.19,
+      "step": 2370,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67763271,
+      "epoch": 1.86,
+      "learning_rate": 9.501059377208062e-05,
+      "loss": 1.02066383,
+      "memory(GiB)": 71.19,
+      "step": 2375,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67828851,
+      "epoch": 1.87,
+      "learning_rate": 9.49823191936084e-05,
+      "loss": 1.03181,
+      "memory(GiB)": 71.19,
+      "step": 2380,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.68185763,
+      "epoch": 1.87,
+      "learning_rate": 9.495396895801504e-05,
+      "loss": 1.01452837,
+      "memory(GiB)": 71.19,
+      "step": 2385,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.69047713,
+      "epoch": 1.87,
+      "learning_rate": 9.492554311298363e-05,
+      "loss": 1.03052311,
+      "memory(GiB)": 71.19,
+      "step": 2390,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67350526,
+      "epoch": 1.88,
+      "learning_rate": 9.489704170632448e-05,
+      "loss": 1.05733795,
+      "memory(GiB)": 71.19,
+      "step": 2395,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66548829,
+      "epoch": 1.88,
+      "learning_rate": 9.486846478597493e-05,
+      "loss": 1.10433922,
+      "memory(GiB)": 71.19,
+      "step": 2400,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.67878256,
+      "epoch": 1.88,
+      "learning_rate": 9.48398123999994e-05,
+      "loss": 1.04339581,
+      "memory(GiB)": 71.19,
+      "step": 2405,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.66891594,
+      "epoch": 1.89,
+      "learning_rate": 9.481108459658918e-05,
+      "loss": 1.06247549,
+      "memory(GiB)": 71.19,
+      "step": 2410,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.67175364,
+      "epoch": 1.89,
+      "learning_rate": 9.478228142406245e-05,
+      "loss": 1.08076143,
+      "memory(GiB)": 71.19,
+      "step": 2415,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66775222,
+      "epoch": 1.9,
+      "learning_rate": 9.475340293086414e-05,
+      "loss": 1.07804575,
+      "memory(GiB)": 71.19,
+      "step": 2420,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.67420983,
+      "epoch": 1.9,
+      "learning_rate": 9.472444916556582e-05,
+      "loss": 1.06133413,
+      "memory(GiB)": 71.19,
+      "step": 2425,
+      "train_speed(iter/s)": 0.024208
+    },
+    {
+      "acc": 0.66697445,
+      "epoch": 1.9,
+      "learning_rate": 9.469542017686574e-05,
+      "loss": 1.08241444,
+      "memory(GiB)": 71.19,
+      "step": 2430,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.65952163,
+      "epoch": 1.91,
+      "learning_rate": 9.466631601358865e-05,
+      "loss": 1.12456112,
+      "memory(GiB)": 71.19,
+      "step": 2435,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.68122878,
+      "epoch": 1.91,
+      "learning_rate": 9.463713672468566e-05,
+      "loss": 1.02802782,
+      "memory(GiB)": 71.19,
+      "step": 2440,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67782488,
+      "epoch": 1.92,
+      "learning_rate": 9.460788235923434e-05,
+      "loss": 1.04834585,
+      "memory(GiB)": 71.19,
+      "step": 2445,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.6813798,
+      "epoch": 1.92,
+      "learning_rate": 9.457855296643847e-05,
+      "loss": 1.03965998,
+      "memory(GiB)": 71.19,
+      "step": 2450,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.6787158,
+      "epoch": 1.92,
+      "learning_rate": 9.454914859562806e-05,
+      "loss": 1.03802996,
+      "memory(GiB)": 71.19,
+      "step": 2455,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.67819457,
+      "epoch": 1.93,
+      "learning_rate": 9.451966929625921e-05,
+      "loss": 1.04884434,
+      "memory(GiB)": 71.19,
+      "step": 2460,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68044524,
+      "epoch": 1.93,
+      "learning_rate": 9.449011511791403e-05,
+      "loss": 1.00336819,
+      "memory(GiB)": 71.19,
+      "step": 2465,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.68315306,
+      "epoch": 1.94,
+      "learning_rate": 9.44604861103006e-05,
+      "loss": 1.0400569,
+      "memory(GiB)": 71.19,
+      "step": 2470,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67919292,
+      "epoch": 1.94,
+      "learning_rate": 9.443078232325283e-05,
+      "loss": 1.0457303,
+      "memory(GiB)": 71.19,
+      "step": 2475,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67261739,
+      "epoch": 1.94,
+      "learning_rate": 9.440100380673041e-05,
+      "loss": 1.08123322,
+      "memory(GiB)": 71.19,
+      "step": 2480,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67910471,
+      "epoch": 1.95,
+      "learning_rate": 9.437115061081873e-05,
+      "loss": 1.06696892,
+      "memory(GiB)": 71.19,
+      "step": 2485,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68002834,
+      "epoch": 1.95,
+      "learning_rate": 9.434122278572881e-05,
+      "loss": 1.01719761,
+      "memory(GiB)": 71.19,
+      "step": 2490,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68830619,
+      "epoch": 1.96,
+      "learning_rate": 9.431122038179713e-05,
+      "loss": 1.02319183,
+      "memory(GiB)": 71.19,
+      "step": 2495,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68742871,
+      "epoch": 1.96,
+      "learning_rate": 9.428114344948566e-05,
+      "loss": 1.01065245,
+      "memory(GiB)": 71.19,
+      "step": 2500,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.67193766,
+      "epoch": 1.96,
+      "learning_rate": 9.425099203938169e-05,
+      "loss": 1.062183,
+      "memory(GiB)": 71.19,
+      "step": 2505,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.66727209,
+      "epoch": 1.97,
+      "learning_rate": 9.422076620219777e-05,
+      "loss": 1.06330957,
+      "memory(GiB)": 71.19,
+      "step": 2510,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68619056,
+      "epoch": 1.97,
+      "learning_rate": 9.419046598877169e-05,
+      "loss": 1.02488794,
+      "memory(GiB)": 71.19,
+      "step": 2515,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.67550688,
+      "epoch": 1.98,
+      "learning_rate": 9.416009145006626e-05,
+      "loss": 1.0441782,
+      "memory(GiB)": 71.19,
+      "step": 2520,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.69182601,
+      "epoch": 1.98,
+      "learning_rate": 9.412964263716934e-05,
+      "loss": 0.99197636,
+      "memory(GiB)": 71.19,
+      "step": 2525,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67208972,
+      "epoch": 1.98,
+      "learning_rate": 9.409911960129373e-05,
+      "loss": 1.04292965,
+      "memory(GiB)": 71.19,
+      "step": 2530,
+      "train_speed(iter/s)": 0.024209
+    },
+    {
+      "acc": 0.69247909,
+      "epoch": 1.99,
+      "learning_rate": 9.406852239377702e-05,
+      "loss": 1.03010435,
+      "memory(GiB)": 71.19,
+      "step": 2535,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.67900863,
+      "epoch": 1.99,
+      "learning_rate": 9.403785106608159e-05,
+      "loss": 1.02731123,
+      "memory(GiB)": 71.19,
+      "step": 2540,
+      "train_speed(iter/s)": 0.024211
+    },
+    {
+      "acc": 0.68802767,
+      "epoch": 1.99,
+      "learning_rate": 9.400710566979446e-05,
+      "loss": 1.02137775,
+      "memory(GiB)": 71.19,
+      "step": 2545,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "acc": 0.68384156,
+      "epoch": 2.0,
+      "learning_rate": 9.397628625662724e-05,
+      "loss": 1.00198641,
+      "memory(GiB)": 71.19,
+      "step": 2550,
+      "train_speed(iter/s)": 0.02421
+    },
+    {
+      "epoch": 2.0,
+      "eval_acc": 0.7063611319809667,
+      "eval_loss": 0.9125774502754211,
+      "eval_runtime": 107.4016,
+      "eval_samples_per_second": 0.866,
+      "eval_steps_per_second": 0.866,
+      "step": 2551
+    },
+    {
+      "acc": 0.69561715,
+      "epoch": 2.0,
+      "learning_rate": 9.394539287841606e-05,
+      "loss": 0.96205349,
+      "memory(GiB)": 71.19,
+      "step": 2555,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.70156784,
+      "epoch": 2.01,
+      "learning_rate": 9.391442558712141e-05,
+      "loss": 0.95558033,
+      "memory(GiB)": 71.19,
+      "step": 2560,
+      "train_speed(iter/s)": 0.024184
+    },
+    {
+      "acc": 0.70194588,
+      "epoch": 2.01,
+      "learning_rate": 9.38833844348281e-05,
+      "loss": 0.94522266,
+      "memory(GiB)": 71.19,
+      "step": 2565,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.69583673,
+      "epoch": 2.01,
+      "learning_rate": 9.385226947374519e-05,
+      "loss": 0.95971365,
+      "memory(GiB)": 71.19,
+      "step": 2570,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.69883165,
+      "epoch": 2.02,
+      "learning_rate": 9.382108075620588e-05,
+      "loss": 0.95136395,
+      "memory(GiB)": 71.19,
+      "step": 2575,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.70211344,
+      "epoch": 2.02,
+      "learning_rate": 9.37898183346674e-05,
+      "loss": 0.95915556,
+      "memory(GiB)": 71.19,
+      "step": 2580,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.6914197,
+      "epoch": 2.03,
+      "learning_rate": 9.375848226171097e-05,
+      "loss": 0.97404881,
+      "memory(GiB)": 71.19,
+      "step": 2585,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.69058862,
+      "epoch": 2.03,
+      "learning_rate": 9.372707259004168e-05,
+      "loss": 0.97601357,
+      "memory(GiB)": 71.19,
+      "step": 2590,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.69468822,
+      "epoch": 2.03,
+      "learning_rate": 9.369558937248841e-05,
+      "loss": 0.94959822,
+      "memory(GiB)": 71.19,
+      "step": 2595,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.69152589,
+      "epoch": 2.04,
+      "learning_rate": 9.366403266200372e-05,
+      "loss": 0.98147879,
+      "memory(GiB)": 71.19,
+      "step": 2600,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.69578485,
+      "epoch": 2.04,
+      "learning_rate": 9.363240251166381e-05,
+      "loss": 0.96815405,
+      "memory(GiB)": 71.19,
+      "step": 2605,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.70924282,
+      "epoch": 2.05,
+      "learning_rate": 9.360069897466837e-05,
+      "loss": 0.92589321,
+      "memory(GiB)": 71.19,
+      "step": 2610,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.69980206,
+      "epoch": 2.05,
+      "learning_rate": 9.356892210434054e-05,
+      "loss": 0.93600454,
+      "memory(GiB)": 71.19,
+      "step": 2615,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.68358474,
+      "epoch": 2.05,
+      "learning_rate": 9.353707195412682e-05,
+      "loss": 0.98476477,
+      "memory(GiB)": 71.19,
+      "step": 2620,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.70984707,
+      "epoch": 2.06,
+      "learning_rate": 9.35051485775969e-05,
+      "loss": 0.92595644,
+      "memory(GiB)": 71.19,
+      "step": 2625,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.70100389,
+      "epoch": 2.06,
+      "learning_rate": 9.34731520284437e-05,
+      "loss": 0.97320604,
+      "memory(GiB)": 71.19,
+      "step": 2630,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.69202828,
+      "epoch": 2.07,
+      "learning_rate": 9.344108236048317e-05,
+      "loss": 0.9583993,
+      "memory(GiB)": 71.19,
+      "step": 2635,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.69916315,
+      "epoch": 2.07,
+      "learning_rate": 9.340893962765427e-05,
+      "loss": 0.97032738,
+      "memory(GiB)": 71.19,
+      "step": 2640,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.69505591,
+      "epoch": 2.07,
+      "learning_rate": 9.337672388401882e-05,
+      "loss": 0.9470933,
+      "memory(GiB)": 71.19,
+      "step": 2645,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.69918957,
+      "epoch": 2.08,
+      "learning_rate": 9.334443518376144e-05,
+      "loss": 0.94485273,
+      "memory(GiB)": 71.19,
+      "step": 2650,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.69890566,
+      "epoch": 2.08,
+      "learning_rate": 9.33120735811895e-05,
+      "loss": 0.95022421,
+      "memory(GiB)": 71.19,
+      "step": 2655,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.70103612,
+      "epoch": 2.08,
+      "learning_rate": 9.327963913073292e-05,
+      "loss": 0.93258381,
+      "memory(GiB)": 71.19,
+      "step": 2660,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.68382578,
+      "epoch": 2.09,
+      "learning_rate": 9.32471318869442e-05,
+      "loss": 0.98318644,
+      "memory(GiB)": 71.19,
+      "step": 2665,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.69849315,
+      "epoch": 2.09,
+      "learning_rate": 9.321455190449826e-05,
+      "loss": 0.95109901,
+      "memory(GiB)": 71.19,
+      "step": 2670,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.69872427,
+      "epoch": 2.1,
+      "learning_rate": 9.318189923819237e-05,
+      "loss": 0.95068798,
+      "memory(GiB)": 71.19,
+      "step": 2675,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.6956666,
+      "epoch": 2.1,
+      "learning_rate": 9.314917394294601e-05,
+      "loss": 0.94351721,
+      "memory(GiB)": 71.19,
+      "step": 2680,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.68996038,
+      "epoch": 2.1,
+      "learning_rate": 9.311637607380087e-05,
+      "loss": 0.98688478,
+      "memory(GiB)": 71.19,
+      "step": 2685,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.69613433,
+      "epoch": 2.11,
+      "learning_rate": 9.308350568592062e-05,
+      "loss": 0.96690464,
+      "memory(GiB)": 71.19,
+      "step": 2690,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.70577445,
+      "epoch": 2.11,
+      "learning_rate": 9.305056283459101e-05,
+      "loss": 0.94153576,
+      "memory(GiB)": 71.19,
+      "step": 2695,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.69807467,
+      "epoch": 2.12,
+      "learning_rate": 9.30175475752196e-05,
+      "loss": 0.95756607,
+      "memory(GiB)": 71.19,
+      "step": 2700,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.69821429,
+      "epoch": 2.12,
+      "learning_rate": 9.298445996333576e-05,
+      "loss": 0.94874229,
+      "memory(GiB)": 71.19,
+      "step": 2705,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.70757422,
+      "epoch": 2.12,
+      "learning_rate": 9.295130005459053e-05,
+      "loss": 0.92731676,
+      "memory(GiB)": 71.19,
+      "step": 2710,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.70590787,
+      "epoch": 2.13,
+      "learning_rate": 9.291806790475659e-05,
+      "loss": 0.91320696,
+      "memory(GiB)": 71.19,
+      "step": 2715,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.69287601,
+      "epoch": 2.13,
+      "learning_rate": 9.288476356972806e-05,
+      "loss": 0.96277952,
+      "memory(GiB)": 71.19,
+      "step": 2720,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.69249158,
+      "epoch": 2.14,
+      "learning_rate": 9.285138710552053e-05,
+      "loss": 0.9557168,
+      "memory(GiB)": 71.19,
+      "step": 2725,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.70517898,
+      "epoch": 2.14,
+      "learning_rate": 9.281793856827086e-05,
+      "loss": 0.95910587,
+      "memory(GiB)": 71.19,
+      "step": 2730,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.68601875,
+      "epoch": 2.14,
+      "learning_rate": 9.278441801423718e-05,
+      "loss": 0.98039818,
+      "memory(GiB)": 71.19,
+      "step": 2735,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.69862165,
+      "epoch": 2.15,
+      "learning_rate": 9.275082549979872e-05,
+      "loss": 0.94497709,
+      "memory(GiB)": 71.19,
+      "step": 2740,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.6958168,
+      "epoch": 2.15,
+      "learning_rate": 9.271716108145574e-05,
+      "loss": 0.95876045,
+      "memory(GiB)": 71.19,
+      "step": 2745,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.69509659,
+      "epoch": 2.16,
+      "learning_rate": 9.268342481582944e-05,
+      "loss": 0.98702393,
+      "memory(GiB)": 71.19,
+      "step": 2750,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.69033999,
+      "epoch": 2.16,
+      "learning_rate": 9.264961675966186e-05,
+      "loss": 0.991891,
+      "memory(GiB)": 71.19,
+      "step": 2755,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.70156898,
+      "epoch": 2.16,
+      "learning_rate": 9.261573696981579e-05,
+      "loss": 0.93319263,
+      "memory(GiB)": 71.19,
+      "step": 2760,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70409079,
+      "epoch": 2.17,
+      "learning_rate": 9.258178550327468e-05,
+      "loss": 0.94536772,
+      "memory(GiB)": 71.19,
+      "step": 2765,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68395271,
+      "epoch": 2.17,
+      "learning_rate": 9.254776241714251e-05,
+      "loss": 0.99986162,
+      "memory(GiB)": 71.19,
+      "step": 2770,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68076348,
+      "epoch": 2.17,
+      "learning_rate": 9.251366776864377e-05,
+      "loss": 1.00616322,
+      "memory(GiB)": 71.19,
+      "step": 2775,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68825445,
+      "epoch": 2.18,
+      "learning_rate": 9.247950161512324e-05,
+      "loss": 0.97691736,
+      "memory(GiB)": 71.19,
+      "step": 2780,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.70971127,
+      "epoch": 2.18,
+      "learning_rate": 9.244526401404604e-05,
+      "loss": 0.92741508,
+      "memory(GiB)": 71.19,
+      "step": 2785,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69368434,
+      "epoch": 2.19,
+      "learning_rate": 9.24109550229974e-05,
+      "loss": 0.97553349,
+      "memory(GiB)": 71.19,
+      "step": 2790,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.71132045,
+      "epoch": 2.19,
+      "learning_rate": 9.237657469968266e-05,
+      "loss": 0.92014847,
+      "memory(GiB)": 71.19,
+      "step": 2795,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69388194,
+      "epoch": 2.19,
+      "learning_rate": 9.234212310192711e-05,
+      "loss": 0.94383049,
+      "memory(GiB)": 71.19,
+      "step": 2800,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.69455805,
+      "epoch": 2.2,
+      "learning_rate": 9.230760028767597e-05,
+      "loss": 0.9620759,
+      "memory(GiB)": 71.19,
+      "step": 2805,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69068451,
+      "epoch": 2.2,
+      "learning_rate": 9.227300631499416e-05,
+      "loss": 0.94675779,
+      "memory(GiB)": 71.19,
+      "step": 2810,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69235501,
+      "epoch": 2.21,
+      "learning_rate": 9.223834124206635e-05,
+      "loss": 0.97269449,
+      "memory(GiB)": 71.19,
+      "step": 2815,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69006801,
+      "epoch": 2.21,
+      "learning_rate": 9.220360512719676e-05,
+      "loss": 0.95500069,
+      "memory(GiB)": 71.19,
+      "step": 2820,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69353352,
+      "epoch": 2.21,
+      "learning_rate": 9.216879802880913e-05,
+      "loss": 0.9617177,
+      "memory(GiB)": 71.19,
+      "step": 2825,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69113631,
+      "epoch": 2.22,
+      "learning_rate": 9.213392000544656e-05,
+      "loss": 0.97730999,
+      "memory(GiB)": 71.19,
+      "step": 2830,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69324679,
+      "epoch": 2.22,
+      "learning_rate": 9.209897111577144e-05,
+      "loss": 0.96714506,
+      "memory(GiB)": 71.19,
+      "step": 2835,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.696696,
+      "epoch": 2.23,
+      "learning_rate": 9.206395141856538e-05,
+      "loss": 0.98762465,
+      "memory(GiB)": 71.19,
+      "step": 2840,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70372343,
+      "epoch": 2.23,
+      "learning_rate": 9.202886097272907e-05,
+      "loss": 0.92046738,
+      "memory(GiB)": 71.19,
+      "step": 2845,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70415306,
+      "epoch": 2.23,
+      "learning_rate": 9.199369983728217e-05,
+      "loss": 0.92664623,
+      "memory(GiB)": 71.19,
+      "step": 2850,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69124141,
+      "epoch": 2.24,
+      "learning_rate": 9.195846807136326e-05,
+      "loss": 0.9966094,
+      "memory(GiB)": 71.19,
+      "step": 2855,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68278403,
+      "epoch": 2.24,
+      "learning_rate": 9.192316573422972e-05,
+      "loss": 1.00215902,
+      "memory(GiB)": 71.19,
+      "step": 2860,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70400138,
+      "epoch": 2.25,
+      "learning_rate": 9.188779288525761e-05,
+      "loss": 0.95005407,
+      "memory(GiB)": 71.19,
+      "step": 2865,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70103235,
