Upload folder using huggingface_hub

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80e513b84eeccc5268e3e022d6b7d961f5d0480f69cdba1d8ca6f4db8a36bc00
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc92917acb0abc2856ec04700e2d4da5dd8a86f4e82838b05b2129daeacbde3c
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:458ad12cedb10e3a955e66056d0f3730143d44b6bb1fb5306ad747bc71bfa20e
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:804b27e3fbff0e564084607818ce25e48977f7fdb57dee9f76594bd3a1e0832c
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:606a4a8472cbe270c7409b19f2959250d5082ba7bf74b2abb6f740079a5409b1
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cec74e94c6ec532a705641951b20514b57883dc4b28c512ba7e88aacf1452720
+size 646064562

qwen_checkpoints/client_states_QWENv2_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario17_new_10000_random0_25_seed1/server_model_round6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12afdbfef398f6d5619205ecfb3b6da949b69f93f8471de3e73c110ec9cc2a0c
+size 646064562

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.0625_seed1/0_trainer_state.json

@@ -0,0 +1,3781 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 1250,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0016,
+      "learning_rate": 2.3755748898855234e-08,
+      "loss": 1.5344,
+      "step": 2
+    },
+    {
+      "epoch": 0.0032,
+      "learning_rate": 2.4118320376915448e-08,
+      "loss": 1.3493,
+      "step": 4
+    },
+    {
+      "epoch": 0.0048,
+      "learning_rate": 2.4483258812684062e-08,
+      "loss": 1.531,
+      "step": 6
+    },
+    {
+      "epoch": 0.0064,
+      "learning_rate": 2.4850552822730343e-08,
+      "loss": 2.7436,
+      "step": 8
+    },
+    {
+      "epoch": 0.008,
+      "learning_rate": 2.5220190950146858e-08,
+      "loss": 1.1871,
+      "step": 10
+    },
+    {
+      "epoch": 0.0096,
+      "learning_rate": 2.5592161664906243e-08,
+      "loss": 1.359,
+      "step": 12
+    },
+    {
+      "epoch": 0.0112,
+      "learning_rate": 2.5966453364222187e-08,
+      "loss": 1.5821,
+      "step": 14
+    },
+    {
+      "epoch": 0.0128,
+      "learning_rate": 2.6343054372909647e-08,
+      "loss": 0.7749,
+      "step": 16
+    },
+    {
+      "epoch": 0.0144,
+      "learning_rate": 2.6721952943750392e-08,
+      "loss": 1.513,
+      "step": 18
+    },
+    {
+      "epoch": 0.016,
+      "learning_rate": 2.7103137257858887e-08,
+      "loss": 1.5182,
+      "step": 20
+    },
+    {
+      "epoch": 0.0176,
+      "learning_rate": 2.7486595425050563e-08,
+      "loss": 1.1992,
+      "step": 22
+    },
+    {
+      "epoch": 0.0192,
+      "learning_rate": 2.787231548421395e-08,
+      "loss": 1.304,
+      "step": 24
+    },
+    {
+      "epoch": 0.0208,
+      "learning_rate": 2.8260285403682116e-08,
+      "loss": 1.3932,
+      "step": 26
+    },
+    {
+      "epoch": 0.0224,
+      "learning_rate": 2.8650493081609305e-08,
+      "loss": 1.2033,
+      "step": 28
+    },
+    {
+      "epoch": 0.024,
+      "learning_rate": 2.904292634634783e-08,
+      "loss": 1.3504,
+      "step": 30
+    },
+    {
+      "epoch": 0.0256,
+      "learning_rate": 2.9437572956827827e-08,
+      "loss": 1.2393,
+      "step": 32
+    },
+    {
+      "epoch": 0.0272,
+      "learning_rate": 2.9834420602939257e-08,
+      "loss": 1.484,
+      "step": 34
+    },
+    {
+      "epoch": 0.0288,
+      "learning_rate": 3.023345690591533e-08,
+      "loss": 1.4129,
+      "step": 36
+    },
+    {
+      "epoch": 0.0304,
+      "learning_rate": 3.063466941871945e-08,
+      "loss": 1.4234,
+      "step": 38
+    },
+    {
+      "epoch": 0.032,
+      "learning_rate": 3.103804562643294e-08,
+      "loss": 1.265,
+      "step": 40
+    },
+    {
+      "epoch": 0.0336,
+      "learning_rate": 3.144357294664568e-08,
+      "loss": 1.4178,
+      "step": 42
+    },
+    {
+      "epoch": 0.0352,
+      "learning_rate": 3.185123872984803e-08,
+      "loss": 1.1136,
+      "step": 44
+    },
+    {
+      "epoch": 0.0368,
+      "learning_rate": 3.226103025982625e-08,
+      "loss": 1.605,
+      "step": 46
+    },
+    {
+      "epoch": 0.0384,
+      "learning_rate": 3.267293475405858e-08,
+      "loss": 1.282,
+      "step": 48
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 3.308693936411411e-08,
+      "loss": 1.1096,
+      "step": 50
+    },
+    {
+      "epoch": 0.0416,
+      "learning_rate": 3.3503031176053686e-08,
+      "loss": 1.4166,
+      "step": 52
+    },
+    {
+      "epoch": 0.0432,
+      "learning_rate": 3.392119721083223e-08,
+      "loss": 1.2273,
+      "step": 54
+    },
+    {
+      "epoch": 0.0448,
+      "learning_rate": 3.434142442470434e-08,
+      "loss": 1.4607,
+      "step": 56
+    },
+    {
+      "epoch": 0.0464,
+      "learning_rate": 3.4763699709630644e-08,
+      "loss": 0.8361,
+      "step": 58
+    },
+    {
+      "epoch": 0.048,
+      "learning_rate": 3.518800989368683e-08,
+      "loss": 1.1233,
+      "step": 60
+    },
+    {
+      "epoch": 0.0496,
+      "learning_rate": 3.5614341741474666e-08,
+      "loss": 0.2855,
+      "step": 62
+    },
+    {
+      "epoch": 0.0512,
+      "learning_rate": 3.60426819545342e-08,
+      "loss": 1.5457,
+      "step": 64
+    },
+    {
+      "epoch": 0.0528,
+      "learning_rate": 3.647301717175955e-08,
+      "loss": 0.367,
+      "step": 66
+    },
+    {
+      "epoch": 0.0544,
+      "learning_rate": 3.690533396981499e-08,
+      "loss": 1.1436,
+      "step": 68
+    },
+    {
+      "epoch": 0.056,
+      "learning_rate": 3.733961886355388e-08,
+      "loss": 1.3108,
+      "step": 70
+    },
+    {
+      "epoch": 0.0576,
+      "learning_rate": 3.77758583064394e-08,
+      "loss": 1.1941,
+      "step": 72
+    },
+    {
+      "epoch": 0.0592,
+      "learning_rate": 3.821403869096643e-08,
+      "loss": 1.2337,
+      "step": 74
+    },
+    {
+      "epoch": 0.0608,
+      "learning_rate": 3.865414634908756e-08,
+      "loss": 1.0858,
+      "step": 76
+    },
+    {
+      "epoch": 0.0624,
+      "learning_rate": 3.909616755263741e-08,
+      "loss": 0.8241,
+      "step": 78
+    },
+    {
+      "epoch": 0.064,
+      "learning_rate": 3.954008851376244e-08,
+      "loss": 1.0609,
+      "step": 80
+    },
+    {
+      "epoch": 0.0656,
+      "learning_rate": 3.998589538535049e-08,
+      "loss": 0.9286,
+      "step": 82
+    },
+    {
+      "epoch": 0.0672,
+      "learning_rate": 4.043357426146209e-08,
+      "loss": 0.9082,
+      "step": 84
+    },
+    {
+      "epoch": 0.0688,
+      "learning_rate": 4.0883111177765794e-08,
+      "loss": 1.3354,
+      "step": 86
+    },
+    {
+      "epoch": 0.0704,
+      "learning_rate": 4.133449211197183e-08,
+      "loss": 1.3678,
+      "step": 88
+    },
+    {
+      "epoch": 0.072,
+      "learning_rate": 4.178770298427113e-08,
+      "loss": 0.9764,
+      "step": 90
+    },
+    {
+      "epoch": 0.0736,
+      "learning_rate": 4.224272965777315e-08,
+      "loss": 1.1008,
+      "step": 92
+    },
+    {
+      "epoch": 0.0752,
+      "learning_rate": 4.269955793894848e-08,
+      "loss": 1.2634,
+      "step": 94
+    },
+    {
+      "epoch": 0.0768,
+      "learning_rate": 4.315817357806969e-08,
+      "loss": 1.2915,
+      "step": 96
+    },
+    {
+      "epoch": 0.0784,
+      "learning_rate": 4.3618562269657274e-08,
+      "loss": 1.2347,
+      "step": 98
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 4.4080709652925255e-08,
+      "loss": 1.1461,
+      "step": 100
+    },
+    {
+      "epoch": 0.0816,
+      "learning_rate": 4.454460131222918e-08,
+      "loss": 1.326,
+      "step": 102
+    },
+    {
+      "epoch": 0.0832,
+      "learning_rate": 4.501022277751605e-08,
+      "loss": 0.7782,
+      "step": 104
+    },
+    {
+      "epoch": 0.0848,
+      "learning_rate": 4.547755952477499e-08,
+      "loss": 0.7964,
+      "step": 106
+    },
+    {
+      "epoch": 0.0864,
+      "learning_rate": 4.594659697649125e-08,
+      "loss": 1.0274,
+      "step": 108
+    },
+    {
+      "epoch": 0.088,
+      "learning_rate": 4.6417320502100274e-08,
+      "loss": 1.2652,
+      "step": 110
+    },
+    {
+      "epoch": 0.0896,
+      "learning_rate": 4.6889715418444245e-08,
+      "loss": 1.3657,
+      "step": 112
+    },
+    {
+      "epoch": 0.0912,
+      "learning_rate": 4.7363766990230224e-08,
+      "loss": 1.1111,
+      "step": 114
+    },
+    {
+      "epoch": 0.0928,
+      "learning_rate": 4.7839460430489206e-08,
+      "loss": 1.2449,
+      "step": 116
+    },
+    {
+      "epoch": 0.0944,
+      "learning_rate": 4.831678090103828e-08,
+      "loss": 0.9763,
+      "step": 118
+    },
+    {
+      "epoch": 0.096,
+      "learning_rate": 4.8795713512942785e-08,
+      "loss": 0.8226,
+      "step": 120
+    },
+    {
+      "epoch": 0.0976,
+      "learning_rate": 4.9276243326980994e-08,
+      "loss": 1.092,
+      "step": 122
+    },
+    {
+      "epoch": 0.0992,
+      "learning_rate": 4.975835535411023e-08,
+      "loss": 0.4841,
+      "step": 124
+    },
+    {
+      "epoch": 0.1008,
+      "learning_rate": 5.024203455593374e-08,
+      "loss": 1.0007,
+      "step": 126
+    },
+    {
+      "epoch": 0.1024,
+      "learning_rate": 5.072726584517082e-08,
+      "loss": 1.2885,
+      "step": 128
+    },
+    {
+      "epoch": 0.104,
+      "learning_rate": 5.1214034086126676e-08,
+      "loss": 0.9553,
+      "step": 130
+    },
+    {
+      "epoch": 0.1056,
+      "learning_rate": 5.170232409516482e-08,
+      "loss": 1.2022,
+      "step": 132
+    },
+    {
+      "epoch": 0.1072,
+      "learning_rate": 5.2192120641180813e-08,
+      "loss": 1.231,
+      "step": 134
+    },
+    {
+      "epoch": 0.1088,
+      "learning_rate": 5.268340844607653e-08,
+      "loss": 1.3282,
+      "step": 136
+    },
+    {
+      "epoch": 0.1104,
+      "learning_rate": 5.317617218523852e-08,
+      "loss": 1.2123,
+      "step": 138
+    },
+    {
+      "epoch": 0.112,
+      "learning_rate": 5.367039648801377e-08,
+      "loss": 0.4693,
+      "step": 140
+    },
+    {
+      "epoch": 0.1136,
+      "learning_rate": 5.4166065938191084e-08,
+      "loss": 1.2777,
+      "step": 142
+    },
+    {
+      "epoch": 0.1152,
+      "learning_rate": 5.466316507448052e-08,
+      "loss": 1.1257,
+      "step": 144
+    },
+    {
+      "epoch": 0.1168,
+      "learning_rate": 5.5161678390996613e-08,
+      "loss": 1.0569,
+      "step": 146
+    },
+    {
+      "epoch": 0.1184,
+      "learning_rate": 5.5661590337742245e-08,
+      "loss": 1.0706,
+      "step": 148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 5.61628853210922e-08,
+      "loss": 0.9219,
+      "step": 150
+    },
+    {
+      "epoch": 0.1216,
+      "learning_rate": 5.666554770428136e-08,
+      "loss": 0.6337,
+      "step": 152
+    },
+    {
+      "epoch": 0.1232,
+      "learning_rate": 5.716956180789085e-08,
