Training in progress, epoch 1

model-00001-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5aa74d3337037b065e2fd242baafdc2ad8d19170a78bdfe792a2d19b0523fb40
+oid sha256:7eed6030b5d7cd6b69380d90ae733e1f6d1cc301cce6766cce0309b9aaab7edf
 size 4877660776

model-00002-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c583809f0b64cec9dae236d607fd496cce984f05a504f0a98ddb05487c48842d
+oid sha256:4e91f0333485bb026ad269d2164745f3a2124956ce00e686d8d8c5c406d16391
 size 4932751008

model-00003-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:aefe10321b5a0fd6dbfd9ecaa5782173d8ca50e5dbd1c12e9a11606b73b0879f
+oid sha256:230d11b0635b62c8d71ce7a18328b0f4f745b7e34e9af7bd4b28b9619a5c17d6
 size 4330865200

model-00004-of-00004.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0c37b838a024b8a14eff4df1bc991a287142d4926bdbb99eda7e8ae8ffcb1f43
+oid sha256:83b66f7d3e3bb1d4802b35177d184e9354bbd64d7275b9c06b66120776856d73
 size 1089994880

trainer_log.jsonl

@@ -188,3 +188,184 @@
 {"current_steps": 188, "total_steps": 540, "loss": 0.4682, "lr": 8.238629983075296e-06, "epoch": 1.040590405904059, "percentage": 34.81, "elapsed_time": "0:55:21", "remaining_time": "1:43:38"}
 {"current_steps": 189, "total_steps": 540, "loss": 0.4198, "lr": 8.213938048432697e-06, "epoch": 1.0461254612546125, "percentage": 35.0, "elapsed_time": "0:55:37", "remaining_time": "1:43:17"}
 {"current_steps": 190, "total_steps": 540, "loss": 0.4754, "lr": 8.18911181775353e-06, "epoch": 1.051660516605166, "percentage": 35.19, "elapsed_time": "0:55:54", "remaining_time": "1:43:00"}
+{"current_steps": 191, "total_steps": 540, "loss": 0.4699, "lr": 8.164152328414476e-06, "epoch": 1.0571955719557196, "percentage": 35.37, "elapsed_time": "0:56:11", "remaining_time": "1:42:41"}
+{"current_steps": 192, "total_steps": 540, "loss": 0.4489, "lr": 8.139060623360494e-06, "epoch": 1.062730627306273, "percentage": 35.56, "elapsed_time": "0:56:28", "remaining_time": "1:42:21"}
+{"current_steps": 193, "total_steps": 540, "loss": 0.4695, "lr": 8.113837751061246e-06, "epoch": 1.0682656826568266, "percentage": 35.74, "elapsed_time": "0:56:46", "remaining_time": "1:42:04"}
+{"current_steps": 194, "total_steps": 540, "loss": 0.4741, "lr": 8.088484765467286e-06, "epoch": 1.07380073800738, "percentage": 35.93, "elapsed_time": "0:57:03", "remaining_time": "1:41:45"}
+{"current_steps": 195, "total_steps": 540, "loss": 0.46, "lr": 8.063002725966014e-06, "epoch": 1.0793357933579335, "percentage": 36.11, "elapsed_time": "0:57:20", "remaining_time": "1:41:26"}
+{"current_steps": 196, "total_steps": 540, "loss": 0.4438, "lr": 8.037392697337418e-06, "epoch": 1.084870848708487, "percentage": 36.3, "elapsed_time": "0:57:38", "remaining_time": "1:41:10"}
+{"current_steps": 197, "total_steps": 540, "loss": 0.451, "lr": 8.011655749709575e-06, "epoch": 1.0904059040590406, "percentage": 36.48, "elapsed_time": "0:57:57", "remaining_time": "1:40:54"}
+{"current_steps": 198, "total_steps": 540, "loss": 0.464, "lr": 7.985792958513932e-06, "epoch": 1.0959409594095941, "percentage": 36.67, "elapsed_time": "0:58:15", "remaining_time": "1:40:36"}
+{"current_steps": 199, "total_steps": 540, "loss": 0.4772, "lr": 7.95980540444038e-06, "epoch": 1.1014760147601477, "percentage": 36.85, "elapsed_time": "0:58:32", "remaining_time": "1:40:19"}
+{"current_steps": 200, "total_steps": 540, "loss": 0.4661, "lr": 7.93369417339209e-06, "epoch": 1.1070110701107012, "percentage": 37.04, "elapsed_time": "0:58:47", "remaining_time": "1:39:57"}
+{"current_steps": 201, "total_steps": 540, "loss": 0.466, "lr": 7.907460356440133e-06, "epoch": 1.1125461254612545, "percentage": 37.22, "elapsed_time": "0:59:05", "remaining_time": "1:39:39"}
+{"current_steps": 202, "total_steps": 540, "loss": 0.4251, "lr": 7.881105049777902e-06, "epoch": 1.118081180811808, "percentage": 37.41, "elapsed_time": "0:59:20", "remaining_time": "1:39:17"}
+{"current_steps": 203, "total_steps": 540, "loss": 0.4477, "lr": 7.854629354675292e-06, "epoch": 1.1236162361623616, "percentage": 37.59, "elapsed_time": "0:59:38", "remaining_time": "1:39:00"}
+{"current_steps": 204, "total_steps": 540, "loss": 0.4522, "lr": 7.828034377432694e-06, "epoch": 1.1291512915129152, "percentage": 37.78, "elapsed_time": "0:59:55", "remaining_time": "1:38:42"}
+{"current_steps": 205, "total_steps": 540, "loss": 0.4284, "lr": 7.801321229334764e-06, "epoch": 1.1346863468634687, "percentage": 37.96, "elapsed_time": "1:00:14", "remaining_time": "1:38:26"}
+{"current_steps": 206, "total_steps": 540, "loss": 0.4468, "lr": 7.774491026603985e-06, "epoch": 1.140221402214022, "percentage": 38.15, "elapsed_time": "1:00:31", "remaining_time": "1:38:07"}