+      "epoch": 2.25,
+      "learning_rate": 9.18523495839416e-05,
+      "loss": 0.93353863,
+      "memory(GiB)": 71.19,
+      "step": 2870,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69932871,
+      "epoch": 2.25,
+      "learning_rate": 9.181683588989485e-05,
+      "loss": 0.94103365,
+      "memory(GiB)": 71.19,
+      "step": 2875,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69865599,
+      "epoch": 2.26,
+      "learning_rate": 9.17812518628489e-05,
+      "loss": 0.96315451,
+      "memory(GiB)": 71.19,
+      "step": 2880,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69689775,
+      "epoch": 2.26,
+      "learning_rate": 9.174559756265361e-05,
+      "loss": 0.95352697,
+      "memory(GiB)": 71.19,
+      "step": 2885,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69051638,
+      "epoch": 2.27,
+      "learning_rate": 9.170987304927704e-05,
+      "loss": 0.97824144,
+      "memory(GiB)": 71.19,
+      "step": 2890,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70052414,
+      "epoch": 2.27,
+      "learning_rate": 9.167407838280531e-05,
+      "loss": 0.93514223,
+      "memory(GiB)": 71.19,
+      "step": 2895,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69622307,
+      "epoch": 2.27,
+      "learning_rate": 9.163821362344254e-05,
+      "loss": 0.96024828,
+      "memory(GiB)": 71.19,
+      "step": 2900,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68836212,
+      "epoch": 2.28,
+      "learning_rate": 9.160227883151077e-05,
+      "loss": 0.97077913,
+      "memory(GiB)": 71.19,
+      "step": 2905,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.6831625,
+      "epoch": 2.28,
+      "learning_rate": 9.15662740674498e-05,
+      "loss": 1.00156384,
+      "memory(GiB)": 71.19,
+      "step": 2910,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69906459,
+      "epoch": 2.28,
+      "learning_rate": 9.153019939181716e-05,
+      "loss": 0.96558504,
+      "memory(GiB)": 71.19,
+      "step": 2915,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69677444,
+      "epoch": 2.29,
+      "learning_rate": 9.149405486528788e-05,
+      "loss": 0.97523499,
+      "memory(GiB)": 71.19,
+      "step": 2920,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68966751,
+      "epoch": 2.29,
+      "learning_rate": 9.145784054865458e-05,
+      "loss": 0.97728767,
+      "memory(GiB)": 71.19,
+      "step": 2925,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.7036571,
+      "epoch": 2.3,
+      "learning_rate": 9.14215565028272e-05,
+      "loss": 0.94556446,
+      "memory(GiB)": 71.19,
+      "step": 2930,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69022222,
+      "epoch": 2.3,
+      "learning_rate": 9.138520278883297e-05,
+      "loss": 0.96314068,
+      "memory(GiB)": 71.19,
+      "step": 2935,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68896809,
+      "epoch": 2.3,
+      "learning_rate": 9.13487794678163e-05,
+      "loss": 0.99049673,
+      "memory(GiB)": 71.19,
+      "step": 2940,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69701419,
+      "epoch": 2.31,
+      "learning_rate": 9.131228660103866e-05,
+      "loss": 0.97968254,
+      "memory(GiB)": 71.19,
+      "step": 2945,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69701881,
+      "epoch": 2.31,
+      "learning_rate": 9.127572424987853e-05,
+      "loss": 0.97570438,
+      "memory(GiB)": 71.19,
+      "step": 2950,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70094166,
+      "epoch": 2.32,
+      "learning_rate": 9.12390924758312e-05,
+      "loss": 0.93014622,
+      "memory(GiB)": 71.19,
+      "step": 2955,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70081906,
+      "epoch": 2.32,
+      "learning_rate": 9.12023913405088e-05,
+      "loss": 0.97871437,
+      "memory(GiB)": 71.19,
+      "step": 2960,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70025148,
+      "epoch": 2.32,
+      "learning_rate": 9.116562090564005e-05,
+      "loss": 0.93432608,
+      "memory(GiB)": 71.19,
+      "step": 2965,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68864193,
+      "epoch": 2.33,
+      "learning_rate": 9.112878123307025e-05,
+      "loss": 0.98558102,
+      "memory(GiB)": 71.19,
+      "step": 2970,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69512825,
+      "epoch": 2.33,
+      "learning_rate": 9.109187238476116e-05,
+      "loss": 0.94346056,
+      "memory(GiB)": 71.19,
+      "step": 2975,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70853047,
+      "epoch": 2.34,
+      "learning_rate": 9.105489442279092e-05,
+      "loss": 0.91953535,
+      "memory(GiB)": 71.19,
+      "step": 2980,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70059996,
+      "epoch": 2.34,
+      "learning_rate": 9.101784740935383e-05,
+      "loss": 0.93242407,
+      "memory(GiB)": 71.19,
+      "step": 2985,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.6832633,
+      "epoch": 2.34,
+      "learning_rate": 9.098073140676042e-05,
+      "loss": 1.00155811,
+      "memory(GiB)": 71.19,
+      "step": 2990,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69065361,
+      "epoch": 2.35,
+      "learning_rate": 9.09435464774372e-05,
+      "loss": 0.98291121,
+      "memory(GiB)": 71.19,
+      "step": 2995,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69865141,
+      "epoch": 2.35,
+      "learning_rate": 9.090629268392661e-05,
+      "loss": 0.96651316,
+      "memory(GiB)": 71.19,
+      "step": 3000,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69779391,
+      "epoch": 2.36,
+      "learning_rate": 9.086897008888697e-05,
+      "loss": 0.96862831,
+      "memory(GiB)": 71.19,
+      "step": 3005,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.7000648,
+      "epoch": 2.36,
+      "learning_rate": 9.083157875509226e-05,
+      "loss": 0.95280199,
+      "memory(GiB)": 71.19,
+      "step": 3010,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69280767,
+      "epoch": 2.36,
+      "learning_rate": 9.079411874543206e-05,
+      "loss": 0.96452456,
+      "memory(GiB)": 71.19,
+      "step": 3015,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70206046,
+      "epoch": 2.37,
+      "learning_rate": 9.075659012291155e-05,
+      "loss": 0.94142323,
+      "memory(GiB)": 71.19,
+      "step": 3020,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69001818,
+      "epoch": 2.37,
+      "learning_rate": 9.071899295065122e-05,
+      "loss": 0.97127142,
+      "memory(GiB)": 71.19,
+      "step": 3025,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69132757,
+      "epoch": 2.37,
+      "learning_rate": 9.068132729188689e-05,
+      "loss": 0.95719604,
+      "memory(GiB)": 71.19,
+      "step": 3030,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69102788,
+      "epoch": 2.38,
+      "learning_rate": 9.064359320996958e-05,
+      "loss": 0.97239933,
+      "memory(GiB)": 71.19,
+      "step": 3035,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68073072,
+      "epoch": 2.38,
+      "learning_rate": 9.060579076836537e-05,
+      "loss": 1.00775642,
+      "memory(GiB)": 71.19,
+      "step": 3040,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69667606,
+      "epoch": 2.39,
+      "learning_rate": 9.056792003065535e-05,
+      "loss": 0.9632431,
+      "memory(GiB)": 71.19,
+      "step": 3045,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70922837,
+      "epoch": 2.39,
+      "learning_rate": 9.052998106053544e-05,
+      "loss": 0.91346598,
+      "memory(GiB)": 71.19,
+      "step": 3050,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68625507,
+      "epoch": 2.39,
+      "learning_rate": 9.049197392181632e-05,
+      "loss": 0.97347794,
+      "memory(GiB)": 71.19,
+      "step": 3055,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69933372,
+      "epoch": 2.4,
+      "learning_rate": 9.045389867842338e-05,
+      "loss": 0.97092943,
+      "memory(GiB)": 71.19,
+      "step": 3060,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70825005,
+      "epoch": 2.4,
+      "learning_rate": 9.041575539439651e-05,
+      "loss": 0.92969103,
+      "memory(GiB)": 71.19,
+      "step": 3065,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69653683,
+      "epoch": 2.41,
+      "learning_rate": 9.037754413389005e-05,
+      "loss": 0.96352663,
+      "memory(GiB)": 71.19,
+      "step": 3070,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69451375,
+      "epoch": 2.41,
+      "learning_rate": 9.033926496117268e-05,
+      "loss": 0.96977873,
+      "memory(GiB)": 71.19,
+      "step": 3075,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.687293,
+      "epoch": 2.41,
+      "learning_rate": 9.030091794062728e-05,
+      "loss": 0.98685484,
+      "memory(GiB)": 71.19,
+      "step": 3080,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69902458,
+      "epoch": 2.42,
+      "learning_rate": 9.026250313675086e-05,
+      "loss": 0.95378513,
+      "memory(GiB)": 71.19,
+      "step": 3085,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69570527,
+      "epoch": 2.42,
+      "learning_rate": 9.022402061415448e-05,
+      "loss": 0.94871998,
+      "memory(GiB)": 71.19,
+      "step": 3090,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69046903,
+      "epoch": 2.43,
+      "learning_rate": 9.018547043756299e-05,
+      "loss": 0.97162466,
+      "memory(GiB)": 71.19,
+      "step": 3095,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69036608,
+      "epoch": 2.43,
+      "learning_rate": 9.014685267181515e-05,
+      "loss": 0.97763748,
+      "memory(GiB)": 71.19,
+      "step": 3100,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69294906,
+      "epoch": 2.43,
+      "learning_rate": 9.010816738186335e-05,
+      "loss": 0.98672915,
+      "memory(GiB)": 71.19,
+      "step": 3105,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69603667,
+      "epoch": 2.44,
+      "learning_rate": 9.006941463277349e-05,
+      "loss": 0.96144562,
+      "memory(GiB)": 71.19,
+      "step": 3110,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69827685,
+      "epoch": 2.44,
+      "learning_rate": 9.003059448972504e-05,
+      "loss": 0.94170532,
+      "memory(GiB)": 71.19,
+      "step": 3115,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69537177,
+      "epoch": 2.45,
+      "learning_rate": 8.999170701801076e-05,
+      "loss": 0.94646502,
+      "memory(GiB)": 71.19,
+      "step": 3120,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69200006,
+      "epoch": 2.45,
+      "learning_rate": 8.995275228303667e-05,
+      "loss": 1.00812445,
+      "memory(GiB)": 71.19,
+      "step": 3125,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69033399,
+      "epoch": 2.45,
+      "learning_rate": 8.99137303503219e-05,
+      "loss": 0.99082174,
+      "memory(GiB)": 71.19,
+      "step": 3130,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69509125,
+      "epoch": 2.46,
+      "learning_rate": 8.987464128549862e-05,
+      "loss": 0.96717873,
+      "memory(GiB)": 71.19,
+      "step": 3135,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69468803,
+      "epoch": 2.46,
+      "learning_rate": 8.98354851543119e-05,
+      "loss": 0.93988543,
+      "memory(GiB)": 71.19,
+      "step": 3140,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70322733,
+      "epoch": 2.46,
+      "learning_rate": 8.979626202261966e-05,
+      "loss": 0.92626677,
+      "memory(GiB)": 71.19,
+      "step": 3145,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69417615,
+      "epoch": 2.47,
+      "learning_rate": 8.97569719563924e-05,
+      "loss": 0.95528698,
+      "memory(GiB)": 71.19,
+      "step": 3150,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69935951,
+      "epoch": 2.47,
+      "learning_rate": 8.971761502171335e-05,
+      "loss": 0.94430065,
+      "memory(GiB)": 71.19,
+      "step": 3155,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69067097,
+      "epoch": 2.48,
+      "learning_rate": 8.967819128477807e-05,
+      "loss": 0.97750225,
+      "memory(GiB)": 71.19,
+      "step": 3160,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69903126,
+      "epoch": 2.48,
+      "learning_rate": 8.963870081189454e-05,
+      "loss": 0.94778118,
+      "memory(GiB)": 71.19,
+      "step": 3165,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70578423,
+      "epoch": 2.48,
+      "learning_rate": 8.959914366948302e-05,
+      "loss": 0.92893019,
+      "memory(GiB)": 71.19,
+      "step": 3170,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69360132,
+      "epoch": 2.49,
+      "learning_rate": 8.95595199240758e-05,
+      "loss": 0.94542074,
+      "memory(GiB)": 71.19,
+      "step": 3175,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70106397,
+      "epoch": 2.49,
+      "learning_rate": 8.951982964231728e-05,
+      "loss": 0.94324751,
+      "memory(GiB)": 71.19,
+      "step": 3180,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70175576,
+      "epoch": 2.5,
+      "learning_rate": 8.948007289096379e-05,
+      "loss": 0.95341234,
+      "memory(GiB)": 71.19,
+      "step": 3185,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70615592,
+      "epoch": 2.5,
+      "learning_rate": 8.944024973688334e-05,
+      "loss": 0.92322388,
+      "memory(GiB)": 71.19,
+      "step": 3190,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70504413,
+      "epoch": 2.5,
+      "learning_rate": 8.940036024705574e-05,
+      "loss": 0.95310001,
+      "memory(GiB)": 71.19,
+      "step": 3195,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69084344,
+      "epoch": 2.51,
+      "learning_rate": 8.93604044885723e-05,
+      "loss": 0.97776327,
+      "memory(GiB)": 71.19,
+      "step": 3200,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.70420504,
+      "epoch": 2.51,
+      "learning_rate": 8.932038252863583e-05,
+      "loss": 0.9348093,
+      "memory(GiB)": 71.19,
+      "step": 3205,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69500103,
+      "epoch": 2.52,
+      "learning_rate": 8.92802944345605e-05,
+      "loss": 0.9558012,
+      "memory(GiB)": 71.19,
+      "step": 3210,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.71531296,
+      "epoch": 2.52,
+      "learning_rate": 8.924014027377164e-05,
+      "loss": 0.9156971,
+      "memory(GiB)": 71.19,
+      "step": 3215,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70045333,
+      "epoch": 2.52,
+      "learning_rate": 8.919992011380576e-05,
+      "loss": 0.93506889,
+      "memory(GiB)": 71.19,
+      "step": 3220,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68789282,
+      "epoch": 2.53,
+      "learning_rate": 8.915963402231038e-05,
+      "loss": 0.97745686,
+      "memory(GiB)": 71.19,
+      "step": 3225,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.7019794,
+      "epoch": 2.53,
+      "learning_rate": 8.911928206704387e-05,
+      "loss": 0.950805,
+      "memory(GiB)": 71.19,
+      "step": 3230,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69622092,
+      "epoch": 2.54,
+      "learning_rate": 8.907886431587543e-05,
+      "loss": 0.96481142,
+      "memory(GiB)": 71.19,
+      "step": 3235,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69668698,
+      "epoch": 2.54,
+      "learning_rate": 8.903838083678486e-05,
+      "loss": 0.95254927,
+      "memory(GiB)": 71.19,
+      "step": 3240,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70273995,
+      "epoch": 2.54,
+      "learning_rate": 8.899783169786257e-05,
+      "loss": 0.93340073,
+      "memory(GiB)": 71.19,
+      "step": 3245,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69511542,
+      "epoch": 2.55,
+      "learning_rate": 8.895721696730939e-05,
+      "loss": 0.96184444,
+      "memory(GiB)": 71.19,
+      "step": 3250,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70252714,
+      "epoch": 2.55,
+      "learning_rate": 8.891653671343643e-05,
+      "loss": 0.94699078,
+      "memory(GiB)": 71.19,
+      "step": 3255,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68809547,
+      "epoch": 2.56,
+      "learning_rate": 8.887579100466508e-05,
+      "loss": 0.96380692,
+      "memory(GiB)": 71.19,
+      "step": 3260,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68375864,
+      "epoch": 2.56,
+      "learning_rate": 8.883497990952674e-05,
+      "loss": 0.98428965,
+      "memory(GiB)": 71.19,
+      "step": 3265,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.68760471,
+      "epoch": 2.56,
+      "learning_rate": 8.879410349666284e-05,
+      "loss": 0.99547338,
+      "memory(GiB)": 71.19,
+      "step": 3270,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.6960813,
+      "epoch": 2.57,
+      "learning_rate": 8.875316183482464e-05,
+      "loss": 0.96283712,
+      "memory(GiB)": 71.19,
+      "step": 3275,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69431438,
+      "epoch": 2.57,
+      "learning_rate": 8.871215499287319e-05,
+      "loss": 0.97081881,
+      "memory(GiB)": 71.19,
+      "step": 3280,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69778886,
+      "epoch": 2.57,
+      "learning_rate": 8.867108303977912e-05,
+      "loss": 0.95300598,
+      "memory(GiB)": 71.19,
+      "step": 3285,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.70309801,
+      "epoch": 2.58,
+      "learning_rate": 8.862994604462256e-05,
+      "loss": 0.93476143,
+      "memory(GiB)": 71.19,
+      "step": 3290,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69512172,
+      "epoch": 2.58,
+      "learning_rate": 8.85887440765931e-05,
+      "loss": 0.96246347,
+      "memory(GiB)": 71.19,
+      "step": 3295,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.71056523,
+      "epoch": 2.59,
+      "learning_rate": 8.854747720498954e-05,
+      "loss": 0.90803175,
+      "memory(GiB)": 71.19,
+      "step": 3300,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69368334,
+      "epoch": 2.59,
+      "learning_rate": 8.850614549921994e-05,
+      "loss": 0.99619999,
+      "memory(GiB)": 71.19,
+      "step": 3305,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68741622,
+      "epoch": 2.59,
+      "learning_rate": 8.846474902880128e-05,
+      "loss": 1.01268291,
+      "memory(GiB)": 71.19,
+      "step": 3310,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68540864,
+      "epoch": 2.6,
+      "learning_rate": 8.842328786335956e-05,
+      "loss": 1.00386658,
+      "memory(GiB)": 71.19,
+      "step": 3315,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69587579,
+      "epoch": 2.6,
+      "learning_rate": 8.838176207262958e-05,
+      "loss": 0.95101843,
+      "memory(GiB)": 71.19,
+      "step": 3320,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69762154,
+      "epoch": 2.61,
+      "learning_rate": 8.834017172645478e-05,
+      "loss": 0.94804831,
+      "memory(GiB)": 71.19,
+      "step": 3325,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69709311,
+      "epoch": 2.61,
+      "learning_rate": 8.829851689478725e-05,
+      "loss": 0.99054155,
+      "memory(GiB)": 71.19,
+      "step": 3330,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.69299603,
+      "epoch": 2.61,
+      "learning_rate": 8.825679764768751e-05,
+      "loss": 0.95626192,
+      "memory(GiB)": 71.19,
+      "step": 3335,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69584079,
+      "epoch": 2.62,
+      "learning_rate": 8.821501405532442e-05,
+      "loss": 0.95285006,
+      "memory(GiB)": 71.19,
+      "step": 3340,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69174972,
+      "epoch": 2.62,
+      "learning_rate": 8.817316618797507e-05,
+      "loss": 0.95945415,
+      "memory(GiB)": 71.19,
+      "step": 3345,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69653916,
+      "epoch": 2.63,
+      "learning_rate": 8.813125411602464e-05,
+      "loss": 0.9533721,
+      "memory(GiB)": 71.19,
+      "step": 3350,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69422874,
+      "epoch": 2.63,
+      "learning_rate": 8.80892779099663e-05,
+      "loss": 0.96425648,
+      "memory(GiB)": 71.19,
+      "step": 3355,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.68755012,
+      "epoch": 2.63,
+      "learning_rate": 8.804723764040112e-05,
+      "loss": 0.97660484,
+      "memory(GiB)": 71.19,
+      "step": 3360,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.6888967,
+      "epoch": 2.64,
+      "learning_rate": 8.800513337803788e-05,
+      "loss": 0.95651054,
+      "memory(GiB)": 71.19,
+      "step": 3365,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69544282,
+      "epoch": 2.64,
+      "learning_rate": 8.796296519369303e-05,
+      "loss": 0.9503314,
+      "memory(GiB)": 71.19,
+      "step": 3370,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.69458733,
+      "epoch": 2.65,