+      "loss": 1.2267,
+      "step": 154
+    },
+    {
+      "epoch": 0.1248,
+      "learning_rate": 5.767491191033907e-08,
+      "loss": 0.8494,
+      "step": 156
+    },
+    {
+      "epoch": 0.1264,
+      "learning_rate": 5.8181582248369823e-08,
+      "loss": 0.7813,
+      "step": 158
+    },
+    {
+      "epoch": 0.128,
+      "learning_rate": 5.8689557017545764e-08,
+      "loss": 0.9805,
+      "step": 160
+    },
+    {
+      "epoch": 0.1296,
+      "learning_rate": 5.919882037274064e-08,
+      "loss": 0.7244,
+      "step": 162
+    },
+    {
+      "epoch": 0.1312,
+      "learning_rate": 5.970935642863361e-08,
+      "loss": 2.4001,
+      "step": 164
+    },
+    {
+      "epoch": 0.1328,
+      "learning_rate": 6.022114926020505e-08,
+      "loss": 1.241,
+      "step": 166
+    },
+    {
+      "epoch": 0.1344,
+      "learning_rate": 6.073418290323246e-08,
+      "loss": 0.7312,
+      "step": 168
+    },
+    {
+      "epoch": 0.136,
+      "learning_rate": 6.124844135478965e-08,
+      "loss": 1.3272,
+      "step": 170
+    },
+    {
+      "epoch": 0.1376,
+      "learning_rate": 6.1763908573745e-08,
+      "loss": 0.7542,
+      "step": 172
+    },
+    {
+      "epoch": 0.1392,
+      "learning_rate": 6.228056848126223e-08,
+      "loss": 1.1428,
+      "step": 174
+    },
+    {
+      "epoch": 0.1408,
+      "learning_rate": 6.279840496130187e-08,
+      "loss": 0.8745,
+      "step": 176
+    },
+    {
+      "epoch": 0.1424,
+      "learning_rate": 6.331740186112357e-08,
+      "loss": 0.925,
+      "step": 178
+    },
+    {
+      "epoch": 0.144,
+      "learning_rate": 6.383754299179072e-08,
+      "loss": 1.2462,
+      "step": 180
+    },
+    {
+      "epoch": 0.1456,
+      "learning_rate": 6.435881212867485e-08,
+      "loss": 0.722,
+      "step": 182
+    },
+    {
+      "epoch": 0.1472,
+      "learning_rate": 6.488119301196189e-08,
+      "loss": 0.4564,
+      "step": 184
+    },
+    {
+      "epoch": 0.1488,
+      "learning_rate": 6.540466934715954e-08,
+      "loss": 1.4409,
+      "step": 186
+    },
+    {
+      "epoch": 0.1504,
+      "learning_rate": 6.592922480560482e-08,
+      "loss": 1.2305,
+      "step": 188
+    },
+    {
+      "epoch": 0.152,
+      "learning_rate": 6.645484302497446e-08,
+      "loss": 0.9539,
+      "step": 190
+    },
+    {
+      "epoch": 0.1536,
+      "learning_rate": 6.698150760979455e-08,
+      "loss": 1.27,
+      "step": 192
+    },
+    {
+      "epoch": 0.1552,
+      "learning_rate": 6.750920213195241e-08,
+      "loss": 0.6927,
+      "step": 194
+    },
+    {
+      "epoch": 0.1568,
+      "learning_rate": 6.803791013120823e-08,
+      "loss": 0.9304,
+      "step": 196
+    },
+    {
+      "epoch": 0.1584,
+      "learning_rate": 6.856761511570943e-08,
+      "loss": 1.1268,
+      "step": 198
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 6.90983005625052e-08,
+      "loss": 1.1436,
+      "step": 200
+    },
+    {
+      "epoch": 0.1616,
+      "learning_rate": 6.96299499180605e-08,
+      "loss": 1.0888,
+      "step": 202
+    },
+    {
+      "epoch": 0.1632,
+      "learning_rate": 7.016254659877403e-08,
+      "loss": 0.7638,
+      "step": 204
+    },
+    {
+      "epoch": 0.1648,
+      "learning_rate": 7.069607399149428e-08,
+      "loss": 1.0044,
+      "step": 206
+    },
+    {
+      "epoch": 0.1664,
+      "learning_rate": 7.123051545403872e-08,
+      "loss": 0.6566,
+      "step": 208
+    },
+    {
+      "epoch": 0.168,
+      "learning_rate": 7.176585431571231e-08,
+      "loss": 0.8715,
+      "step": 210
+    },
+    {
+      "epoch": 0.1696,
+      "learning_rate": 7.23020738778277e-08,
+      "loss": 0.8417,
+      "step": 212
+    },
+    {
+      "epoch": 0.1712,
+      "learning_rate": 7.28391574142262e-08,
+      "loss": 0.9313,
+      "step": 214
+    },
+    {
+      "epoch": 0.1728,
+      "learning_rate": 7.337708817179874e-08,
+      "loss": 1.0465,
+      "step": 216
+    },
+    {
+      "epoch": 0.1744,
+      "learning_rate": 7.39158493710103e-08,
+      "loss": 0.7281,
+      "step": 218
+    },
+    {
+      "epoch": 0.176,
+      "learning_rate": 7.44554242064209e-08,
+      "loss": 0.827,
+      "step": 220
+    },
+    {
+      "epoch": 0.1776,
+      "learning_rate": 7.499579584721171e-08,
+      "loss": 0.6899,
+      "step": 222
+    },
+    {
+      "epoch": 0.1792,
+      "learning_rate": 7.553694743770916e-08,
+      "loss": 0.7605,
+      "step": 224
+    },
+    {
+      "epoch": 0.1808,
+      "learning_rate": 7.607886209791094e-08,
+      "loss": 0.4112,
+      "step": 226
+    },
+    {
+      "epoch": 0.1824,
+      "learning_rate": 7.662152292401263e-08,
+      "loss": 0.47,
+      "step": 228
+    },
+    {
+      "epoch": 0.184,
+      "learning_rate": 7.716491298893439e-08,
+      "loss": 1.0721,
+      "step": 230
+    },
+    {
+      "epoch": 0.1856,
+      "learning_rate": 7.77090153428499e-08,
+      "loss": 0.8861,
+      "step": 232
+    },
+    {
+      "epoch": 0.1872,
+      "learning_rate": 7.825381301371443e-08,
+      "loss": 0.7312,
+      "step": 234
+    },
+    {
+      "epoch": 0.1888,
+      "learning_rate": 7.87992890077944e-08,
+      "loss": 0.873,
+      "step": 236
+    },
+    {
+      "epoch": 0.1904,
+      "learning_rate": 7.934542631019766e-08,
+      "loss": 1.0387,
+      "step": 238
+    },
+    {
+      "epoch": 0.192,
+      "learning_rate": 7.989220788540351e-08,
+      "loss": 0.8897,
+      "step": 240
+    },
+    {
+      "epoch": 0.1936,
+      "learning_rate": 8.043961667779511e-08,
+      "loss": 0.4891,
+      "step": 242
+    },
+    {
+      "epoch": 0.1952,
+      "learning_rate": 8.098763561219088e-08,
+      "loss": 0.8573,
+      "step": 244
+    },
+    {
+      "epoch": 0.1968,
+      "learning_rate": 8.153624759437717e-08,
+      "loss": 0.9298,
+      "step": 246
+    },
+    {
+      "epoch": 0.1984,
+      "learning_rate": 8.208543551164178e-08,
+      "loss": 0.7362,
+      "step": 248
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 8.263518223330694e-08,
+      "loss": 0.3852,
+      "step": 250
+    },
+    {
+      "epoch": 0.2016,
+      "learning_rate": 8.318547061126479e-08,
+      "loss": 0.2213,
+      "step": 252
+    },
+    {
+      "epoch": 0.2032,
+      "learning_rate": 8.373628348051154e-08,
+      "loss": 0.7391,
+      "step": 254
+    },
+    {
+      "epoch": 0.2048,
+      "learning_rate": 8.428760365968327e-08,
+      "loss": 1.1032,
+      "step": 256
+    },
+    {
+      "epoch": 0.2064,
+      "learning_rate": 8.483941395159112e-08,
+      "loss": 0.4072,
+      "step": 258
+    },
+    {
+      "epoch": 0.208,
+      "learning_rate": 8.539169714375883e-08,
+      "loss": 0.549,
+      "step": 260
+    },
+    {
+      "epoch": 0.2096,
+      "learning_rate": 8.594443600895884e-08,
+      "loss": 0.8849,
+      "step": 262
+    },
+    {
+      "epoch": 0.2112,
+      "learning_rate": 8.649761330575e-08,
+      "loss": 0.6731,
+      "step": 264
+    },
+    {
+      "epoch": 0.2128,
+      "learning_rate": 8.705121177901536e-08,
+      "loss": 0.9314,
+      "step": 266
+    },
+    {
+      "epoch": 0.2144,
+      "learning_rate": 8.760521416049984e-08,
+      "loss": 1.3233,
+      "step": 268
+    },
+    {
+      "epoch": 0.216,
+      "learning_rate": 8.815960316934988e-08,
+      "loss": 0.6171,
+      "step": 270
+    },
+    {
+      "epoch": 0.2176,
+      "learning_rate": 8.871436151265178e-08,
+      "loss": 1.0752,
+      "step": 272
+    },
+    {
+      "epoch": 0.2192,
+      "learning_rate": 8.926947188597124e-08,
+      "loss": 0.8997,
+      "step": 274
+    },
+    {
+      "epoch": 0.2208,
+      "learning_rate": 8.982491697389343e-08,
+      "loss": 0.7845,
+      "step": 276
+    },
+    {
+      "epoch": 0.2224,
+      "learning_rate": 9.038067945056208e-08,
+      "loss": 0.979,
+      "step": 278
+    },
+    {
+      "epoch": 0.224,
+      "learning_rate": 9.093674198022196e-08,
+      "loss": 1.2292,
+      "step": 280
+    },
+    {
+      "epoch": 0.2256,
+      "learning_rate": 9.149308721775714e-08,
+      "loss": 0.4902,
+      "step": 282
+    },
+    {
+      "epoch": 0.2272,
+      "learning_rate": 9.204969780923394e-08,
+      "loss": 0.7539,
+      "step": 284
+    },
+    {
+      "epoch": 0.2288,
+      "learning_rate": 9.260655639244138e-08,
+      "loss": 0.7636,
+      "step": 286
+    },
+    {
+      "epoch": 0.2304,
+      "learning_rate": 9.316364559743297e-08,
+      "loss": 0.6472,
+      "step": 288
+    },
+    {
+      "epoch": 0.232,
+      "learning_rate": 9.372094804706865e-08,
+      "loss": 0.7449,
+      "step": 290
+    },
+    {
+      "epoch": 0.2336,
+      "learning_rate": 9.427844635755613e-08,
+      "loss": 1.1085,
+      "step": 292
+    },
+    {
+      "epoch": 0.2352,
+      "learning_rate": 9.483612313899444e-08,
+      "loss": 0.5536,
+      "step": 294
+    },
+    {
+      "epoch": 0.2368,
+      "learning_rate": 9.539396099591467e-08,
+      "loss": 0.781,
+      "step": 296
+    },
+    {
+      "epoch": 0.2384,
+      "learning_rate": 9.59519425278246e-08,
+      "loss": 0.6727,
+      "step": 298
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.651005032974989e-08,
+      "loss": 0.9434,
+      "step": 300
+    },
+    {
+      "epoch": 0.2416,
+      "learning_rate": 9.706826699277712e-08,
+      "loss": 1.0254,
+      "step": 302
+    },
+    {
+      "epoch": 0.2432,
+      "learning_rate": 9.762657510459771e-08,
+      "loss": 0.8204,
+      "step": 304
+    },
+    {
+      "epoch": 0.2448,
+      "learning_rate": 9.818495725005042e-08,
+      "loss": 0.9419,
+      "step": 306
+    },
+    {
+      "epoch": 0.2464,
+      "learning_rate": 9.874339601166478e-08,
+      "loss": 0.9943,
+      "step": 308
+    },
+    {
+      "epoch": 0.248,
+      "learning_rate": 9.930187397020385e-08,
+      "loss": 0.8739,
+      "step": 310
+    },
+    {
+      "epoch": 0.2496,
+      "learning_rate": 9.986037370520853e-08,
+      "loss": 0.8851,
+      "step": 312
+    },
+    {
+      "epoch": 0.2512,
+      "learning_rate": 1.0041887779554033e-07,
+      "loss": 1.134,
+      "step": 314
+    },
+    {
+      "epoch": 0.2528,
+      "learning_rate": 1.009773688199248e-07,
+      "loss": 0.4205,
+      "step": 316
+    },
+    {
+      "epoch": 0.2544,
+      "learning_rate": 1.0153582935749532e-07,
+      "loss": 0.6858,
+      "step": 318
+    },
+    {
+      "epoch": 0.256,
+      "learning_rate": 1.020942419883357e-07,
+      "loss": 0.8924,
+      "step": 320
+    },
+    {
+      "epoch": 0.2576,
+      "learning_rate": 1.0265258929402456e-07,
+      "loss": 0.8844,
+      "step": 322
+    },
+    {
+      "epoch": 0.2592,
+      "learning_rate": 1.032108538581781e-07,
+      "loss": 0.8018,
+      "step": 324
+    },
+    {
+      "epoch": 0.2608,
+      "learning_rate": 1.0376901826699335e-07,
+      "loss": 0.8816,
+      "step": 326
