Training in progress, epoch 1

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:63856058fa8651f4545fdd1951eac4e69def138dd3638485b5c6b0fbeadf512f
+oid sha256:682e324df29bab7ba1f448be187564d8e0b4fdfc12c597d06dda3b1465999484
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a37da6ec0243d7506685ec8b73fd28f3997b312ed1ab67fabd66455db7aabd0e
+oid sha256:acbdcc136dca480f1c8f135fa2d6369e270e511b9181b888571326595ef510af
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:690d0d9b8338ae94ab8bed163c0fca58bb0eeb04af1f343cbd20e142a84446f6
+oid sha256:ea5f17881874218445b14c78053e08eb7d0707c556c6b5930904171b33af2f3c
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5e4dfcae80c9bfbb340f031583ab55db821b71440c91e1a731c997b146cced5a
+oid sha256:096c0b5f12bff1cacae308936236b1a192d1173545cb8bbd28deaff837b0e795
 size 1089994880

trainer_log.jsonl

@@ -183,3 +183,187 @@
 {"current_steps": 183, "total_steps": 915, "loss": 0.9922, "lr": 7.761086094280291e-05, "epoch": 0.9986357435197817, "percentage": 20.0, "elapsed_time": "2:39:32", "remaining_time": "10:38:08"}
 {"current_steps": 184, "total_steps": 915, "loss": 1.6748, "lr": 7.755860756011531e-05, "epoch": 1.004092769440655, "percentage": 20.11, "elapsed_time": "2:41:39", "remaining_time": "10:42:14"}
 {"current_steps": 185, "total_steps": 915, "loss": 0.9345, "lr": 7.750580689811683e-05, "epoch": 1.009549795361528, "percentage": 20.22, "elapsed_time": "2:42:30", "remaining_time": "10:41:15"}
+{"current_steps": 186, "total_steps": 915, "loss": 0.9733, "lr": 7.745245972618396e-05, "epoch": 1.015006821282401, "percentage": 20.33, "elapsed_time": "2:43:21", "remaining_time": "10:40:16"}
+{"current_steps": 187, "total_steps": 915, "loss": 0.9699, "lr": 7.739856682165661e-05, "epoch": 1.0204638472032743, "percentage": 20.44, "elapsed_time": "2:44:13", "remaining_time": "10:39:18"}
+{"current_steps": 188, "total_steps": 915, "loss": 0.9756, "lr": 7.73441289698267e-05, "epoch": 1.0259208731241474, "percentage": 20.55, "elapsed_time": "2:45:04", "remaining_time": "10:38:21"}
+{"current_steps": 189, "total_steps": 915, "loss": 0.9735, "lr": 7.728914696392677e-05, "epoch": 1.0313778990450204, "percentage": 20.66, "elapsed_time": "2:45:55", "remaining_time": "10:37:23"}
+{"current_steps": 190, "total_steps": 915, "loss": 0.9362, "lr": 7.723362160511844e-05, "epoch": 1.0368349249658937, "percentage": 20.77, "elapsed_time": "2:46:47", "remaining_time": "10:36:25"}
+{"current_steps": 191, "total_steps": 915, "loss": 0.938, "lr": 7.717755370248069e-05, "epoch": 1.0422919508867667, "percentage": 20.87, "elapsed_time": "2:47:39", "remaining_time": "10:35:29"}
+{"current_steps": 192, "total_steps": 915, "loss": 0.9506, "lr": 7.712094407299809e-05, "epoch": 1.0477489768076398, "percentage": 20.98, "elapsed_time": "2:48:47", "remaining_time": "10:35:35"}
+{"current_steps": 193, "total_steps": 915, "loss": 0.963, "lr": 7.706379354154891e-05, "epoch": 1.053206002728513, "percentage": 21.09, "elapsed_time": "2:49:38", "remaining_time": "10:34:37"}
+{"current_steps": 194, "total_steps": 915, "loss": 0.93, "lr": 7.700610294089309e-05, "epoch": 1.058663028649386, "percentage": 21.2, "elapsed_time": "2:50:29", "remaining_time": "10:33:38"}
+{"current_steps": 195, "total_steps": 915, "loss": 0.9758, "lr": 7.694787311166009e-05, "epoch": 1.0641200545702592, "percentage": 21.31, "elapsed_time": "2:51:20", "remaining_time": "10:32:40"}
+{"current_steps": 196, "total_steps": 915, "loss": 0.9634, "lr": 7.688910490233664e-05, "epoch": 1.0695770804911324, "percentage": 21.42, "elapsed_time": "2:52:12", "remaining_time": "10:31:42"}
+{"current_steps": 197, "total_steps": 915, "loss": 0.9435, "lr": 7.682979916925445e-05, "epoch": 1.0750341064120055, "percentage": 21.53, "elapsed_time": "2:53:03", "remaining_time": "10:30:44"}
+{"current_steps": 198, "total_steps": 915, "loss": 0.9471, "lr": 7.676995677657759e-05, "epoch": 1.0804911323328785, "percentage": 21.64, "elapsed_time": "2:53:54", "remaining_time": "10:29:45"}
+{"current_steps": 199, "total_steps": 915, "loss": 0.9668, "lr": 7.670957859629003e-05, "epoch": 1.0859481582537518, "percentage": 21.75, "elapsed_time": "2:54:45", "remaining_time": "10:28:47"}
+{"current_steps": 200, "total_steps": 915, "loss": 0.9465, "lr": 7.664866550818289e-05, "epoch": 1.0914051841746248, "percentage": 21.86, "elapsed_time": "2:55:36", "remaining_time": "10:27:49"}
+{"current_steps": 201, "total_steps": 915, "loss": 0.9431, "lr": 7.65872183998416e-05, "epoch": 1.096862210095498, "percentage": 21.97, "elapsed_time": "2:56:28", "remaining_time": "10:26:51"}