+{"current_steps": 207, "total_steps": 540, "loss": 0.4914, "lr": 7.747544890354031e-06, "epoch": 1.1457564575645756, "percentage": 38.33, "elapsed_time": "1:00:49", "remaining_time": "1:37:51"}
+{"current_steps": 208, "total_steps": 540, "loss": 0.4527, "lr": 7.720483946542913e-06, "epoch": 1.151291512915129, "percentage": 38.52, "elapsed_time": "1:01:04", "remaining_time": "1:37:29"}
+{"current_steps": 209, "total_steps": 540, "loss": 0.4537, "lr": 7.69330932592594e-06, "epoch": 1.1568265682656826, "percentage": 38.7, "elapsed_time": "1:01:22", "remaining_time": "1:37:12"}
+{"current_steps": 210, "total_steps": 540, "loss": 0.4454, "lr": 7.666022164008458e-06, "epoch": 1.1623616236162362, "percentage": 38.89, "elapsed_time": "1:01:41", "remaining_time": "1:36:56"}
+{"current_steps": 211, "total_steps": 540, "loss": 0.4547, "lr": 7.638623600998409e-06, "epoch": 1.1678966789667897, "percentage": 39.07, "elapsed_time": "1:01:56", "remaining_time": "1:36:35"}
+{"current_steps": 212, "total_steps": 540, "loss": 0.4915, "lr": 7.6111147817586925e-06, "epoch": 1.1734317343173433, "percentage": 39.26, "elapsed_time": "1:02:15", "remaining_time": "1:36:18"}
+{"current_steps": 213, "total_steps": 540, "loss": 0.4359, "lr": 7.5834968557593155e-06, "epoch": 1.1789667896678966, "percentage": 39.44, "elapsed_time": "1:02:33", "remaining_time": "1:36:02"}
+{"current_steps": 214, "total_steps": 540, "loss": 0.4775, "lr": 7.5557709770293664e-06, "epoch": 1.1845018450184501, "percentage": 39.63, "elapsed_time": "1:02:51", "remaining_time": "1:35:44"}
+{"current_steps": 215, "total_steps": 540, "loss": 0.4603, "lr": 7.527938304108795e-06, "epoch": 1.1900369003690037, "percentage": 39.81, "elapsed_time": "1:03:08", "remaining_time": "1:35:26"}
+{"current_steps": 216, "total_steps": 540, "loss": 0.4806, "lr": 7.500000000000001e-06, "epoch": 1.1955719557195572, "percentage": 40.0, "elapsed_time": "1:03:25", "remaining_time": "1:35:07"}
+{"current_steps": 217, "total_steps": 540, "loss": 0.4672, "lr": 7.471957232119235e-06, "epoch": 1.2011070110701108, "percentage": 40.19, "elapsed_time": "1:03:42", "remaining_time": "1:34:49"}
+{"current_steps": 218, "total_steps": 540, "loss": 0.4939, "lr": 7.443811172247822e-06, "epoch": 1.2066420664206643, "percentage": 40.37, "elapsed_time": "1:03:59", "remaining_time": "1:34:31"}
+{"current_steps": 219, "total_steps": 540, "loss": 0.4407, "lr": 7.415562996483193e-06, "epoch": 1.2121771217712176, "percentage": 40.56, "elapsed_time": "1:04:17", "remaining_time": "1:34:13"}
+{"current_steps": 220, "total_steps": 540, "loss": 0.4655, "lr": 7.387213885189746e-06, "epoch": 1.2177121771217712, "percentage": 40.74, "elapsed_time": "1:04:34", "remaining_time": "1:33:54"}
+{"current_steps": 221, "total_steps": 540, "loss": 0.4476, "lr": 7.358765022949519e-06, "epoch": 1.2232472324723247, "percentage": 40.93, "elapsed_time": "1:04:52", "remaining_time": "1:33:38"}
+{"current_steps": 222, "total_steps": 540, "loss": 0.4544, "lr": 7.330217598512696e-06, "epoch": 1.2287822878228782, "percentage": 41.11, "elapsed_time": "1:05:09", "remaining_time": "1:33:20"}
+{"current_steps": 223, "total_steps": 540, "loss": 0.4688, "lr": 7.30157280474793e-06, "epoch": 1.2343173431734318, "percentage": 41.3, "elapsed_time": "1:05:26", "remaining_time": "1:33:01"}
+{"current_steps": 224, "total_steps": 540, "loss": 0.4788, "lr": 7.2728318385925035e-06, "epoch": 1.2398523985239853, "percentage": 41.48, "elapsed_time": "1:05:45", "remaining_time": "1:32:45"}
+{"current_steps": 225, "total_steps": 540, "loss": 0.4684, "lr": 7.243995901002312e-06, "epoch": 1.2453874538745389, "percentage": 41.67, "elapsed_time": "1:06:03", "remaining_time": "1:32:28"}
+{"current_steps": 226, "total_steps": 540, "loss": 0.4922, "lr": 7.215066196901676e-06, "epoch": 1.2509225092250922, "percentage": 41.85, "elapsed_time": "1:06:21", "remaining_time": "1:32:11"}
+{"current_steps": 227, "total_steps": 540, "loss": 0.4348, "lr": 7.186043935133005e-06, "epoch": 1.2564575645756457, "percentage": 42.04, "elapsed_time": "1:06:38", "remaining_time": "1:31:53"}
+{"current_steps": 228, "total_steps": 540, "loss": 0.5033, "lr": 7.156930328406268e-06, "epoch": 1.2619926199261993, "percentage": 42.22, "elapsed_time": "1:06:55", "remaining_time": "1:31:34"}
+{"current_steps": 229, "total_steps": 540, "loss": 0.4462, "lr": 7.127726593248337e-06, "epoch": 1.2675276752767528, "percentage": 42.41, "elapsed_time": "1:07:13", "remaining_time": "1:31:17"}