+      "learning_rate": 8.792073315829052e-05,
+      "loss": 0.940135,
+      "memory(GiB)": 71.19,
+      "step": 3375,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69739385,
+      "epoch": 2.65,
+      "learning_rate": 8.787843734286165e-05,
+      "loss": 0.9286274,
+      "memory(GiB)": 71.19,
+      "step": 3380,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.71626267,
+      "epoch": 2.65,
+      "learning_rate": 8.783607781854506e-05,
+      "loss": 0.91786966,
+      "memory(GiB)": 71.19,
+      "step": 3385,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70292225,
+      "epoch": 2.66,
+      "learning_rate": 8.77936546565865e-05,
+      "loss": 0.94578028,
+      "memory(GiB)": 71.19,
+      "step": 3390,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70064354,
+      "epoch": 2.66,
+      "learning_rate": 8.775116792833878e-05,
+      "loss": 0.95401373,
+      "memory(GiB)": 71.19,
+      "step": 3395,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68651481,
+      "epoch": 2.66,
+      "learning_rate": 8.770861770526158e-05,
+      "loss": 0.96929026,
+      "memory(GiB)": 71.19,
+      "step": 3400,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.68540168,
+      "epoch": 2.67,
+      "learning_rate": 8.766600405892145e-05,
+      "loss": 1.01124077,
+      "memory(GiB)": 71.19,
+      "step": 3405,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70259171,
+      "epoch": 2.67,
+      "learning_rate": 8.762332706099153e-05,
+      "loss": 0.94904728,
+      "memory(GiB)": 71.19,
+      "step": 3410,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69669948,
+      "epoch": 2.68,
+      "learning_rate": 8.758058678325156e-05,
+      "loss": 0.94466686,
+      "memory(GiB)": 71.19,
+      "step": 3415,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.699084,
+      "epoch": 2.68,
+      "learning_rate": 8.753778329758773e-05,
+      "loss": 0.96061077,
+      "memory(GiB)": 71.19,
+      "step": 3420,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70037179,
+      "epoch": 2.68,
+      "learning_rate": 8.74949166759925e-05,
+      "loss": 0.95514784,
+      "memory(GiB)": 71.19,
+      "step": 3425,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69292159,
+      "epoch": 2.69,
+      "learning_rate": 8.745198699056451e-05,
+      "loss": 0.98498917,
+      "memory(GiB)": 71.19,
+      "step": 3430,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70176616,
+      "epoch": 2.69,
+      "learning_rate": 8.740899431350852e-05,
+      "loss": 0.95132113,
+      "memory(GiB)": 71.19,
+      "step": 3435,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70519481,
+      "epoch": 2.7,
+      "learning_rate": 8.736593871713523e-05,
+      "loss": 0.95861454,
+      "memory(GiB)": 71.19,
+      "step": 3440,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.69500475,
+      "epoch": 2.7,
+      "learning_rate": 8.732282027386111e-05,
+      "loss": 0.96890717,
+      "memory(GiB)": 71.19,
+      "step": 3445,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69789162,
+      "epoch": 2.7,
+      "learning_rate": 8.72796390562084e-05,
+      "loss": 0.93789511,
+      "memory(GiB)": 71.19,
+      "step": 3450,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.6822947,
+      "epoch": 2.71,
+      "learning_rate": 8.723639513680486e-05,
+      "loss": 0.99033756,
+      "memory(GiB)": 71.19,
+      "step": 3455,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.70193062,
+      "epoch": 2.71,
+      "learning_rate": 8.719308858838377e-05,
+      "loss": 0.95832653,
+      "memory(GiB)": 71.19,
+      "step": 3460,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70000944,
+      "epoch": 2.72,
+      "learning_rate": 8.714971948378374e-05,
+      "loss": 0.94106884,
+      "memory(GiB)": 71.19,
+      "step": 3465,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69467239,
+      "epoch": 2.72,
+      "learning_rate": 8.710628789594855e-05,
+      "loss": 0.97685585,
+      "memory(GiB)": 71.19,
+      "step": 3470,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69117222,
+      "epoch": 2.72,
+      "learning_rate": 8.706279389792708e-05,
+      "loss": 0.9753437,
+      "memory(GiB)": 71.19,
+      "step": 3475,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.7065135,
+      "epoch": 2.73,
+      "learning_rate": 8.701923756287325e-05,
+      "loss": 0.92029209,
+      "memory(GiB)": 71.19,
+      "step": 3480,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68631201,
+      "epoch": 2.73,
+      "learning_rate": 8.697561896404573e-05,
+      "loss": 0.97980461,
+      "memory(GiB)": 71.19,
+      "step": 3485,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69038978,
+      "epoch": 2.74,
+      "learning_rate": 8.693193817480798e-05,
+      "loss": 0.96484947,
+      "memory(GiB)": 71.19,
+      "step": 3490,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68747773,
+      "epoch": 2.74,
+      "learning_rate": 8.688819526862803e-05,
+      "loss": 0.97741976,
+      "memory(GiB)": 71.19,
+      "step": 3495,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69307537,
+      "epoch": 2.74,
+      "learning_rate": 8.684439031907843e-05,
+      "loss": 0.97401991,
+      "memory(GiB)": 71.19,
+      "step": 3500,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68593907,
+      "epoch": 2.75,
+      "learning_rate": 8.6800523399836e-05,
+      "loss": 0.98575306,
+      "memory(GiB)": 71.19,
+      "step": 3505,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.69273987,
+      "epoch": 2.75,
+      "learning_rate": 8.675659458468184e-05,
+      "loss": 0.97432585,
+      "memory(GiB)": 71.19,
+      "step": 3510,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69315062,
+      "epoch": 2.75,
+      "learning_rate": 8.671260394750119e-05,
+      "loss": 0.95112991,
+      "memory(GiB)": 71.19,
+      "step": 3515,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69168601,
+      "epoch": 2.76,
+      "learning_rate": 8.66685515622832e-05,
+      "loss": 0.98360157,
+      "memory(GiB)": 71.19,
+      "step": 3520,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69696198,
+      "epoch": 2.76,
+      "learning_rate": 8.66244375031209e-05,
+      "loss": 0.95998116,
+      "memory(GiB)": 71.19,
+      "step": 3525,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70211124,
+      "epoch": 2.77,
+      "learning_rate": 8.658026184421108e-05,
+      "loss": 0.91666279,
+      "memory(GiB)": 71.19,
+      "step": 3530,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70166306,
+      "epoch": 2.77,
+      "learning_rate": 8.653602465985411e-05,
+      "loss": 0.94730425,
+      "memory(GiB)": 71.19,
+      "step": 3535,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69360576,
+      "epoch": 2.77,
+      "learning_rate": 8.649172602445384e-05,
+      "loss": 0.96196995,
+      "memory(GiB)": 71.19,
+      "step": 3540,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69113541,
+      "epoch": 2.78,
+      "learning_rate": 8.644736601251749e-05,
+      "loss": 0.97629557,
+      "memory(GiB)": 71.19,
+      "step": 3545,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69766083,
+      "epoch": 2.78,
+      "learning_rate": 8.640294469865549e-05,
+      "loss": 0.97009659,
+      "memory(GiB)": 71.19,
+      "step": 3550,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.7054462,
+      "epoch": 2.79,
+      "learning_rate": 8.635846215758141e-05,
+      "loss": 0.94846039,
+      "memory(GiB)": 71.19,
+      "step": 3555,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68731637,
+      "epoch": 2.79,
+      "learning_rate": 8.631391846411177e-05,
+      "loss": 1.00905914,
+      "memory(GiB)": 71.19,
+      "step": 3560,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.7051527,
+      "epoch": 2.79,
+      "learning_rate": 8.626931369316594e-05,
+      "loss": 0.93751678,
+      "memory(GiB)": 71.19,
+      "step": 3565,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69043989,
+      "epoch": 2.8,
+      "learning_rate": 8.622464791976604e-05,
+      "loss": 0.97965469,
+      "memory(GiB)": 71.19,
+      "step": 3570,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.68530822,
+      "epoch": 2.8,
+      "learning_rate": 8.617992121903679e-05,
+      "loss": 1.03002996,
+      "memory(GiB)": 71.19,
+      "step": 3575,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69462538,
+      "epoch": 2.81,
+      "learning_rate": 8.613513366620538e-05,
+      "loss": 0.958325,
+      "memory(GiB)": 71.19,
+      "step": 3580,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70790119,
+      "epoch": 2.81,
+      "learning_rate": 8.609028533660135e-05,
+      "loss": 0.91673079,
+      "memory(GiB)": 71.19,
+      "step": 3585,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69034023,
+      "epoch": 2.81,
+      "learning_rate": 8.604537630565644e-05,
+      "loss": 0.9714694,
+      "memory(GiB)": 71.19,
+      "step": 3590,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.69581952,
+      "epoch": 2.82,
+      "learning_rate": 8.600040664890453e-05,
+      "loss": 0.96100874,
+      "memory(GiB)": 71.19,
+      "step": 3595,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.70398974,
+      "epoch": 2.82,
+      "learning_rate": 8.595537644198141e-05,
+      "loss": 0.9286829,
+      "memory(GiB)": 71.19,
+      "step": 3600,
+      "train_speed(iter/s)": 0.024199
+    },
+    {
+      "acc": 0.68837748,
+      "epoch": 2.83,
+      "learning_rate": 8.591028576062478e-05,
+      "loss": 0.99620714,
+      "memory(GiB)": 71.19,
+      "step": 3605,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.69134026,
+      "epoch": 2.83,
+      "learning_rate": 8.586513468067404e-05,
+      "loss": 0.97660503,
+      "memory(GiB)": 71.19,
+      "step": 3610,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.69011774,
+      "epoch": 2.83,
+      "learning_rate": 8.581992327807007e-05,
+      "loss": 0.9757412,
+      "memory(GiB)": 71.19,
+      "step": 3615,
+      "train_speed(iter/s)": 0.0242
+    },
+    {
+      "acc": 0.6966249,
+      "epoch": 2.84,
+      "learning_rate": 8.577465162885538e-05,
+      "loss": 0.96154041,
+      "memory(GiB)": 71.19,
+      "step": 3620,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.70414891,
+      "epoch": 2.84,
+      "learning_rate": 8.572931980917366e-05,
+      "loss": 0.95618954,
+      "memory(GiB)": 71.19,
+      "step": 3625,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.69498696,
+      "epoch": 2.85,
+      "learning_rate": 8.568392789526991e-05,
+      "loss": 0.9771987,
+      "memory(GiB)": 71.19,
+      "step": 3630,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.69978442,
+      "epoch": 2.85,
+      "learning_rate": 8.563847596349015e-05,
+      "loss": 0.93518782,
+      "memory(GiB)": 71.19,
+      "step": 3635,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.70075121,
+      "epoch": 2.85,
+      "learning_rate": 8.559296409028134e-05,
+      "loss": 0.95782223,
+      "memory(GiB)": 71.19,
+      "step": 3640,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.69935665,
+      "epoch": 2.86,
+      "learning_rate": 8.554739235219129e-05,
+      "loss": 0.96709738,
+      "memory(GiB)": 71.19,
+      "step": 3645,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.70329223,
+      "epoch": 2.86,
+      "learning_rate": 8.550176082586847e-05,
+      "loss": 0.92794161,
+      "memory(GiB)": 71.19,
+      "step": 3650,
+      "train_speed(iter/s)": 0.024201
+    },
+    {
+      "acc": 0.69627366,
+      "epoch": 2.86,
+      "learning_rate": 8.545606958806195e-05,
+      "loss": 0.96420841,
+      "memory(GiB)": 71.19,
+      "step": 3655,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.69576621,
+      "epoch": 2.87,
+      "learning_rate": 8.541031871562118e-05,
+      "loss": 0.97613201,
+      "memory(GiB)": 71.19,
+      "step": 3660,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.69630737,
+      "epoch": 2.87,
+      "learning_rate": 8.536450828549593e-05,
+      "loss": 0.93979073,
+      "memory(GiB)": 71.19,
+      "step": 3665,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.68497229,
+      "epoch": 2.88,
+      "learning_rate": 8.531863837473617e-05,
+      "loss": 0.98675127,
+      "memory(GiB)": 71.19,
+      "step": 3670,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.68888111,
+      "epoch": 2.88,
+      "learning_rate": 8.527270906049185e-05,
+      "loss": 0.99011154,
+      "memory(GiB)": 71.19,
+      "step": 3675,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.70074887,
+      "epoch": 2.88,
+      "learning_rate": 8.522672042001291e-05,
+      "loss": 0.95177927,
+      "memory(GiB)": 71.19,
+      "step": 3680,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69165268,
+      "epoch": 2.89,
+      "learning_rate": 8.5180672530649e-05,
+      "loss": 0.97305288,
+      "memory(GiB)": 71.19,
+      "step": 3685,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69212561,
+      "epoch": 2.89,
+      "learning_rate": 8.51345654698495e-05,
+      "loss": 0.96181002,
+      "memory(GiB)": 71.19,
+      "step": 3690,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.7015995,
+      "epoch": 2.9,
+      "learning_rate": 8.508839931516322e-05,
+      "loss": 0.94027262,
+      "memory(GiB)": 71.19,
+      "step": 3695,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.6916369,
+      "epoch": 2.9,
+      "learning_rate": 8.504217414423843e-05,
+      "loss": 0.96168337,
+      "memory(GiB)": 71.19,
+      "step": 3700,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69469385,
+      "epoch": 2.9,
+      "learning_rate": 8.499589003482264e-05,
+      "loss": 0.96138353,
+      "memory(GiB)": 71.19,
+      "step": 3705,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69942169,
+      "epoch": 2.91,
+      "learning_rate": 8.49495470647625e-05,
+      "loss": 0.96778059,
+      "memory(GiB)": 71.19,
+      "step": 3710,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.69427686,
+      "epoch": 2.91,
+      "learning_rate": 8.490314531200365e-05,
+      "loss": 0.97816887,
+      "memory(GiB)": 71.19,
+      "step": 3715,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.7088768,
+      "epoch": 2.92,
+      "learning_rate": 8.485668485459057e-05,
+      "loss": 0.94817715,
+      "memory(GiB)": 71.19,
+      "step": 3720,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69802999,
+      "epoch": 2.92,
+      "learning_rate": 8.481016577066654e-05,
+      "loss": 0.97512226,
+      "memory(GiB)": 71.19,
+      "step": 3725,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.70213175,
+      "epoch": 2.92,
+      "learning_rate": 8.47635881384734e-05,
+      "loss": 0.94330521,
+      "memory(GiB)": 71.19,
+      "step": 3730,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.68532333,
+      "epoch": 2.93,
+      "learning_rate": 8.47169520363515e-05,
+      "loss": 0.96857815,
+      "memory(GiB)": 71.19,
+      "step": 3735,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.69682803,
+      "epoch": 2.93,
+      "learning_rate": 8.467025754273947e-05,
+      "loss": 0.98179436,
+      "memory(GiB)": 71.19,
+      "step": 3740,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.69457374,
+      "epoch": 2.94,
+      "learning_rate": 8.46235047361742e-05,
+      "loss": 0.97254868,
+      "memory(GiB)": 71.19,
+      "step": 3745,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.69419403,
+      "epoch": 2.94,
+      "learning_rate": 8.457669369529066e-05,
+      "loss": 0.95455761,
+      "memory(GiB)": 71.19,
+      "step": 3750,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.70040722,
+      "epoch": 2.94,
+      "learning_rate": 8.452982449882175e-05,
+      "loss": 0.92892199,
+      "memory(GiB)": 71.19,
+      "step": 3755,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.70409389,
+      "epoch": 2.95,
+      "learning_rate": 8.448289722559816e-05,
+      "loss": 0.93102856,
+      "memory(GiB)": 71.19,
+      "step": 3760,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69685011,
+      "epoch": 2.95,
+      "learning_rate": 8.443591195454834e-05,
+      "loss": 0.96960402,
+      "memory(GiB)": 71.19,
+      "step": 3765,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69049578,
+      "epoch": 2.95,
+      "learning_rate": 8.438886876469818e-05,
+      "loss": 0.9981617,
+      "memory(GiB)": 71.19,
+      "step": 3770,
+      "train_speed(iter/s)": 0.024204
+    },
+    {
+      "acc": 0.70177493,
+      "epoch": 2.96,
+      "learning_rate": 8.43417677351711e-05,
+      "loss": 0.91999292,
+      "memory(GiB)": 71.19,
+      "step": 3775,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69396958,
+      "epoch": 2.96,
+      "learning_rate": 8.429460894518771e-05,
+      "loss": 0.98489656,
+      "memory(GiB)": 71.19,
+      "step": 3780,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69775419,
+      "epoch": 2.97,
+      "learning_rate": 8.424739247406579e-05,
+      "loss": 0.95434799,
+      "memory(GiB)": 71.19,
+      "step": 3785,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.69874148,
+      "epoch": 2.97,
+      "learning_rate": 8.420011840122016e-05,
+      "loss": 0.94881783,
+      "memory(GiB)": 71.19,
+      "step": 3790,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.70053377,
+      "epoch": 2.97,
+      "learning_rate": 8.415278680616252e-05,
+      "loss": 0.95457497,
+      "memory(GiB)": 71.19,
+      "step": 3795,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "acc": 0.68822541,
+      "epoch": 2.98,
+      "learning_rate": 8.410539776850133e-05,
+      "loss": 0.97871866,
+      "memory(GiB)": 71.19,
+      "step": 3800,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.69143753,
+      "epoch": 2.98,
+      "learning_rate": 8.405795136794159e-05,
+      "loss": 0.9598156,
+      "memory(GiB)": 71.19,
+      "step": 3805,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.68940425,
+      "epoch": 2.99,
+      "learning_rate": 8.401044768428487e-05,
+      "loss": 0.98408442,
+      "memory(GiB)": 71.19,
+      "step": 3810,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.71102018,
+      "epoch": 2.99,
+      "learning_rate": 8.396288679742905e-05,
+      "loss": 0.91080713,
+      "memory(GiB)": 71.19,
+      "step": 3815,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.69605985,
+      "epoch": 2.99,
+      "learning_rate": 8.391526878736822e-05,
+      "loss": 0.96173496,
+      "memory(GiB)": 71.19,
+      "step": 3820,
+      "train_speed(iter/s)": 0.024202
+    },
+    {
+      "acc": 0.68129706,
+      "epoch": 3.0,
+      "learning_rate": 8.386759373419252e-05,
+      "loss": 1.0301609,
+      "memory(GiB)": 71.19,
+      "step": 3825,
+      "train_speed(iter/s)": 0.024203
+    },
+    {
+      "epoch": 3.0,
+      "eval_acc": 0.7299023290758828,
+      "eval_loss": 0.811944305896759,
+      "eval_runtime": 107.6655,
+      "eval_samples_per_second": 0.864,
+      "eval_steps_per_second": 0.864,
+      "step": 3827
+    },
+    {
+      "acc": 0.70716329,
+      "epoch": 3.0,
+      "learning_rate": 8.381986171808811e-05,
+      "loss": 0.90910578,
+      "memory(GiB)": 71.19,
+      "step": 3830,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.73387895,
+      "epoch": 3.01,
+      "learning_rate": 8.377207281933687e-05,
+      "loss": 0.82784376,
+      "memory(GiB)": 71.19,
+      "step": 3835,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.72814317,
+      "epoch": 3.01,
+      "learning_rate": 8.372422711831644e-05,
+      "loss": 0.83022747,
+      "memory(GiB)": 71.19,
+      "step": 3840,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.72625008,
+      "epoch": 3.01,
+      "learning_rate": 8.367632469549989e-05,
+      "loss": 0.82957611,
+      "memory(GiB)": 71.19,
+      "step": 3845,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.7326128,
+      "epoch": 3.02,
+      "learning_rate": 8.362836563145578e-05,
+      "loss": 0.82128582,
+      "memory(GiB)": 71.19,
+      "step": 3850,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.7478919,
+      "epoch": 3.02,
+      "learning_rate": 8.358035000684791e-05,
+      "loss": 0.77482996,
+      "memory(GiB)": 71.19,
+      "step": 3855,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.73634334,
+      "epoch": 3.03,
+      "learning_rate": 8.353227790243521e-05,
+      "loss": 0.804041,
+      "memory(GiB)": 71.19,
+      "step": 3860,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.73621793,
+      "epoch": 3.03,
+      "learning_rate": 8.34841493990716e-05,
+      "loss": 0.80979652,
+      "memory(GiB)": 71.19,
+      "step": 3865,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.7203692,
+      "epoch": 3.03,
+      "learning_rate": 8.343596457770587e-05,
+      "loss": 0.84263067,
+      "memory(GiB)": 71.19,