+{"current_steps": 202, "total_steps": 915, "loss": 0.9467, "lr": 7.652523816663298e-05, "epoch": 1.1023192360163712, "percentage": 22.08, "elapsed_time": "2:57:19", "remaining_time": "10:25:54"}
+{"current_steps": 203, "total_steps": 915, "loss": 0.9426, "lr": 7.646272571169221e-05, "epoch": 1.1077762619372442, "percentage": 22.19, "elapsed_time": "2:58:10", "remaining_time": "10:24:56"}
+{"current_steps": 204, "total_steps": 915, "loss": 0.9429, "lr": 7.639968194590961e-05, "epoch": 1.1132332878581173, "percentage": 22.3, "elapsed_time": "2:59:02", "remaining_time": "10:23:59"}
+{"current_steps": 205, "total_steps": 915, "loss": 0.9375, "lr": 7.633610778791746e-05, "epoch": 1.1186903137789905, "percentage": 22.4, "elapsed_time": "2:59:53", "remaining_time": "10:23:01"}
+{"current_steps": 206, "total_steps": 915, "loss": 0.9592, "lr": 7.627200416407656e-05, "epoch": 1.1241473396998636, "percentage": 22.51, "elapsed_time": "3:00:46", "remaining_time": "10:22:10"}
+{"current_steps": 207, "total_steps": 915, "loss": 0.9535, "lr": 7.620737200846271e-05, "epoch": 1.1296043656207366, "percentage": 22.62, "elapsed_time": "3:01:53", "remaining_time": "10:22:07"}
+{"current_steps": 208, "total_steps": 915, "loss": 0.9586, "lr": 7.614221226285317e-05, "epoch": 1.13506139154161, "percentage": 22.73, "elapsed_time": "3:02:44", "remaining_time": "10:21:08"}
+{"current_steps": 209, "total_steps": 915, "loss": 0.9449, "lr": 7.607652587671284e-05, "epoch": 1.140518417462483, "percentage": 22.84, "elapsed_time": "3:03:35", "remaining_time": "10:20:10"}
+{"current_steps": 210, "total_steps": 915, "loss": 0.9328, "lr": 7.601031380718053e-05, "epoch": 1.145975443383356, "percentage": 22.95, "elapsed_time": "3:04:26", "remaining_time": "10:19:12"}
+{"current_steps": 211, "total_steps": 915, "loss": 0.9559, "lr": 7.594357701905493e-05, "epoch": 1.1514324693042293, "percentage": 23.06, "elapsed_time": "3:05:17", "remaining_time": "10:18:15"}
+{"current_steps": 212, "total_steps": 915, "loss": 0.9624, "lr": 7.58763164847806e-05, "epoch": 1.1568894952251023, "percentage": 23.17, "elapsed_time": "3:06:09", "remaining_time": "10:17:17"}
+{"current_steps": 213, "total_steps": 915, "loss": 0.9551, "lr": 7.580853318443378e-05, "epoch": 1.1623465211459754, "percentage": 23.28, "elapsed_time": "3:07:00", "remaining_time": "10:16:20"}
+{"current_steps": 214, "total_steps": 915, "loss": 0.9638, "lr": 7.574022810570811e-05, "epoch": 1.1678035470668486, "percentage": 23.39, "elapsed_time": "3:07:51", "remaining_time": "10:15:23"}
+{"current_steps": 215, "total_steps": 915, "loss": 0.9428, "lr": 7.567140224390026e-05, "epoch": 1.1732605729877217, "percentage": 23.5, "elapsed_time": "3:08:43", "remaining_time": "10:14:25"}
+{"current_steps": 216, "total_steps": 915, "loss": 0.9538, "lr": 7.560205660189536e-05, "epoch": 1.1787175989085947, "percentage": 23.61, "elapsed_time": "3:09:34", "remaining_time": "10:13:28"}
+{"current_steps": 217, "total_steps": 915, "loss": 0.9405, "lr": 7.553219219015247e-05, "epoch": 1.184174624829468, "percentage": 23.72, "elapsed_time": "3:10:26", "remaining_time": "10:12:32"}
+{"current_steps": 218, "total_steps": 915, "loss": 0.9457, "lr": 7.546181002668985e-05, "epoch": 1.189631650750341, "percentage": 23.83, "elapsed_time": "3:11:17", "remaining_time": "10:11:35"}
+{"current_steps": 219, "total_steps": 915, "loss": 0.952, "lr": 7.539091113707003e-05, "epoch": 1.195088676671214, "percentage": 23.93, "elapsed_time": "3:12:08", "remaining_time": "10:10:38"}
+{"current_steps": 220, "total_steps": 915, "loss": 0.9188, "lr": 7.5319496554385e-05, "epoch": 1.2005457025920874, "percentage": 24.04, "elapsed_time": "3:12:59", "remaining_time": "10:09:41"}
+{"current_steps": 221, "total_steps": 915, "loss": 0.9644, "lr": 7.524756731924103e-05, "epoch": 1.2060027285129604, "percentage": 24.15, "elapsed_time": "3:13:52", "remaining_time": "10:08:49"}
+{"current_steps": 222, "total_steps": 915, "loss": 0.9628, "lr": 7.517512447974361e-05, "epoch": 1.2114597544338335, "percentage": 24.26, "elapsed_time": "3:14:57", "remaining_time": "10:08:36"}
+{"current_steps": 223, "total_steps": 915, "loss": 0.9674, "lr": 7.510216909148214e-05, "epoch": 1.2169167803547067, "percentage": 24.37, "elapsed_time": "3:15:48", "remaining_time": "10:07:38"}
+{"current_steps": 224, "total_steps": 915, "loss": 0.9603, "lr": 7.502870221751448e-05, "epoch": 1.2223738062755798, "percentage": 24.48, "elapsed_time": "3:16:40", "remaining_time": "10:06:41"}
+{"current_steps": 225, "total_steps": 915, "loss": 0.9567, "lr": 7.495472492835159e-05, "epoch": 1.2278308321964528, "percentage": 24.59, "elapsed_time": "3:17:31", "remaining_time": "10:05:44"}