+{"current_steps": 230, "total_steps": 540, "loss": 0.4212, "lr": 7.098433949952146e-06, "epoch": 1.2730627306273063, "percentage": 42.59, "elapsed_time": "1:07:29", "remaining_time": "1:30:58"}
+{"current_steps": 231, "total_steps": 540, "loss": 0.4898, "lr": 7.069053622525697e-06, "epoch": 1.2785977859778597, "percentage": 42.78, "elapsed_time": "1:07:45", "remaining_time": "1:30:38"}
+{"current_steps": 232, "total_steps": 540, "loss": 0.4382, "lr": 7.039586838640918e-06, "epoch": 1.2841328413284132, "percentage": 42.96, "elapsed_time": "1:08:01", "remaining_time": "1:30:18"}
+{"current_steps": 233, "total_steps": 540, "loss": 0.4584, "lr": 7.0100348295823706e-06, "epoch": 1.2896678966789668, "percentage": 43.15, "elapsed_time": "1:08:17", "remaining_time": "1:29:59"}
+{"current_steps": 234, "total_steps": 540, "loss": 0.4319, "lr": 6.980398830195785e-06, "epoch": 1.2952029520295203, "percentage": 43.33, "elapsed_time": "1:08:36", "remaining_time": "1:29:43"}
+{"current_steps": 235, "total_steps": 540, "loss": 0.5255, "lr": 6.950680078836475e-06, "epoch": 1.3007380073800738, "percentage": 43.52, "elapsed_time": "1:08:53", "remaining_time": "1:29:25"}
+{"current_steps": 236, "total_steps": 540, "loss": 0.4637, "lr": 6.920879817317588e-06, "epoch": 1.3062730627306274, "percentage": 43.7, "elapsed_time": "1:09:10", "remaining_time": "1:29:06"}
+{"current_steps": 237, "total_steps": 540, "loss": 0.432, "lr": 6.890999290858213e-06, "epoch": 1.311808118081181, "percentage": 43.89, "elapsed_time": "1:09:27", "remaining_time": "1:28:48"}
+{"current_steps": 238, "total_steps": 540, "loss": 0.4521, "lr": 6.861039748031351e-06, "epoch": 1.3173431734317342, "percentage": 44.07, "elapsed_time": "1:09:45", "remaining_time": "1:28:31"}
+{"current_steps": 239, "total_steps": 540, "loss": 0.4404, "lr": 6.8310024407117405e-06, "epoch": 1.3228782287822878, "percentage": 44.26, "elapsed_time": "1:10:03", "remaining_time": "1:28:13"}
+{"current_steps": 240, "total_steps": 540, "loss": 0.4531, "lr": 6.800888624023552e-06, "epoch": 1.3284132841328413, "percentage": 44.44, "elapsed_time": "1:10:20", "remaining_time": "1:27:55"}
+{"current_steps": 241, "total_steps": 540, "loss": 0.4654, "lr": 6.770699556287939e-06, "epoch": 1.3339483394833949, "percentage": 44.63, "elapsed_time": "1:10:38", "remaining_time": "1:27:38"}
+{"current_steps": 242, "total_steps": 540, "loss": 0.4599, "lr": 6.740436498970453e-06, "epoch": 1.3394833948339484, "percentage": 44.81, "elapsed_time": "1:10:56", "remaining_time": "1:27:21"}
+{"current_steps": 243, "total_steps": 540, "loss": 0.4631, "lr": 6.710100716628345e-06, "epoch": 1.3450184501845017, "percentage": 45.0, "elapsed_time": "1:11:14", "remaining_time": "1:27:04"}
+{"current_steps": 244, "total_steps": 540, "loss": 0.4619, "lr": 6.679693476857712e-06, "epoch": 1.3505535055350553, "percentage": 45.19, "elapsed_time": "1:11:31", "remaining_time": "1:26:45"}
+{"current_steps": 245, "total_steps": 540, "loss": 0.4487, "lr": 6.649216050240539e-06, "epoch": 1.3560885608856088, "percentage": 45.37, "elapsed_time": "1:11:48", "remaining_time": "1:26:28"}
+{"current_steps": 246, "total_steps": 540, "loss": 0.4701, "lr": 6.618669710291607e-06, "epoch": 1.3616236162361623, "percentage": 45.56, "elapsed_time": "1:12:05", "remaining_time": "1:26:10"}
+{"current_steps": 247, "total_steps": 540, "loss": 0.4441, "lr": 6.588055733405266e-06, "epoch": 1.367158671586716, "percentage": 45.74, "elapsed_time": "1:12:22", "remaining_time": "1:25:50"}
+{"current_steps": 248, "total_steps": 540, "loss": 0.4388, "lr": 6.557375398802124e-06, "epoch": 1.3726937269372694, "percentage": 45.93, "elapsed_time": "1:12:38", "remaining_time": "1:25:31"}
+{"current_steps": 249, "total_steps": 540, "loss": 0.5193, "lr": 6.526629988475567e-06, "epoch": 1.378228782287823, "percentage": 46.11, "elapsed_time": "1:12:56", "remaining_time": "1:25:15"}
+{"current_steps": 250, "total_steps": 540, "loss": 0.4264, "lr": 6.495820787138209e-06, "epoch": 1.3837638376383765, "percentage": 46.3, "elapsed_time": "1:13:12", "remaining_time": "1:24:54"}
+{"current_steps": 251, "total_steps": 540, "loss": 0.5035, "lr": 6.4649490821682035e-06, "epoch": 1.3892988929889298, "percentage": 46.48, "elapsed_time": "1:13:30", "remaining_time": "1:24:38"}
+{"current_steps": 252, "total_steps": 540, "loss": 0.4319, "lr": 6.434016163555452e-06, "epoch": 1.3948339483394834, "percentage": 46.67, "elapsed_time": "1:13:49", "remaining_time": "1:24:21"}
+{"current_steps": 253, "total_steps": 540, "loss": 0.4477, "lr": 6.403023323847695e-06, "epoch": 1.400369003690037, "percentage": 46.85, "elapsed_time": "1:14:05", "remaining_time": "1:24:03"}