+      "step": 3870,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.71783953,
+      "epoch": 3.04,
+      "learning_rate": 8.338772351938148e-05,
+      "loss": 0.87986736,
+      "memory(GiB)": 71.19,
+      "step": 3875,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.74515467,
+      "epoch": 3.04,
+      "learning_rate": 8.333942630523662e-05,
+      "loss": 0.78914156,
+      "memory(GiB)": 71.19,
+      "step": 3880,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.71863422,
+      "epoch": 3.04,
+      "learning_rate": 8.329107301650374e-05,
+      "loss": 0.85929756,
+      "memory(GiB)": 71.19,
+      "step": 3885,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.73198423,
+      "epoch": 3.05,
+      "learning_rate": 8.324266373450974e-05,
+      "loss": 0.84285383,
+      "memory(GiB)": 71.19,
+      "step": 3890,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.71086202,
+      "epoch": 3.05,
+      "learning_rate": 8.319419854067564e-05,
+      "loss": 0.89417582,
+      "memory(GiB)": 71.19,
+      "step": 3895,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.73028541,
+      "epoch": 3.06,
+      "learning_rate": 8.314567751651654e-05,
+      "loss": 0.84059143,
+      "memory(GiB)": 71.19,
+      "step": 3900,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.73362865,
+      "epoch": 3.06,
+      "learning_rate": 8.309710074364138e-05,
+      "loss": 0.82103605,
+      "memory(GiB)": 71.19,
+      "step": 3905,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.74105625,
+      "epoch": 3.06,
+      "learning_rate": 8.304846830375292e-05,
+      "loss": 0.7890769,
+      "memory(GiB)": 71.19,
+      "step": 3910,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.73808637,
+      "epoch": 3.07,
+      "learning_rate": 8.299978027864752e-05,
+      "loss": 0.81043692,
+      "memory(GiB)": 71.19,
+      "step": 3915,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.71901126,
+      "epoch": 3.07,
+      "learning_rate": 8.295103675021505e-05,
+      "loss": 0.85230951,
+      "memory(GiB)": 71.19,
+      "step": 3920,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7334156,
+      "epoch": 3.08,
+      "learning_rate": 8.290223780043874e-05,
+      "loss": 0.8069334,
+      "memory(GiB)": 71.19,
+      "step": 3925,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.72944226,
+      "epoch": 3.08,
+      "learning_rate": 8.285338351139496e-05,
+      "loss": 0.84712934,
+      "memory(GiB)": 71.19,
+      "step": 3930,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.73505478,
+      "epoch": 3.08,
+      "learning_rate": 8.280447396525328e-05,
+      "loss": 0.79626627,
+      "memory(GiB)": 71.19,
+      "step": 3935,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.72741809,
+      "epoch": 3.09,
+      "learning_rate": 8.275550924427609e-05,
+      "loss": 0.82367277,
+      "memory(GiB)": 71.19,
+      "step": 3940,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.72524977,
+      "epoch": 3.09,
+      "learning_rate": 8.270648943081867e-05,
+      "loss": 0.84687901,
+      "memory(GiB)": 71.19,
+      "step": 3945,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.73152761,
+      "epoch": 3.1,
+      "learning_rate": 8.26574146073289e-05,
+      "loss": 0.84828215,
+      "memory(GiB)": 71.19,
+      "step": 3950,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.72471776,
+      "epoch": 3.1,
+      "learning_rate": 8.260828485634722e-05,
+      "loss": 0.86218882,
+      "memory(GiB)": 71.19,
+      "step": 3955,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.7207871,
+      "epoch": 3.1,
+      "learning_rate": 8.255910026050643e-05,
+      "loss": 0.84007759,
+      "memory(GiB)": 71.19,
+      "step": 3960,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.73610811,
+      "epoch": 3.11,
+      "learning_rate": 8.250986090253163e-05,
+      "loss": 0.82142134,
+      "memory(GiB)": 71.19,
+      "step": 3965,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73013797,
+      "epoch": 3.11,
+      "learning_rate": 8.246056686523994e-05,
+      "loss": 0.83745384,
+      "memory(GiB)": 71.19,
+      "step": 3970,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73723321,
+      "epoch": 3.12,
+      "learning_rate": 8.241121823154047e-05,
+      "loss": 0.81523333,
+      "memory(GiB)": 71.19,
+      "step": 3975,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7233716,
+      "epoch": 3.12,
+      "learning_rate": 8.236181508443424e-05,
+      "loss": 0.83602371,
+      "memory(GiB)": 71.19,
+      "step": 3980,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.72590818,
+      "epoch": 3.12,
+      "learning_rate": 8.231235750701385e-05,
+      "loss": 0.8545907,
+      "memory(GiB)": 71.19,
+      "step": 3985,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.7300528,
+      "epoch": 3.13,
+      "learning_rate": 8.226284558246351e-05,
+      "loss": 0.82412634,
+      "memory(GiB)": 71.19,
+      "step": 3990,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.72257438,
+      "epoch": 3.13,
+      "learning_rate": 8.221327939405881e-05,
+      "loss": 0.85074911,
+      "memory(GiB)": 71.19,
+      "step": 3995,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73442616,
+      "epoch": 3.14,
+      "learning_rate": 8.216365902516664e-05,
+      "loss": 0.80528517,
+      "memory(GiB)": 71.19,
+      "step": 4000,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.72549343,
+      "epoch": 3.14,
+      "learning_rate": 8.211398455924497e-05,
+      "loss": 0.86022482,
+      "memory(GiB)": 71.19,
+      "step": 4005,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72123246,
+      "epoch": 3.14,
+      "learning_rate": 8.206425607984282e-05,
+      "loss": 0.86917925,
+      "memory(GiB)": 71.19,
+      "step": 4010,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73181915,
+      "epoch": 3.15,
+      "learning_rate": 8.201447367059998e-05,
+      "loss": 0.82419157,
+      "memory(GiB)": 71.19,
+      "step": 4015,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72473426,
+      "epoch": 3.15,
+      "learning_rate": 8.196463741524701e-05,
+      "loss": 0.85585718,
+      "memory(GiB)": 71.19,
+      "step": 4020,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72140322,
+      "epoch": 3.15,
+      "learning_rate": 8.1914747397605e-05,
+      "loss": 0.85199308,
+      "memory(GiB)": 71.19,
+      "step": 4025,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73175788,
+      "epoch": 3.16,
+      "learning_rate": 8.186480370158551e-05,
+      "loss": 0.85048962,
+      "memory(GiB)": 71.19,
+      "step": 4030,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72417698,
+      "epoch": 3.16,
+      "learning_rate": 8.181480641119032e-05,
+      "loss": 0.84959297,
+      "memory(GiB)": 71.19,
+      "step": 4035,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72861757,
+      "epoch": 3.17,
+      "learning_rate": 8.176475561051137e-05,
+      "loss": 0.83446856,
+      "memory(GiB)": 71.19,
+      "step": 4040,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73830481,
+      "epoch": 3.17,
+      "learning_rate": 8.171465138373067e-05,
+      "loss": 0.79196048,
+      "memory(GiB)": 71.19,
+      "step": 4045,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73525729,
+      "epoch": 3.17,
+      "learning_rate": 8.166449381511998e-05,
+      "loss": 0.80135098,
+      "memory(GiB)": 71.19,
+      "step": 4050,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72653871,
+      "epoch": 3.18,
+      "learning_rate": 8.161428298904085e-05,
+      "loss": 0.84726629,
+      "memory(GiB)": 71.19,
+      "step": 4055,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73992734,
+      "epoch": 3.18,
+      "learning_rate": 8.156401898994436e-05,
+      "loss": 0.80075979,
+      "memory(GiB)": 71.19,
+      "step": 4060,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7188323,
+      "epoch": 3.19,
+      "learning_rate": 8.151370190237108e-05,
+      "loss": 0.88397989,
+      "memory(GiB)": 71.19,
+      "step": 4065,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.73720374,
+      "epoch": 3.19,
+      "learning_rate": 8.146333181095084e-05,
+      "loss": 0.82127218,
+      "memory(GiB)": 71.19,
+      "step": 4070,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72629747,
+      "epoch": 3.19,
+      "learning_rate": 8.14129088004026e-05,
+      "loss": 0.86305199,
+      "memory(GiB)": 71.19,
+      "step": 4075,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.73932276,
+      "epoch": 3.2,
+      "learning_rate": 8.136243295553434e-05,
+      "loss": 0.82476034,
+      "memory(GiB)": 71.19,
+      "step": 4080,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72464457,
+      "epoch": 3.2,
+      "learning_rate": 8.131190436124294e-05,
+      "loss": 0.86362638,
+      "memory(GiB)": 71.19,
+      "step": 4085,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.72220616,
+      "epoch": 3.21,
+      "learning_rate": 8.126132310251393e-05,
+      "loss": 0.83337736,
+      "memory(GiB)": 71.19,
+      "step": 4090,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73724484,
+      "epoch": 3.21,
+      "learning_rate": 8.121068926442148e-05,
+      "loss": 0.82069883,
+      "memory(GiB)": 71.19,
+      "step": 4095,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7229681,
+      "epoch": 3.21,
+      "learning_rate": 8.116000293212815e-05,
+      "loss": 0.85246668,
+      "memory(GiB)": 71.19,
+      "step": 4100,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73643503,
+      "epoch": 3.22,
+      "learning_rate": 8.110926419088485e-05,
+      "loss": 0.83178596,
+      "memory(GiB)": 71.19,
+      "step": 4105,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73539133,
+      "epoch": 3.22,
+      "learning_rate": 8.105847312603056e-05,
+      "loss": 0.82203569,
+      "memory(GiB)": 71.19,
+      "step": 4110,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73065267,
+      "epoch": 3.23,
+      "learning_rate": 8.100762982299232e-05,
+      "loss": 0.79565067,
+      "memory(GiB)": 71.19,
+      "step": 4115,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72940273,
+      "epoch": 3.23,
+      "learning_rate": 8.095673436728504e-05,
+      "loss": 0.83768406,
+      "memory(GiB)": 71.19,
+      "step": 4120,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72964163,
+      "epoch": 3.23,
+      "learning_rate": 8.090578684451131e-05,
+      "loss": 0.84069805,
+      "memory(GiB)": 71.19,
+      "step": 4125,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73201489,
+      "epoch": 3.24,
+      "learning_rate": 8.085478734036129e-05,
+      "loss": 0.82388248,
+      "memory(GiB)": 71.19,
+      "step": 4130,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72238297,
+      "epoch": 3.24,
+      "learning_rate": 8.080373594061261e-05,
+      "loss": 0.85109663,
+      "memory(GiB)": 71.19,
+      "step": 4135,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.74465365,
+      "epoch": 3.24,
+      "learning_rate": 8.075263273113013e-05,
+      "loss": 0.78524365,
+      "memory(GiB)": 71.19,
+      "step": 4140,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71745272,
+      "epoch": 3.25,
+      "learning_rate": 8.070147779786593e-05,
+      "loss": 0.87544708,
+      "memory(GiB)": 71.19,
+      "step": 4145,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71826987,
+      "epoch": 3.25,
+      "learning_rate": 8.0650271226859e-05,
+      "loss": 0.8916872,
+      "memory(GiB)": 71.19,
+      "step": 4150,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72508607,
+      "epoch": 3.26,
+      "learning_rate": 8.05990131042352e-05,
+      "loss": 0.83632765,
+      "memory(GiB)": 71.19,
+      "step": 4155,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72774878,
+      "epoch": 3.26,
+      "learning_rate": 8.054770351620718e-05,
+      "loss": 0.84630947,
+      "memory(GiB)": 71.19,
+      "step": 4160,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71050835,
+      "epoch": 3.26,
+      "learning_rate": 8.049634254907404e-05,
+      "loss": 0.88207722,
+      "memory(GiB)": 71.19,
+      "step": 4165,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73926072,
+      "epoch": 3.27,
+      "learning_rate": 8.044493028922133e-05,
+      "loss": 0.80433979,
+      "memory(GiB)": 71.19,
+      "step": 4170,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7169281,
+      "epoch": 3.27,
+      "learning_rate": 8.03934668231209e-05,
+      "loss": 0.85596704,
+      "memory(GiB)": 71.19,
+      "step": 4175,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7407104,
+      "epoch": 3.28,
+      "learning_rate": 8.034195223733074e-05,
+      "loss": 0.80006132,
+      "memory(GiB)": 71.19,
+      "step": 4180,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72462053,
+      "epoch": 3.28,
+      "learning_rate": 8.029038661849472e-05,
+      "loss": 0.85594482,
+      "memory(GiB)": 71.19,
+      "step": 4185,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72585344,
+      "epoch": 3.28,
+      "learning_rate": 8.023877005334268e-05,
+      "loss": 0.84595318,
+      "memory(GiB)": 71.19,
+      "step": 4190,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7315063,
+      "epoch": 3.29,
+      "learning_rate": 8.018710262869005e-05,
+      "loss": 0.82566252,
+      "memory(GiB)": 71.19,
+      "step": 4195,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7300137,
+      "epoch": 3.29,
+      "learning_rate": 8.013538443143782e-05,
+      "loss": 0.8305336,
+      "memory(GiB)": 71.19,
+      "step": 4200,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72604132,
+      "epoch": 3.3,
+      "learning_rate": 8.008361554857237e-05,
+      "loss": 0.85777216,
+      "memory(GiB)": 71.19,
+      "step": 4205,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71913915,
+      "epoch": 3.3,
+      "learning_rate": 8.003179606716543e-05,
+      "loss": 0.85751801,
+      "memory(GiB)": 71.19,
+      "step": 4210,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71904831,
+      "epoch": 3.3,
+      "learning_rate": 7.997992607437365e-05,
+      "loss": 0.86428804,
+      "memory(GiB)": 71.19,
+      "step": 4215,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71747947,
+      "epoch": 3.31,
+      "learning_rate": 7.992800565743882e-05,
+      "loss": 0.86572142,
+      "memory(GiB)": 71.19,
+      "step": 4220,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72012987,
+      "epoch": 3.31,
+      "learning_rate": 7.987603490368741e-05,
+      "loss": 0.85723791,
+      "memory(GiB)": 71.19,
+      "step": 4225,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73201785,
+      "epoch": 3.32,
+      "learning_rate": 7.98240139005306e-05,
+      "loss": 0.8323741,
+      "memory(GiB)": 71.19,
+      "step": 4230,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72473407,
+      "epoch": 3.32,
+      "learning_rate": 7.977194273546411e-05,
+      "loss": 0.85458097,
+      "memory(GiB)": 71.19,
+      "step": 4235,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73437066,
+      "epoch": 3.32,
+      "learning_rate": 7.971982149606799e-05,
+      "loss": 0.80146246,
+      "memory(GiB)": 71.19,
+      "step": 4240,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71682711,
+      "epoch": 3.33,
+      "learning_rate": 7.966765027000654e-05,
+      "loss": 0.86212845,
+      "memory(GiB)": 71.19,
+      "step": 4245,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72666965,
+      "epoch": 3.33,
+      "learning_rate": 7.961542914502808e-05,
+      "loss": 0.84145832,
+      "memory(GiB)": 71.19,
+      "step": 4250,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72355342,
+      "epoch": 3.33,
+      "learning_rate": 7.956315820896496e-05,
+      "loss": 0.85306997,
+      "memory(GiB)": 71.19,
+      "step": 4255,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73941717,
+      "epoch": 3.34,
+      "learning_rate": 7.951083754973321e-05,
+      "loss": 0.82038488,
+      "memory(GiB)": 71.19,
+      "step": 4260,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71963987,
+      "epoch": 3.34,
+      "learning_rate": 7.945846725533251e-05,
+      "loss": 0.8572114,
+      "memory(GiB)": 71.19,
+      "step": 4265,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73555474,
+      "epoch": 3.35,
+      "learning_rate": 7.940604741384607e-05,
+      "loss": 0.82101889,
+      "memory(GiB)": 71.19,
+      "step": 4270,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.74150233,
+      "epoch": 3.35,
+      "learning_rate": 7.935357811344042e-05,
+      "loss": 0.80582638,
+      "memory(GiB)": 71.19,
+      "step": 4275,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72068758,
+      "epoch": 3.35,
+      "learning_rate": 7.93010594423652e-05,
+      "loss": 0.8544672,
+      "memory(GiB)": 71.19,
+      "step": 4280,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73440638,
+      "epoch": 3.36,
+      "learning_rate": 7.924849148895321e-05,
+      "loss": 0.80622263,
+      "memory(GiB)": 71.19,
+      "step": 4285,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72935481,
+      "epoch": 3.36,
+      "learning_rate": 7.919587434162004e-05,
+      "loss": 0.82818518,
+      "memory(GiB)": 71.19,
+      "step": 4290,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72351742,
+      "epoch": 3.37,
+      "learning_rate": 7.914320808886409e-05,
+      "loss": 0.85957403,
+      "memory(GiB)": 71.19,
+      "step": 4295,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7248076,
+      "epoch": 3.37,
+      "learning_rate": 7.909049281926629e-05,
+      "loss": 0.8453124,
+      "memory(GiB)": 71.19,
+      "step": 4300,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72555737,
+      "epoch": 3.37,
+      "learning_rate": 7.903772862149004e-05,
+      "loss": 0.8440258,
+      "memory(GiB)": 71.19,
+      "step": 4305,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72598963,
+      "epoch": 3.38,
+      "learning_rate": 7.898491558428108e-05,
+      "loss": 0.83704453,
+      "memory(GiB)": 71.19,
+      "step": 4310,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71948795,
+      "epoch": 3.38,
+      "learning_rate": 7.893205379646724e-05,
+      "loss": 0.86024466,
+      "memory(GiB)": 71.19,
+      "step": 4315,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73115282,
+      "epoch": 3.39,
+      "learning_rate": 7.887914334695831e-05,
+      "loss": 0.83252287,
+      "memory(GiB)": 71.19,
+      "step": 4320,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72599154,
+      "epoch": 3.39,
+      "learning_rate": 7.882618432474604e-05,
+      "loss": 0.83888359,
+      "memory(GiB)": 71.19,
+      "step": 4325,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73242888,
+      "epoch": 3.39,
+      "learning_rate": 7.877317681890376e-05,
+      "loss": 0.81837931,
+      "memory(GiB)": 71.19,
+      "step": 4330,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.742377,
+      "epoch": 3.4,
+      "learning_rate": 7.87201209185864e-05,
+      "loss": 0.80096016,
+      "memory(GiB)": 71.19,
+      "step": 4335,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7232306,
+      "epoch": 3.4,
+      "learning_rate": 7.866701671303032e-05,
+      "loss": 0.86032343,
+      "memory(GiB)": 71.19,
+      "step": 4340,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72954855,
+      "epoch": 3.41,
+      "learning_rate": 7.861386429155304e-05,
+      "loss": 0.82318716,
+      "memory(GiB)": 71.19,
+      "step": 4345,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71807418,
+      "epoch": 3.41,
+      "learning_rate": 7.856066374355326e-05,
+      "loss": 0.88005505,
+      "memory(GiB)": 71.19,
+      "step": 4350,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73638568,
+      "epoch": 3.41,
+      "learning_rate": 7.850741515851057e-05,
+      "loss": 0.81349354,
+      "memory(GiB)": 71.19,
+      "step": 4355,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73490205,
+      "epoch": 3.42,
+      "learning_rate": 7.845411862598537e-05,
+      "loss": 0.83284054,
+      "memory(GiB)": 71.19,
+      "step": 4360,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73105078,
+      "epoch": 3.42,
+      "learning_rate": 7.840077423561871e-05,
+      "loss": 0.84245071,
+      "memory(GiB)": 71.19,
+      "step": 4365,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7196713,
+      "epoch": 3.43,
+      "learning_rate": 7.834738207713213e-05,
+      "loss": 0.86376743,
+      "memory(GiB)": 71.19,
+      "step": 4370,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72638869,
+      "epoch": 3.43,
+      "learning_rate": 7.829394224032753e-05,
+      "loss": 0.82913513,
+      "memory(GiB)": 71.19,
+      "step": 4375,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73445344,