+{"current_steps": 226, "total_steps": 915, "loss": 0.9626, "lr": 7.488023830194185e-05, "epoch": 1.233287858117326, "percentage": 24.7, "elapsed_time": "3:18:22", "remaining_time": "10:04:47"}
+{"current_steps": 227, "total_steps": 915, "loss": 0.93, "lr": 7.480524342365535e-05, "epoch": 1.2387448840381992, "percentage": 24.81, "elapsed_time": "3:19:13", "remaining_time": "10:03:50"}
+{"current_steps": 228, "total_steps": 915, "loss": 0.9731, "lr": 7.472974138626809e-05, "epoch": 1.2442019099590724, "percentage": 24.92, "elapsed_time": "3:20:05", "remaining_time": "10:02:53"}
+{"current_steps": 229, "total_steps": 915, "loss": 0.9556, "lr": 7.465373328994608e-05, "epoch": 1.2496589358799455, "percentage": 25.03, "elapsed_time": "3:20:56", "remaining_time": "10:01:56"}
+{"current_steps": 230, "total_steps": 915, "loss": 0.9586, "lr": 7.457722024222927e-05, "epoch": 1.2551159618008185, "percentage": 25.14, "elapsed_time": "3:21:47", "remaining_time": "10:00:59"}
+{"current_steps": 231, "total_steps": 915, "loss": 0.9566, "lr": 7.450020335801544e-05, "epoch": 1.2605729877216918, "percentage": 25.25, "elapsed_time": "3:22:38", "remaining_time": "10:00:03"}
+{"current_steps": 232, "total_steps": 915, "loss": 0.9483, "lr": 7.442268375954394e-05, "epoch": 1.2660300136425648, "percentage": 25.36, "elapsed_time": "3:23:30", "remaining_time": "9:59:07"}
+{"current_steps": 233, "total_steps": 915, "loss": 0.9561, "lr": 7.434466257637933e-05, "epoch": 1.271487039563438, "percentage": 25.46, "elapsed_time": "3:24:21", "remaining_time": "9:58:10"}
+{"current_steps": 234, "total_steps": 915, "loss": 0.9484, "lr": 7.426614094539496e-05, "epoch": 1.2769440654843112, "percentage": 25.57, "elapsed_time": "3:25:13", "remaining_time": "9:57:13"}
+{"current_steps": 235, "total_steps": 915, "loss": 0.9407, "lr": 7.418712001075635e-05, "epoch": 1.2824010914051842, "percentage": 25.68, "elapsed_time": "3:26:04", "remaining_time": "9:56:17"}
+{"current_steps": 236, "total_steps": 915, "loss": 0.923, "lr": 7.410760092390456e-05, "epoch": 1.2878581173260573, "percentage": 25.79, "elapsed_time": "3:26:57", "remaining_time": "9:55:27"}
+{"current_steps": 237, "total_steps": 915, "loss": 0.9452, "lr": 7.402758484353938e-05, "epoch": 1.2933151432469305, "percentage": 25.9, "elapsed_time": "3:28:01", "remaining_time": "9:55:06"}
+{"current_steps": 238, "total_steps": 915, "loss": 0.9396, "lr": 7.394707293560248e-05, "epoch": 1.2987721691678036, "percentage": 26.01, "elapsed_time": "3:28:52", "remaining_time": "9:54:09"}
+{"current_steps": 239, "total_steps": 915, "loss": 0.9493, "lr": 7.386606637326038e-05, "epoch": 1.3042291950886766, "percentage": 26.12, "elapsed_time": "3:29:43", "remaining_time": "9:53:13"}
+{"current_steps": 240, "total_steps": 915, "loss": 0.9428, "lr": 7.378456633688742e-05, "epoch": 1.30968622100955, "percentage": 26.23, "elapsed_time": "3:30:35", "remaining_time": "9:52:16"}
+{"current_steps": 241, "total_steps": 915, "loss": 0.9572, "lr": 7.370257401404846e-05, "epoch": 1.315143246930423, "percentage": 26.34, "elapsed_time": "3:31:26", "remaining_time": "9:51:19"}
+{"current_steps": 242, "total_steps": 915, "loss": 0.9633, "lr": 7.36200905994817e-05, "epoch": 1.320600272851296, "percentage": 26.45, "elapsed_time": "3:32:17", "remaining_time": "9:50:22"}
+{"current_steps": 243, "total_steps": 915, "loss": 0.9511, "lr": 7.353711729508114e-05, "epoch": 1.3260572987721693, "percentage": 26.56, "elapsed_time": "3:33:08", "remaining_time": "9:49:25"}
+{"current_steps": 244, "total_steps": 915, "loss": 0.9806, "lr": 7.34536553098792e-05, "epoch": 1.3315143246930423, "percentage": 26.67, "elapsed_time": "3:33:59", "remaining_time": "9:48:29"}
+{"current_steps": 245, "total_steps": 915, "loss": 0.9836, "lr": 7.336970586002896e-05, "epoch": 1.3369713506139154, "percentage": 26.78, "elapsed_time": "3:34:50", "remaining_time": "9:47:32"}
+{"current_steps": 246, "total_steps": 915, "loss": 0.9507, "lr": 7.328527016878658e-05, "epoch": 1.3424283765347886, "percentage": 26.89, "elapsed_time": "3:35:42", "remaining_time": "9:46:36"}
+{"current_steps": 247, "total_steps": 915, "loss": 0.9444, "lr": 7.320034946649339e-05, "epoch": 1.3478854024556617, "percentage": 26.99, "elapsed_time": "3:36:33", "remaining_time": "9:45:40"}
+{"current_steps": 248, "total_steps": 915, "loss": 0.9612, "lr": 7.311494499055794e-05, "epoch": 1.3533424283765347, "percentage": 27.1, "elapsed_time": "3:37:24", "remaining_time": "9:44:44"}
+{"current_steps": 249, "total_steps": 915, "loss": 0.945, "lr": 7.302905798543809e-05, "epoch": 1.358799454297408, "percentage": 27.21, "elapsed_time": "3:38:15", "remaining_time": "9:43:47"}