+{"current_steps": 254, "total_steps": 540, "loss": 0.4683, "lr": 6.371971858096509e-06, "epoch": 1.4059040590405905, "percentage": 47.04, "elapsed_time": "1:14:22", "remaining_time": "1:23:44"}
+{"current_steps": 255, "total_steps": 540, "loss": 0.4231, "lr": 6.340863063803187e-06, "epoch": 1.4114391143911438, "percentage": 47.22, "elapsed_time": "1:14:40", "remaining_time": "1:23:27"}
+{"current_steps": 256, "total_steps": 540, "loss": 0.4909, "lr": 6.30969824086453e-06, "epoch": 1.4169741697416973, "percentage": 47.41, "elapsed_time": "1:14:57", "remaining_time": "1:23:09"}
+{"current_steps": 257, "total_steps": 540, "loss": 0.415, "lr": 6.278478691518519e-06, "epoch": 1.4225092250922509, "percentage": 47.59, "elapsed_time": "1:15:14", "remaining_time": "1:22:51"}
+{"current_steps": 258, "total_steps": 540, "loss": 0.4629, "lr": 6.247205720289907e-06, "epoch": 1.4280442804428044, "percentage": 47.78, "elapsed_time": "1:15:32", "remaining_time": "1:22:33"}
+{"current_steps": 259, "total_steps": 540, "loss": 0.4492, "lr": 6.215880633935709e-06, "epoch": 1.433579335793358, "percentage": 47.96, "elapsed_time": "1:15:49", "remaining_time": "1:22:16"}
+{"current_steps": 260, "total_steps": 540, "loss": 0.4728, "lr": 6.184504741390596e-06, "epoch": 1.4391143911439115, "percentage": 48.15, "elapsed_time": "1:16:07", "remaining_time": "1:21:59"}
+{"current_steps": 261, "total_steps": 540, "loss": 0.4414, "lr": 6.153079353712201e-06, "epoch": 1.444649446494465, "percentage": 48.33, "elapsed_time": "1:16:25", "remaining_time": "1:21:42"}
+{"current_steps": 262, "total_steps": 540, "loss": 0.4498, "lr": 6.121605784026339e-06, "epoch": 1.4501845018450186, "percentage": 48.52, "elapsed_time": "1:16:44", "remaining_time": "1:21:25"}
+{"current_steps": 263, "total_steps": 540, "loss": 0.478, "lr": 6.09008534747213e-06, "epoch": 1.455719557195572, "percentage": 48.7, "elapsed_time": "1:17:01", "remaining_time": "1:21:07"}
+{"current_steps": 264, "total_steps": 540, "loss": 0.4332, "lr": 6.058519361147055e-06, "epoch": 1.4612546125461254, "percentage": 48.89, "elapsed_time": "1:17:19", "remaining_time": "1:20:50"}
+{"current_steps": 265, "total_steps": 540, "loss": 0.4561, "lr": 6.02690914405191e-06, "epoch": 1.466789667896679, "percentage": 49.07, "elapsed_time": "1:17:38", "remaining_time": "1:20:34"}
+{"current_steps": 266, "total_steps": 540, "loss": 0.4646, "lr": 5.995256017035703e-06, "epoch": 1.4723247232472325, "percentage": 49.26, "elapsed_time": "1:17:56", "remaining_time": "1:20:16"}
+{"current_steps": 267, "total_steps": 540, "loss": 0.4608, "lr": 5.9635613027404495e-06, "epoch": 1.477859778597786, "percentage": 49.44, "elapsed_time": "1:18:12", "remaining_time": "1:19:57"}
+{"current_steps": 268, "total_steps": 540, "loss": 0.4662, "lr": 5.931826325545912e-06, "epoch": 1.4833948339483394, "percentage": 49.63, "elapsed_time": "1:18:28", "remaining_time": "1:19:38"}
+{"current_steps": 269, "total_steps": 540, "loss": 0.4452, "lr": 5.900052411514257e-06, "epoch": 1.488929889298893, "percentage": 49.81, "elapsed_time": "1:18:45", "remaining_time": "1:19:20"}
+{"current_steps": 270, "total_steps": 540, "loss": 0.4638, "lr": 5.8682408883346535e-06, "epoch": 1.4944649446494465, "percentage": 50.0, "elapsed_time": "1:19:04", "remaining_time": "1:19:04"}
+{"current_steps": 271, "total_steps": 540, "loss": 0.4176, "lr": 5.836393085267777e-06, "epoch": 1.5, "percentage": 50.19, "elapsed_time": "1:19:21", "remaining_time": "1:18:46"}
+{"current_steps": 272, "total_steps": 540, "loss": 0.4638, "lr": 5.804510333090287e-06, "epoch": 1.5055350553505535, "percentage": 50.37, "elapsed_time": "1:19:38", "remaining_time": "1:18:28"}
+{"current_steps": 273, "total_steps": 540, "loss": 0.4859, "lr": 5.772593964039203e-06, "epoch": 1.511070110701107, "percentage": 50.56, "elapsed_time": "1:19:57", "remaining_time": "1:18:11"}
+{"current_steps": 274, "total_steps": 540, "loss": 0.4353, "lr": 5.740645311756246e-06, "epoch": 1.5166051660516606, "percentage": 50.74, "elapsed_time": "1:20:13", "remaining_time": "1:17:52"}
+{"current_steps": 275, "total_steps": 540, "loss": 0.476, "lr": 5.708665711232103e-06, "epoch": 1.5221402214022142, "percentage": 50.93, "elapsed_time": "1:20:30", "remaining_time": "1:17:34"}
+{"current_steps": 276, "total_steps": 540, "loss": 0.4369, "lr": 5.6766564987506564e-06, "epoch": 1.5276752767527675, "percentage": 51.11, "elapsed_time": "1:20:45", "remaining_time": "1:17:14"}
+{"current_steps": 277, "total_steps": 540, "loss": 0.4562, "lr": 5.644619011833134e-06, "epoch": 1.533210332103321, "percentage": 51.3, "elapsed_time": "1:21:01", "remaining_time": "1:16:55"}