+      "epoch": 3.43,
+      "learning_rate": 7.824045481508696e-05,
+      "loss": 0.8318285,
+      "memory(GiB)": 71.19,
+      "step": 4380,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7262754,
+      "epoch": 3.44,
+      "learning_rate": 7.818691989137255e-05,
+      "loss": 0.84009447,
+      "memory(GiB)": 71.19,
+      "step": 4385,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7386354,
+      "epoch": 3.44,
+      "learning_rate": 7.813333755922632e-05,
+      "loss": 0.81151896,
+      "memory(GiB)": 71.19,
+      "step": 4390,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72992892,
+      "epoch": 3.44,
+      "learning_rate": 7.807970790876997e-05,
+      "loss": 0.81546345,
+      "memory(GiB)": 71.19,
+      "step": 4395,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73798876,
+      "epoch": 3.45,
+      "learning_rate": 7.802603103020487e-05,
+      "loss": 0.80516491,
+      "memory(GiB)": 71.19,
+      "step": 4400,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72618403,
+      "epoch": 3.45,
+      "learning_rate": 7.797230701381177e-05,
+      "loss": 0.83551912,
+      "memory(GiB)": 71.19,
+      "step": 4405,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71417842,
+      "epoch": 3.46,
+      "learning_rate": 7.791853594995072e-05,
+      "loss": 0.84696274,
+      "memory(GiB)": 71.19,
+      "step": 4410,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7323916,
+      "epoch": 3.46,
+      "learning_rate": 7.78647179290609e-05,
+      "loss": 0.82821932,
+      "memory(GiB)": 71.19,
+      "step": 4415,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72546659,
+      "epoch": 3.46,
+      "learning_rate": 7.781085304166042e-05,
+      "loss": 0.84084616,
+      "memory(GiB)": 71.19,
+      "step": 4420,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73983116,
+      "epoch": 3.47,
+      "learning_rate": 7.775694137834632e-05,
+      "loss": 0.80149288,
+      "memory(GiB)": 71.19,
+      "step": 4425,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72577829,
+      "epoch": 3.47,
+      "learning_rate": 7.770298302979421e-05,
+      "loss": 0.85075827,
+      "memory(GiB)": 71.19,
+      "step": 4430,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.731323,
+      "epoch": 3.48,
+      "learning_rate": 7.764897808675831e-05,
+      "loss": 0.83516245,
+      "memory(GiB)": 71.19,
+      "step": 4435,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72064219,
+      "epoch": 3.48,
+      "learning_rate": 7.759492664007114e-05,
+      "loss": 0.87227879,
+      "memory(GiB)": 71.19,
+      "step": 4440,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73189783,
+      "epoch": 3.48,
+      "learning_rate": 7.754082878064346e-05,
+      "loss": 0.82408819,
+      "memory(GiB)": 71.19,
+      "step": 4445,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7397872,
+      "epoch": 3.49,
+      "learning_rate": 7.748668459946408e-05,
+      "loss": 0.8035758,
+      "memory(GiB)": 71.19,
+      "step": 4450,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73634014,
+      "epoch": 3.49,
+      "learning_rate": 7.743249418759976e-05,
+      "loss": 0.8225769,
+      "memory(GiB)": 71.19,
+      "step": 4455,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72766705,
+      "epoch": 3.5,
+      "learning_rate": 7.7378257636195e-05,
+      "loss": 0.83043652,
+      "memory(GiB)": 71.19,
+      "step": 4460,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72869205,
+      "epoch": 3.5,
+      "learning_rate": 7.732397503647184e-05,
+      "loss": 0.83498526,
+      "memory(GiB)": 71.19,
+      "step": 4465,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71786594,
+      "epoch": 3.5,
+      "learning_rate": 7.726964647972988e-05,
+      "loss": 0.86003532,
+      "memory(GiB)": 71.19,
+      "step": 4470,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73369732,
+      "epoch": 3.51,
+      "learning_rate": 7.721527205734593e-05,
+      "loss": 0.82165928,
+      "memory(GiB)": 71.19,
+      "step": 4475,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71339808,
+      "epoch": 3.51,
+      "learning_rate": 7.716085186077398e-05,
+      "loss": 0.86701927,
+      "memory(GiB)": 71.19,
+      "step": 4480,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.72527127,
+      "epoch": 3.52,
+      "learning_rate": 7.7106385981545e-05,
+      "loss": 0.8485733,
+      "memory(GiB)": 71.19,
+      "step": 4485,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.71741381,
+      "epoch": 3.52,
+      "learning_rate": 7.70518745112668e-05,
+      "loss": 0.87526913,
+      "memory(GiB)": 71.19,
+      "step": 4490,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.73170438,
+      "epoch": 3.52,
+      "learning_rate": 7.699731754162388e-05,
+      "loss": 0.84059429,
+      "memory(GiB)": 71.19,
+      "step": 4495,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72498207,
+      "epoch": 3.53,
+      "learning_rate": 7.694271516437723e-05,
+      "loss": 0.84459229,
+      "memory(GiB)": 71.19,
+      "step": 4500,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7239903,
+      "epoch": 3.53,
+      "learning_rate": 7.688806747136426e-05,
+      "loss": 0.84521217,
+      "memory(GiB)": 71.19,
+      "step": 4505,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73728113,
+      "epoch": 3.53,
+      "learning_rate": 7.683337455449856e-05,
+      "loss": 0.82545519,
+      "memory(GiB)": 71.19,
+      "step": 4510,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72692027,
+      "epoch": 3.54,
+      "learning_rate": 7.677863650576979e-05,
+      "loss": 0.84334087,
+      "memory(GiB)": 71.19,
+      "step": 4515,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72908607,
+      "epoch": 3.54,
+      "learning_rate": 7.672385341724355e-05,
+      "loss": 0.82269535,
+      "memory(GiB)": 71.19,
+      "step": 4520,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72978554,
+      "epoch": 3.55,
+      "learning_rate": 7.666902538106118e-05,
+      "loss": 0.84637251,
+      "memory(GiB)": 71.19,
+      "step": 4525,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.71525011,
+      "epoch": 3.55,
+      "learning_rate": 7.661415248943958e-05,
+      "loss": 0.87230482,
+      "memory(GiB)": 71.19,
+      "step": 4530,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72793851,
+      "epoch": 3.55,
+      "learning_rate": 7.655923483467114e-05,
+      "loss": 0.81432896,
+      "memory(GiB)": 71.19,
+      "step": 4535,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.73044486,
+      "epoch": 3.56,
+      "learning_rate": 7.650427250912351e-05,
+      "loss": 0.84025393,
+      "memory(GiB)": 71.19,
+      "step": 4540,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72327971,
+      "epoch": 3.56,
+      "learning_rate": 7.644926560523952e-05,
+      "loss": 0.84944382,
+      "memory(GiB)": 71.19,
+      "step": 4545,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72713952,
+      "epoch": 3.57,
+      "learning_rate": 7.639421421553687e-05,
+      "loss": 0.83744144,
+      "memory(GiB)": 71.19,
+      "step": 4550,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7387239,
+      "epoch": 3.57,
+      "learning_rate": 7.633911843260825e-05,
+      "loss": 0.82736578,
+      "memory(GiB)": 71.19,
+      "step": 4555,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.72111745,
+      "epoch": 3.57,
+      "learning_rate": 7.628397834912085e-05,
+      "loss": 0.8412631,
+      "memory(GiB)": 71.19,
+      "step": 4560,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.71875992,
+      "epoch": 3.58,
+      "learning_rate": 7.622879405781645e-05,
+      "loss": 0.85906324,
+      "memory(GiB)": 71.19,
+      "step": 4565,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72346978,
+      "epoch": 3.58,
+      "learning_rate": 7.617356565151122e-05,
+      "loss": 0.84273167,
+      "memory(GiB)": 71.19,
+      "step": 4570,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72537384,
+      "epoch": 3.59,
+      "learning_rate": 7.611829322309544e-05,
+      "loss": 0.86372871,
+      "memory(GiB)": 71.19,
+      "step": 4575,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72613001,
+      "epoch": 3.59,
+      "learning_rate": 7.60629768655335e-05,
+      "loss": 0.83706751,
+      "memory(GiB)": 71.19,
+      "step": 4580,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72427754,
+      "epoch": 3.59,
+      "learning_rate": 7.600761667186362e-05,
+      "loss": 0.83796959,
+      "memory(GiB)": 71.19,
+      "step": 4585,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.73405333,
+      "epoch": 3.6,
+      "learning_rate": 7.595221273519783e-05,
+      "loss": 0.84556036,
+      "memory(GiB)": 71.19,
+      "step": 4590,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72501197,
+      "epoch": 3.6,
+      "learning_rate": 7.589676514872165e-05,
+      "loss": 0.83936701,
+      "memory(GiB)": 71.19,
+      "step": 4595,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.71823664,
+      "epoch": 3.61,
+      "learning_rate": 7.584127400569408e-05,
+      "loss": 0.85496521,
+      "memory(GiB)": 71.19,
+      "step": 4600,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72726426,
+      "epoch": 3.61,
+      "learning_rate": 7.57857393994473e-05,
+      "loss": 0.84609032,
+      "memory(GiB)": 71.19,
+      "step": 4605,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72315588,
+      "epoch": 3.61,
+      "learning_rate": 7.573016142338668e-05,
+      "loss": 0.85114698,
+      "memory(GiB)": 71.19,
+      "step": 4610,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.71871891,
+      "epoch": 3.62,
+      "learning_rate": 7.56745401709905e-05,
+      "loss": 0.8591588,
+      "memory(GiB)": 71.19,
+      "step": 4615,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73322353,
+      "epoch": 3.62,
+      "learning_rate": 7.56188757358098e-05,
+      "loss": 0.83172417,
+      "memory(GiB)": 71.19,
+      "step": 4620,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72749004,
+      "epoch": 3.62,
+      "learning_rate": 7.55631682114683e-05,
+      "loss": 0.84570827,
+      "memory(GiB)": 71.19,
+      "step": 4625,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73905587,
+      "epoch": 3.63,
+      "learning_rate": 7.550741769166214e-05,
+      "loss": 0.80478392,
+      "memory(GiB)": 71.19,
+      "step": 4630,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.73352895,
+      "epoch": 3.63,
+      "learning_rate": 7.545162427015981e-05,
+      "loss": 0.81894255,
+      "memory(GiB)": 71.19,
+      "step": 4635,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.73099337,
+      "epoch": 3.64,
+      "learning_rate": 7.539578804080198e-05,
+      "loss": 0.84308519,
+      "memory(GiB)": 71.19,
+      "step": 4640,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.71498265,
+      "epoch": 3.64,
+      "learning_rate": 7.533990909750125e-05,
+      "loss": 0.87364149,
+      "memory(GiB)": 71.19,
+      "step": 4645,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.71505318,
+      "epoch": 3.64,
+      "learning_rate": 7.528398753424213e-05,
+      "loss": 0.86194353,
+      "memory(GiB)": 71.19,
+      "step": 4650,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7217392,
+      "epoch": 3.65,
+      "learning_rate": 7.522802344508078e-05,
+      "loss": 0.8368371,
+      "memory(GiB)": 71.19,
+      "step": 4655,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.74326339,
+      "epoch": 3.65,
+      "learning_rate": 7.517201692414488e-05,
+      "loss": 0.79576321,
+      "memory(GiB)": 71.19,
+      "step": 4660,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72128105,
+      "epoch": 3.66,
+      "learning_rate": 7.51159680656335e-05,
+      "loss": 0.84689484,
+      "memory(GiB)": 71.19,
+      "step": 4665,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.73218131,
+      "epoch": 3.66,
+      "learning_rate": 7.505987696381691e-05,
+      "loss": 0.80757494,
+      "memory(GiB)": 71.19,
+      "step": 4670,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72402844,
+      "epoch": 3.66,
+      "learning_rate": 7.500374371303643e-05,
+      "loss": 0.83772345,
+      "memory(GiB)": 71.19,
+      "step": 4675,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73203745,
+      "epoch": 3.67,
+      "learning_rate": 7.494756840770425e-05,
+      "loss": 0.82900801,
+      "memory(GiB)": 71.19,
+      "step": 4680,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73576927,
+      "epoch": 3.67,
+      "learning_rate": 7.489135114230333e-05,
+      "loss": 0.82100172,
+      "memory(GiB)": 71.19,
+      "step": 4685,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72248497,
+      "epoch": 3.68,
+      "learning_rate": 7.483509201138717e-05,
+      "loss": 0.83817987,
+      "memory(GiB)": 71.19,
+      "step": 4690,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72471027,
+      "epoch": 3.68,
+      "learning_rate": 7.477879110957972e-05,
+      "loss": 0.83272848,
+      "memory(GiB)": 71.19,
+      "step": 4695,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72231956,
+      "epoch": 3.68,
+      "learning_rate": 7.472244853157517e-05,
+      "loss": 0.85366192,
+      "memory(GiB)": 71.19,
+      "step": 4700,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72966657,
+      "epoch": 3.69,
+      "learning_rate": 7.46660643721378e-05,
+      "loss": 0.84410038,
+      "memory(GiB)": 71.19,
+      "step": 4705,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.71471176,
+      "epoch": 3.69,
+      "learning_rate": 7.460963872610181e-05,
+      "loss": 0.86216116,
+      "memory(GiB)": 71.19,
+      "step": 4710,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.73702974,
+      "epoch": 3.7,
+      "learning_rate": 7.455317168837122e-05,
+      "loss": 0.83383703,
+      "memory(GiB)": 71.19,
+      "step": 4715,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.70991087,
+      "epoch": 3.7,
+      "learning_rate": 7.449666335391963e-05,
+      "loss": 0.8940443,
+      "memory(GiB)": 71.19,
+      "step": 4720,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71537218,
+      "epoch": 3.7,
+      "learning_rate": 7.444011381779013e-05,
+      "loss": 0.86577148,
+      "memory(GiB)": 71.19,
+      "step": 4725,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72070432,
+      "epoch": 3.71,
+      "learning_rate": 7.438352317509508e-05,
+      "loss": 0.84208603,
+      "memory(GiB)": 71.19,
+      "step": 4730,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73044333,
+      "epoch": 3.71,
+      "learning_rate": 7.4326891521016e-05,
+      "loss": 0.82190485,
+      "memory(GiB)": 71.19,
+      "step": 4735,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72588086,
+      "epoch": 3.72,
+      "learning_rate": 7.427021895080339e-05,
+      "loss": 0.85073795,
+      "memory(GiB)": 71.19,
+      "step": 4740,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7267787,
+      "epoch": 3.72,
+      "learning_rate": 7.421350555977653e-05,
+      "loss": 0.83103905,
+      "memory(GiB)": 71.19,
+      "step": 4745,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72804618,
+      "epoch": 3.72,
+      "learning_rate": 7.415675144332339e-05,
+      "loss": 0.83636494,
+      "memory(GiB)": 71.19,
+      "step": 4750,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72480588,
+      "epoch": 3.73,
+      "learning_rate": 7.409995669690046e-05,
+      "loss": 0.84966698,
+      "memory(GiB)": 71.19,
+      "step": 4755,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.72015243,
+      "epoch": 3.73,
+      "learning_rate": 7.404312141603251e-05,
+      "loss": 0.86545382,
+      "memory(GiB)": 71.19,
+      "step": 4760,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.7205214,
+      "epoch": 3.73,
+      "learning_rate": 7.398624569631254e-05,
+      "loss": 0.86042595,
+      "memory(GiB)": 71.19,
+      "step": 4765,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.734021,
+      "epoch": 3.74,
+      "learning_rate": 7.392932963340151e-05,
+      "loss": 0.82725487,
+      "memory(GiB)": 71.19,
+      "step": 4770,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71987934,
+      "epoch": 3.74,
+      "learning_rate": 7.38723733230283e-05,
+      "loss": 0.85444155,
+      "memory(GiB)": 71.19,
+      "step": 4775,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7245378,
+      "epoch": 3.75,
+      "learning_rate": 7.381537686098942e-05,
+      "loss": 0.85145273,
+      "memory(GiB)": 71.19,
+      "step": 4780,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71786847,
+      "epoch": 3.75,
+      "learning_rate": 7.375834034314895e-05,
+      "loss": 0.88409557,
+      "memory(GiB)": 71.19,
+      "step": 4785,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71265817,
+      "epoch": 3.75,
+      "learning_rate": 7.370126386543833e-05,
+      "loss": 0.89585934,
+      "memory(GiB)": 71.19,
+      "step": 4790,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7173708,
+      "epoch": 3.76,
+      "learning_rate": 7.364414752385622e-05,
+      "loss": 0.88221331,
+      "memory(GiB)": 71.19,
+      "step": 4795,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73161063,
+      "epoch": 3.76,
+      "learning_rate": 7.358699141446833e-05,
+      "loss": 0.8258852,
+      "memory(GiB)": 71.19,
+      "step": 4800,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72865162,
+      "epoch": 3.77,
+      "learning_rate": 7.35297956334072e-05,
+      "loss": 0.84000244,
+      "memory(GiB)": 71.19,
+      "step": 4805,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72472777,
+      "epoch": 3.77,
+      "learning_rate": 7.34725602768722e-05,
+      "loss": 0.84447889,
+      "memory(GiB)": 71.19,
+      "step": 4810,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7250689,
+      "epoch": 3.77,
+      "learning_rate": 7.341528544112915e-05,
+      "loss": 0.866607,
+      "memory(GiB)": 71.19,
+      "step": 4815,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.72776718,
+      "epoch": 3.78,
+      "learning_rate": 7.335797122251038e-05,
+      "loss": 0.84377356,
+      "memory(GiB)": 71.19,
+      "step": 4820,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.72724218,
+      "epoch": 3.78,
+      "learning_rate": 7.330061771741436e-05,
+      "loss": 0.85711832,
+      "memory(GiB)": 71.19,
+      "step": 4825,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.72393856,
+      "epoch": 3.79,
+      "learning_rate": 7.324322502230571e-05,
+      "loss": 0.8321475,
+      "memory(GiB)": 71.19,
+      "step": 4830,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73015528,
+      "epoch": 3.79,
+      "learning_rate": 7.318579323371493e-05,
+      "loss": 0.84410172,
+      "memory(GiB)": 71.19,
+      "step": 4835,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73202195,
+      "epoch": 3.79,
+      "learning_rate": 7.312832244823827e-05,
+      "loss": 0.81751471,
+      "memory(GiB)": 71.19,
+      "step": 4840,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.71658878,
+      "epoch": 3.8,
+      "learning_rate": 7.307081276253761e-05,
+      "loss": 0.87323961,
+      "memory(GiB)": 71.19,
+      "step": 4845,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73308387,
+      "epoch": 3.8,
+      "learning_rate": 7.301326427334019e-05,
+      "loss": 0.82617254,
+      "memory(GiB)": 71.19,
+      "step": 4850,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.725809,
+      "epoch": 3.81,
+      "learning_rate": 7.295567707743856e-05,
+      "loss": 0.8599144,
+      "memory(GiB)": 71.19,
+      "step": 4855,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71556239,
+      "epoch": 3.81,
+      "learning_rate": 7.289805127169038e-05,
+      "loss": 0.8742794,
+      "memory(GiB)": 71.19,
+      "step": 4860,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73000875,
+      "epoch": 3.81,
+      "learning_rate": 7.284038695301823e-05,
+      "loss": 0.83999195,
+      "memory(GiB)": 71.19,
+      "step": 4865,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72756143,
+      "epoch": 3.82,
+      "learning_rate": 7.278268421840944e-05,
+      "loss": 0.84345179,
+      "memory(GiB)": 71.19,
+      "step": 4870,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72185001,
+      "epoch": 3.82,
+      "learning_rate": 7.272494316491602e-05,
+      "loss": 0.84758148,
+      "memory(GiB)": 71.19,
+      "step": 4875,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72773247,
+      "epoch": 3.82,
+      "learning_rate": 7.266716388965437e-05,
+      "loss": 0.83907547,
+      "memory(GiB)": 71.19,
+      "step": 4880,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7324759,
+      "epoch": 3.83,