+{"current_steps": 250, "total_steps": 915, "loss": 0.9515, "lr": 7.294268970262271e-05, "epoch": 1.364256480218281, "percentage": 27.32, "elapsed_time": "3:39:06", "remaining_time": "9:42:50"}
+{"current_steps": 251, "total_steps": 915, "loss": 0.9434, "lr": 7.28558414006136e-05, "epoch": 1.369713506139154, "percentage": 27.43, "elapsed_time": "3:40:00", "remaining_time": "9:42:02"}
+{"current_steps": 252, "total_steps": 915, "loss": 0.9505, "lr": 7.276851434490706e-05, "epoch": 1.3751705320600274, "percentage": 27.54, "elapsed_time": "3:41:05", "remaining_time": "9:41:39"}
+{"current_steps": 253, "total_steps": 915, "loss": 0.939, "lr": 7.268070980797547e-05, "epoch": 1.3806275579809004, "percentage": 27.65, "elapsed_time": "3:41:56", "remaining_time": "9:40:43"}
+{"current_steps": 254, "total_steps": 915, "loss": 0.9612, "lr": 7.259242906924876e-05, "epoch": 1.3860845839017735, "percentage": 27.76, "elapsed_time": "3:42:47", "remaining_time": "9:39:46"}
+{"current_steps": 255, "total_steps": 915, "loss": 0.9465, "lr": 7.250367341509573e-05, "epoch": 1.3915416098226467, "percentage": 27.87, "elapsed_time": "3:43:38", "remaining_time": "9:38:50"}
+{"current_steps": 256, "total_steps": 915, "loss": 0.9464, "lr": 7.241444413880539e-05, "epoch": 1.3969986357435198, "percentage": 27.98, "elapsed_time": "3:44:29", "remaining_time": "9:37:54"}
+{"current_steps": 257, "total_steps": 915, "loss": 0.9413, "lr": 7.232474254056801e-05, "epoch": 1.4024556616643928, "percentage": 28.09, "elapsed_time": "3:45:21", "remaining_time": "9:36:58"}
+{"current_steps": 258, "total_steps": 915, "loss": 0.9373, "lr": 7.223456992745626e-05, "epoch": 1.407912687585266, "percentage": 28.2, "elapsed_time": "3:46:12", "remaining_time": "9:36:01"}
+{"current_steps": 259, "total_steps": 915, "loss": 0.9417, "lr": 7.214392761340611e-05, "epoch": 1.4133697135061392, "percentage": 28.31, "elapsed_time": "3:47:03", "remaining_time": "9:35:05"}
+{"current_steps": 260, "total_steps": 915, "loss": 0.9644, "lr": 7.205281691919769e-05, "epoch": 1.4188267394270122, "percentage": 28.42, "elapsed_time": "3:47:54", "remaining_time": "9:34:09"}
+{"current_steps": 261, "total_steps": 915, "loss": 0.9554, "lr": 7.196123917243609e-05, "epoch": 1.4242837653478855, "percentage": 28.52, "elapsed_time": "3:48:46", "remaining_time": "9:33:14"}
+{"current_steps": 262, "total_steps": 915, "loss": 0.9384, "lr": 7.186919570753194e-05, "epoch": 1.4297407912687585, "percentage": 28.63, "elapsed_time": "3:49:37", "remaining_time": "9:32:17"}
+{"current_steps": 263, "total_steps": 915, "loss": 0.9518, "lr": 7.177668786568207e-05, "epoch": 1.4351978171896316, "percentage": 28.74, "elapsed_time": "3:50:28", "remaining_time": "9:31:21"}
+{"current_steps": 264, "total_steps": 915, "loss": 0.9266, "lr": 7.168371699484984e-05, "epoch": 1.4406548431105048, "percentage": 28.85, "elapsed_time": "3:51:19", "remaining_time": "9:30:25"}
+{"current_steps": 265, "total_steps": 915, "loss": 0.9504, "lr": 7.159028444974562e-05, "epoch": 1.446111869031378, "percentage": 28.96, "elapsed_time": "3:52:10", "remaining_time": "9:29:29"}
+{"current_steps": 266, "total_steps": 915, "loss": 0.9312, "lr": 7.149639159180695e-05, "epoch": 1.451568894952251, "percentage": 29.07, "elapsed_time": "3:53:05", "remaining_time": "9:28:43"}
+{"current_steps": 267, "total_steps": 915, "loss": 0.947, "lr": 7.140203978917875e-05, "epoch": 1.4570259208731242, "percentage": 29.18, "elapsed_time": "3:54:10", "remaining_time": "9:28:20"}
+{"current_steps": 268, "total_steps": 915, "loss": 0.9422, "lr": 7.130723041669342e-05, "epoch": 1.4624829467939973, "percentage": 29.29, "elapsed_time": "3:55:01", "remaining_time": "9:27:23"}
+{"current_steps": 269, "total_steps": 915, "loss": 0.9286, "lr": 7.121196485585071e-05, "epoch": 1.4679399727148703, "percentage": 29.4, "elapsed_time": "3:55:52", "remaining_time": "9:26:27"}
+{"current_steps": 270, "total_steps": 915, "loss": 0.9566, "lr": 7.111624449479769e-05, "epoch": 1.4733969986357436, "percentage": 29.51, "elapsed_time": "3:56:44", "remaining_time": "9:25:32"}
+{"current_steps": 271, "total_steps": 915, "loss": 0.9522, "lr": 7.102007072830847e-05, "epoch": 1.4788540245566166, "percentage": 29.62, "elapsed_time": "3:57:35", "remaining_time": "9:24:35"}
+{"current_steps": 272, "total_steps": 915, "loss": 0.9313, "lr": 7.092344495776387e-05, "epoch": 1.4843110504774897, "percentage": 29.73, "elapsed_time": "3:58:26", "remaining_time": "9:23:39"}
+{"current_steps": 273, "total_steps": 915, "loss": 0.9248, "lr": 7.082636859113104e-05, "epoch": 1.489768076398363, "percentage": 29.84, "elapsed_time": "3:59:17", "remaining_time": "9:22:43"}