+{"current_steps": 278, "total_steps": 540, "loss": 0.4634, "lr": 5.612554589182228e-06, "epoch": 1.5387453874538746, "percentage": 51.48, "elapsed_time": "1:21:19", "remaining_time": "1:16:38"}
+{"current_steps": 279, "total_steps": 540, "loss": 0.4841, "lr": 5.5804645706261515e-06, "epoch": 1.5442804428044279, "percentage": 51.67, "elapsed_time": "1:21:38", "remaining_time": "1:16:22"}
+{"current_steps": 280, "total_steps": 540, "loss": 0.4295, "lr": 5.548350297062659e-06, "epoch": 1.5498154981549814, "percentage": 51.85, "elapsed_time": "1:21:57", "remaining_time": "1:16:05"}
+{"current_steps": 281, "total_steps": 540, "loss": 0.4829, "lr": 5.516213110403009e-06, "epoch": 1.555350553505535, "percentage": 52.04, "elapsed_time": "1:22:14", "remaining_time": "1:15:48"}
+{"current_steps": 282, "total_steps": 540, "loss": 0.4284, "lr": 5.484054353515896e-06, "epoch": 1.5608856088560885, "percentage": 52.22, "elapsed_time": "1:22:31", "remaining_time": "1:15:30"}
+{"current_steps": 283, "total_steps": 540, "loss": 0.4779, "lr": 5.451875370171341e-06, "epoch": 1.566420664206642, "percentage": 52.41, "elapsed_time": "1:22:48", "remaining_time": "1:15:11"}
+{"current_steps": 284, "total_steps": 540, "loss": 0.4608, "lr": 5.419677504984534e-06, "epoch": 1.5719557195571956, "percentage": 52.59, "elapsed_time": "1:23:05", "remaining_time": "1:14:54"}
+{"current_steps": 285, "total_steps": 540, "loss": 0.4556, "lr": 5.387462103359655e-06, "epoch": 1.5774907749077491, "percentage": 52.78, "elapsed_time": "1:23:23", "remaining_time": "1:14:36"}
+{"current_steps": 286, "total_steps": 540, "loss": 0.4789, "lr": 5.3552305114336515e-06, "epoch": 1.5830258302583027, "percentage": 52.96, "elapsed_time": "1:23:39", "remaining_time": "1:14:18"}
+{"current_steps": 287, "total_steps": 540, "loss": 0.4469, "lr": 5.32298407601999e-06, "epoch": 1.5885608856088562, "percentage": 53.15, "elapsed_time": "1:23:56", "remaining_time": "1:13:59"}
+{"current_steps": 288, "total_steps": 540, "loss": 0.4452, "lr": 5.290724144552379e-06, "epoch": 1.5940959409594095, "percentage": 53.33, "elapsed_time": "1:24:13", "remaining_time": "1:13:42"}
+{"current_steps": 289, "total_steps": 540, "loss": 0.4391, "lr": 5.258452065028473e-06, "epoch": 1.599630996309963, "percentage": 53.52, "elapsed_time": "1:24:32", "remaining_time": "1:13:25"}
+{"current_steps": 290, "total_steps": 540, "loss": 0.4741, "lr": 5.2261691859535325e-06, "epoch": 1.6051660516605166, "percentage": 53.7, "elapsed_time": "1:24:50", "remaining_time": "1:13:08"}
+{"current_steps": 291, "total_steps": 540, "loss": 0.4814, "lr": 5.193876856284085e-06, "epoch": 1.6107011070110702, "percentage": 53.89, "elapsed_time": "1:25:07", "remaining_time": "1:12:50"}
+{"current_steps": 292, "total_steps": 540, "loss": 0.4112, "lr": 5.161576425371554e-06, "epoch": 1.6162361623616235, "percentage": 54.07, "elapsed_time": "1:25:23", "remaining_time": "1:12:31"}
+{"current_steps": 293, "total_steps": 540, "loss": 0.4432, "lr": 5.1292692429058824e-06, "epoch": 1.621771217712177, "percentage": 54.26, "elapsed_time": "1:25:41", "remaining_time": "1:12:14"}
+{"current_steps": 294, "total_steps": 540, "loss": 0.4759, "lr": 5.096956658859122e-06, "epoch": 1.6273062730627306, "percentage": 54.44, "elapsed_time": "1:25:59", "remaining_time": "1:11:56"}
+{"current_steps": 295, "total_steps": 540, "loss": 0.4357, "lr": 5.064640023429042e-06, "epoch": 1.632841328413284, "percentage": 54.63, "elapsed_time": "1:26:14", "remaining_time": "1:11:37"}
+{"current_steps": 296, "total_steps": 540, "loss": 0.4726, "lr": 5.032320686982697e-06, "epoch": 1.6383763837638377, "percentage": 54.81, "elapsed_time": "1:26:32", "remaining_time": "1:11:20"}
+{"current_steps": 297, "total_steps": 540, "loss": 0.4206, "lr": 5e-06, "epoch": 1.6439114391143912, "percentage": 55.0, "elapsed_time": "1:26:51", "remaining_time": "1:11:03"}
+{"current_steps": 298, "total_steps": 540, "loss": 0.4485, "lr": 4.967679313017304e-06, "epoch": 1.6494464944649447, "percentage": 55.19, "elapsed_time": "1:27:08", "remaining_time": "1:10:46"}
+{"current_steps": 299, "total_steps": 540, "loss": 0.4423, "lr": 4.9353599765709585e-06, "epoch": 1.6549815498154983, "percentage": 55.37, "elapsed_time": "1:27:25", "remaining_time": "1:10:28"}
+{"current_steps": 300, "total_steps": 540, "loss": 0.4525, "lr": 4.903043341140879e-06, "epoch": 1.6605166051660518, "percentage": 55.56, "elapsed_time": "1:27:42", "remaining_time": "1:10:10"}