+    },
+    {
+      "epoch": 0.2624,
+      "learning_rate": 1.0432706510979175e-07,
+      "loss": 0.9749,
+      "step": 328
+    },
+    {
+      "epoch": 0.264,
+      "learning_rate": 1.0488497697956133e-07,
+      "loss": 0.5533,
+      "step": 330
+    },
+    {
+      "epoch": 0.2656,
+      "learning_rate": 1.0544273647350088e-07,
+      "loss": 0.9202,
+      "step": 332
+    },
+    {
+      "epoch": 0.2672,
+      "learning_rate": 1.0600032619356201e-07,
+      "loss": 0.9548,
+      "step": 334
+    },
+    {
+      "epoch": 0.2688,
+      "learning_rate": 1.0655772874699204e-07,
+      "loss": 0.6163,
+      "step": 336
+    },
+    {
+      "epoch": 0.2704,
+      "learning_rate": 1.0711492674687673e-07,
+      "loss": 0.3942,
+      "step": 338
+    },
+    {
+      "epoch": 0.272,
+      "learning_rate": 1.076719028126817e-07,
+      "loss": 0.593,
+      "step": 340
+    },
+    {
+      "epoch": 0.2736,
+      "learning_rate": 1.0822863957079653e-07,
+      "loss": 0.7285,
+      "step": 342
+    },
+    {
+      "epoch": 0.2752,
+      "learning_rate": 1.0878511965507426e-07,
+      "loss": 0.5399,
+      "step": 344
+    },
+    {
+      "epoch": 0.2768,
+      "learning_rate": 1.0934132570737495e-07,
+      "loss": 0.7572,
+      "step": 346
+    },
+    {
+      "epoch": 0.2784,
+      "learning_rate": 1.0989724037810637e-07,
+      "loss": 0.3596,
+      "step": 348
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.104528463267652e-07,
+      "loss": 0.7278,
+      "step": 350
+    },
+    {
+      "epoch": 0.2816,
+      "learning_rate": 1.110081262224782e-07,
+      "loss": 1.0649,
+      "step": 352
+    },
+    {
+      "epoch": 0.2832,
+      "learning_rate": 1.115630627445421e-07,
+      "loss": 0.4888,
+      "step": 354
+    },
+    {
+      "epoch": 0.2848,
+      "learning_rate": 1.1211763858296514e-07,
+      "loss": 0.6159,
+      "step": 356
+    },
+    {
+      "epoch": 0.2864,
+      "learning_rate": 1.1267183643900532e-07,
+      "loss": 0.5885,
+      "step": 358
+    },
+    {
+      "epoch": 0.288,
+      "learning_rate": 1.1322563902571225e-07,
+      "loss": 0.558,
+      "step": 360
+    },
+    {
+      "epoch": 0.2896,
+      "learning_rate": 1.137790290684638e-07,
+      "loss": 0.795,
+      "step": 362
+    },
+    {
+      "epoch": 0.2912,
+      "learning_rate": 1.1433198930550687e-07,
+      "loss": 0.7102,
+      "step": 364
+    },
+    {
+      "epoch": 0.2928,
+      "learning_rate": 1.1488450248849513e-07,
+      "loss": 0.7162,
+      "step": 366
+    },
+    {
+      "epoch": 0.2944,
+      "learning_rate": 1.1543655138302699e-07,
+      "loss": 0.608,
+      "step": 368
+    },
+    {
+      "epoch": 0.296,
+      "learning_rate": 1.1598811876918351e-07,
+      "loss": 0.7869,
+      "step": 370
+    },
+    {
+      "epoch": 0.2976,
+      "learning_rate": 1.1653918744206477e-07,
+      "loss": 0.4868,
+      "step": 372
+    },
+    {
+      "epoch": 0.2992,
+      "learning_rate": 1.1708974021232761e-07,
+      "loss": 0.6533,
+      "step": 374
+    },
+    {
+      "epoch": 0.3008,
+      "learning_rate": 1.1763975990672115e-07,
+      "loss": 0.4112,
+      "step": 376
+    },
+    {
+      "epoch": 0.3024,
+      "learning_rate": 1.1818922936862257e-07,
+      "loss": 0.6011,
+      "step": 378
+    },
+    {
+      "epoch": 0.304,
+      "learning_rate": 1.1873813145857248e-07,
+      "loss": 0.6559,
+      "step": 380
+    },
+    {
+      "epoch": 0.3056,
+      "learning_rate": 1.192864490548089e-07,
+      "loss": 0.4218,
+      "step": 382
+    },
+    {
+      "epoch": 0.3072,
+      "learning_rate": 1.198341650538023e-07,
+      "loss": 0.5647,
+      "step": 384
+    },
+    {
+      "epoch": 0.3088,
+      "learning_rate": 1.2038126237078838e-07,
+      "loss": 0.7457,
+      "step": 386
+    },
+    {
+      "epoch": 0.3104,
+      "learning_rate": 1.209277239403014e-07,
+      "loss": 0.4972,
+      "step": 388
+    },
+    {
+      "epoch": 0.312,
+      "learning_rate": 1.2147353271670636e-07,
+      "loss": 0.4324,
+      "step": 390
+    },
+    {
+      "epoch": 0.3136,
+      "learning_rate": 1.2201867167473014e-07,
+      "loss": 0.5187,
+      "step": 392
+    },
+    {
+      "epoch": 0.3152,
+      "learning_rate": 1.225631238099937e-07,
+      "loss": 0.5617,
+      "step": 394
+    },
+    {
+      "epoch": 0.3168,
+      "learning_rate": 1.2310687213954172e-07,
+      "loss": 0.5243,
+      "step": 396
+    },
+    {
+      "epoch": 0.3184,
+      "learning_rate": 1.2364989970237238e-07,
+      "loss": 0.5335,
+      "step": 398
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.2419218955996679e-07,
+      "loss": 0.6022,
+      "step": 400
+    },
+    {
+      "epoch": 0.3216,
+      "learning_rate": 1.247337247968165e-07,
+      "loss": 0.7804,
+      "step": 402
+    },
+    {
+      "epoch": 0.3232,
+      "learning_rate": 1.252744885209529e-07,
+      "loss": 0.4939,
+      "step": 404
+    },
+    {
+      "epoch": 0.3248,
+      "learning_rate": 1.258144638644717e-07,
+      "loss": 0.5223,
+      "step": 406
+    },
+    {
+      "epoch": 0.3264,
+      "learning_rate": 1.2635363398406132e-07,
+      "loss": 0.534,
+      "step": 408
+    },
+    {
+      "epoch": 0.328,
+      "learning_rate": 1.2689198206152642e-07,
+      "loss": 0.3608,
+      "step": 410
+    },
+    {
+      "epoch": 0.3296,
+      "learning_rate": 1.2742949130431467e-07,
+      "loss": 0.6405,
+      "step": 412
+    },
+    {
+      "epoch": 0.3312,
+      "learning_rate": 1.2796614494603796e-07,
+      "loss": 0.4897,
+      "step": 414
+    },
+    {
+      "epoch": 0.3328,
+      "learning_rate": 1.2850192624699755e-07,
+      "loss": 0.436,
+      "step": 416
+    },
+    {
+      "epoch": 0.3344,
+      "learning_rate": 1.2903681849470533e-07,
+      "loss": 0.4304,
+      "step": 418
+    },
+    {
+      "epoch": 0.336,
+      "learning_rate": 1.2957080500440454e-07,
+      "loss": 0.3124,
+      "step": 420
+    },
+    {
+      "epoch": 0.3376,
+      "learning_rate": 1.3010386911959204e-07,
+      "loss": 0.6727,
+      "step": 422
+    },
+    {
+      "epoch": 0.3392,
+      "learning_rate": 1.3063599421253556e-07,
+      "loss": 0.5108,
+      "step": 424
+    },
+    {
+      "epoch": 0.3408,
+      "learning_rate": 1.3116716368479412e-07,
+      "loss": 0.3579,
+      "step": 426
+    },
+    {
+      "epoch": 0.3424,
+      "learning_rate": 1.3169736096773506e-07,
+      "loss": 0.3544,
+      "step": 428
+    },
+    {
+      "epoch": 0.344,
+      "learning_rate": 1.3222656952305098e-07,
+      "loss": 0.4325,
+      "step": 430
+    },
+    {
+      "epoch": 0.3456,
+      "learning_rate": 1.327547728432757e-07,
+      "loss": 0.3743,
+      "step": 432
+    },
+    {
+      "epoch": 0.3472,
+      "learning_rate": 1.3328195445229863e-07,
+      "loss": 0.1725,
+      "step": 434
+    },
+    {
+      "epoch": 0.3488,
+      "learning_rate": 1.3380809790587967e-07,
+      "loss": 0.4319,
+      "step": 436
+    },
+    {
+      "epoch": 0.3504,
+      "learning_rate": 1.3433318679216144e-07,
+      "loss": 0.5034,
+      "step": 438
+    },
+    {
+      "epoch": 0.352,
+      "learning_rate": 1.348572047321814e-07,
+      "loss": 1.4124,
+      "step": 440
+    },
+    {
+      "epoch": 0.3536,
+      "learning_rate": 1.3538013538038296e-07,
+      "loss": 1.4059,
+      "step": 442
+    },
+    {
+      "epoch": 0.3552,
+      "learning_rate": 1.3590196242512458e-07,
+      "loss": 0.4279,
+      "step": 444
+    },
+    {
+      "epoch": 0.3568,
+      "learning_rate": 1.3642266958918979e-07,
+      "loss": 0.3227,
+      "step": 446
+    },
+    {
+      "epoch": 0.3584,
+      "learning_rate": 1.3694224063029385e-07,
+      "loss": 0.3695,
+      "step": 448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.3746065934159107e-07,
+      "loss": 0.3335,
+      "step": 450
+    },
+    {
+      "epoch": 0.3616,
+      "learning_rate": 1.3797790955218012e-07,
+      "loss": 0.3231,
+      "step": 452
+    },
+    {
+      "epoch": 0.3632,
+      "learning_rate": 1.3849397512760793e-07,
+      "loss": 0.312,
+      "step": 454
+    },
+    {
+      "epoch": 0.3648,
+      "learning_rate": 1.390088399703739e-07,
+      "loss": 0.3441,
+      "step": 456
+    },
+    {
+      "epoch": 0.3664,
+      "learning_rate": 1.3952248802043157e-07,
+      "loss": 0.4827,
+      "step": 458
+    },
+    {
+      "epoch": 0.368,
+      "learning_rate": 1.4003490325568954e-07,
+      "loss": 0.284,
+      "step": 460
+    },
+    {
+      "epoch": 0.3696,
+      "learning_rate": 1.4054606969251096e-07,
+      "loss": 0.2959,
+      "step": 462
+    },
+    {
+      "epoch": 0.3712,
+      "learning_rate": 1.410559713862128e-07,
+      "loss": 0.4057,
+      "step": 464
+    },
+    {
+      "epoch": 0.3728,
+      "learning_rate": 1.4156459243156274e-07,
+      "loss": 0.3438,
+      "step": 466
+    },
+    {
+      "epoch": 0.3744,
+      "learning_rate": 1.420719169632754e-07,
+      "loss": 0.4746,
+      "step": 468
+    },
+    {
+      "epoch": 0.376,
+      "learning_rate": 1.4257792915650734e-07,
+      "loss": 0.3085,
+      "step": 470
+    },
+    {
+      "epoch": 0.3776,
+      "learning_rate": 1.430826132273499e-07,
+      "loss": 0.4413,
+      "step": 472
+    },
+    {
+      "epoch": 0.3792,
+      "learning_rate": 1.435859534333234e-07,
+      "loss": 0.553,
+      "step": 474
+    },
+    {
+      "epoch": 0.3808,
+      "learning_rate": 1.440879340738658e-07,
+      "loss": 0.5621,
+      "step": 476
+    },
+    {
+      "epoch": 0.3824,
+      "learning_rate": 1.4458853949082433e-07,
+      "loss": 0.7384,
+      "step": 478
+    },
+    {
+      "epoch": 0.384,
+      "learning_rate": 1.4508775406894312e-07,
+      "loss": 1.5034,
+      "step": 480
+    },
+    {
+      "epoch": 0.3856,
+      "learning_rate": 1.4558556223634987e-07,
+      "loss": 0.548,
+      "step": 482
+    },
+    {
+      "epoch": 0.3872,
+      "learning_rate": 1.4608194846504308e-07,
+      "loss": 0.3735,
+      "step": 484
+    },
+    {
+      "epoch": 0.3888,
+      "learning_rate": 1.465768972713744e-07,
+      "loss": 0.3027,
+      "step": 486
+    },
+    {
+      "epoch": 0.3904,
+      "learning_rate": 1.4707039321653318e-07,
+      "loss": 0.3443,
+      "step": 488
+    },
+    {
+      "epoch": 0.392,
+      "learning_rate": 1.4756242090702741e-07,
+      "loss": 0.2778,
+      "step": 490
+    },
+    {
+      "epoch": 0.3936,
+      "learning_rate": 1.4805296499516396e-07,
+      "loss": 0.3164,
+      "step": 492
+    },
+    {
+      "epoch": 0.3952,
+      "learning_rate": 1.4854201017952738e-07,
+      "loss": 0.2379,
+      "step": 494
+    },
+    {
+      "epoch": 0.3968,
+      "learning_rate": 1.4902954120545682e-07,
+      "loss": 0.3995,
+      "step": 496
+    },
+    {
+      "epoch": 0.3984,
+      "learning_rate": 1.495155428655226e-07,