+{"current_steps": 274, "total_steps": 915, "loss": 0.9454, "lr": 7.072884304294289e-05, "epoch": 1.495225102319236, "percentage": 29.95, "elapsed_time": "4:00:08", "remaining_time": "9:21:47"}
+{"current_steps": 275, "total_steps": 915, "loss": 0.9563, "lr": 7.063086973427754e-05, "epoch": 1.500682128240109, "percentage": 30.05, "elapsed_time": "4:00:59", "remaining_time": "9:20:52"}
+{"current_steps": 276, "total_steps": 915, "loss": 0.9431, "lr": 7.053245009273753e-05, "epoch": 1.5061391541609823, "percentage": 30.16, "elapsed_time": "4:01:51", "remaining_time": "9:19:56"}
+{"current_steps": 277, "total_steps": 915, "loss": 0.9554, "lr": 7.043358555242914e-05, "epoch": 1.5115961800818554, "percentage": 30.27, "elapsed_time": "4:02:42", "remaining_time": "9:19:01"}
+{"current_steps": 278, "total_steps": 915, "loss": 0.9251, "lr": 7.033427755394137e-05, "epoch": 1.5170532060027284, "percentage": 30.38, "elapsed_time": "4:03:33", "remaining_time": "9:18:05"}
+{"current_steps": 279, "total_steps": 915, "loss": 0.9545, "lr": 7.023452754432503e-05, "epoch": 1.5225102319236017, "percentage": 30.49, "elapsed_time": "4:04:25", "remaining_time": "9:17:10"}
+{"current_steps": 280, "total_steps": 915, "loss": 0.9688, "lr": 7.013433697707159e-05, "epoch": 1.5279672578444747, "percentage": 30.6, "elapsed_time": "4:05:16", "remaining_time": "9:16:14"}
+{"current_steps": 281, "total_steps": 915, "loss": 0.9388, "lr": 7.003370731209207e-05, "epoch": 1.5334242837653478, "percentage": 30.71, "elapsed_time": "4:06:12", "remaining_time": "9:15:29"}
+{"current_steps": 282, "total_steps": 915, "loss": 0.9566, "lr": 6.993264001569574e-05, "epoch": 1.538881309686221, "percentage": 30.82, "elapsed_time": "4:07:17", "remaining_time": "9:15:06"}
+{"current_steps": 283, "total_steps": 915, "loss": 0.9353, "lr": 6.983113656056867e-05, "epoch": 1.544338335607094, "percentage": 30.93, "elapsed_time": "4:08:08", "remaining_time": "9:14:10"}
+{"current_steps": 284, "total_steps": 915, "loss": 0.939, "lr": 6.972919842575247e-05, "epoch": 1.5497953615279672, "percentage": 31.04, "elapsed_time": "4:09:00", "remaining_time": "9:13:14"}
+{"current_steps": 285, "total_steps": 915, "loss": 0.918, "lr": 6.96268270966225e-05, "epoch": 1.5552523874488404, "percentage": 31.15, "elapsed_time": "4:09:51", "remaining_time": "9:12:18"}
+{"current_steps": 286, "total_steps": 915, "loss": 0.945, "lr": 6.952402406486644e-05, "epoch": 1.5607094133697135, "percentage": 31.26, "elapsed_time": "4:10:42", "remaining_time": "9:11:22"}
+{"current_steps": 287, "total_steps": 915, "loss": 0.9629, "lr": 6.942079082846239e-05, "epoch": 1.5661664392905865, "percentage": 31.37, "elapsed_time": "4:11:33", "remaining_time": "9:10:27"}
+{"current_steps": 288, "total_steps": 915, "loss": 0.9436, "lr": 6.931712889165711e-05, "epoch": 1.5716234652114598, "percentage": 31.48, "elapsed_time": "4:12:24", "remaining_time": "9:09:31"}
+{"current_steps": 289, "total_steps": 915, "loss": 0.9562, "lr": 6.921303976494416e-05, "epoch": 1.5770804911323328, "percentage": 31.58, "elapsed_time": "4:13:16", "remaining_time": "9:08:35"}
+{"current_steps": 290, "total_steps": 915, "loss": 0.916, "lr": 6.910852496504175e-05, "epoch": 1.5825375170532059, "percentage": 31.69, "elapsed_time": "4:14:07", "remaining_time": "9:07:40"}
+{"current_steps": 291, "total_steps": 915, "loss": 0.939, "lr": 6.900358601487079e-05, "epoch": 1.5879945429740792, "percentage": 31.8, "elapsed_time": "4:14:58", "remaining_time": "9:06:46"}
+{"current_steps": 292, "total_steps": 915, "loss": 0.9423, "lr": 6.889822444353258e-05, "epoch": 1.5934515688949522, "percentage": 31.91, "elapsed_time": "4:15:50", "remaining_time": "9:05:50"}
+{"current_steps": 293, "total_steps": 915, "loss": 0.9548, "lr": 6.87924417862866e-05, "epoch": 1.5989085948158253, "percentage": 32.02, "elapsed_time": "4:16:41", "remaining_time": "9:04:54"}
+{"current_steps": 294, "total_steps": 915, "loss": 0.9482, "lr": 6.868623958452812e-05, "epoch": 1.6043656207366985, "percentage": 32.13, "elapsed_time": "4:17:32", "remaining_time": "9:03:59"}
+{"current_steps": 295, "total_steps": 915, "loss": 0.9384, "lr": 6.857961938576573e-05, "epoch": 1.6098226466575716, "percentage": 32.24, "elapsed_time": "4:18:23", "remaining_time": "9:03:04"}
+{"current_steps": 296, "total_steps": 915, "loss": 0.9333, "lr": 6.84725827435988e-05, "epoch": 1.6152796725784446, "percentage": 32.35, "elapsed_time": "4:19:20", "remaining_time": "9:02:20"}
+{"current_steps": 297, "total_steps": 915, "loss": 0.9413, "lr": 6.83651312176948e-05, "epoch": 1.620736698499318, "percentage": 32.46, "elapsed_time": "4:20:24", "remaining_time": "9:01:52"}