+{"current_steps": 301, "total_steps": 540, "loss": 0.4705, "lr": 4.870730757094121e-06, "epoch": 1.6660516605166051, "percentage": 55.74, "elapsed_time": "1:27:58", "remaining_time": "1:09:51"}
+{"current_steps": 302, "total_steps": 540, "loss": 0.4523, "lr": 4.838423574628447e-06, "epoch": 1.6715867158671587, "percentage": 55.93, "elapsed_time": "1:28:15", "remaining_time": "1:09:33"}
+{"current_steps": 303, "total_steps": 540, "loss": 0.4841, "lr": 4.806123143715916e-06, "epoch": 1.6771217712177122, "percentage": 56.11, "elapsed_time": "1:28:33", "remaining_time": "1:09:16"}
+{"current_steps": 304, "total_steps": 540, "loss": 0.3898, "lr": 4.773830814046469e-06, "epoch": 1.6826568265682655, "percentage": 56.3, "elapsed_time": "1:28:50", "remaining_time": "1:08:58"}
+{"current_steps": 305, "total_steps": 540, "loss": 0.4716, "lr": 4.741547934971528e-06, "epoch": 1.688191881918819, "percentage": 56.48, "elapsed_time": "1:29:07", "remaining_time": "1:08:40"}
+{"current_steps": 306, "total_steps": 540, "loss": 0.4246, "lr": 4.7092758554476215e-06, "epoch": 1.6937269372693726, "percentage": 56.67, "elapsed_time": "1:29:25", "remaining_time": "1:08:22"}
+{"current_steps": 307, "total_steps": 540, "loss": 0.4659, "lr": 4.677015923980012e-06, "epoch": 1.6992619926199262, "percentage": 56.85, "elapsed_time": "1:29:42", "remaining_time": "1:08:05"}
+{"current_steps": 308, "total_steps": 540, "loss": 0.4739, "lr": 4.644769488566351e-06, "epoch": 1.7047970479704797, "percentage": 57.04, "elapsed_time": "1:30:01", "remaining_time": "1:07:48"}
+{"current_steps": 309, "total_steps": 540, "loss": 0.4604, "lr": 4.6125378966403465e-06, "epoch": 1.7103321033210332, "percentage": 57.22, "elapsed_time": "1:30:19", "remaining_time": "1:07:31"}
+{"current_steps": 310, "total_steps": 540, "loss": 0.4506, "lr": 4.580322495015466e-06, "epoch": 1.7158671586715868, "percentage": 57.41, "elapsed_time": "1:30:37", "remaining_time": "1:07:14"}
+{"current_steps": 311, "total_steps": 540, "loss": 0.4048, "lr": 4.548124629828661e-06, "epoch": 1.7214022140221403, "percentage": 57.59, "elapsed_time": "1:30:53", "remaining_time": "1:06:55"}
+{"current_steps": 312, "total_steps": 540, "loss": 0.4654, "lr": 4.515945646484105e-06, "epoch": 1.7269372693726939, "percentage": 57.78, "elapsed_time": "1:31:10", "remaining_time": "1:06:37"}
+{"current_steps": 313, "total_steps": 540, "loss": 0.4789, "lr": 4.483786889596993e-06, "epoch": 1.7324723247232472, "percentage": 57.96, "elapsed_time": "1:31:27", "remaining_time": "1:06:19"}
+{"current_steps": 314, "total_steps": 540, "loss": 0.4604, "lr": 4.451649702937343e-06, "epoch": 1.7380073800738007, "percentage": 58.15, "elapsed_time": "1:31:44", "remaining_time": "1:06:01"}
+{"current_steps": 315, "total_steps": 540, "loss": 0.4698, "lr": 4.4195354293738484e-06, "epoch": 1.7435424354243543, "percentage": 58.33, "elapsed_time": "1:32:02", "remaining_time": "1:05:44"}
+{"current_steps": 316, "total_steps": 540, "loss": 0.4445, "lr": 4.387445410817774e-06, "epoch": 1.7490774907749076, "percentage": 58.52, "elapsed_time": "1:32:19", "remaining_time": "1:05:26"}
+{"current_steps": 317, "total_steps": 540, "loss": 0.4216, "lr": 4.355380988166867e-06, "epoch": 1.7546125461254611, "percentage": 58.7, "elapsed_time": "1:32:37", "remaining_time": "1:05:09"}
+{"current_steps": 318, "total_steps": 540, "loss": 0.4716, "lr": 4.323343501249346e-06, "epoch": 1.7601476014760147, "percentage": 58.89, "elapsed_time": "1:32:53", "remaining_time": "1:04:51"}
+{"current_steps": 319, "total_steps": 540, "loss": 0.4426, "lr": 4.291334288767899e-06, "epoch": 1.7656826568265682, "percentage": 59.07, "elapsed_time": "1:33:10", "remaining_time": "1:04:33"}
+{"current_steps": 320, "total_steps": 540, "loss": 0.4416, "lr": 4.259354688243758e-06, "epoch": 1.7712177121771218, "percentage": 59.26, "elapsed_time": "1:33:27", "remaining_time": "1:04:14"}
+{"current_steps": 321, "total_steps": 540, "loss": 0.4487, "lr": 4.227406035960798e-06, "epoch": 1.7767527675276753, "percentage": 59.44, "elapsed_time": "1:33:44", "remaining_time": "1:03:57"}
+{"current_steps": 322, "total_steps": 540, "loss": 0.421, "lr": 4.195489666909714e-06, "epoch": 1.7822878228782288, "percentage": 59.63, "elapsed_time": "1:34:02", "remaining_time": "1:03:39"}
+{"current_steps": 323, "total_steps": 540, "loss": 0.4879, "lr": 4.163606914732224e-06, "epoch": 1.7878228782287824, "percentage": 59.81, "elapsed_time": "1:34:18", "remaining_time": "1:03:21"}
+{"current_steps": 324, "total_steps": 540, "loss": 0.4564, "lr": 4.131759111665349e-06, "epoch": 1.793357933579336, "percentage": 60.0, "elapsed_time": "1:34:36", "remaining_time": "1:03:04"}