+      "loss": 0.3256,
+      "step": 498
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.4999999999999991e-07,
+      "loss": 0.3324,
+      "step": 500
+    },
+    {
+      "epoch": 0.4016,
+      "learning_rate": 1.5048289749734205e-07,
+      "loss": 0.4055,
+      "step": 502
+    },
+    {
+      "epoch": 0.4032,
+      "learning_rate": 1.5096422029465177e-07,
+      "loss": 0.4987,
+      "step": 504
+    },
+    {
+      "epoch": 0.4048,
+      "learning_rate": 1.514439533781506e-07,
+      "loss": 0.352,
+      "step": 506
+    },
+    {
+      "epoch": 0.4064,
+      "learning_rate": 1.5192208178364807e-07,
+      "loss": 0.2759,
+      "step": 508
+    },
+    {
+      "epoch": 0.408,
+      "learning_rate": 1.523985905970078e-07,
+      "loss": 0.3225,
+      "step": 510
+    },
+    {
+      "epoch": 0.4096,
+      "learning_rate": 1.5287346495461318e-07,
+      "loss": 0.3325,
+      "step": 512
+    },
+    {
+      "epoch": 0.4112,
+      "learning_rate": 1.5334669004383024e-07,
+      "loss": 0.2828,
+      "step": 514
+    },
+    {
+      "epoch": 0.4128,
+      "learning_rate": 1.5381825110347068e-07,
+      "loss": 0.3241,
+      "step": 516
+    },
+    {
+      "epoch": 0.4144,
+      "learning_rate": 1.542881334242517e-07,
+      "loss": 0.2396,
+      "step": 518
+    },
+    {
+      "epoch": 0.416,
+      "learning_rate": 1.5475632234925494e-07,
+      "loss": 0.3409,
+      "step": 520
+    },
+    {
+      "epoch": 0.4176,
+      "learning_rate": 1.552228032743839e-07,
+      "loss": 0.2867,
+      "step": 522
+    },
+    {
+      "epoch": 0.4192,
+      "learning_rate": 1.556875616488188e-07,
+      "loss": 0.5007,
+      "step": 524
+    },
+    {
+      "epoch": 0.4208,
+      "learning_rate": 1.561505829754714e-07,
+      "loss": 0.2759,
+      "step": 526
+    },
+    {
+      "epoch": 0.4224,
+      "learning_rate": 1.5661185281143662e-07,
+      "loss": 0.3979,
+      "step": 528
+    },
+    {
+      "epoch": 0.424,
+      "learning_rate": 1.570713567684431e-07,
+      "loss": 0.3604,
+      "step": 530
+    },
+    {
+      "epoch": 0.4256,
+      "learning_rate": 1.575290805133023e-07,
+      "loss": 0.4861,
+      "step": 532
+    },
+    {
+      "epoch": 0.4272,
+      "learning_rate": 1.5798500976835478e-07,
+      "loss": 0.3699,
+      "step": 534
+    },
+    {
+      "epoch": 0.4288,
+      "learning_rate": 1.584391303119172e-07,
+      "loss": 0.2808,
+      "step": 536
+    },
+    {
+      "epoch": 0.4304,
+      "learning_rate": 1.588914279787238e-07,
+      "loss": 0.2899,
+      "step": 538
+    },
+    {
+      "epoch": 0.432,
+      "learning_rate": 1.5934188866037006e-07,
+      "loss": 0.254,
+      "step": 540
+    },
+    {
+      "epoch": 0.4336,
+      "learning_rate": 1.597904983057519e-07,
+      "loss": 0.3573,
+      "step": 542
+    },
+    {
+      "epoch": 0.4352,
+      "learning_rate": 1.6023724292150373e-07,
+      "loss": 0.3805,
+      "step": 544
+    },
+    {
+      "epoch": 0.4368,
+      "learning_rate": 1.606821085724362e-07,
+      "loss": 0.2926,
+      "step": 546
+    },
+    {
+      "epoch": 0.4384,
+      "learning_rate": 1.6112508138196912e-07,
+      "loss": 0.2481,
+      "step": 548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6156614753256575e-07,
+      "loss": 0.3013,
+      "step": 550
+    },
+    {
+      "epoch": 0.4416,
+      "learning_rate": 1.6200529326616317e-07,
+      "loss": 0.3073,
+      "step": 552
+    },
+    {
+      "epoch": 0.4432,
+      "learning_rate": 1.6244250488460145e-07,
+      "loss": 0.3699,
+      "step": 554
+    },
+    {
+      "epoch": 0.4448,
+      "learning_rate": 1.6287776875005126e-07,
+      "loss": 0.3286,
+      "step": 556
+    },
+    {
+      "epoch": 0.4464,
+      "learning_rate": 1.6331107128543853e-07,
+      "loss": 0.3042,
+      "step": 558
+    },
+    {
+      "epoch": 0.448,
+      "learning_rate": 1.6374239897486904e-07,
+      "loss": 0.2656,
+      "step": 560
+    },
+    {
+      "epoch": 0.4496,
+      "learning_rate": 1.6417173836404875e-07,
+      "loss": 0.3882,
+      "step": 562
+    },
+    {
+      "epoch": 0.4512,
+      "learning_rate": 1.645990760607051e-07,
+      "loss": 0.2785,
+      "step": 564
+    },
+    {
+      "epoch": 0.4528,
+      "learning_rate": 1.6502439873500286e-07,
+      "loss": 0.2969,
+      "step": 566
+    },
+    {
+      "epoch": 0.4544,
+      "learning_rate": 1.6544769311996145e-07,
+      "loss": 0.3323,
+      "step": 568
+    },
+    {
+      "epoch": 0.456,
+      "learning_rate": 1.65868946011868e-07,
+      "loss": 0.3305,
+      "step": 570
+    },
+    {
+      "epoch": 0.4576,
+      "learning_rate": 1.6628814427068942e-07,
+      "loss": 0.3192,
+      "step": 572
+    },
+    {
+      "epoch": 0.4592,
+      "learning_rate": 1.6670527482048246e-07,
+      "loss": 0.3215,
+      "step": 574
+    },
+    {
+      "epoch": 0.4608,
+      "learning_rate": 1.6712032464980092e-07,
+      "loss": 0.2576,
+      "step": 576
+    },
+    {
+      "epoch": 0.4624,
+      "learning_rate": 1.675332808121024e-07,
+      "loss": 0.3976,
+      "step": 578
+    },
+    {
+      "epoch": 0.464,
+      "learning_rate": 1.679441304261516e-07,
+      "loss": 0.2681,
+      "step": 580
+    },
+    {
+      "epoch": 0.4656,
+      "learning_rate": 1.683528606764222e-07,
+      "loss": 0.3189,
+      "step": 582
+    },
+    {
+      "epoch": 0.4672,
+      "learning_rate": 1.6875945881349675e-07,
+      "loss": 0.2911,
+      "step": 584
+    },
+    {
+      "epoch": 0.4688,
+      "learning_rate": 1.69163912154464e-07,
+      "loss": 0.2781,
+      "step": 586
+    },
+    {
+      "epoch": 0.4704,
+      "learning_rate": 1.6956620808331502e-07,
+      "loss": 0.2579,
+      "step": 588
+    },
+    {
+      "epoch": 0.472,
+      "learning_rate": 1.6996633405133647e-07,
+      "loss": 0.2747,
+      "step": 590
+    },
+    {
+      "epoch": 0.4736,
+      "learning_rate": 1.7036427757750196e-07,
+      "loss": 0.2441,
+      "step": 592
+    },
+    {
+      "epoch": 0.4752,
+      "learning_rate": 1.7076002624886155e-07,
+      "loss": 0.3047,
+      "step": 594
+    },
+    {
+      "epoch": 0.4768,
+      "learning_rate": 1.7115356772092842e-07,
+      "loss": 0.2947,
+      "step": 596
+    },
+    {
+      "epoch": 0.4784,
+      "learning_rate": 1.7154488971806515e-07,
+      "loss": 0.441,
+      "step": 598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.7193398003386506e-07,
+      "loss": 0.3017,
+      "step": 600
+    },
+    {
+      "epoch": 0.4816,
+      "learning_rate": 1.7232082653153411e-07,
+      "loss": 0.3059,
+      "step": 602
+    },
+    {
+      "epoch": 0.4832,
+      "learning_rate": 1.727054171442692e-07,
+      "loss": 0.246,
+      "step": 604
+    },
+    {
+      "epoch": 0.4848,
+      "learning_rate": 1.730877398756339e-07,
+      "loss": 0.2272,
+      "step": 606
+    },
+    {
+      "epoch": 0.4864,
+      "learning_rate": 1.734677827999341e-07,
+      "loss": 0.2579,
+      "step": 608
+    },
+    {
+      "epoch": 0.488,
+      "learning_rate": 1.7384553406258833e-07,
+      "loss": 0.3192,
+      "step": 610
+    },
+    {
+      "epoch": 0.4896,
+      "learning_rate": 1.7422098188049884e-07,
+      "loss": 0.2299,
+      "step": 612
+    },
+    {
+      "epoch": 0.4912,
+      "learning_rate": 1.745941145424181e-07,
+      "loss": 0.3176,
+      "step": 614
+    },
+    {
+      "epoch": 0.4928,
+      "learning_rate": 1.749649204093154e-07,
+      "loss": 0.2942,
+      "step": 616
+    },
+    {
+      "epoch": 0.4944,
+      "learning_rate": 1.7533338791473867e-07,
+      "loss": 0.3185,
+      "step": 618
+    },
+    {
+      "epoch": 0.496,
+      "learning_rate": 1.7569950556517558e-07,
+      "loss": 0.2766,
+      "step": 620
+    },
+    {
+      "epoch": 0.4976,
+      "learning_rate": 1.7606326194041277e-07,
+      "loss": 0.2714,
+      "step": 622
+    },
+    {
+      "epoch": 0.4992,
+      "learning_rate": 1.7642464569389082e-07,
+      "loss": 0.268,
+      "step": 624
+    },
+    {
+      "epoch": 0.5008,
+      "learning_rate": 1.7678364555305975e-07,
+      "loss": 0.3564,
+      "step": 626
+    },
+    {
+      "epoch": 0.5024,
+      "learning_rate": 1.77140250319729e-07,
+      "loss": 0.3194,
+      "step": 628
+    },
+    {
+      "epoch": 0.504,
+      "learning_rate": 1.774944488704179e-07,
+      "loss": 0.218,
+      "step": 630
+    },
+    {
+      "epoch": 0.5056,
+      "learning_rate": 1.778462301567023e-07,
+      "loss": 0.3733,
+      "step": 632
+    },
+    {
+      "epoch": 0.5072,
+      "learning_rate": 1.7819558320555892e-07,
+      "loss": 0.2487,
+      "step": 634
+    },
+    {
+      "epoch": 0.5088,
+      "learning_rate": 1.785424971197082e-07,
+      "loss": 0.3369,
+      "step": 636
+    },
+    {
+      "epoch": 0.5104,
+      "learning_rate": 1.788869610779534e-07,
+      "loss": 0.2447,
+      "step": 638
+    },
+    {
+      "epoch": 0.512,
+      "learning_rate": 1.79228964335519e-07,
+      "loss": 0.271,
+      "step": 640
+    },
+    {
+      "epoch": 0.5136,
+      "learning_rate": 1.7956849622438547e-07,
+      "loss": 0.2709,
+      "step": 642
+    },
+    {
+      "epoch": 0.5152,
+      "learning_rate": 1.799055461536219e-07,
+      "loss": 0.2729,
+      "step": 644
+    },
+    {
+      "epoch": 0.5168,
+      "learning_rate": 1.8024010360971666e-07,
+      "loss": 0.2243,
+      "step": 646
+    },
+    {
+      "epoch": 0.5184,
+      "learning_rate": 1.805721581569049e-07,
+      "loss": 0.3112,
+      "step": 648
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.809016994374947e-07,
+      "loss": 0.3436,
+      "step": 650
+    },
+    {
+      "epoch": 0.5216,
+      "learning_rate": 1.8122871717218967e-07,
+      "loss": 0.3182,
+      "step": 652
+    },
+    {
+      "epoch": 0.5232,
+      "learning_rate": 1.8155320116040972e-07,
+      "loss": 0.2448,
+      "step": 654
+    },
+    {
+      "epoch": 0.5248,
+      "learning_rate": 1.8187514128060946e-07,
+      "loss": 0.2993,
+      "step": 656
+    },
+    {
+      "epoch": 0.5264,
+      "learning_rate": 1.821945274905932e-07,
+      "loss": 0.2361,
+      "step": 658
+    },
+    {
+      "epoch": 0.528,
+      "learning_rate": 1.825113498278295e-07,
+      "loss": 0.2564,
+      "step": 660
+    },
+    {
+      "epoch": 0.5296,
+      "learning_rate": 1.8282559840976036e-07,
+      "loss": 0.2588,
+      "step": 662
+    },
+    {
+      "epoch": 0.5312,
+      "learning_rate": 1.8313726343411088e-07,
+      "loss": 0.2208,
+      "step": 664
+    },
+    {
+      "epoch": 0.5328,
+      "learning_rate": 1.8344633517919393e-07,
+      "loss": 0.3172,
+      "step": 666
+    },
+    {
+      "epoch": 0.5344,
+      "learning_rate": 1.8375280400421406e-07,
+      "loss": 0.2993,
+      "step": 668
+    },
+    {
+      "epoch": 0.536,