+{"current_steps": 298, "total_steps": 915, "loss": 0.9541, "lr": 6.825726637376669e-05, "epoch": 1.626193724420191, "percentage": 32.57, "elapsed_time": "4:21:16", "remaining_time": "9:00:56"}
+{"current_steps": 299, "total_steps": 915, "loss": 0.9484, "lr": 6.814898978354997e-05, "epoch": 1.631650750341064, "percentage": 32.68, "elapsed_time": "4:22:07", "remaining_time": "9:00:01"}
+{"current_steps": 300, "total_steps": 915, "loss": 0.9247, "lr": 6.804030302477985e-05, "epoch": 1.6371077762619373, "percentage": 32.79, "elapsed_time": "4:22:58", "remaining_time": "8:59:05"}
+{"current_steps": 301, "total_steps": 915, "loss": 0.9295, "lr": 6.793120768116825e-05, "epoch": 1.6425648021828103, "percentage": 32.9, "elapsed_time": "4:23:49", "remaining_time": "8:58:10"}
+{"current_steps": 302, "total_steps": 915, "loss": 0.921, "lr": 6.782170534238073e-05, "epoch": 1.6480218281036834, "percentage": 33.01, "elapsed_time": "4:24:40", "remaining_time": "8:57:14"}
+{"current_steps": 303, "total_steps": 915, "loss": 0.9316, "lr": 6.771179760401329e-05, "epoch": 1.6534788540245566, "percentage": 33.11, "elapsed_time": "4:25:32", "remaining_time": "8:56:19"}
+{"current_steps": 304, "total_steps": 915, "loss": 0.9388, "lr": 6.760148606756916e-05, "epoch": 1.65893587994543, "percentage": 33.22, "elapsed_time": "4:26:23", "remaining_time": "8:55:24"}
+{"current_steps": 305, "total_steps": 915, "loss": 0.9235, "lr": 6.749077234043545e-05, "epoch": 1.6643929058663027, "percentage": 33.33, "elapsed_time": "4:27:14", "remaining_time": "8:54:29"}
+{"current_steps": 306, "total_steps": 915, "loss": 0.9526, "lr": 6.737965803585975e-05, "epoch": 1.669849931787176, "percentage": 33.44, "elapsed_time": "4:28:06", "remaining_time": "8:53:34"}
+{"current_steps": 307, "total_steps": 915, "loss": 0.9412, "lr": 6.726814477292654e-05, "epoch": 1.6753069577080493, "percentage": 33.55, "elapsed_time": "4:28:57", "remaining_time": "8:52:39"}
+{"current_steps": 308, "total_steps": 915, "loss": 0.9349, "lr": 6.715623417653372e-05, "epoch": 1.680763983628922, "percentage": 33.66, "elapsed_time": "4:29:48", "remaining_time": "8:51:44"}
+{"current_steps": 309, "total_steps": 915, "loss": 0.945, "lr": 6.704392787736884e-05, "epoch": 1.6862210095497954, "percentage": 33.77, "elapsed_time": "4:30:39", "remaining_time": "8:50:49"}
+{"current_steps": 310, "total_steps": 915, "loss": 0.9372, "lr": 6.69312275118854e-05, "epoch": 1.6916780354706686, "percentage": 33.88, "elapsed_time": "4:31:31", "remaining_time": "8:49:54"}
+{"current_steps": 311, "total_steps": 915, "loss": 0.9131, "lr": 6.681813472227893e-05, "epoch": 1.6971350613915415, "percentage": 33.99, "elapsed_time": "4:32:27", "remaining_time": "8:49:09"}
+{"current_steps": 312, "total_steps": 915, "loss": 0.9651, "lr": 6.670465115646313e-05, "epoch": 1.7025920873124147, "percentage": 34.1, "elapsed_time": "4:33:31", "remaining_time": "8:48:38"}
+{"current_steps": 313, "total_steps": 915, "loss": 0.9601, "lr": 6.659077846804586e-05, "epoch": 1.708049113233288, "percentage": 34.21, "elapsed_time": "4:34:22", "remaining_time": "8:47:42"}
+{"current_steps": 314, "total_steps": 915, "loss": 0.9388, "lr": 6.647651831630496e-05, "epoch": 1.7135061391541608, "percentage": 34.32, "elapsed_time": "4:35:13", "remaining_time": "8:46:47"}
+{"current_steps": 315, "total_steps": 915, "loss": 0.9694, "lr": 6.63618723661642e-05, "epoch": 1.718963165075034, "percentage": 34.43, "elapsed_time": "4:36:05", "remaining_time": "8:45:53"}
+{"current_steps": 316, "total_steps": 915, "loss": 0.9503, "lr": 6.624684228816887e-05, "epoch": 1.7244201909959074, "percentage": 34.54, "elapsed_time": "4:36:56", "remaining_time": "8:44:57"}
+{"current_steps": 317, "total_steps": 915, "loss": 0.9741, "lr": 6.61314297584616e-05, "epoch": 1.7298772169167802, "percentage": 34.64, "elapsed_time": "4:37:47", "remaining_time": "8:44:02"}
+{"current_steps": 318, "total_steps": 915, "loss": 0.9626, "lr": 6.601563645875779e-05, "epoch": 1.7353342428376535, "percentage": 34.75, "elapsed_time": "4:38:38", "remaining_time": "8:43:06"}
+{"current_steps": 319, "total_steps": 915, "loss": 0.956, "lr": 6.58994640763212e-05, "epoch": 1.7407912687585267, "percentage": 34.86, "elapsed_time": "4:39:29", "remaining_time": "8:42:11"}
+{"current_steps": 320, "total_steps": 915, "loss": 0.9449, "lr": 6.578291430393935e-05, "epoch": 1.7462482946793996, "percentage": 34.97, "elapsed_time": "4:40:21", "remaining_time": "8:41:17"}
+{"current_steps": 321, "total_steps": 915, "loss": 0.941, "lr": 6.566598883989879e-05, "epoch": 1.7517053206002728, "percentage": 35.08, "elapsed_time": "4:41:12", "remaining_time": "8:40:21"}