+      "learning_rate": 7.260934648980521e-05,
+      "loss": 0.81709414,
+      "memory(GiB)": 71.19,
+      "step": 4885,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73284607,
+      "epoch": 3.83,
+      "learning_rate": 7.255149106261339e-05,
+      "loss": 0.81797581,
+      "memory(GiB)": 71.19,
+      "step": 4890,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72334728,
+      "epoch": 3.84,
+      "learning_rate": 7.249359770538764e-05,
+      "loss": 0.83954468,
+      "memory(GiB)": 71.19,
+      "step": 4895,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72269335,
+      "epoch": 3.84,
+      "learning_rate": 7.24356665155006e-05,
+      "loss": 0.84613752,
+      "memory(GiB)": 71.19,
+      "step": 4900,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72381783,
+      "epoch": 3.84,
+      "learning_rate": 7.237769759038846e-05,
+      "loss": 0.83016071,
+      "memory(GiB)": 71.19,
+      "step": 4905,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72248049,
+      "epoch": 3.85,
+      "learning_rate": 7.231969102755093e-05,
+      "loss": 0.85668192,
+      "memory(GiB)": 71.19,
+      "step": 4910,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73048677,
+      "epoch": 3.85,
+      "learning_rate": 7.226164692455098e-05,
+      "loss": 0.84103241,
+      "memory(GiB)": 71.19,
+      "step": 4915,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72759271,
+      "epoch": 3.86,
+      "learning_rate": 7.220356537901474e-05,
+      "loss": 0.82246981,
+      "memory(GiB)": 71.19,
+      "step": 4920,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72148042,
+      "epoch": 3.86,
+      "learning_rate": 7.214544648863131e-05,
+      "loss": 0.85546913,
+      "memory(GiB)": 71.19,
+      "step": 4925,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72802148,
+      "epoch": 3.86,
+      "learning_rate": 7.208729035115264e-05,
+      "loss": 0.84229441,
+      "memory(GiB)": 71.19,
+      "step": 4930,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73225174,
+      "epoch": 3.87,
+      "learning_rate": 7.202909706439326e-05,
+      "loss": 0.82962236,
+      "memory(GiB)": 71.19,
+      "step": 4935,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72838197,
+      "epoch": 3.87,
+      "learning_rate": 7.197086672623023e-05,
+      "loss": 0.84410496,
+      "memory(GiB)": 71.19,
+      "step": 4940,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72434216,
+      "epoch": 3.88,
+      "learning_rate": 7.191259943460292e-05,
+      "loss": 0.85847845,
+      "memory(GiB)": 71.19,
+      "step": 4945,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.71965685,
+      "epoch": 3.88,
+      "learning_rate": 7.185429528751285e-05,
+      "loss": 0.85441608,
+      "memory(GiB)": 71.19,
+      "step": 4950,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.713869,
+      "epoch": 3.88,
+      "learning_rate": 7.179595438302348e-05,
+      "loss": 0.8926239,
+      "memory(GiB)": 71.19,
+      "step": 4955,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72364521,
+      "epoch": 3.89,
+      "learning_rate": 7.173757681926021e-05,
+      "loss": 0.84683857,
+      "memory(GiB)": 71.19,
+      "step": 4960,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.73327866,
+      "epoch": 3.89,
+      "learning_rate": 7.167916269440998e-05,
+      "loss": 0.8079627,
+      "memory(GiB)": 71.19,
+      "step": 4965,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7266232,
+      "epoch": 3.9,
+      "learning_rate": 7.162071210672128e-05,
+      "loss": 0.83985033,
+      "memory(GiB)": 71.19,
+      "step": 4970,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72303486,
+      "epoch": 3.9,
+      "learning_rate": 7.156222515450393e-05,
+      "loss": 0.85938587,
+      "memory(GiB)": 71.19,
+      "step": 4975,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72731833,
+      "epoch": 3.9,
+      "learning_rate": 7.150370193612889e-05,
+      "loss": 0.81108408,
+      "memory(GiB)": 71.19,
+      "step": 4980,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.72412114,
+      "epoch": 3.91,
+      "learning_rate": 7.144514255002813e-05,
+      "loss": 0.86580133,
+      "memory(GiB)": 71.19,
+      "step": 4985,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73388848,
+      "epoch": 3.91,
+      "learning_rate": 7.138654709469446e-05,
+      "loss": 0.82623234,
+      "memory(GiB)": 71.19,
+      "step": 4990,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73632493,
+      "epoch": 3.91,
+      "learning_rate": 7.132791566868133e-05,
+      "loss": 0.83154993,
+      "memory(GiB)": 71.19,
+      "step": 4995,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.71437707,
+      "epoch": 3.92,
+      "learning_rate": 7.126924837060271e-05,
+      "loss": 0.88094139,
+      "memory(GiB)": 71.19,
+      "step": 5000,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.71930361,
+      "epoch": 3.92,
+      "learning_rate": 7.121054529913292e-05,
+      "loss": 0.85584583,
+      "memory(GiB)": 71.19,
+      "step": 5005,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73194895,
+      "epoch": 3.93,
+      "learning_rate": 7.11518065530064e-05,
+      "loss": 0.83459311,
+      "memory(GiB)": 71.19,
+      "step": 5010,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.73548174,
+      "epoch": 3.93,
+      "learning_rate": 7.109303223101765e-05,
+      "loss": 0.82189837,
+      "memory(GiB)": 71.19,
+      "step": 5015,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.72127552,
+      "epoch": 3.93,
+      "learning_rate": 7.103422243202096e-05,
+      "loss": 0.86382265,
+      "memory(GiB)": 71.19,
+      "step": 5020,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.73075724,
+      "epoch": 3.94,
+      "learning_rate": 7.09753772549303e-05,
+      "loss": 0.81052485,
+      "memory(GiB)": 71.19,
+      "step": 5025,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.72500548,
+      "epoch": 3.94,
+      "learning_rate": 7.091649679871914e-05,
+      "loss": 0.84569874,
+      "memory(GiB)": 71.19,
+      "step": 5030,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.72286496,
+      "epoch": 3.95,
+      "learning_rate": 7.085758116242036e-05,
+      "loss": 0.84847412,
+      "memory(GiB)": 71.19,
+      "step": 5035,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.72470217,
+      "epoch": 3.95,
+      "learning_rate": 7.079863044512588e-05,
+      "loss": 0.85448866,
+      "memory(GiB)": 71.19,
+      "step": 5040,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.71964459,
+      "epoch": 3.95,
+      "learning_rate": 7.07396447459867e-05,
+      "loss": 0.86957884,
+      "memory(GiB)": 71.19,
+      "step": 5045,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.72210011,
+      "epoch": 3.96,
+      "learning_rate": 7.06806241642127e-05,
+      "loss": 0.85360498,
+      "memory(GiB)": 71.19,
+      "step": 5050,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.71635909,
+      "epoch": 3.96,
+      "learning_rate": 7.062156879907234e-05,
+      "loss": 0.88525906,
+      "memory(GiB)": 71.19,
+      "step": 5055,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.71934958,
+      "epoch": 3.97,
+      "learning_rate": 7.05624787498926e-05,
+      "loss": 0.86156912,
+      "memory(GiB)": 71.19,
+      "step": 5060,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.7274035,
+      "epoch": 3.97,
+      "learning_rate": 7.050335411605888e-05,
+      "loss": 0.84397354,
+      "memory(GiB)": 71.19,
+      "step": 5065,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.72633142,
+      "epoch": 3.97,
+      "learning_rate": 7.044419499701462e-05,
+      "loss": 0.85779991,
+      "memory(GiB)": 71.19,
+      "step": 5070,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.7241158,
+      "epoch": 3.98,
+      "learning_rate": 7.038500149226138e-05,
+      "loss": 0.84771671,
+      "memory(GiB)": 71.19,
+      "step": 5075,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.71850133,
+      "epoch": 3.98,
+      "learning_rate": 7.032577370135846e-05,
+      "loss": 0.85709019,
+      "memory(GiB)": 71.19,
+      "step": 5080,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.71291666,
+      "epoch": 3.99,
+      "learning_rate": 7.026651172392293e-05,
+      "loss": 0.88760433,
+      "memory(GiB)": 71.19,
+      "step": 5085,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.72612734,
+      "epoch": 3.99,
+      "learning_rate": 7.020721565962925e-05,
+      "loss": 0.8476985,
+      "memory(GiB)": 71.19,
+      "step": 5090,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.72394257,
+      "epoch": 3.99,
+      "learning_rate": 7.014788560820928e-05,
+      "loss": 0.83070583,
+      "memory(GiB)": 71.19,
+      "step": 5095,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "acc": 0.72538056,
+      "epoch": 4.0,
+      "learning_rate": 7.0088521669452e-05,
+      "loss": 0.83622665,
+      "memory(GiB)": 71.19,
+      "step": 5100,
+      "train_speed(iter/s)": 0.024198
+    },
+    {
+      "epoch": 4.0,
+      "eval_acc": 0.7649636864512898,
+      "eval_loss": 0.6868141293525696,
+      "eval_runtime": 107.3501,
+      "eval_samples_per_second": 0.866,
+      "eval_steps_per_second": 0.866,
+      "step": 5103
+    },
+    {
+      "acc": 0.74643955,
+      "epoch": 4.0,
+      "learning_rate": 7.002912394320344e-05,
+      "loss": 0.78384089,
+      "memory(GiB)": 71.19,
+      "step": 5105,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76987772,
+      "epoch": 4.01,
+      "learning_rate": 6.996969252936645e-05,
+      "loss": 0.70066395,
+      "memory(GiB)": 71.19,
+      "step": 5110,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76714759,
+      "epoch": 4.01,
+      "learning_rate": 6.991022752790045e-05,
+      "loss": 0.69841785,
+      "memory(GiB)": 71.19,
+      "step": 5115,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76072993,
+      "epoch": 4.01,
+      "learning_rate": 6.985072903882149e-05,
+      "loss": 0.7248908,
+      "memory(GiB)": 71.19,
+      "step": 5120,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76630473,
+      "epoch": 4.02,
+      "learning_rate": 6.979119716220184e-05,
+      "loss": 0.71821766,
+      "memory(GiB)": 71.19,
+      "step": 5125,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.75970073,
+      "epoch": 4.02,
+      "learning_rate": 6.973163199816998e-05,
+      "loss": 0.71612158,
+      "memory(GiB)": 71.19,
+      "step": 5130,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.77745566,
+      "epoch": 4.02,
+      "learning_rate": 6.967203364691035e-05,
+      "loss": 0.65998745,
+      "memory(GiB)": 71.19,
+      "step": 5135,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.77100053,
+      "epoch": 4.03,
+      "learning_rate": 6.961240220866321e-05,
+      "loss": 0.68110933,
+      "memory(GiB)": 71.19,
+      "step": 5140,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76853023,
+      "epoch": 4.03,
+      "learning_rate": 6.955273778372448e-05,
+      "loss": 0.69694996,
+      "memory(GiB)": 71.19,
+      "step": 5145,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.77470398,
+      "epoch": 4.04,
+      "learning_rate": 6.949304047244557e-05,
+      "loss": 0.67210259,
+      "memory(GiB)": 71.19,
+      "step": 5150,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.78012462,
+      "epoch": 4.04,
+      "learning_rate": 6.943331037523318e-05,
+      "loss": 0.68669949,
+      "memory(GiB)": 71.19,
+      "step": 5155,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.7726408,
+      "epoch": 4.04,
+      "learning_rate": 6.937354759254915e-05,
+      "loss": 0.6830338,
+      "memory(GiB)": 71.19,
+      "step": 5160,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.78194818,
+      "epoch": 4.05,
+      "learning_rate": 6.931375222491035e-05,
+      "loss": 0.66556716,
+      "memory(GiB)": 71.19,
+      "step": 5165,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.7694963,
+      "epoch": 4.05,
+      "learning_rate": 6.925392437288837e-05,
+      "loss": 0.70391464,
+      "memory(GiB)": 71.19,
+      "step": 5170,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76015844,
+      "epoch": 4.06,
+      "learning_rate": 6.919406413710951e-05,
+      "loss": 0.70853,
+      "memory(GiB)": 71.19,
+      "step": 5175,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.78214989,
+      "epoch": 4.06,
+      "learning_rate": 6.91341716182545e-05,
+      "loss": 0.64797435,
+      "memory(GiB)": 71.19,
+      "step": 5180,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76053729,
+      "epoch": 4.06,
+      "learning_rate": 6.907424691705836e-05,
+      "loss": 0.71844401,
+      "memory(GiB)": 71.19,
+      "step": 5185,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76250482,
+      "epoch": 4.07,
+      "learning_rate": 6.90142901343103e-05,
+      "loss": 0.69147439,
+      "memory(GiB)": 71.19,
+      "step": 5190,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.76390204,
+      "epoch": 4.07,
+      "learning_rate": 6.89543013708534e-05,
+      "loss": 0.69992247,
+      "memory(GiB)": 71.19,
+      "step": 5195,
+      "train_speed(iter/s)": 0.024185
+    },
+    {
+      "acc": 0.77699156,
+      "epoch": 4.08,
+      "learning_rate": 6.889428072758458e-05,
+      "loss": 0.65943475,
+      "memory(GiB)": 71.19,
+      "step": 5200,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.77530951,
+      "epoch": 4.08,
+      "learning_rate": 6.883422830545437e-05,
+      "loss": 0.68648615,
+      "memory(GiB)": 71.19,
+      "step": 5205,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.7692956,
+      "epoch": 4.08,
+      "learning_rate": 6.87741442054668e-05,
+      "loss": 0.70791373,
+      "memory(GiB)": 71.19,
+      "step": 5210,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.75982862,
+      "epoch": 4.09,
+      "learning_rate": 6.871402852867906e-05,
+      "loss": 0.72244291,
+      "memory(GiB)": 71.19,
+      "step": 5215,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76754866,
+      "epoch": 4.09,
+      "learning_rate": 6.865388137620156e-05,
+      "loss": 0.71881785,
+      "memory(GiB)": 71.19,
+      "step": 5220,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76008248,
+      "epoch": 4.1,
+      "learning_rate": 6.859370284919762e-05,
+      "loss": 0.72455072,
+      "memory(GiB)": 71.19,
+      "step": 5225,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76211185,
+      "epoch": 4.1,
+      "learning_rate": 6.853349304888331e-05,
+      "loss": 0.71970272,
+      "memory(GiB)": 71.19,
+      "step": 5230,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76919765,
+      "epoch": 4.1,
+      "learning_rate": 6.847325207652733e-05,
+      "loss": 0.69645362,
+      "memory(GiB)": 71.19,
+      "step": 5235,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.78420725,
+      "epoch": 4.11,
+      "learning_rate": 6.841298003345075e-05,
+      "loss": 0.66271205,
+      "memory(GiB)": 71.19,
+      "step": 5240,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.77155137,
+      "epoch": 4.11,
+      "learning_rate": 6.835267702102697e-05,
+      "loss": 0.69231195,
+      "memory(GiB)": 71.19,
+      "step": 5245,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.76921053,
+      "epoch": 4.11,
+      "learning_rate": 6.829234314068143e-05,
+      "loss": 0.69602065,
+      "memory(GiB)": 71.19,
+      "step": 5250,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.7555491,
+      "epoch": 4.12,
+      "learning_rate": 6.823197849389152e-05,
+      "loss": 0.74989166,
+      "memory(GiB)": 71.19,
+      "step": 5255,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.77402582,
+      "epoch": 4.12,
+      "learning_rate": 6.817158318218638e-05,
+      "loss": 0.68813601,
+      "memory(GiB)": 71.19,
+      "step": 5260,
+      "train_speed(iter/s)": 0.024186
+    },
+    {
+      "acc": 0.77681746,
+      "epoch": 4.13,
+      "learning_rate": 6.811115730714665e-05,
+      "loss": 0.66130342,
+      "memory(GiB)": 71.19,
+      "step": 5265,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76329417,
+      "epoch": 4.13,
+      "learning_rate": 6.80507009704045e-05,
+      "loss": 0.72811098,
+      "memory(GiB)": 71.19,
+      "step": 5270,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76749864,
+      "epoch": 4.13,
+      "learning_rate": 6.799021427364324e-05,
+      "loss": 0.69163623,
+      "memory(GiB)": 71.19,
+      "step": 5275,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77502966,
+      "epoch": 4.14,
+      "learning_rate": 6.792969731859727e-05,
+      "loss": 0.68422966,
+      "memory(GiB)": 71.19,
+      "step": 5280,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76088901,
+      "epoch": 4.14,
+      "learning_rate": 6.786915020705189e-05,
+      "loss": 0.72609415,
+      "memory(GiB)": 71.19,
+      "step": 5285,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77280045,
+      "epoch": 4.15,
+      "learning_rate": 6.780857304084309e-05,
+      "loss": 0.67699676,
+      "memory(GiB)": 71.19,
+      "step": 5290,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76591234,
+      "epoch": 4.15,
+      "learning_rate": 6.774796592185746e-05,
+      "loss": 0.71467628,
+      "memory(GiB)": 71.19,
+      "step": 5295,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76299758,
+      "epoch": 4.15,
+      "learning_rate": 6.768732895203196e-05,
+      "loss": 0.71544986,
+      "memory(GiB)": 71.19,
+      "step": 5300,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77150006,
+      "epoch": 4.16,
+      "learning_rate": 6.762666223335372e-05,
+      "loss": 0.69964447,
+      "memory(GiB)": 71.19,
+      "step": 5305,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75832853,
+      "epoch": 4.16,
+      "learning_rate": 6.756596586785992e-05,
+      "loss": 0.74018202,
+      "memory(GiB)": 71.19,
+      "step": 5310,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77024693,
+      "epoch": 4.17,
+      "learning_rate": 6.750523995763762e-05,
+      "loss": 0.70159526,
+      "memory(GiB)": 71.19,
+      "step": 5315,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77265477,
+      "epoch": 4.17,
+      "learning_rate": 6.744448460482357e-05,
+      "loss": 0.65708079,
+      "memory(GiB)": 71.19,
+      "step": 5320,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.7672987,
+      "epoch": 4.17,
+      "learning_rate": 6.738369991160402e-05,
+      "loss": 0.70357451,
+      "memory(GiB)": 71.19,
+      "step": 5325,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.78470359,
+      "epoch": 4.18,
+      "learning_rate": 6.732288598021458e-05,
+      "loss": 0.66129317,
+      "memory(GiB)": 71.19,
+      "step": 5330,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75318675,
+      "epoch": 4.18,
+      "learning_rate": 6.726204291294004e-05,
+      "loss": 0.7573597,
+      "memory(GiB)": 71.19,
+      "step": 5335,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76494775,
+      "epoch": 4.19,
+      "learning_rate": 6.720117081211419e-05,
+      "loss": 0.72770386,
+      "memory(GiB)": 71.19,
+      "step": 5340,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77200723,
+      "epoch": 4.19,
+      "learning_rate": 6.714026978011967e-05,
+      "loss": 0.68737392,
+      "memory(GiB)": 71.19,
+      "step": 5345,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75903749,
+      "epoch": 4.19,
+      "learning_rate": 6.707933991938776e-05,
+      "loss": 0.70887847,
+      "memory(GiB)": 71.19,
+      "step": 5350,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76615915,
+      "epoch": 4.2,
+      "learning_rate": 6.701838133239822e-05,
+      "loss": 0.70868678,
+      "memory(GiB)": 71.19,
+      "step": 5355,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76639304,
+      "epoch": 4.2,
+      "learning_rate": 6.695739412167916e-05,
+      "loss": 0.71408496,
+      "memory(GiB)": 71.19,
+      "step": 5360,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.7674984,
+      "epoch": 4.2,
+      "learning_rate": 6.689637838980678e-05,
+      "loss": 0.69706955,
+      "memory(GiB)": 71.19,
+      "step": 5365,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77202821,
+      "epoch": 4.21,
+      "learning_rate": 6.683533423940531e-05,
+      "loss": 0.69185266,
+      "memory(GiB)": 71.19,
+      "step": 5370,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77602949,
+      "epoch": 4.21,
+      "learning_rate": 6.677426177314675e-05,
+      "loss": 0.66611705,
+      "memory(GiB)": 71.19,
+      "step": 5375,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76406932,
+      "epoch": 4.22,
+      "learning_rate": 6.67131610937507e-05,
+      "loss": 0.70172687,
+      "memory(GiB)": 71.19,