+      "learning_rate": 1.8405666034956838e-07,
+      "loss": 0.3097,
+      "step": 670
+    },
+    {
+      "epoch": 0.5376,
+      "learning_rate": 1.8435789473714382e-07,
+      "loss": 0.2616,
+      "step": 672
+    },
+    {
+      "epoch": 0.5392,
+      "learning_rate": 1.846564977706138e-07,
+      "loss": 0.233,
+      "step": 674
+    },
+    {
+      "epoch": 0.5408,
+      "learning_rate": 1.8495246013573044e-07,
+      "loss": 0.3568,
+      "step": 676
+    },
+    {
+      "epoch": 0.5424,
+      "learning_rate": 1.8524577260061625e-07,
+      "loss": 0.2176,
+      "step": 678
+    },
+    {
+      "epoch": 0.544,
+      "learning_rate": 1.8553642601605065e-07,
+      "loss": 0.2635,
+      "step": 680
+    },
+    {
+      "epoch": 0.5456,
+      "learning_rate": 1.8582441131575654e-07,
+      "loss": 0.2222,
+      "step": 682
+    },
+    {
+      "epoch": 0.5472,
+      "learning_rate": 1.8610971951668266e-07,
+      "loss": 0.2565,
+      "step": 684
+    },
+    {
+      "epoch": 0.5488,
+      "learning_rate": 1.8639234171928348e-07,
+      "loss": 0.223,
+      "step": 686
+    },
+    {
+      "epoch": 0.5504,
+      "learning_rate": 1.8667226910779764e-07,
+      "loss": 0.2811,
+      "step": 688
+    },
+    {
+      "epoch": 0.552,
+      "learning_rate": 1.869494929505219e-07,
+      "loss": 0.2565,
+      "step": 690
+    },
+    {
+      "epoch": 0.5536,
+      "learning_rate": 1.8722400460008433e-07,
+      "loss": 0.2452,
+      "step": 692
+    },
+    {
+      "epoch": 0.5552,
+      "learning_rate": 1.874957954937137e-07,
+      "loss": 0.3375,
+      "step": 694
+    },
+    {
+      "epoch": 0.5568,
+      "learning_rate": 1.8776485715350664e-07,
+      "loss": 0.2803,
+      "step": 696
+    },
+    {
+      "epoch": 0.5584,
+      "learning_rate": 1.88031181186692e-07,
+      "loss": 0.3566,
+      "step": 698
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.8829475928589266e-07,
+      "loss": 0.259,
+      "step": 700
+    },
+    {
+      "epoch": 0.5616,
+      "learning_rate": 1.8855558322938484e-07,
+      "loss": 0.2542,
+      "step": 702
+    },
+    {
+      "epoch": 0.5632,
+      "learning_rate": 1.888136448813544e-07,
+      "loss": 0.2347,
+      "step": 704
+    },
+    {
+      "epoch": 0.5648,
+      "learning_rate": 1.890689361921506e-07,
+      "loss": 0.306,
+      "step": 706
+    },
+    {
+      "epoch": 0.5664,
+      "learning_rate": 1.8932144919853737e-07,
+      "loss": 0.3177,
+      "step": 708
+    },
+    {
+      "epoch": 0.568,
+      "learning_rate": 1.8957117602394125e-07,
+      "loss": 0.4839,
+      "step": 710
+    },
+    {
+      "epoch": 0.5696,
+      "learning_rate": 1.8981810887869783e-07,
+      "loss": 0.2289,
+      "step": 712
+    },
+    {
+      "epoch": 0.5712,
+      "learning_rate": 1.90062240060294e-07,
+      "loss": 0.2714,
+      "step": 714
+    },
+    {
+      "epoch": 0.5728,
+      "learning_rate": 1.9030356195360865e-07,
+      "loss": 0.2621,
+      "step": 716
+    },
+    {
+      "epoch": 0.5744,
+      "learning_rate": 1.9054206703115016e-07,
+      "loss": 0.3599,
+      "step": 718
+    },
+    {
+      "epoch": 0.576,
+      "learning_rate": 1.9077774785329077e-07,
+      "loss": 0.3381,
+      "step": 720
+    },
+    {
+      "epoch": 0.5776,
+      "learning_rate": 1.9101059706849954e-07,
+      "loss": 0.2561,
+      "step": 722
+    },
+    {
+      "epoch": 0.5792,
+      "learning_rate": 1.9124060741357058e-07,
+      "loss": 0.2973,
+      "step": 724
+    },
+    {
+      "epoch": 0.5808,
+      "learning_rate": 1.914677717138505e-07,
+      "loss": 0.3187,
+      "step": 726
+    },
+    {
+      "epoch": 0.5824,
+      "learning_rate": 1.9169208288346164e-07,
+      "loss": 0.2653,
+      "step": 728
+    },
+    {
+      "epoch": 0.584,
+      "learning_rate": 1.9191353392552342e-07,
+      "loss": 0.2954,
+      "step": 730
+    },
+    {
+      "epoch": 0.5856,
+      "learning_rate": 1.921321179323705e-07,
+      "loss": 0.2759,
+      "step": 732
+    },
+    {
+      "epoch": 0.5872,
+      "learning_rate": 1.9234782808576818e-07,
+      "loss": 0.3331,
+      "step": 734
+    },
+    {
+      "epoch": 0.5888,
+      "learning_rate": 1.9256065765712522e-07,
+      "loss": 0.2122,
+      "step": 736
+    },
+    {
+      "epoch": 0.5904,
+      "learning_rate": 1.9277060000770335e-07,
+      "loss": 0.3301,
+      "step": 738
+    },
+    {
+      "epoch": 0.592,
+      "learning_rate": 1.9297764858882513e-07,
+      "loss": 0.2914,
+      "step": 740
+    },
+    {
+      "epoch": 0.5936,
+      "learning_rate": 1.931817969420772e-07,
+      "loss": 0.2877,
+      "step": 742
+    },
+    {
+      "epoch": 0.5952,
+      "learning_rate": 1.9338303869951264e-07,
+      "loss": 0.2913,
+      "step": 744
+    },
+    {
+      "epoch": 0.5968,
+      "learning_rate": 1.935813675838491e-07,
+      "loss": 0.2441,
+      "step": 746
+    },
+    {
+      "epoch": 0.5984,
+      "learning_rate": 1.9377677740866453e-07,
+      "loss": 0.2351,
+      "step": 748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.9396926207859085e-07,
+      "loss": 0.3086,
+      "step": 750
+    },
+    {
+      "epoch": 0.6016,
+      "learning_rate": 1.94158815589503e-07,
+      "loss": 0.317,
+      "step": 752
+    },
+    {
+      "epoch": 0.6032,
+      "learning_rate": 1.9434543202870718e-07,
+      "loss": 0.2617,
+      "step": 754
+    },
+    {
+      "epoch": 0.6048,
+      "learning_rate": 1.945291055751249e-07,
+      "loss": 0.3554,
+      "step": 756
+    },
+    {
+      "epoch": 0.6064,
+      "learning_rate": 1.9470983049947436e-07,
+      "loss": 0.2201,
+      "step": 758
+    },
+    {
+      "epoch": 0.608,
+      "learning_rate": 1.9488760116444965e-07,
+      "loss": 0.2903,
+      "step": 760
+    },
+    {
+      "epoch": 0.6096,
+      "learning_rate": 1.95062412024896e-07,
+      "loss": 0.3373,
+      "step": 762
+    },
+    {
+      "epoch": 0.6112,
+      "learning_rate": 1.9523425762798328e-07,
+      "loss": 0.2149,
+      "step": 764
+    },
+    {
+      "epoch": 0.6128,
+      "learning_rate": 1.9540313261337575e-07,
+      "loss": 0.2301,
+      "step": 766
+    },
+    {
+      "epoch": 0.6144,
+      "learning_rate": 1.9556903171339955e-07,
+      "loss": 0.2814,
+      "step": 768
+    },
+    {
+      "epoch": 0.616,
+      "learning_rate": 1.957319497532067e-07,
+      "loss": 0.2988,
+      "step": 770
+    },
+    {
+      "epoch": 0.6176,
+      "learning_rate": 1.9589188165093668e-07,
+      "loss": 0.3393,
+      "step": 772
+    },
+    {
+      "epoch": 0.6192,
+      "learning_rate": 1.9604882241787494e-07,
+      "loss": 0.2766,
+      "step": 774
+    },
+    {
+      "epoch": 0.6208,
+      "learning_rate": 1.9620276715860856e-07,
+      "loss": 0.2593,
+      "step": 776
+    },
+    {
+      "epoch": 0.6224,
+      "learning_rate": 1.9635371107117887e-07,
+      "loss": 0.3004,
+      "step": 778
+    },
+    {
+      "epoch": 0.624,
+      "learning_rate": 1.9650164944723114e-07,
+      "loss": 0.2291,
+      "step": 780
+    },
+    {
+      "epoch": 0.6256,
+      "learning_rate": 1.9664657767216175e-07,
+      "loss": 0.2913,
+      "step": 782
+    },
+    {
+      "epoch": 0.6272,
+      "learning_rate": 1.9678849122526183e-07,
+      "loss": 0.2778,
+      "step": 784
+    },
+    {
+      "epoch": 0.6288,
+      "learning_rate": 1.9692738567985847e-07,
+      "loss": 0.2436,
+      "step": 786
+    },
+    {
+      "epoch": 0.6304,
+      "learning_rate": 1.9706325670345273e-07,
+      "loss": 0.2479,
+      "step": 788
+    },
+    {
+      "epoch": 0.632,
+      "learning_rate": 1.9719610005785463e-07,
+      "loss": 0.2173,
+      "step": 790
+    },
+    {
+      "epoch": 0.6336,
+      "learning_rate": 1.973259115993156e-07,
+      "loss": 0.2387,
+      "step": 792
+    },
+    {
+      "epoch": 0.6352,
+      "learning_rate": 1.9745268727865768e-07,
+      "loss": 0.1918,
+      "step": 794
+    },
+    {
+      "epoch": 0.6368,
+      "learning_rate": 1.9757642314139973e-07,
+      "loss": 0.2392,
+      "step": 796
+    },
+    {
+      "epoch": 0.6384,
+      "learning_rate": 1.976971153278808e-07,
+      "loss": 0.2002,
+      "step": 798
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.978147600733805e-07,
+      "loss": 0.2776,
+      "step": 800
+    },
+    {
+      "epoch": 0.6416,
+      "learning_rate": 1.979293537082367e-07,
+      "loss": 0.3198,
+      "step": 802
+    },
+    {
+      "epoch": 0.6432,
+      "learning_rate": 1.9804089265795955e-07,
+      "loss": 0.2648,
+      "step": 804
+    },
+    {
+      "epoch": 0.6448,
+      "learning_rate": 1.9814937344334325e-07,
+      "loss": 0.3544,
+      "step": 806
+    },
+    {
+      "epoch": 0.6464,
+      "learning_rate": 1.9825479268057468e-07,
+      "loss": 0.2656,
+      "step": 808
+    },
+    {
+      "epoch": 0.648,
+      "learning_rate": 1.9835714708133855e-07,
+      "loss": 0.2537,
+      "step": 810
+    },
+    {
+      "epoch": 0.6496,
+      "learning_rate": 1.9845643345292053e-07,
+      "loss": 0.2683,
+      "step": 812
+    },
+    {
+      "epoch": 0.6512,
+      "learning_rate": 1.9855264869830626e-07,
+      "loss": 0.2453,
+      "step": 814
+    },
+    {
+      "epoch": 0.6528,
+      "learning_rate": 1.986457898162784e-07,
+      "loss": 0.2596,
+      "step": 816
+    },
+    {
+      "epoch": 0.6544,
+      "learning_rate": 1.9873585390151e-07,
+      "loss": 0.2432,
+      "step": 818
+    },
+    {
+      "epoch": 0.656,
+      "learning_rate": 1.9882283814465523e-07,
+      "loss": 0.2039,
+      "step": 820
+    },
+    {
+      "epoch": 0.6576,
+      "learning_rate": 1.98906739832437e-07,
+      "loss": 0.2387,
+      "step": 822
+    },
+    {
+      "epoch": 0.6592,
+      "learning_rate": 1.9898755634773155e-07,
+      "loss": 0.2987,
+      "step": 824
+    },
+    {
+      "epoch": 0.6608,
+      "learning_rate": 1.990652851696501e-07,
+      "loss": 0.1978,
+      "step": 826
+    },
+    {
+      "epoch": 0.6624,
+      "learning_rate": 1.9913992387361743e-07,
+      "loss": 0.2961,
+      "step": 828
+    },
+    {
+      "epoch": 0.664,
+      "learning_rate": 1.992114701314478e-07,
+      "loss": 0.3095,
+      "step": 830
+    },
+    {
+      "epoch": 0.6656,
+      "learning_rate": 1.9927992171141704e-07,
+      "loss": 0.3039,
+      "step": 832
+    },
+    {
+      "epoch": 0.6672,
+      "learning_rate": 1.9934527647833275e-07,
+      "loss": 0.2034,
+      "step": 834
+    },
+    {
+      "epoch": 0.6688,
+      "learning_rate": 1.9940753239360044e-07,
+      "loss": 0.2204,
+      "step": 836
+    },
+    {
+      "epoch": 0.6704,
+      "learning_rate": 1.9946668751528738e-07,
+      "loss": 0.2657,
+      "step": 838
+    },
+    {
+      "epoch": 0.672,
+      "learning_rate": 1.995227399981831e-07,
+      "loss": 0.2192,
+      "step": 840
+    },
+    {