+{"current_steps": 322, "total_steps": 915, "loss": 0.9339, "lr": 6.554868938796044e-05, "epoch": 1.7571623465211461, "percentage": 35.19, "elapsed_time": "4:42:03", "remaining_time": "8:39:26"}
+{"current_steps": 323, "total_steps": 915, "loss": 0.9279, "lr": 6.543101765733473e-05, "epoch": 1.762619372442019, "percentage": 35.3, "elapsed_time": "4:42:55", "remaining_time": "8:38:32"}
+{"current_steps": 324, "total_steps": 915, "loss": 0.9322, "lr": 6.531297536265668e-05, "epoch": 1.7680763983628922, "percentage": 35.41, "elapsed_time": "4:43:46", "remaining_time": "8:37:37"}
+{"current_steps": 325, "total_steps": 915, "loss": 0.9479, "lr": 6.519456422396089e-05, "epoch": 1.7735334242837655, "percentage": 35.52, "elapsed_time": "4:44:37", "remaining_time": "8:36:42"}
+{"current_steps": 326, "total_steps": 915, "loss": 0.9596, "lr": 6.507578596665655e-05, "epoch": 1.7789904502046383, "percentage": 35.63, "elapsed_time": "4:45:35", "remaining_time": "8:36:00"}
+{"current_steps": 327, "total_steps": 915, "loss": 0.942, "lr": 6.495664232150227e-05, "epoch": 1.7844474761255116, "percentage": 35.74, "elapsed_time": "4:46:37", "remaining_time": "8:35:23"}
+{"current_steps": 328, "total_steps": 915, "loss": 0.9455, "lr": 6.483713502458082e-05, "epoch": 1.7899045020463848, "percentage": 35.85, "elapsed_time": "4:47:28", "remaining_time": "8:34:28"}
+{"current_steps": 329, "total_steps": 915, "loss": 0.9505, "lr": 6.471726581727391e-05, "epoch": 1.795361527967258, "percentage": 35.96, "elapsed_time": "4:48:19", "remaining_time": "8:33:33"}
+{"current_steps": 330, "total_steps": 915, "loss": 0.9251, "lr": 6.459703644623669e-05, "epoch": 1.800818553888131, "percentage": 36.07, "elapsed_time": "4:49:10", "remaining_time": "8:32:37"}
+{"current_steps": 331, "total_steps": 915, "loss": 0.9215, "lr": 6.447644866337246e-05, "epoch": 1.8062755798090042, "percentage": 36.17, "elapsed_time": "4:50:01", "remaining_time": "8:31:43"}
+{"current_steps": 332, "total_steps": 915, "loss": 0.9449, "lr": 6.435550422580703e-05, "epoch": 1.8117326057298773, "percentage": 36.28, "elapsed_time": "4:50:53", "remaining_time": "8:30:48"}
+{"current_steps": 333, "total_steps": 915, "loss": 0.914, "lr": 6.423420489586312e-05, "epoch": 1.8171896316507503, "percentage": 36.39, "elapsed_time": "4:51:44", "remaining_time": "8:29:53"}
+{"current_steps": 334, "total_steps": 915, "loss": 0.944, "lr": 6.411255244103476e-05, "epoch": 1.8226466575716236, "percentage": 36.5, "elapsed_time": "4:52:35", "remaining_time": "8:28:58"}
+{"current_steps": 335, "total_steps": 915, "loss": 0.9413, "lr": 6.399054863396147e-05, "epoch": 1.8281036834924966, "percentage": 36.61, "elapsed_time": "4:53:27", "remaining_time": "8:28:04"}
+{"current_steps": 336, "total_steps": 915, "loss": 0.9366, "lr": 6.38681952524024e-05, "epoch": 1.8335607094133697, "percentage": 36.72, "elapsed_time": "4:54:18", "remaining_time": "8:27:09"}
+{"current_steps": 337, "total_steps": 915, "loss": 0.9357, "lr": 6.374549407921052e-05, "epoch": 1.839017735334243, "percentage": 36.83, "elapsed_time": "4:55:09", "remaining_time": "8:26:14"}
+{"current_steps": 338, "total_steps": 915, "loss": 0.9307, "lr": 6.362244690230658e-05, "epoch": 1.844474761255116, "percentage": 36.94, "elapsed_time": "4:56:00", "remaining_time": "8:25:19"}
+{"current_steps": 339, "total_steps": 915, "loss": 0.9343, "lr": 6.349905551465302e-05, "epoch": 1.849931787175989, "percentage": 37.05, "elapsed_time": "4:56:52", "remaining_time": "8:24:25"}
+{"current_steps": 340, "total_steps": 915, "loss": 0.949, "lr": 6.3375321714228e-05, "epoch": 1.8553888130968623, "percentage": 37.16, "elapsed_time": "4:57:43", "remaining_time": "8:23:30"}
+{"current_steps": 341, "total_steps": 915, "loss": 0.9436, "lr": 6.325124730399897e-05, "epoch": 1.8608458390177354, "percentage": 37.27, "elapsed_time": "4:58:41", "remaining_time": "8:22:47"}
+{"current_steps": 342, "total_steps": 915, "loss": 0.9331, "lr": 6.312683409189659e-05, "epoch": 1.8663028649386084, "percentage": 37.38, "elapsed_time": "4:59:44", "remaining_time": "8:22:11"}
+{"current_steps": 343, "total_steps": 915, "loss": 0.9286, "lr": 6.300208389078834e-05, "epoch": 1.8717598908594817, "percentage": 37.49, "elapsed_time": "5:00:35", "remaining_time": "8:21:16"}
+{"current_steps": 344, "total_steps": 915, "loss": 0.9557, "lr": 6.2876998518452e-05, "epoch": 1.8772169167803547, "percentage": 37.6, "elapsed_time": "5:01:26", "remaining_time": "8:20:21"}
+{"current_steps": 345, "total_steps": 915, "loss": 0.9658, "lr": 6.275157979754932e-05, "epoch": 1.8826739427012278, "percentage": 37.7, "elapsed_time": "5:02:17", "remaining_time": "8:19:26"}