+{"current_steps": 325, "total_steps": 540, "loss": 0.4521, "lr": 4.099947588485744e-06, "epoch": 1.7988929889298892, "percentage": 60.19, "elapsed_time": "1:34:53", "remaining_time": "1:02:46"}
+{"current_steps": 326, "total_steps": 540, "loss": 0.4192, "lr": 4.06817367445409e-06, "epoch": 1.8044280442804428, "percentage": 60.37, "elapsed_time": "1:35:09", "remaining_time": "1:02:27"}
+{"current_steps": 327, "total_steps": 540, "loss": 0.444, "lr": 4.036438697259551e-06, "epoch": 1.8099630996309963, "percentage": 60.56, "elapsed_time": "1:35:26", "remaining_time": "1:02:10"}
+{"current_steps": 328, "total_steps": 540, "loss": 0.4835, "lr": 4.004743982964298e-06, "epoch": 1.8154981549815496, "percentage": 60.74, "elapsed_time": "1:35:45", "remaining_time": "1:01:53"}
+{"current_steps": 329, "total_steps": 540, "loss": 0.4262, "lr": 3.9730908559480904e-06, "epoch": 1.8210332103321032, "percentage": 60.93, "elapsed_time": "1:36:02", "remaining_time": "1:01:35"}
+{"current_steps": 330, "total_steps": 540, "loss": 0.4501, "lr": 3.941480638852948e-06, "epoch": 1.8265682656826567, "percentage": 61.11, "elapsed_time": "1:36:19", "remaining_time": "1:01:17"}
+{"current_steps": 331, "total_steps": 540, "loss": 0.4524, "lr": 3.909914652527872e-06, "epoch": 1.8321033210332103, "percentage": 61.3, "elapsed_time": "1:36:37", "remaining_time": "1:01:00"}
+{"current_steps": 332, "total_steps": 540, "loss": 0.4637, "lr": 3.878394215973663e-06, "epoch": 1.8376383763837638, "percentage": 61.48, "elapsed_time": "1:36:55", "remaining_time": "1:00:43"}
+{"current_steps": 333, "total_steps": 540, "loss": 0.4351, "lr": 3.8469206462878e-06, "epoch": 1.8431734317343174, "percentage": 61.67, "elapsed_time": "1:37:10", "remaining_time": "1:00:24"}
+{"current_steps": 334, "total_steps": 540, "loss": 0.4466, "lr": 3.815495258609404e-06, "epoch": 1.848708487084871, "percentage": 61.85, "elapsed_time": "1:37:28", "remaining_time": "1:00:07"}
+{"current_steps": 335, "total_steps": 540, "loss": 0.4597, "lr": 3.784119366064293e-06, "epoch": 1.8542435424354244, "percentage": 62.04, "elapsed_time": "1:37:46", "remaining_time": "0:59:49"}
+{"current_steps": 336, "total_steps": 540, "loss": 0.4845, "lr": 3.752794279710094e-06, "epoch": 1.859778597785978, "percentage": 62.22, "elapsed_time": "1:38:04", "remaining_time": "0:59:32"}
+{"current_steps": 337, "total_steps": 540, "loss": 0.4046, "lr": 3.721521308481483e-06, "epoch": 1.8653136531365315, "percentage": 62.41, "elapsed_time": "1:38:21", "remaining_time": "0:59:14"}
+{"current_steps": 338, "total_steps": 540, "loss": 0.4541, "lr": 3.690301759135471e-06, "epoch": 1.8708487084870848, "percentage": 62.59, "elapsed_time": "1:38:38", "remaining_time": "0:58:56"}
+{"current_steps": 339, "total_steps": 540, "loss": 0.4605, "lr": 3.6591369361968127e-06, "epoch": 1.8763837638376384, "percentage": 62.78, "elapsed_time": "1:38:56", "remaining_time": "0:58:39"}
+{"current_steps": 340, "total_steps": 540, "loss": 0.4486, "lr": 3.6280281419034934e-06, "epoch": 1.881918819188192, "percentage": 62.96, "elapsed_time": "1:39:15", "remaining_time": "0:58:22"}
+{"current_steps": 341, "total_steps": 540, "loss": 0.4538, "lr": 3.596976676152306e-06, "epoch": 1.8874538745387452, "percentage": 63.15, "elapsed_time": "1:39:33", "remaining_time": "0:58:05"}
+{"current_steps": 342, "total_steps": 540, "loss": 0.4683, "lr": 3.5659838364445505e-06, "epoch": 1.8929889298892988, "percentage": 63.33, "elapsed_time": "1:39:51", "remaining_time": "0:57:48"}
+{"current_steps": 343, "total_steps": 540, "loss": 0.4309, "lr": 3.535050917831797e-06, "epoch": 1.8985239852398523, "percentage": 63.52, "elapsed_time": "1:40:08", "remaining_time": "0:57:30"}
+{"current_steps": 344, "total_steps": 540, "loss": 0.4885, "lr": 3.504179212861793e-06, "epoch": 1.9040590405904059, "percentage": 63.7, "elapsed_time": "1:40:26", "remaining_time": "0:57:13"}
+{"current_steps": 345, "total_steps": 540, "loss": 0.4154, "lr": 3.473370011524435e-06, "epoch": 1.9095940959409594, "percentage": 63.89, "elapsed_time": "1:40:43", "remaining_time": "0:56:55"}
+{"current_steps": 346, "total_steps": 540, "loss": 0.4392, "lr": 3.442624601197877e-06, "epoch": 1.915129151291513, "percentage": 64.07, "elapsed_time": "1:41:00", "remaining_time": "0:56:38"}
+{"current_steps": 347, "total_steps": 540, "loss": 0.4866, "lr": 3.4119442665947346e-06, "epoch": 1.9206642066420665, "percentage": 64.26, "elapsed_time": "1:41:16", "remaining_time": "0:56:19"}