+      "epoch": 0.6736,
+      "learning_rate": 1.995756880938569e-07,
+      "loss": 0.2496,
+      "step": 842
+    },
+    {
+      "epoch": 0.6752,
+      "learning_rate": 1.9962553015071247e-07,
+      "loss": 0.463,
+      "step": 844
+    },
+    {
+      "epoch": 0.6768,
+      "learning_rate": 1.9967226461403932e-07,
+      "loss": 0.3489,
+      "step": 846
+    },
+    {
+      "epoch": 0.6784,
+      "learning_rate": 1.9971589002606136e-07,
+      "loss": 0.2598,
+      "step": 848
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.9975640502598244e-07,
+      "loss": 0.2318,
+      "step": 850
+    },
+    {
+      "epoch": 0.6816,
+      "learning_rate": 1.9979380835002844e-07,
+      "loss": 0.2423,
+      "step": 852
+    },
+    {
+      "epoch": 0.6832,
+      "learning_rate": 1.9982809883148719e-07,
+      "loss": 0.2055,
+      "step": 854
+    },
+    {
+      "epoch": 0.6848,
+      "learning_rate": 1.9985927540074453e-07,
+      "loss": 0.1924,
+      "step": 856
+    },
+    {
+      "epoch": 0.6864,
+      "learning_rate": 1.9988733708531768e-07,
+      "loss": 0.2353,
+      "step": 858
+    },
+    {
+      "epoch": 0.688,
+      "learning_rate": 1.9991228300988582e-07,
+      "loss": 0.2445,
+      "step": 860
+    },
+    {
+      "epoch": 0.6896,
+      "learning_rate": 1.999341123963171e-07,
+      "loss": 0.3359,
+      "step": 862
+    },
+    {
+      "epoch": 0.6912,
+      "learning_rate": 1.9995282456369308e-07,
+      "loss": 0.2623,
+      "step": 864
+    },
+    {
+      "epoch": 0.6928,
+      "learning_rate": 1.9996841892833e-07,
+      "loss": 0.2137,
+      "step": 866
+    },
+    {
+      "epoch": 0.6944,
+      "learning_rate": 1.9998089500379676e-07,
+      "loss": 0.2465,
+      "step": 868
+    },
+    {
+      "epoch": 0.696,
+      "learning_rate": 1.9999025240093041e-07,
+      "loss": 0.2112,
+      "step": 870
+    },
+    {
+      "epoch": 0.6976,
+      "learning_rate": 1.9999649082784806e-07,
+      "loss": 0.217,
+      "step": 872
+    },
+    {
+      "epoch": 0.6992,
+      "learning_rate": 1.9999961008995604e-07,
+      "loss": 0.2505,
+      "step": 874
+    },
+    {
+      "epoch": 0.7008,
+      "learning_rate": 1.9999961008995604e-07,
+      "loss": 0.2525,
+      "step": 876
+    },
+    {
+      "epoch": 0.7024,
+      "learning_rate": 1.9999649082784806e-07,
+      "loss": 0.4038,
+      "step": 878
+    },
+    {
+      "epoch": 0.704,
+      "learning_rate": 1.9999025240093041e-07,
+      "loss": 0.2722,
+      "step": 880
+    },
+    {
+      "epoch": 0.7056,
+      "learning_rate": 1.9998089500379676e-07,
+      "loss": 0.2191,
+      "step": 882
+    },
+    {
+      "epoch": 0.7072,
+      "learning_rate": 1.9996841892833e-07,
+      "loss": 0.2937,
+      "step": 884
+    },
+    {
+      "epoch": 0.7088,
+      "learning_rate": 1.9995282456369308e-07,
+      "loss": 0.3898,
+      "step": 886
+    },
+    {
+      "epoch": 0.7104,
+      "learning_rate": 1.999341123963171e-07,
+      "loss": 0.2623,
+      "step": 888
+    },
+    {
+      "epoch": 0.712,
+      "learning_rate": 1.9991228300988582e-07,
+      "loss": 0.2572,
+      "step": 890
+    },
+    {
+      "epoch": 0.7136,
+      "learning_rate": 1.9988733708531768e-07,
+      "loss": 0.2713,
+      "step": 892
+    },
+    {
+      "epoch": 0.7152,
+      "learning_rate": 1.9985927540074453e-07,
+      "loss": 0.2392,
+      "step": 894
+    },
+    {
+      "epoch": 0.7168,
+      "learning_rate": 1.9982809883148719e-07,
+      "loss": 0.2192,
+      "step": 896
+    },
+    {
+      "epoch": 0.7184,
+      "learning_rate": 1.9979380835002844e-07,
+      "loss": 0.2,
+      "step": 898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.9975640502598244e-07,
+      "loss": 0.275,
+      "step": 900
+    },
+    {
+      "epoch": 0.7216,
+      "learning_rate": 1.9971589002606136e-07,
+      "loss": 0.3037,
+      "step": 902
+    },
+    {
+      "epoch": 0.7232,
+      "learning_rate": 1.9967226461403932e-07,
+      "loss": 0.3342,
+      "step": 904
+    },
+    {
+      "epoch": 0.7248,
+      "learning_rate": 1.9962553015071247e-07,
+      "loss": 0.2816,
+      "step": 906
+    },
+    {
+      "epoch": 0.7264,
+      "learning_rate": 1.995756880938569e-07,
+      "loss": 0.2631,
+      "step": 908
+    },
+    {
+      "epoch": 0.728,
+      "learning_rate": 1.995227399981831e-07,
+      "loss": 0.2649,
+      "step": 910
+    },
+    {
+      "epoch": 0.7296,
+      "learning_rate": 1.994666875152874e-07,
+      "loss": 0.2909,
+      "step": 912
+    },
+    {
+      "epoch": 0.7312,
+      "learning_rate": 1.9940753239360047e-07,
+      "loss": 0.2901,
+      "step": 914
+    },
+    {
+      "epoch": 0.7328,
+      "learning_rate": 1.9934527647833278e-07,
+      "loss": 0.2052,
+      "step": 916
+    },
+    {
+      "epoch": 0.7344,
+      "learning_rate": 1.9927992171141704e-07,
+      "loss": 0.1885,
+      "step": 918
+    },
+    {
+      "epoch": 0.736,
+      "learning_rate": 1.992114701314478e-07,
+      "loss": 0.3081,
+      "step": 920
+    },
+    {
+      "epoch": 0.7376,
+      "learning_rate": 1.9913992387361745e-07,
+      "loss": 0.3075,
+      "step": 922
+    },
+    {
+      "epoch": 0.7392,
+      "learning_rate": 1.9906528516965008e-07,
+      "loss": 0.2252,
+      "step": 924
+    },
+    {
+      "epoch": 0.7408,
+      "learning_rate": 1.9898755634773155e-07,
+      "loss": 0.2522,
+      "step": 926
+    },
+    {
+      "epoch": 0.7424,
+      "learning_rate": 1.98906739832437e-07,
+      "loss": 0.2393,
+      "step": 928
+    },
+    {
+      "epoch": 0.744,
+      "learning_rate": 1.9882283814465528e-07,
+      "loss": 0.2741,
+      "step": 930
+    },
+    {
+      "epoch": 0.7456,
+      "learning_rate": 1.9873585390151004e-07,
+      "loss": 0.3561,
+      "step": 932
+    },
+    {
+      "epoch": 0.7472,
+      "learning_rate": 1.9864578981627842e-07,
+      "loss": 0.2442,
+      "step": 934
+    },
+    {
+      "epoch": 0.7488,
+      "learning_rate": 1.9855264869830628e-07,
+      "loss": 0.2441,
+      "step": 936
+    },
+    {
+      "epoch": 0.7504,
+      "learning_rate": 1.9845643345292053e-07,
+      "loss": 0.2302,
+      "step": 938
+    },
+    {
+      "epoch": 0.752,
+      "learning_rate": 1.9835714708133857e-07,
+      "loss": 0.2032,
+      "step": 940
+    },
+    {
+      "epoch": 0.7536,
+      "learning_rate": 1.9825479268057465e-07,
+      "loss": 0.3126,
+      "step": 942
+    },
+    {
+      "epoch": 0.7552,
+      "learning_rate": 1.9814937344334328e-07,
+      "loss": 0.216,
+      "step": 944
+    },
+    {
+      "epoch": 0.7568,
+      "learning_rate": 1.980408926579596e-07,
+      "loss": 0.1926,
+      "step": 946
+    },
+    {
+      "epoch": 0.7584,
+      "learning_rate": 1.9792935370823676e-07,
+      "loss": 0.238,
+      "step": 948
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.9781476007338059e-07,
+      "loss": 0.251,
+      "step": 950
+    },
+    {
+      "epoch": 0.7616,
+      "learning_rate": 1.9769711532788084e-07,
+      "loss": 0.2139,
+      "step": 952
+    },
+    {
+      "epoch": 0.7632,
+      "learning_rate": 1.9757642314139976e-07,
+      "loss": 0.2855,
+      "step": 954
+    },
+    {
+      "epoch": 0.7648,
+      "learning_rate": 1.974526872786577e-07,
+      "loss": 0.2021,
+      "step": 956
+    },
+    {
+      "epoch": 0.7664,
+      "learning_rate": 1.9732591159931564e-07,
+      "loss": 0.1877,
+      "step": 958
+    },
+    {
+      "epoch": 0.768,
+      "learning_rate": 1.9719610005785458e-07,
+      "loss": 0.2033,
+      "step": 960
+    },
+    {
+      "epoch": 0.7696,
+      "learning_rate": 1.970632567034527e-07,
+      "loss": 0.2733,
+      "step": 962
+    },
+    {
+      "epoch": 0.7712,
+      "learning_rate": 1.9692738567985855e-07,
+      "loss": 0.2136,
+      "step": 964
+    },
+    {
+      "epoch": 0.7728,
+      "learning_rate": 1.967884912252619e-07,
+      "loss": 0.2421,
+      "step": 966
+    },
+    {
+      "epoch": 0.7744,
+      "learning_rate": 1.9664657767216178e-07,
+      "loss": 0.2565,
+      "step": 968
+    },
+    {
+      "epoch": 0.776,
+      "learning_rate": 1.9650164944723116e-07,
+      "loss": 0.2527,
+      "step": 970
+    },
+    {
+      "epoch": 0.7776,
+      "learning_rate": 1.9635371107117887e-07,
+      "loss": 0.2322,
+      "step": 972
+    },
+    {
+      "epoch": 0.7792,
+      "learning_rate": 1.9620276715860856e-07,
+      "loss": 0.233,
+      "step": 974
+    },
+    {
+      "epoch": 0.7808,
+      "learning_rate": 1.9604882241787497e-07,
+      "loss": 0.2353,
+      "step": 976
+    },
+    {
+      "epoch": 0.7824,
+      "learning_rate": 1.9589188165093665e-07,
+      "loss": 0.1941,
+      "step": 978
+    },
+    {
+      "epoch": 0.784,
+      "learning_rate": 1.957319497532067e-07,
+      "loss": 0.213,
+      "step": 980
+    },
+    {
+      "epoch": 0.7856,
+      "learning_rate": 1.9556903171339963e-07,
+      "loss": 0.2822,
+      "step": 982
+    },
+    {
+      "epoch": 0.7872,
+      "learning_rate": 1.9540313261337583e-07,
+      "loss": 0.2186,
+      "step": 984
+    },
+    {
+      "epoch": 0.7888,
+      "learning_rate": 1.9523425762798334e-07,
+      "loss": 0.2368,
+      "step": 986
+    },
+    {
+      "epoch": 0.7904,
+      "learning_rate": 1.95062412024896e-07,
+      "loss": 0.2158,
+      "step": 988
+    },
+    {
+      "epoch": 0.792,
+      "learning_rate": 1.9488760116444967e-07,
+      "loss": 0.1976,
+      "step": 990
+    },
+    {
+      "epoch": 0.7936,
+      "learning_rate": 1.947098304994744e-07,
+      "loss": 0.1895,
+      "step": 992
+    },
+    {
+      "epoch": 0.7952,
+      "learning_rate": 1.945291055751249e-07,
+      "loss": 0.2151,
+      "step": 994
+    },
+    {
+      "epoch": 0.7968,
+      "learning_rate": 1.9434543202870723e-07,
+      "loss": 0.2703,
+      "step": 996
+    },
+    {
+      "epoch": 0.7984,
+      "learning_rate": 1.9415881558950296e-07,
+      "loss": 0.2249,
+      "step": 998
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.9396926207859085e-07,
+      "loss": 0.2279,
+      "step": 1000
+    },
+    {
+      "epoch": 0.8016,
+      "learning_rate": 1.9377677740866464e-07,
+      "loss": 0.257,
+      "step": 1002
+    },
+    {
+      "epoch": 0.8032,
+      "learning_rate": 1.9358136758384914e-07,
+      "loss": 0.2405,
+      "step": 1004
+    },
+    {
+      "epoch": 0.8048,
+      "learning_rate": 1.933830386995127e-07,
+      "loss": 0.2457,
+      "step": 1006
+    },
+    {
+      "epoch": 0.8064,
+      "learning_rate": 1.9318179694207726e-07,
+      "loss": 0.3022,
+      "step": 1008
+    },
+    {
+      "epoch": 0.808,
+      "learning_rate": 1.9297764858882516e-07,
+      "loss": 0.261,
+      "step": 1010
+    },
+    {
+      "epoch": 0.8096,
+      "learning_rate": 1.927706000077034e-07,
+      "loss": 0.2684,