+{"current_steps": 346, "total_steps": 915, "loss": 0.9643, "lr": 6.262582955559936e-05, "epoch": 1.888130968622101, "percentage": 37.81, "elapsed_time": "5:03:08", "remaining_time": "8:18:31"}
+{"current_steps": 347, "total_steps": 915, "loss": 0.9592, "lr": 6.249974962495187e-05, "epoch": 1.893587994542974, "percentage": 37.92, "elapsed_time": "5:03:59", "remaining_time": "8:17:36"}
+{"current_steps": 348, "total_steps": 915, "loss": 0.9253, "lr": 6.237334184276066e-05, "epoch": 1.8990450204638472, "percentage": 38.03, "elapsed_time": "5:04:51", "remaining_time": "8:16:42"}
+{"current_steps": 349, "total_steps": 915, "loss": 0.9724, "lr": 6.224660805095669e-05, "epoch": 1.9045020463847204, "percentage": 38.14, "elapsed_time": "5:05:42", "remaining_time": "8:15:47"}
+{"current_steps": 350, "total_steps": 915, "loss": 0.9471, "lr": 6.21195500962214e-05, "epoch": 1.9099590723055935, "percentage": 38.25, "elapsed_time": "5:06:33", "remaining_time": "8:14:52"}
+{"current_steps": 351, "total_steps": 915, "loss": 0.9424, "lr": 6.199216982995966e-05, "epoch": 1.9154160982264665, "percentage": 38.36, "elapsed_time": "5:07:25", "remaining_time": "8:13:58"}
+{"current_steps": 352, "total_steps": 915, "loss": 0.9239, "lr": 6.18644691082729e-05, "epoch": 1.9208731241473398, "percentage": 38.47, "elapsed_time": "5:08:16", "remaining_time": "8:13:03"}
+{"current_steps": 353, "total_steps": 915, "loss": 0.9513, "lr": 6.173644979193199e-05, "epoch": 1.9263301500682128, "percentage": 38.58, "elapsed_time": "5:09:07", "remaining_time": "8:12:08"}
+{"current_steps": 354, "total_steps": 915, "loss": 0.9333, "lr": 6.160811374635015e-05, "epoch": 1.931787175989086, "percentage": 38.69, "elapsed_time": "5:09:58", "remaining_time": "8:11:13"}
+{"current_steps": 355, "total_steps": 915, "loss": 0.9537, "lr": 6.147946284155576e-05, "epoch": 1.9372442019099592, "percentage": 38.8, "elapsed_time": "5:10:49", "remaining_time": "8:10:19"}
+{"current_steps": 356, "total_steps": 915, "loss": 0.9206, "lr": 6.135049895216516e-05, "epoch": 1.9427012278308322, "percentage": 38.91, "elapsed_time": "5:11:49", "remaining_time": "8:09:38"}
+{"current_steps": 357, "total_steps": 915, "loss": 0.9188, "lr": 6.122122395735525e-05, "epoch": 1.9481582537517053, "percentage": 39.02, "elapsed_time": "5:12:51", "remaining_time": "8:09:00"}
+{"current_steps": 358, "total_steps": 915, "loss": 0.934, "lr": 6.10916397408362e-05, "epoch": 1.9536152796725785, "percentage": 39.13, "elapsed_time": "5:13:42", "remaining_time": "8:08:05"}
+{"current_steps": 359, "total_steps": 915, "loss": 0.9502, "lr": 6.096174819082389e-05, "epoch": 1.9590723055934516, "percentage": 39.23, "elapsed_time": "5:14:33", "remaining_time": "8:07:10"}
+{"current_steps": 360, "total_steps": 915, "loss": 0.9535, "lr": 6.0831551200012534e-05, "epoch": 1.9645293315143246, "percentage": 39.34, "elapsed_time": "5:15:24", "remaining_time": "8:06:15"}
+{"current_steps": 361, "total_steps": 915, "loss": 0.9036, "lr": 6.070105066554698e-05, "epoch": 1.969986357435198, "percentage": 39.45, "elapsed_time": "5:16:15", "remaining_time": "8:05:21"}
+{"current_steps": 362, "total_steps": 915, "loss": 0.9248, "lr": 6.057024848899515e-05, "epoch": 1.975443383356071, "percentage": 39.56, "elapsed_time": "5:17:07", "remaining_time": "8:04:26"}
+{"current_steps": 363, "total_steps": 915, "loss": 0.9553, "lr": 6.0439146576320256e-05, "epoch": 1.980900409276944, "percentage": 39.67, "elapsed_time": "5:17:58", "remaining_time": "8:03:31"}
+{"current_steps": 364, "total_steps": 915, "loss": 0.9333, "lr": 6.0307746837853087e-05, "epoch": 1.9863574351978173, "percentage": 39.78, "elapsed_time": "5:18:49", "remaining_time": "8:02:37"}
+{"current_steps": 365, "total_steps": 915, "loss": 0.9545, "lr": 6.017605118826417e-05, "epoch": 1.9918144611186903, "percentage": 39.89, "elapsed_time": "5:19:40", "remaining_time": "8:01:42"}
+{"current_steps": 366, "total_steps": 915, "loss": 0.9338, "lr": 6.0044061546535815e-05, "epoch": 1.9972714870395634, "percentage": 40.0, "elapsed_time": "5:20:31", "remaining_time": "8:00:47"}
+{"current_steps": 367, "total_steps": 915, "loss": 1.6363, "lr": 5.991177983593423e-05, "epoch": 2.0027285129604366, "percentage": 40.11, "elapsed_time": "5:22:43", "remaining_time": "8:01:53"}
+{"current_steps": 368, "total_steps": 915, "loss": 0.9085, "lr": 5.9779207983981435e-05, "epoch": 2.00818553888131, "percentage": 40.22, "elapsed_time": "5:23:34", "remaining_time": "8:00:58"}
+{"current_steps": 369, "total_steps": 915, "loss": 0.9226, "lr": 5.964634792242723e-05, "epoch": 2.0136425648021827, "percentage": 40.33, "elapsed_time": "5:24:25", "remaining_time": "8:00:03"}