+{"current_steps": 348, "total_steps": 540, "loss": 0.4619, "lr": 3.3813302897083955e-06, "epoch": 1.92619926199262, "percentage": 64.44, "elapsed_time": "1:41:35", "remaining_time": "0:56:03"}
+{"current_steps": 349, "total_steps": 540, "loss": 0.416, "lr": 3.350783949759462e-06, "epoch": 1.9317343173431736, "percentage": 64.63, "elapsed_time": "1:41:53", "remaining_time": "0:55:45"}
+{"current_steps": 350, "total_steps": 540, "loss": 0.4499, "lr": 3.3203065231422904e-06, "epoch": 1.937269372693727, "percentage": 64.81, "elapsed_time": "1:42:11", "remaining_time": "0:55:28"}
+{"current_steps": 351, "total_steps": 540, "loss": 0.4638, "lr": 3.289899283371657e-06, "epoch": 1.9428044280442804, "percentage": 65.0, "elapsed_time": "1:42:28", "remaining_time": "0:55:10"}
+{"current_steps": 352, "total_steps": 540, "loss": 0.4605, "lr": 3.259563501029548e-06, "epoch": 1.948339483394834, "percentage": 65.19, "elapsed_time": "1:42:46", "remaining_time": "0:54:53"}
+{"current_steps": 353, "total_steps": 540, "loss": 0.4383, "lr": 3.2293004437120622e-06, "epoch": 1.9538745387453873, "percentage": 65.37, "elapsed_time": "1:43:03", "remaining_time": "0:54:35"}
+{"current_steps": 354, "total_steps": 540, "loss": 0.5036, "lr": 3.1991113759764493e-06, "epoch": 1.9594095940959408, "percentage": 65.56, "elapsed_time": "1:43:22", "remaining_time": "0:54:18"}
+{"current_steps": 355, "total_steps": 540, "loss": 0.3884, "lr": 3.1689975592882603e-06, "epoch": 1.9649446494464944, "percentage": 65.74, "elapsed_time": "1:43:36", "remaining_time": "0:53:59"}
+{"current_steps": 356, "total_steps": 540, "loss": 0.4737, "lr": 3.1389602519686515e-06, "epoch": 1.970479704797048, "percentage": 65.93, "elapsed_time": "1:43:53", "remaining_time": "0:53:41"}
+{"current_steps": 357, "total_steps": 540, "loss": 0.4183, "lr": 3.1090007091417884e-06, "epoch": 1.9760147601476015, "percentage": 66.11, "elapsed_time": "1:44:10", "remaining_time": "0:53:23"}
+{"current_steps": 358, "total_steps": 540, "loss": 0.4142, "lr": 3.0791201826824117e-06, "epoch": 1.981549815498155, "percentage": 66.3, "elapsed_time": "1:44:27", "remaining_time": "0:53:06"}
+{"current_steps": 359, "total_steps": 540, "loss": 0.4617, "lr": 3.049319921163526e-06, "epoch": 1.9870848708487086, "percentage": 66.48, "elapsed_time": "1:44:45", "remaining_time": "0:52:49"}
+{"current_steps": 360, "total_steps": 540, "loss": 0.4714, "lr": 3.019601169804216e-06, "epoch": 1.992619926199262, "percentage": 66.67, "elapsed_time": "1:45:02", "remaining_time": "0:52:31"}
+{"current_steps": 361, "total_steps": 540, "loss": 0.4929, "lr": 2.9899651704176324e-06, "epoch": 1.9981549815498156, "percentage": 66.85, "elapsed_time": "1:45:20", "remaining_time": "0:52:14"}
+{"current_steps": 362, "total_steps": 540, "loss": 0.5202, "lr": 2.9604131613590825e-06, "epoch": 2.003690036900369, "percentage": 67.04, "elapsed_time": "1:46:56", "remaining_time": "0:52:35"}
+{"current_steps": 363, "total_steps": 540, "loss": 0.4115, "lr": 2.9309463774743047e-06, "epoch": 2.0092250922509227, "percentage": 67.22, "elapsed_time": "1:47:13", "remaining_time": "0:52:16"}
+{"current_steps": 364, "total_steps": 540, "loss": 0.4151, "lr": 2.901566050047855e-06, "epoch": 2.014760147601476, "percentage": 67.41, "elapsed_time": "1:47:29", "remaining_time": "0:51:58"}
+{"current_steps": 365, "total_steps": 540, "loss": 0.4199, "lr": 2.8722734067516637e-06, "epoch": 2.0202952029520294, "percentage": 67.59, "elapsed_time": "1:47:48", "remaining_time": "0:51:41"}
+{"current_steps": 366, "total_steps": 540, "loss": 0.4601, "lr": 2.843069671593734e-06, "epoch": 2.025830258302583, "percentage": 67.78, "elapsed_time": "1:48:06", "remaining_time": "0:51:23"}
+{"current_steps": 367, "total_steps": 540, "loss": 0.3952, "lr": 2.813956064866996e-06, "epoch": 2.0313653136531364, "percentage": 67.96, "elapsed_time": "1:48:24", "remaining_time": "0:51:06"}
+{"current_steps": 368, "total_steps": 540, "loss": 0.4301, "lr": 2.784933803098326e-06, "epoch": 2.03690036900369, "percentage": 68.15, "elapsed_time": "1:48:41", "remaining_time": "0:50:47"}
+{"current_steps": 369, "total_steps": 540, "loss": 0.4316, "lr": 2.7560040989976894e-06, "epoch": 2.0424354243542435, "percentage": 68.33, "elapsed_time": "1:48:57", "remaining_time": "0:50:29"}
+{"current_steps": 370, "total_steps": 540, "loss": 0.4271, "lr": 2.7271681614074973e-06, "epoch": 2.047970479704797, "percentage": 68.52, "elapsed_time": "1:49:14", "remaining_time": "0:50:11"}
+{"current_steps": 371, "total_steps": 540, "loss": 0.4226, "lr": 2.6984271952520723e-06, "epoch": 2.0535055350553506, "percentage": 68.7, "elapsed_time": "1:49:32", "remaining_time": "0:49:54"}