+      "step": 1012
+    },
+    {
+      "epoch": 0.8112,
+      "learning_rate": 1.925606576571252e-07,
+      "loss": 0.3317,
+      "step": 1014
+    },
+    {
+      "epoch": 0.8128,
+      "learning_rate": 1.9234782808576818e-07,
+      "loss": 0.2756,
+      "step": 1016
+    },
+    {
+      "epoch": 0.8144,
+      "learning_rate": 1.9213211793237063e-07,
+      "loss": 0.2387,
+      "step": 1018
+    },
+    {
+      "epoch": 0.816,
+      "learning_rate": 1.9191353392552347e-07,
+      "loss": 0.2796,
+      "step": 1020
+    },
+    {
+      "epoch": 0.8176,
+      "learning_rate": 1.916920828834617e-07,
+      "loss": 0.1965,
+      "step": 1022
+    },
+    {
+      "epoch": 0.8192,
+      "learning_rate": 1.9146777171385056e-07,
+      "loss": 0.1846,
+      "step": 1024
+    },
+    {
+      "epoch": 0.8208,
+      "learning_rate": 1.9124060741357063e-07,
+      "loss": 0.2228,
+      "step": 1026
+    },
+    {
+      "epoch": 0.8224,
+      "learning_rate": 1.910105970684996e-07,
+      "loss": 0.2156,
+      "step": 1028
+    },
+    {
+      "epoch": 0.824,
+      "learning_rate": 1.9077774785329082e-07,
+      "loss": 0.2583,
+      "step": 1030
+    },
+    {
+      "epoch": 0.8256,
+      "learning_rate": 1.9054206703115013e-07,
+      "loss": 0.1962,
+      "step": 1032
+    },
+    {
+      "epoch": 0.8272,
+      "learning_rate": 1.9030356195360867e-07,
+      "loss": 0.2498,
+      "step": 1034
+    },
+    {
+      "epoch": 0.8288,
+      "learning_rate": 1.9006224006029413e-07,
+      "loss": 0.2349,
+      "step": 1036
+    },
+    {
+      "epoch": 0.8304,
+      "learning_rate": 1.8981810887869793e-07,
+      "loss": 0.214,
+      "step": 1038
+    },
+    {
+      "epoch": 0.832,
+      "learning_rate": 1.895711760239413e-07,
+      "loss": 0.2335,
+      "step": 1040
+    },
+    {
+      "epoch": 0.8336,
+      "learning_rate": 1.8932144919853742e-07,
+      "loss": 0.2339,
+      "step": 1042
+    },
+    {
+      "epoch": 0.8352,
+      "learning_rate": 1.8906893619215066e-07,
+      "loss": 0.236,
+      "step": 1044
+    },
+    {
+      "epoch": 0.8368,
+      "learning_rate": 1.8881364488135442e-07,
+      "loss": 0.2198,
+      "step": 1046
+    },
+    {
+      "epoch": 0.8384,
+      "learning_rate": 1.885555832293849e-07,
+      "loss": 0.3242,
+      "step": 1048
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.8829475928589266e-07,
+      "loss": 0.2632,
+      "step": 1050
+    },
+    {
+      "epoch": 0.8416,
+      "learning_rate": 1.8803118118669196e-07,
+      "loss": 0.2865,
+      "step": 1052
+    },
+    {
+      "epoch": 0.8432,
+      "learning_rate": 1.8776485715350677e-07,
+      "loss": 0.2837,
+      "step": 1054
+    },
+    {
+      "epoch": 0.8448,
+      "learning_rate": 1.8749579549371385e-07,
+      "loss": 0.2462,
+      "step": 1056
+    },
+    {
+      "epoch": 0.8464,
+      "learning_rate": 1.8722400460008435e-07,
+      "loss": 0.3001,
+      "step": 1058
+    },
+    {
+      "epoch": 0.848,
+      "learning_rate": 1.8694949295052192e-07,
+      "loss": 0.268,
+      "step": 1060
+    },
+    {
+      "epoch": 0.8496,
+      "learning_rate": 1.8667226910779766e-07,
+      "loss": 0.2089,
+      "step": 1062
+    },
+    {
+      "epoch": 0.8512,
+      "learning_rate": 1.863923417192835e-07,
+      "loss": 0.2341,
+      "step": 1064
+    },
+    {
+      "epoch": 0.8528,
+      "learning_rate": 1.861097195166826e-07,
+      "loss": 0.4218,
+      "step": 1066
+    },
+    {
+      "epoch": 0.8544,
+      "learning_rate": 1.858244113157566e-07,
+      "loss": 0.3543,
+      "step": 1068
+    },
+    {
+      "epoch": 0.856,
+      "learning_rate": 1.8553642601605078e-07,
+      "loss": 0.2358,
+      "step": 1070
+    },
+    {
+      "epoch": 0.8576,
+      "learning_rate": 1.852457726006163e-07,
+      "loss": 0.2376,
+      "step": 1072
+    },
+    {
+      "epoch": 0.8592,
+      "learning_rate": 1.849524601357306e-07,
+      "loss": 0.2616,
+      "step": 1074
+    },
+    {
+      "epoch": 0.8608,
+      "learning_rate": 1.8465649777061386e-07,
+      "loss": 0.2106,
+      "step": 1076
+    },
+    {
+      "epoch": 0.8624,
+      "learning_rate": 1.843578947371439e-07,
+      "loss": 0.2305,
+      "step": 1078
+    },
+    {
+      "epoch": 0.864,
+      "learning_rate": 1.8405666034956844e-07,
+      "loss": 0.2159,
+      "step": 1080
+    },
+    {
+      "epoch": 0.8656,
+      "learning_rate": 1.8375280400421412e-07,
+      "loss": 0.2616,
+      "step": 1082
+    },
+    {
+      "epoch": 0.8672,
+      "learning_rate": 1.8344633517919388e-07,
+      "loss": 0.2485,
+      "step": 1084
+    },
+    {
+      "epoch": 0.8688,
+      "learning_rate": 1.8313726343411085e-07,
+      "loss": 0.2046,
+      "step": 1086
+    },
+    {
+      "epoch": 0.8704,
+      "learning_rate": 1.828255984097605e-07,
+      "loss": 0.2305,
+      "step": 1088
+    },
+    {
+      "epoch": 0.872,
+      "learning_rate": 1.8251134982782963e-07,
+      "loss": 0.2428,
+      "step": 1090
+    },
+    {
+      "epoch": 0.8736,
+      "learning_rate": 1.8219452749059333e-07,
+      "loss": 0.1977,
+      "step": 1092
+    },
+    {
+      "epoch": 0.8752,
+      "learning_rate": 1.8187514128060954e-07,
+      "loss": 0.2048,
+      "step": 1094
+    },
+    {
+      "epoch": 0.8768,
+      "learning_rate": 1.815532011604098e-07,
+      "loss": 0.2709,
+      "step": 1096
+    },
+    {
+      "epoch": 0.8784,
+      "learning_rate": 1.8122871717218972e-07,
+      "loss": 0.2154,
+      "step": 1098
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.8090169943749475e-07,
+      "loss": 0.2909,
+      "step": 1100
+    },
+    {
+      "epoch": 0.8816,
+      "learning_rate": 1.8057215815690486e-07,
+      "loss": 0.2283,
+      "step": 1102
+    },
+    {
+      "epoch": 0.8832,
+      "learning_rate": 1.802401036097166e-07,
+      "loss": 0.2326,
+      "step": 1104
+    },
+    {
+      "epoch": 0.8848,
+      "learning_rate": 1.7990554615362207e-07,
+      "loss": 0.2652,
+      "step": 1106
+    },
+    {
+      "epoch": 0.8864,
+      "learning_rate": 1.7956849622438565e-07,
+      "loss": 0.2498,
+      "step": 1108
+    },
+    {
+      "epoch": 0.888,
+      "learning_rate": 1.7922896433551912e-07,
+      "loss": 0.223,
+      "step": 1110
+    },
+    {
+      "epoch": 0.8896,
+      "learning_rate": 1.7888696107795344e-07,
+      "loss": 0.2787,
+      "step": 1112
+    },
+    {
+      "epoch": 0.8912,
+      "learning_rate": 1.7854249711970827e-07,
+      "loss": 0.2278,
+      "step": 1114
+    },
+    {
+      "epoch": 0.8928,
+      "learning_rate": 1.7819558320555897e-07,
+      "loss": 0.2196,
+      "step": 1116
+    },
+    {
+      "epoch": 0.8944,
+      "learning_rate": 1.7784623015670234e-07,
+      "loss": 0.3112,
+      "step": 1118
+    },
+    {
+      "epoch": 0.896,
+      "learning_rate": 1.77494448870418e-07,
+      "loss": 0.2557,
+      "step": 1120
+    },
+    {
+      "epoch": 0.8976,
+      "learning_rate": 1.7714025031972893e-07,
+      "loss": 0.3011,
+      "step": 1122
+    },
+    {
+      "epoch": 0.8992,
+      "learning_rate": 1.767836455530598e-07,
+      "loss": 0.2202,
+      "step": 1124
+    },
+    {
+      "epoch": 0.9008,
+      "learning_rate": 1.76424645693891e-07,
+      "loss": 0.2767,
+      "step": 1126
+    },
+    {
+      "epoch": 0.9024,
+      "learning_rate": 1.7606326194041282e-07,
+      "loss": 0.2585,
+      "step": 1128
+    },
+    {
+      "epoch": 0.904,
+      "learning_rate": 1.756995055651757e-07,
+      "loss": 0.2949,
+      "step": 1130
+    },
+    {
+      "epoch": 0.9056,
+      "learning_rate": 1.7533338791473873e-07,
+      "loss": 0.2458,
+      "step": 1132
+    },
+    {
+      "epoch": 0.9072,
+      "learning_rate": 1.7496492040931544e-07,
+      "loss": 0.2316,
+      "step": 1134
+    },
+    {
+      "epoch": 0.9088,
+      "learning_rate": 1.7459411454241817e-07,
+      "loss": 0.2565,
+      "step": 1136
+    },
+    {
+      "epoch": 0.9104,
+      "learning_rate": 1.7422098188049881e-07,
+      "loss": 0.2546,
+      "step": 1138
+    },
+    {
+      "epoch": 0.912,
+      "learning_rate": 1.7384553406258825e-07,
+      "loss": 0.218,
+      "step": 1140
+    },
+    {
+      "epoch": 0.9136,
+      "learning_rate": 1.734677827999343e-07,
+      "loss": 0.2779,
+      "step": 1142
+    },
+    {
+      "epoch": 0.9152,
+      "learning_rate": 1.7308773987563407e-07,
+      "loss": 0.2232,
+      "step": 1144
+    },
+    {
+      "epoch": 0.9168,
+      "learning_rate": 1.7270541714426926e-07,
+      "loss": 0.2643,
+      "step": 1146
+    },
+    {
+      "epoch": 0.9184,
+      "learning_rate": 1.723208265315342e-07,
+      "loss": 0.4043,
+      "step": 1148
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.7193398003386513e-07,
+      "loss": 0.2473,
+      "step": 1150
+    },
+    {
+      "epoch": 0.9216,
+      "learning_rate": 1.7154488971806523e-07,
+      "loss": 0.2704,
+      "step": 1152
+    },
+    {
+      "epoch": 0.9232,
+      "learning_rate": 1.7115356772092847e-07,
+      "loss": 0.2787,
+      "step": 1154
+    },
+    {
+      "epoch": 0.9248,
+      "learning_rate": 1.707600262488615e-07,
+      "loss": 0.2381,
+      "step": 1156
+    },
+    {
+      "epoch": 0.9264,
+      "learning_rate": 1.70364277577502e-07,
+      "loss": 0.2083,
+      "step": 1158
+    },
+    {
+      "epoch": 0.928,
+      "learning_rate": 1.699663340513367e-07,
+      "loss": 0.1927,
+      "step": 1160
+    },
+    {
+      "epoch": 0.9296,
+      "learning_rate": 1.6956620808331512e-07,
+      "loss": 0.196,
+      "step": 1162
+    },
+    {
+      "epoch": 0.9312,
+      "learning_rate": 1.6916391215446407e-07,
+      "loss": 0.3946,
+      "step": 1164
+    },
+    {
+      "epoch": 0.9328,
+      "learning_rate": 1.6875945881349682e-07,
+      "loss": 0.2263,
+      "step": 1166
+    },
+    {
+      "epoch": 0.9344,
+      "learning_rate": 1.6835286067642224e-07,
+      "loss": 0.2346,
+      "step": 1168
+    },
+    {
+      "epoch": 0.936,
+      "learning_rate": 1.6794413042615165e-07,
+      "loss": 0.2413,
+      "step": 1170
+    },
+    {
+      "epoch": 0.9376,
+      "learning_rate": 1.6753328081210248e-07,
+      "loss": 0.2885,
+      "step": 1172
+    },
+    {
+      "epoch": 0.9392,
+      "learning_rate": 1.671203246498009e-07,
+      "loss": 0.1995,
+      "step": 1174
+    },
+    {
+      "epoch": 0.9408,
+      "learning_rate": 1.667052748204824e-07,
+      "loss": 0.2294,
+      "step": 1176
+    },
+    {
+      "epoch": 0.9424,
+      "learning_rate": 1.6628814427068963e-07,
+      "loss": 0.2662,
+      "step": 1178
+    },
+    {
+      "epoch": 0.944,
+      "learning_rate": 1.658689460118682e-07,
+      "loss": 0.2065,
+      "step": 1180
+    },
+    {
+      "epoch": 0.9456,
+      "learning_rate": 1.6544769311996153e-07,
+      "loss": 0.204,
+      "step": 1182