+      "step": 5380,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76476321,
+      "epoch": 4.22,
+      "learning_rate": 6.665203230398425e-05,
+      "loss": 0.6902421,
+      "memory(GiB)": 71.19,
+      "step": 5385,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75160317,
+      "epoch": 4.22,
+      "learning_rate": 6.659087550666177e-05,
+      "loss": 0.74526944,
+      "memory(GiB)": 71.19,
+      "step": 5390,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76145148,
+      "epoch": 4.23,
+      "learning_rate": 6.652969080464472e-05,
+      "loss": 0.73384714,
+      "memory(GiB)": 71.19,
+      "step": 5395,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76717629,
+      "epoch": 4.23,
+      "learning_rate": 6.646847830084148e-05,
+      "loss": 0.69845848,
+      "memory(GiB)": 71.19,
+      "step": 5400,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.78010674,
+      "epoch": 4.24,
+      "learning_rate": 6.640723809820724e-05,
+      "loss": 0.67031679,
+      "memory(GiB)": 71.19,
+      "step": 5405,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76649265,
+      "epoch": 4.24,
+      "learning_rate": 6.634597029974373e-05,
+      "loss": 0.72474589,
+      "memory(GiB)": 71.19,
+      "step": 5410,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77016811,
+      "epoch": 4.24,
+      "learning_rate": 6.628467500849909e-05,
+      "loss": 0.69438682,
+      "memory(GiB)": 71.19,
+      "step": 5415,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76155553,
+      "epoch": 4.25,
+      "learning_rate": 6.622335232756773e-05,
+      "loss": 0.70451875,
+      "memory(GiB)": 71.19,
+      "step": 5420,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75910053,
+      "epoch": 4.25,
+      "learning_rate": 6.616200236009016e-05,
+      "loss": 0.73806357,
+      "memory(GiB)": 71.19,
+      "step": 5425,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76222086,
+      "epoch": 4.26,
+      "learning_rate": 6.61006252092527e-05,
+      "loss": 0.7077971,
+      "memory(GiB)": 71.19,
+      "step": 5430,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76169763,
+      "epoch": 4.26,
+      "learning_rate": 6.603922097828745e-05,
+      "loss": 0.71544447,
+      "memory(GiB)": 71.19,
+      "step": 5435,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77774611,
+      "epoch": 4.26,
+      "learning_rate": 6.597778977047205e-05,
+      "loss": 0.68002701,
+      "memory(GiB)": 71.19,
+      "step": 5440,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76104169,
+      "epoch": 4.27,
+      "learning_rate": 6.591633168912947e-05,
+      "loss": 0.72893152,
+      "memory(GiB)": 71.19,
+      "step": 5445,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77122054,
+      "epoch": 4.27,
+      "learning_rate": 6.585484683762794e-05,
+      "loss": 0.69275894,
+      "memory(GiB)": 71.19,
+      "step": 5450,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77282248,
+      "epoch": 4.28,
+      "learning_rate": 6.57933353193807e-05,
+      "loss": 0.69376159,
+      "memory(GiB)": 71.19,
+      "step": 5455,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7599041,
+      "epoch": 4.28,
+      "learning_rate": 6.57317972378458e-05,
+      "loss": 0.7232058,
+      "memory(GiB)": 71.19,
+      "step": 5460,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.77883954,
+      "epoch": 4.28,
+      "learning_rate": 6.567023269652602e-05,
+      "loss": 0.66660323,
+      "memory(GiB)": 71.19,
+      "step": 5465,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76177578,
+      "epoch": 4.29,
+      "learning_rate": 6.56086417989686e-05,
+      "loss": 0.71664767,
+      "memory(GiB)": 71.19,
+      "step": 5470,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76507282,
+      "epoch": 4.29,
+      "learning_rate": 6.554702464876514e-05,
+      "loss": 0.71534958,
+      "memory(GiB)": 71.19,
+      "step": 5475,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.75966554,
+      "epoch": 4.3,
+      "learning_rate": 6.54853813495514e-05,
+      "loss": 0.71393213,
+      "memory(GiB)": 71.19,
+      "step": 5480,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76525397,
+      "epoch": 4.3,
+      "learning_rate": 6.54237120050071e-05,
+      "loss": 0.70895991,
+      "memory(GiB)": 71.19,
+      "step": 5485,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77907438,
+      "epoch": 4.3,
+      "learning_rate": 6.536201671885575e-05,
+      "loss": 0.69699798,
+      "memory(GiB)": 71.19,
+      "step": 5490,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77636795,
+      "epoch": 4.31,
+      "learning_rate": 6.530029559486455e-05,
+      "loss": 0.67650938,
+      "memory(GiB)": 71.19,
+      "step": 5495,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.75787897,
+      "epoch": 4.31,
+      "learning_rate": 6.523854873684409e-05,
+      "loss": 0.73933854,
+      "memory(GiB)": 71.19,
+      "step": 5500,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75708508,
+      "epoch": 4.31,
+      "learning_rate": 6.517677624864831e-05,
+      "loss": 0.72797713,
+      "memory(GiB)": 71.19,
+      "step": 5505,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76517878,
+      "epoch": 4.32,
+      "learning_rate": 6.511497823417418e-05,
+      "loss": 0.71284537,
+      "memory(GiB)": 71.19,
+      "step": 5510,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77976227,
+      "epoch": 4.32,
+      "learning_rate": 6.50531547973617e-05,
+      "loss": 0.66220131,
+      "memory(GiB)": 71.19,
+      "step": 5515,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.7598443,
+      "epoch": 4.33,
+      "learning_rate": 6.499130604219354e-05,
+      "loss": 0.71064129,
+      "memory(GiB)": 71.19,
+      "step": 5520,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76225853,
+      "epoch": 4.33,
+      "learning_rate": 6.492943207269498e-05,
+      "loss": 0.74577045,
+      "memory(GiB)": 71.19,
+      "step": 5525,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.78424692,
+      "epoch": 4.33,
+      "learning_rate": 6.486753299293375e-05,
+      "loss": 0.66039305,
+      "memory(GiB)": 71.19,
+      "step": 5530,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76662159,
+      "epoch": 4.34,
+      "learning_rate": 6.480560890701976e-05,
+      "loss": 0.71364965,
+      "memory(GiB)": 71.19,
+      "step": 5535,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76090045,
+      "epoch": 4.34,
+      "learning_rate": 6.474365991910501e-05,
+      "loss": 0.70872817,
+      "memory(GiB)": 71.19,
+      "step": 5540,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.75455337,
+      "epoch": 4.35,
+      "learning_rate": 6.468168613338339e-05,
+      "loss": 0.74482007,
+      "memory(GiB)": 71.19,
+      "step": 5545,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.76986208,
+      "epoch": 4.35,
+      "learning_rate": 6.461968765409041e-05,
+      "loss": 0.69489803,
+      "memory(GiB)": 71.19,
+      "step": 5550,
+      "train_speed(iter/s)": 0.024187
+    },
+    {
+      "acc": 0.77070737,
+      "epoch": 4.35,
+      "learning_rate": 6.455766458550329e-05,
+      "loss": 0.71167393,
+      "memory(GiB)": 71.19,
+      "step": 5555,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76613202,
+      "epoch": 4.36,
+      "learning_rate": 6.449561703194042e-05,
+      "loss": 0.70667567,
+      "memory(GiB)": 71.19,
+      "step": 5560,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76189098,
+      "epoch": 4.36,
+      "learning_rate": 6.44335450977615e-05,
+      "loss": 0.71174026,
+      "memory(GiB)": 71.19,
+      "step": 5565,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7586081,
+      "epoch": 4.37,
+      "learning_rate": 6.437144888736715e-05,
+      "loss": 0.7246048,
+      "memory(GiB)": 71.19,
+      "step": 5570,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7720696,
+      "epoch": 4.37,
+      "learning_rate": 6.43093285051989e-05,
+      "loss": 0.69544797,
+      "memory(GiB)": 71.19,
+      "step": 5575,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77282968,
+      "epoch": 4.37,
+      "learning_rate": 6.424718405573888e-05,
+      "loss": 0.6709199,
+      "memory(GiB)": 71.19,
+      "step": 5580,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76675258,
+      "epoch": 4.38,
+      "learning_rate": 6.418501564350972e-05,
+      "loss": 0.70900588,
+      "memory(GiB)": 71.19,
+      "step": 5585,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77312188,
+      "epoch": 4.38,
+      "learning_rate": 6.412282337307436e-05,
+      "loss": 0.66980944,
+      "memory(GiB)": 71.19,
+      "step": 5590,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76313424,
+      "epoch": 4.39,
+      "learning_rate": 6.406060734903582e-05,
+      "loss": 0.71873169,
+      "memory(GiB)": 71.19,
+      "step": 5595,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76170006,
+      "epoch": 4.39,
+      "learning_rate": 6.399836767603715e-05,
+      "loss": 0.73406935,
+      "memory(GiB)": 71.19,
+      "step": 5600,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77519426,
+      "epoch": 4.39,
+      "learning_rate": 6.393610445876113e-05,
+      "loss": 0.70191274,
+      "memory(GiB)": 71.19,
+      "step": 5605,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76001482,
+      "epoch": 4.4,
+      "learning_rate": 6.387381780193014e-05,
+      "loss": 0.70856462,
+      "memory(GiB)": 71.19,
+      "step": 5610,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76338596,
+      "epoch": 4.4,
+      "learning_rate": 6.381150781030597e-05,
+      "loss": 0.70342155,
+      "memory(GiB)": 71.19,
+      "step": 5615,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.77544284,
+      "epoch": 4.4,
+      "learning_rate": 6.37491745886897e-05,
+      "loss": 0.68927674,
+      "memory(GiB)": 71.19,
+      "step": 5620,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7679841,
+      "epoch": 4.41,
+      "learning_rate": 6.368681824192147e-05,
+      "loss": 0.69150701,
+      "memory(GiB)": 71.19,
+      "step": 5625,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.7673996,
+      "epoch": 4.41,
+      "learning_rate": 6.362443887488026e-05,
+      "loss": 0.7154283,
+      "memory(GiB)": 71.19,
+      "step": 5630,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76109505,
+      "epoch": 4.42,
+      "learning_rate": 6.356203659248386e-05,
+      "loss": 0.72742548,
+      "memory(GiB)": 71.19,
+      "step": 5635,
+      "train_speed(iter/s)": 0.024188
+    },
+    {
+      "acc": 0.76917353,
+      "epoch": 4.42,
+      "learning_rate": 6.349961149968849e-05,
+      "loss": 0.71016984,
+      "memory(GiB)": 71.19,
+      "step": 5640,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.77111497,
+      "epoch": 4.42,
+      "learning_rate": 6.343716370148887e-05,
+      "loss": 0.70852461,
+      "memory(GiB)": 71.19,
+      "step": 5645,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76087008,
+      "epoch": 4.43,
+      "learning_rate": 6.337469330291778e-05,
+      "loss": 0.71220889,
+      "memory(GiB)": 71.19,
+      "step": 5650,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7703217,
+      "epoch": 4.43,
+      "learning_rate": 6.331220040904612e-05,
+      "loss": 0.69431052,
+      "memory(GiB)": 71.19,
+      "step": 5655,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76420479,
+      "epoch": 4.44,
+      "learning_rate": 6.324968512498255e-05,
+      "loss": 0.71943445,
+      "memory(GiB)": 71.19,
+      "step": 5660,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76351399,
+      "epoch": 4.44,
+      "learning_rate": 6.318714755587341e-05,
+      "loss": 0.72373419,
+      "memory(GiB)": 71.19,
+      "step": 5665,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76989708,
+      "epoch": 4.44,
+      "learning_rate": 6.312458780690253e-05,
+      "loss": 0.70568027,
+      "memory(GiB)": 71.19,
+      "step": 5670,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.7618772,
+      "epoch": 4.45,
+      "learning_rate": 6.306200598329105e-05,
+      "loss": 0.72609386,
+      "memory(GiB)": 71.19,
+      "step": 5675,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76857285,
+      "epoch": 4.45,
+      "learning_rate": 6.299940219029722e-05,
+      "loss": 0.7043541,
+      "memory(GiB)": 71.19,
+      "step": 5680,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.7670013,
+      "epoch": 4.46,
+      "learning_rate": 6.293677653321624e-05,
+      "loss": 0.69272251,
+      "memory(GiB)": 71.19,
+      "step": 5685,
+      "train_speed(iter/s)": 0.024189
+    },
+    {
+      "acc": 0.76531138,
+      "epoch": 4.46,
+      "learning_rate": 6.287412911738013e-05,
+      "loss": 0.72677112,
+      "memory(GiB)": 71.19,
+      "step": 5690,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76662335,
+      "epoch": 4.46,
+      "learning_rate": 6.281146004815743e-05,
+      "loss": 0.71060648,
+      "memory(GiB)": 71.19,
+      "step": 5695,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76985221,
+      "epoch": 4.47,
+      "learning_rate": 6.274876943095316e-05,
+      "loss": 0.69214835,
+      "memory(GiB)": 71.19,
+      "step": 5700,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.75513039,
+      "epoch": 4.47,
+      "learning_rate": 6.268605737120856e-05,
+      "loss": 0.73119974,
+      "memory(GiB)": 71.19,
+      "step": 5705,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7597374,
+      "epoch": 4.48,
+      "learning_rate": 6.262332397440094e-05,
+      "loss": 0.71646504,
+      "memory(GiB)": 71.19,
+      "step": 5710,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76583333,
+      "epoch": 4.48,
+      "learning_rate": 6.256056934604348e-05,
+      "loss": 0.71394172,
+      "memory(GiB)": 71.19,
+      "step": 5715,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.77272115,
+      "epoch": 4.48,
+      "learning_rate": 6.24977935916851e-05,
+      "loss": 0.69063115,
+      "memory(GiB)": 71.19,
+      "step": 5720,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76995649,
+      "epoch": 4.49,
+      "learning_rate": 6.243499681691024e-05,
+      "loss": 0.695227,
+      "memory(GiB)": 71.19,
+      "step": 5725,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.75672503,
+      "epoch": 4.49,
+      "learning_rate": 6.237217912733869e-05,
+      "loss": 0.73567729,
+      "memory(GiB)": 71.19,
+      "step": 5730,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76756783,
+      "epoch": 4.49,
+      "learning_rate": 6.23093406286254e-05,
+      "loss": 0.69406104,
+      "memory(GiB)": 71.19,
+      "step": 5735,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.78130131,
+      "epoch": 4.5,
+      "learning_rate": 6.224648142646037e-05,
+      "loss": 0.68319197,
+      "memory(GiB)": 71.19,
+      "step": 5740,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75013237,
+      "epoch": 4.5,
+      "learning_rate": 6.218360162656836e-05,
+      "loss": 0.73892965,
+      "memory(GiB)": 71.19,
+      "step": 5745,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76402731,
+      "epoch": 4.51,
+      "learning_rate": 6.212070133470884e-05,
+      "loss": 0.71493678,
+      "memory(GiB)": 71.19,
+      "step": 5750,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76769323,
+      "epoch": 4.51,
+      "learning_rate": 6.205778065667566e-05,
+      "loss": 0.71724548,
+      "memory(GiB)": 71.19,
+      "step": 5755,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75835562,
+      "epoch": 4.51,
+      "learning_rate": 6.199483969829705e-05,
+      "loss": 0.7360487,
+      "memory(GiB)": 71.19,
+      "step": 5760,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76376677,
+      "epoch": 4.52,
+      "learning_rate": 6.19318785654353e-05,
+      "loss": 0.71627126,
+      "memory(GiB)": 71.19,
+      "step": 5765,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76958847,
+      "epoch": 4.52,
+      "learning_rate": 6.186889736398664e-05,
+      "loss": 0.71689034,
+      "memory(GiB)": 71.19,
+      "step": 5770,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75499663,
+      "epoch": 4.53,
+      "learning_rate": 6.180589619988103e-05,
+      "loss": 0.74792767,
+      "memory(GiB)": 71.19,
+      "step": 5775,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7811924,
+      "epoch": 4.53,
+      "learning_rate": 6.174287517908207e-05,
+      "loss": 0.6795465,
+      "memory(GiB)": 71.19,
+      "step": 5780,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76298246,
+      "epoch": 4.53,
+      "learning_rate": 6.167983440758672e-05,
+      "loss": 0.71815233,
+      "memory(GiB)": 71.19,
+      "step": 5785,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7640779,
+      "epoch": 4.54,
+      "learning_rate": 6.16167739914251e-05,
+      "loss": 0.71062479,
+      "memory(GiB)": 71.19,
+      "step": 5790,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76794705,
+      "epoch": 4.54,
+      "learning_rate": 6.15536940366605e-05,
+      "loss": 0.70243373,
+      "memory(GiB)": 71.19,
+      "step": 5795,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76325617,
+      "epoch": 4.55,
+      "learning_rate": 6.149059464938893e-05,
+      "loss": 0.70105352,
+      "memory(GiB)": 71.19,
+      "step": 5800,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.77031732,
+      "epoch": 4.55,
+      "learning_rate": 6.142747593573922e-05,
+      "loss": 0.70099783,
+      "memory(GiB)": 71.19,
+      "step": 5805,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76253133,
+      "epoch": 4.55,
+      "learning_rate": 6.136433800187262e-05,
+      "loss": 0.72461586,
+      "memory(GiB)": 71.19,
+      "step": 5810,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.77606282,
+      "epoch": 4.56,
+      "learning_rate": 6.130118095398269e-05,
+      "loss": 0.67780962,
+      "memory(GiB)": 71.19,
+      "step": 5815,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7617002,
+      "epoch": 4.56,
+      "learning_rate": 6.123800489829523e-05,
+      "loss": 0.70110459,
+      "memory(GiB)": 71.19,
+      "step": 5820,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76943331,
+      "epoch": 4.57,
+      "learning_rate": 6.117480994106793e-05,
+      "loss": 0.71336999,
+      "memory(GiB)": 71.19,
+      "step": 5825,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75526834,
+      "epoch": 4.57,
+      "learning_rate": 6.111159618859031e-05,
+      "loss": 0.74314675,
+      "memory(GiB)": 71.19,
+      "step": 5830,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76729212,
+      "epoch": 4.57,
+      "learning_rate": 6.104836374718347e-05,
+      "loss": 0.6993835,
+      "memory(GiB)": 71.19,
+      "step": 5835,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.77066221,
+      "epoch": 4.58,
+      "learning_rate": 6.0985112723199976e-05,
+      "loss": 0.68562093,
+      "memory(GiB)": 71.19,
+      "step": 5840,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76890574,
+      "epoch": 4.58,
+      "learning_rate": 6.0921843223023634e-05,
+      "loss": 0.70060396,
+      "memory(GiB)": 71.19,
+      "step": 5845,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76268287,
+      "epoch": 4.59,
+      "learning_rate": 6.085855535306931e-05,
+      "loss": 0.71980524,
+      "memory(GiB)": 71.19,
+      "step": 5850,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.7607635,
+      "epoch": 4.59,
+      "learning_rate": 6.0795249219782814e-05,
+      "loss": 0.7061645,
+      "memory(GiB)": 71.19,
+      "step": 5855,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76794925,
+      "epoch": 4.59,
+      "learning_rate": 6.0731924929640614e-05,
+      "loss": 0.70129228,
+      "memory(GiB)": 71.19,
+      "step": 5860,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76963058,
+      "epoch": 4.6,
+      "learning_rate": 6.066858258914978e-05,
+      "loss": 0.71304092,
+      "memory(GiB)": 71.19,
+      "step": 5865,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76559944,
+      "epoch": 4.6,
+      "learning_rate": 6.060522230484769e-05,
+      "loss": 0.70504079,
+      "memory(GiB)": 71.19,
+      "step": 5870,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.77082806,
+      "epoch": 4.6,
+      "learning_rate": 6.054184418330191e-05,
+      "loss": 0.6952466,
+      "memory(GiB)": 71.19,
+      "step": 5875,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76205645,
+      "epoch": 4.61,
+      "learning_rate": 6.0478448331110015e-05,
+      "loss": 0.71116371,
+      "memory(GiB)": 71.19,
+      "step": 5880,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.77297888,
+      "epoch": 4.61,
+      "learning_rate": 6.041503485489942e-05,
+      "loss": 0.68239374,
+      "memory(GiB)": 71.19,
+      "step": 5885,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76356063,
+      "epoch": 4.62,
+      "learning_rate": 6.035160386132718e-05,
+      "loss": 0.70998626,
+      "memory(GiB)": 71.19,
+      "step": 5890,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76024551,
+      "epoch": 4.62,
+      "learning_rate": 6.0288155457079754e-05,
+      "loss": 0.72426705,
+      "memory(GiB)": 71.19,
+      "step": 5895,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76642303,
+      "epoch": 4.62,
+      "learning_rate": 6.022468974887295e-05,
+      "loss": 0.72213707,
+      "memory(GiB)": 71.19,
+      "step": 5900,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.77551522,
+      "epoch": 4.63,
+      "learning_rate": 6.016120684345167e-05,
+      "loss": 0.68373032,
+      "memory(GiB)": 71.19,
+      "step": 5905,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76769862,
+      "epoch": 4.63,
+      "learning_rate": 6.009770684758973e-05,
+      "loss": 0.71322379,
+      "memory(GiB)": 71.19,
+      "step": 5910,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76745868,
+      "epoch": 4.64,
+      "learning_rate": 6.0034189868089677e-05,
+      "loss": 0.71255426,
+      "memory(GiB)": 71.19,
+      "step": 5915,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76841478,
+      "epoch": 4.64,
+      "learning_rate": 5.9970656011782646e-05,
+      "loss": 0.70099401,
+      "memory(GiB)": 71.19,
+      "step": 5920,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76973267,
+      "epoch": 4.64,
+      "learning_rate": 5.990710538552813e-05,
+      "loss": 0.6992341,
+      "memory(GiB)": 71.19,
+      "step": 5925,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.74826236,
+      "epoch": 4.65,
+      "learning_rate": 5.984353809621388e-05,
+      "loss": 0.74642177,
+      "memory(GiB)": 71.19,
+      "step": 5930,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7560431,
+      "epoch": 4.65,
+      "learning_rate": 5.977995425075562e-05,
+      "loss": 0.72930694,
+      "memory(GiB)": 71.19,
+      "step": 5935,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76061263,
+      "epoch": 4.66,
+      "learning_rate": 5.971635395609694e-05,
+      "loss": 0.73713555,
+      "memory(GiB)": 71.19,
+      "step": 5940,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.75073433,
+      "epoch": 4.66,
+      "learning_rate": 5.965273731920908e-05,
+      "loss": 0.74310646,
+      "memory(GiB)": 71.19,
+      "step": 5945,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76606636,
+      "epoch": 4.66,
+      "learning_rate": 5.958910444709083e-05,
+      "loss": 0.7026166,
+      "memory(GiB)": 71.19,
+      "step": 5950,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.74374471,
+      "epoch": 4.67,
+      "learning_rate": 5.9525455446768194e-05,
+      "loss": 0.77414222,
+      "memory(GiB)": 71.19,
+      "step": 5955,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7663578,
+      "epoch": 4.67,
+      "learning_rate": 5.9461790425294375e-05,
+      "loss": 0.71838512,
+      "memory(GiB)": 71.19,
+      "step": 5960,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.76450734,
+      "epoch": 4.68,
+      "learning_rate": 5.939810948974948e-05,
+      "loss": 0.72742691,
+      "memory(GiB)": 71.19,
+      "step": 5965,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.7788063,
+      "epoch": 4.68,
+      "learning_rate": 5.933441274724041e-05,
+      "loss": 0.66779647,
+      "memory(GiB)": 71.19,
+      "step": 5970,
+      "train_speed(iter/s)": 0.02419
+    },
+    {
+      "acc": 0.77166514,
+      "epoch": 4.68,
+      "learning_rate": 5.927070030490062e-05,
+      "loss": 0.68094988,
+      "memory(GiB)": 71.19,
+      "step": 5975,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76896586,
+      "epoch": 4.69,
+      "learning_rate": 5.9206972269890014e-05,
+      "loss": 0.68136163,
+      "memory(GiB)": 71.19,
+      "step": 5980,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76952062,
+      "epoch": 4.69,
+      "learning_rate": 5.914322874939466e-05,
+      "loss": 0.7040164,
+      "memory(GiB)": 71.19,
+      "step": 5985,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75679116,
+      "epoch": 4.69,
+      "learning_rate": 5.907946985062678e-05,
+      "loss": 0.73868618,
+      "memory(GiB)": 71.19,
+      "step": 5990,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76209521,
+      "epoch": 4.7,
+      "learning_rate": 5.9015695680824325e-05,
+      "loss": 0.72485847,
+      "memory(GiB)": 71.19,
+      "step": 5995,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76353059,
+      "epoch": 4.7,
+      "learning_rate": 5.8951906347251027e-05,
+      "loss": 0.71688752,
+      "memory(GiB)": 71.19,
+      "step": 6000,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75670443,
+      "epoch": 4.71,
+      "learning_rate": 5.888810195719609e-05,
+      "loss": 0.73053293,
+      "memory(GiB)": 71.19,
+      "step": 6005,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75863943,
+      "epoch": 4.71,
+      "learning_rate": 5.8824282617974045e-05,
+      "loss": 0.70380878,
+      "memory(GiB)": 71.19,
+      "step": 6010,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76119242,
+      "epoch": 4.71,
+      "learning_rate": 5.876044843692456e-05,
+      "loss": 0.7133338,
+      "memory(GiB)": 71.19,
+      "step": 6015,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76129284,
+      "epoch": 4.72,
+      "learning_rate": 5.869659952141228e-05,
+      "loss": 0.71880956,
+      "memory(GiB)": 71.19,
+      "step": 6020,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76461582,
+      "epoch": 4.72,
+      "learning_rate": 5.8632735978826626e-05,
+      "loss": 0.71644011,
+      "memory(GiB)": 71.19,
+      "step": 6025,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76767044,
+      "epoch": 4.73,
+      "learning_rate": 5.856885791658158e-05,
+      "loss": 0.69806213,
+      "memory(GiB)": 71.19,
+      "step": 6030,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.75022593,
+      "epoch": 4.73,
+      "learning_rate": 5.8504965442115644e-05,
+      "loss": 0.75385232,
+      "memory(GiB)": 71.19,
+      "step": 6035,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.75788932,
+      "epoch": 4.73,
+      "learning_rate": 5.844105866289147e-05,
+      "loss": 0.73194909,
+      "memory(GiB)": 71.19,
+      "step": 6040,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.75529494,
+      "epoch": 4.74,
+      "learning_rate": 5.83771376863958e-05,
+      "loss": 0.72789454,
+      "memory(GiB)": 71.19,
+      "step": 6045,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.77266655,
+      "epoch": 4.74,
+      "learning_rate": 5.831320262013926e-05,
+      "loss": 0.6826066,
+      "memory(GiB)": 71.19,
+      "step": 6050,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.75812349,
+      "epoch": 4.75,
+      "learning_rate": 5.824925357165617e-05,
+      "loss": 0.73104687,
+      "memory(GiB)": 71.19,
+      "step": 6055,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.76626301,
+      "epoch": 4.75,
+      "learning_rate": 5.818529064850436e-05,
+      "loss": 0.7082016,
+      "memory(GiB)": 71.19,
+      "step": 6060,
+      "train_speed(iter/s)": 0.024191
+    },
+    {
+      "acc": 0.7526711,
+      "epoch": 4.75,
+      "learning_rate": 5.8121313958265e-05,
+      "loss": 0.74757395,
+      "memory(GiB)": 71.19,
+      "step": 6065,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76689634,
+      "epoch": 4.76,
+      "learning_rate": 5.8057323608542425e-05,
+      "loss": 0.70414853,
+      "memory(GiB)": 71.19,
+      "step": 6070,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.75855989,
+      "epoch": 4.76,
+      "learning_rate": 5.799331970696393e-05,
+      "loss": 0.73491292,
+      "memory(GiB)": 71.19,
+      "step": 6075,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.74774814,
+      "epoch": 4.77,
+      "learning_rate": 5.792930236117964e-05,
+      "loss": 0.76034083,
+      "memory(GiB)": 71.19,
+      "step": 6080,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.74793382,
+      "epoch": 4.77,
+      "learning_rate": 5.786527167886221e-05,
+      "loss": 0.76183257,
+      "memory(GiB)": 71.19,
+      "step": 6085,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76219697,
+      "epoch": 4.77,
+      "learning_rate": 5.78012277677068e-05,
+      "loss": 0.73324904,
+      "memory(GiB)": 71.19,
+      "step": 6090,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.76384172,
+      "epoch": 4.78,
+      "learning_rate": 5.7737170735430825e-05,
+      "loss": 0.71691432,
+      "memory(GiB)": 71.19,
+      "step": 6095,
+      "train_speed(iter/s)": 0.024192
+    },
+    {
+      "acc": 0.74610209,
+      "epoch": 4.78,
+      "learning_rate": 5.7673100689773707e-05,
+      "loss": 0.75250793,
+      "memory(GiB)": 71.19,
+      "step": 6100,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.75925684,
+      "epoch": 4.78,
+      "learning_rate": 5.760901773849682e-05,
+      "loss": 0.71436172,
+      "memory(GiB)": 71.19,
+      "step": 6105,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.76955056,
+      "epoch": 4.79,
+      "learning_rate": 5.754492198938318e-05,
+      "loss": 0.69148941,
+      "memory(GiB)": 71.19,
+      "step": 6110,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.75973935,
+      "epoch": 4.79,
+      "learning_rate": 5.748081355023739e-05,
+      "loss": 0.7391295,
+      "memory(GiB)": 71.19,
+      "step": 6115,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.75714288,
+      "epoch": 4.8,
+      "learning_rate": 5.741669252888535e-05,
+      "loss": 0.71268778,
+      "memory(GiB)": 71.19,
+      "step": 6120,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.77134218,
+      "epoch": 4.8,
+      "learning_rate": 5.735255903317417e-05,
+      "loss": 0.69459882,
+      "memory(GiB)": 71.19,
+      "step": 6125,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.76627364,
+      "epoch": 4.8,
+      "learning_rate": 5.72884131709719e-05,
+      "loss": 0.69015708,
+      "memory(GiB)": 71.19,
+      "step": 6130,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.75454583,
+      "epoch": 4.81,
+      "learning_rate": 5.7224255050167394e-05,
+      "loss": 0.75072398,
+      "memory(GiB)": 71.19,
+      "step": 6135,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.77098956,
+      "epoch": 4.81,
+      "learning_rate": 5.7160084778670156e-05,
+      "loss": 0.69697561,
+      "memory(GiB)": 71.19,
+      "step": 6140,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.76857743,
+      "epoch": 4.82,
+      "learning_rate": 5.709590246441007e-05,
+      "loss": 0.71755748,
+      "memory(GiB)": 71.19,
+      "step": 6145,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.75801721,
+      "epoch": 4.82,
+      "learning_rate": 5.703170821533733e-05,
+      "loss": 0.71778822,
+      "memory(GiB)": 71.19,
+      "step": 6150,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.76996355,
+      "epoch": 4.82,
+      "learning_rate": 5.696750213942218e-05,
+      "loss": 0.67777882,
+      "memory(GiB)": 71.19,
+      "step": 6155,
+      "train_speed(iter/s)": 0.024193
+    },
+    {
+      "acc": 0.7731328,
+      "epoch": 4.83,
+      "learning_rate": 5.690328434465475e-05,
+      "loss": 0.70121741,
+      "memory(GiB)": 71.19,
+      "step": 6160,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.75167289,
+      "epoch": 4.83,
+      "learning_rate": 5.68390549390449e-05,
+      "loss": 0.73404212,
+      "memory(GiB)": 71.19,
+      "step": 6165,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76081247,
+      "epoch": 4.84,
+      "learning_rate": 5.677481403062199e-05,
+      "loss": 0.72303152,
+      "memory(GiB)": 71.19,
+      "step": 6170,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77228723,
+      "epoch": 4.84,
+      "learning_rate": 5.671056172743479e-05,
+      "loss": 0.68236284,
+      "memory(GiB)": 71.19,
+      "step": 6175,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76532092,
+      "epoch": 4.84,
+      "learning_rate": 5.6646298137551125e-05,
+      "loss": 0.70234327,
+      "memory(GiB)": 71.19,
+      "step": 6180,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76404386,
+      "epoch": 4.85,
+      "learning_rate": 5.658202336905791e-05,
+      "loss": 0.71822534,
+      "memory(GiB)": 71.19,
+      "step": 6185,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.75237103,
+      "epoch": 4.85,
+      "learning_rate": 5.651773753006081e-05,
+      "loss": 0.74845624,
+      "memory(GiB)": 71.19,
+      "step": 6190,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7637733,
+      "epoch": 4.86,
+      "learning_rate": 5.64534407286841e-05,
+      "loss": 0.71927757,
+      "memory(GiB)": 71.19,
+      "step": 6195,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.7700985,
+      "epoch": 4.86,
+      "learning_rate": 5.638913307307055e-05,
+      "loss": 0.71858625,
+      "memory(GiB)": 71.19,
+      "step": 6200,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76865525,
+      "epoch": 4.86,
+      "learning_rate": 5.632481467138111e-05,
+      "loss": 0.70387897,
+      "memory(GiB)": 71.19,
+      "step": 6205,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76575408,
+      "epoch": 4.87,
+      "learning_rate": 5.626048563179487e-05,
+      "loss": 0.6937983,
+      "memory(GiB)": 71.19,
+      "step": 6210,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76595235,
+      "epoch": 4.87,
+      "learning_rate": 5.619614606250877e-05,
+      "loss": 0.70782719,
+      "memory(GiB)": 71.19,
+      "step": 6215,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77544599,
+      "epoch": 4.88,
+      "learning_rate": 5.613179607173748e-05,
+      "loss": 0.70953703,
+      "memory(GiB)": 71.19,
+      "step": 6220,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76125436,
+      "epoch": 4.88,
+      "learning_rate": 5.6067435767713184e-05,
+      "loss": 0.71401358,
+      "memory(GiB)": 71.19,
+      "step": 6225,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77986512,
+      "epoch": 4.88,
+      "learning_rate": 5.600306525868544e-05,
+      "loss": 0.66382799,
+      "memory(GiB)": 71.19,
+      "step": 6230,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77460809,
+      "epoch": 4.89,
+      "learning_rate": 5.593868465292094e-05,
+      "loss": 0.70628066,
+      "memory(GiB)": 71.19,
+      "step": 6235,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77852488,
+      "epoch": 4.89,
+      "learning_rate": 5.5874294058703346e-05,
+      "loss": 0.68490057,
+      "memory(GiB)": 71.19,
+      "step": 6240,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76132755,
+      "epoch": 4.89,
+      "learning_rate": 5.58098935843332e-05,
+      "loss": 0.71025805,
+      "memory(GiB)": 71.19,
+      "step": 6245,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76081877,
+      "epoch": 4.9,
+      "learning_rate": 5.574548333812753e-05,
+      "loss": 0.70690475,
+      "memory(GiB)": 71.19,
+      "step": 6250,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.76902518,
+      "epoch": 4.9,
+      "learning_rate": 5.5681063428419944e-05,
+      "loss": 0.69533243,
+      "memory(GiB)": 71.19,
+      "step": 6255,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.77235093,
+      "epoch": 4.91,
+      "learning_rate": 5.5616633963560194e-05,
+      "loss": 0.69402099,
+      "memory(GiB)": 71.19,
+      "step": 6260,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.75550814,
+      "epoch": 4.91,
+      "learning_rate": 5.555219505191417e-05,
+      "loss": 0.7272469,
+      "memory(GiB)": 71.19,
+      "step": 6265,
+      "train_speed(iter/s)": 0.024194
+    },
+    {
+      "acc": 0.75596237,
+      "epoch": 4.91,
+      "learning_rate": 5.548774680186358e-05,
+      "loss": 0.74081788,
+      "memory(GiB)": 71.19,
+      "step": 6270,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.7628664,
+      "epoch": 4.92,
+      "learning_rate": 5.5423289321805914e-05,
+      "loss": 0.71786399,
+      "memory(GiB)": 71.19,
+      "step": 6275,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.76951499,
+      "epoch": 4.92,
+      "learning_rate": 5.535882272015417e-05,
+      "loss": 0.70130796,
+      "memory(GiB)": 71.19,
+      "step": 6280,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.75305843,
+      "epoch": 4.93,
+      "learning_rate": 5.529434710533664e-05,
+      "loss": 0.74882407,
+      "memory(GiB)": 71.19,
+      "step": 6285,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.76734796,
+      "epoch": 4.93,
+      "learning_rate": 5.522986258579681e-05,
+      "loss": 0.71578069,
+      "memory(GiB)": 71.19,
+      "step": 6290,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.76613975,
+      "epoch": 4.93,
+      "learning_rate": 5.516536926999316e-05,
+      "loss": 0.70828838,
+      "memory(GiB)": 71.19,
+      "step": 6295,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.77031693,
+      "epoch": 4.94,
+      "learning_rate": 5.510086726639894e-05,
+      "loss": 0.70443668,
+      "memory(GiB)": 71.19,
+      "step": 6300,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.75785489,
+      "epoch": 4.94,
+      "learning_rate": 5.5036356683502e-05,
+      "loss": 0.7272397,
+      "memory(GiB)": 71.19,
+      "step": 6305,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.77772474,
+      "epoch": 4.95,
+      "learning_rate": 5.497183762980467e-05,
+      "loss": 0.67617226,
+      "memory(GiB)": 71.19,
+      "step": 6310,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.76613221,
+      "epoch": 4.95,
+      "learning_rate": 5.490731021382346e-05,
+      "loss": 0.71000295,
+      "memory(GiB)": 71.19,
+      "step": 6315,
+      "train_speed(iter/s)": 0.024195
+    },
+    {
+      "acc": 0.74996147,
+      "epoch": 4.95,
+      "learning_rate": 5.4842774544089e-05,
+      "loss": 0.75001779,
+      "memory(GiB)": 71.19,
+      "step": 6320,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.76159739,
+      "epoch": 4.96,
+      "learning_rate": 5.477823072914579e-05,
+      "loss": 0.71828237,
+      "memory(GiB)": 71.19,
+      "step": 6325,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.74771967,
+      "epoch": 4.96,
+      "learning_rate": 5.4713678877552e-05,
+      "loss": 0.77577434,
+      "memory(GiB)": 71.19,
+      "step": 6330,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.76517348,
+      "epoch": 4.97,
+      "learning_rate": 5.464911909787934e-05,
+      "loss": 0.72100554,
+      "memory(GiB)": 71.19,
+      "step": 6335,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.76637516,
+      "epoch": 4.97,
+      "learning_rate": 5.4584551498712865e-05,
+      "loss": 0.70268941,
+      "memory(GiB)": 71.19,
+      "step": 6340,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.76800013,
+      "epoch": 4.97,
+      "learning_rate": 5.451997618865078e-05,
+      "loss": 0.72028384,
+      "memory(GiB)": 71.19,
+      "step": 6345,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.76661825,
+      "epoch": 4.98,
+      "learning_rate": 5.445539327630422e-05,
+      "loss": 0.69725647,
+      "memory(GiB)": 71.19,
+      "step": 6350,
+      "train_speed(iter/s)": 0.024196
+    },
+    {
+      "acc": 0.75148797,
+      "epoch": 4.98,
+      "learning_rate": 5.439080287029716e-05,
+      "loss": 0.73803744,
+      "memory(GiB)": 71.19,
+      "step": 6355,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.75281067,
+      "epoch": 4.98,
+      "learning_rate": 5.432620507926616e-05,
+      "loss": 0.75406656,
+      "memory(GiB)": 71.19,
+      "step": 6360,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.76611757,
+      "epoch": 4.99,
+      "learning_rate": 5.426160001186016e-05,
+      "loss": 0.70822902,
+      "memory(GiB)": 71.19,
+      "step": 6365,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.77433181,
+      "epoch": 4.99,
+      "learning_rate": 5.41969877767404e-05,
+      "loss": 0.68499122,
+      "memory(GiB)": 71.19,
+      "step": 6370,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "acc": 0.7715888,
+      "epoch": 5.0,
+      "learning_rate": 5.413236848258015e-05,
+      "loss": 0.70902772,
+      "memory(GiB)": 71.19,
+      "step": 6375,
+      "train_speed(iter/s)": 0.024197
+    },
+    {
+      "epoch": 5.0,
+      "eval_acc": 0.7930127723516154,
+      "eval_loss": 0.5855495929718018,
+      "eval_runtime": 107.3505,
+      "eval_samples_per_second": 0.866,
+      "eval_steps_per_second": 0.866,
+      "step": 6379
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 12750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 1,
+  "total_flos": 2.781799283988169e+22,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}