+    },
+    {
+      "epoch": 0.9472,
+      "learning_rate": 1.6502439873500294e-07,
+      "loss": 0.244,
+      "step": 1184
+    },
+    {
+      "epoch": 0.9488,
+      "learning_rate": 1.6459907606070517e-07,
+      "loss": 0.2435,
+      "step": 1186
+    },
+    {
+      "epoch": 0.9504,
+      "learning_rate": 1.641717383640488e-07,
+      "loss": 0.2202,
+      "step": 1188
+    },
+    {
+      "epoch": 0.952,
+      "learning_rate": 1.6374239897486896e-07,
+      "loss": 0.3329,
+      "step": 1190
+    },
+    {
+      "epoch": 0.9536,
+      "learning_rate": 1.6331107128543847e-07,
+      "loss": 0.2756,
+      "step": 1192
+    },
+    {
+      "epoch": 0.9552,
+      "learning_rate": 1.6287776875005145e-07,
+      "loss": 0.311,
+      "step": 1194
+    },
+    {
+      "epoch": 0.9568,
+      "learning_rate": 1.6244250488460166e-07,
+      "loss": 0.2488,
+      "step": 1196
+    },
+    {
+      "epoch": 0.9584,
+      "learning_rate": 1.6200529326616338e-07,
+      "loss": 0.3105,
+      "step": 1198
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.6156614753256583e-07,
+      "loss": 0.3013,
+      "step": 1200
+    },
+    {
+      "epoch": 0.9616,
+      "learning_rate": 1.611250813819692e-07,
+      "loss": 0.3105,
+      "step": 1202
+    },
+    {
+      "epoch": 0.9632,
+      "learning_rate": 1.6068210857243628e-07,
+      "loss": 0.2465,
+      "step": 1204
+    },
+    {
+      "epoch": 0.9648,
+      "learning_rate": 1.602372429215038e-07,
+      "loss": 0.2062,
+      "step": 1206
+    },
+    {
+      "epoch": 0.9664,
+      "learning_rate": 1.5979049830575185e-07,
+      "loss": 0.1815,
+      "step": 1208
+    },
+    {
+      "epoch": 0.968,
+      "learning_rate": 1.5934188866037012e-07,
+      "loss": 0.1858,
+      "step": 1210
+    },
+    {
+      "epoch": 0.9696,
+      "learning_rate": 1.5889142797872405e-07,
+      "loss": 0.2574,
+      "step": 1212
+    },
+    {
+      "epoch": 0.9712,
+      "learning_rate": 1.5843913031191726e-07,
+      "loss": 0.2405,
+      "step": 1214
+    },
+    {
+      "epoch": 0.9728,
+      "learning_rate": 1.57985009768355e-07,
+      "loss": 0.2637,
+      "step": 1216
+    },
+    {
+      "epoch": 0.9744,
+      "learning_rate": 1.5752908051330237e-07,
+      "loss": 0.2814,
+      "step": 1218
+    },
+    {
+      "epoch": 0.976,
+      "learning_rate": 1.5707135676844315e-07,
+      "loss": 0.2238,
+      "step": 1220
+    },
+    {
+      "epoch": 0.9776,
+      "learning_rate": 1.5661185281143667e-07,
+      "loss": 0.2017,
+      "step": 1222
+    },
+    {
+      "epoch": 0.9792,
+      "learning_rate": 1.561505829754715e-07,
+      "loss": 0.1962,
+      "step": 1224
+    },
+    {
+      "epoch": 0.9808,
+      "learning_rate": 1.5568756164881872e-07,
+      "loss": 0.2456,
+      "step": 1226
+    },
+    {
+      "epoch": 0.9824,
+      "learning_rate": 1.5522280327438382e-07,
+      "loss": 0.2372,
+      "step": 1228
+    },
+    {
+      "epoch": 0.984,
+      "learning_rate": 1.5475632234925518e-07,
+      "loss": 0.2173,
+      "step": 1230
+    },
+    {
+      "epoch": 0.9856,
+      "learning_rate": 1.5428813342425193e-07,
+      "loss": 0.2561,
+      "step": 1232
+    },
+    {
+      "epoch": 0.9872,
+      "learning_rate": 1.5381825110347078e-07,
+      "loss": 0.2815,
+      "step": 1234
+    },
+    {
+      "epoch": 0.9888,
+      "learning_rate": 1.5334669004383035e-07,
+      "loss": 0.2162,
+      "step": 1236
+    },
+    {
+      "epoch": 0.9904,
+      "learning_rate": 1.528734649546133e-07,
+      "loss": 0.2158,
+      "step": 1238
+    },
+    {
+      "epoch": 0.992,
+      "learning_rate": 1.5239859059700791e-07,
+      "loss": 0.2662,
+      "step": 1240
+    },
+    {
+      "epoch": 0.9936,
+      "learning_rate": 1.5192208178364815e-07,
+      "loss": 0.1906,
+      "step": 1242
+    },
+    {
+      "epoch": 0.9952,
+      "learning_rate": 1.5144395337815056e-07,
+      "loss": 0.2795,
+      "step": 1244
+    },
+    {
+      "epoch": 0.9968,
+      "learning_rate": 1.509642202946517e-07,
+      "loss": 0.2255,
+      "step": 1246
+    },
+    {
+      "epoch": 0.9984,
+      "learning_rate": 1.5048289749734229e-07,
+      "loss": 0.2472,
+      "step": 1248
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.5000000000000013e-07,
+      "loss": 0.2388,
+      "step": 1250
+    },
+    {
+      "epoch": 1.0,
+      "step": 1250,
+      "total_flos": 5169738430808064.0,
+      "train_loss": 0.5022303078651428,
+      "train_runtime": 2434.8082,
+      "train_samples_per_second": 8.214,
+      "train_steps_per_second": 0.513
+    }
+  ],
+  "logging_steps": 2,
+  "max_steps": 1250,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "stateful_callbacks": {},
+  "total_flos": 5169738430808064.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.0625_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a395983148e841925f34c7c0cfd8b63187a310e2969bf1e20c41ddee0c63c606
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.0625_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4f0b400b3082889966a834d6f67f8d4fb8e943ed2009f4f98f817e7bcbbd366
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.0625_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c98c7a4c1c2c2013fb3e22d876e005cea39e2661d2ec5fc0bfe0883a603c157a
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.0625_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f0707a245d11105843b3615a186addfe72a7a46bc49ce56cae1aa937a89979a
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.125_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7ca4d185778a1af4fa291f41f9a19e1959bec13b785586e0e4475bff2d2aea7
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.125_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6be9ba8883d0cab04414ab090ce2fe33a336d2e08ae2e2966ec72904bb861288
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.125_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0eeb1f0905d128a760044b5073c4f1bca845bc00cd2183b8cf1121e6a6990878
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.125_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3fa23db415283ebb5d2d01ac1af7b36e84697f685208fee9cb41d6f31fe93dfc
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.25_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0127d501b09a52600d44d3dd289cb26c6841b67deb44f452900d1260a4e97b1c
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.25_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fba84de3e232fc340d36a35a5f2375b09e00f0d08d8b9551244ab6d24652d7c
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.25_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c9552c26ab82d3b433fbb6ca93b18103414faf16cdc363daf24ccd1c51a5f79
+size 685911174

qwen_checkpoints/client_states_QWENv2_v7_NEURIPS_3_2_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_infoBatch_scenario18_10000_random0.25_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6316e9f4e5a23ebf467229c3461183e9f14ca77867c63697a2a0b1b0c6d4ce83
+size 685911174

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_0625_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6220b636661eb24bf0c3f7012564cee68e8aaa142cf78809c9d1613f9bae13b
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_0625_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7380ffc05da3fdde19971f9c268440231a5a3b2c4709596e81f660a398abe3ac
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_0625_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:967c60d8f41d55c1f7fa8f492a7e950d00869e2da050446b960fe148fb885d4f
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_0625_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a28adf7dc1eef172aed023b5008ede9860a6f9732b3541cabc55b6abcbbc8b59
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_125_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3098a2cf2721e33edb2a408d120cf5289190249b2a6088b8f80ef67c59533cb
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_125_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6c162d32950f37fc1f4b083f969f8f1f00ce9d1612ea3bcef9202d3778870eb
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_125_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9a0b934f7e66383fb1f004b3bb9cbeb88525985541523537c1d20e0446243d9
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_125_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e4cf128b39d34f8832efd158aa817cef2b6d0986a8d509c43dfe228ff54b004
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_25_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aaba2be0eefc7eb8087b2763cb033b02380bc8c85990ac2d3c901558d3b0808d
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_25_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bf47b6f98f209af2521c3d7cee21119f77f50dfc9f6d112cbc777bd71d16c76
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_25_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e61bb3d1e73a092314b8ba052e7c2a445db95e431f7ab8e19368bb8bd724b6d
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario16_10000_random0_25_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b20c264e1ce7ba668a99fbf68b3e7af2b58b7c1d7c78120f7e932a66579b6d8d
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_0625_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c461b38936d41eeccc96e340a2f135fbb355881f3542bae3d4d7e0631f53b52c
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_0625_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9e003c3f140ffe8b4073432544f7459106978435af7cf356cf05486f2947883
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_0625_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ae611a6931710bd083f44304b0b2fec55b06079eea3cb6d936efbece5c404d1
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_0625_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86b24341c0663ab0379f46d9ae08e35b0945bba0b79cd1fc109d06d263df78ed
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_125_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d32f8c0b424d18e60e6c99d9fe60b2870909414791d5f7bd66a588316612a37
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_125_seed1/server_model_round1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9aa8e6f25bf978323c5956c2c06f4c4ee4d800e7b68b54f6b1b9652420e5728a
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_125_seed1/server_model_round2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82147e018cc6c50c1f5311bb5ea933552a92e3845d6e5041aaad2ddc6b1dbe94
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_125_seed1/server_model_round3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47b1bad9b830bcf69899a571479fb0e5ed5d037fc5905b4a0f805c4ee7c217d3
+size 646064562

qwen_checkpoints/client_states_QWENv4_NEURIPS_DISJOINT_Memonly_LORA_llava_lr2e-5_bs1_gradacc32_iter0_125_divbs_scenario20_10000_random0_25_seed1/server_model_round0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a05404dc62e1782d46ce66e2ee8ec4192db0b3236ee0319c6f92b47ecfc62b4a
+size 646064562