Training in progress, step 1000

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b8c9c801ad0e0026117f5694d056da71527f0dad31739ebcac9a2abc96a1b8c4
+oid sha256:bb25e3407e67ac82e39a2c7bbdce7d708ef473fd168953791d03a359b9079c35
 size 65309632

trainer_log.jsonl

@@ -49,3 +49,54 @@
 {"current_steps": 490, "total_steps": 1854, "loss": 1.4904, "accuracy": 0.5, "learning_rate": 4.186861743633911e-06, "epoch": 0.7920792079207921, "percentage": 26.43, "elapsed_time": "1:16:26", "remaining_time": "3:32:46"}
 {"current_steps": 500, "total_steps": 1854, "loss": 1.5005, "accuracy": 0.46875, "learning_rate": 4.155353766456497e-06, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "1:18:04", "remaining_time": "3:31:26"}
 {"current_steps": 500, "total_steps": 1854, "eval_loss": 1.5202080011367798, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "1:21:32", "remaining_time": "3:40:48"}
+{"current_steps": 510, "total_steps": 1854, "loss": 1.4927, "accuracy": 0.5, "learning_rate": 4.123370445773134e-06, "epoch": 0.8244089715093958, "percentage": 27.51, "elapsed_time": "1:23:09", "remaining_time": "3:39:08"}
+{"current_steps": 520, "total_steps": 1854, "loss": 1.559, "accuracy": 0.5375000238418579, "learning_rate": 4.090920965761906e-06, "epoch": 0.8405738533036977, "percentage": 28.05, "elapsed_time": "1:24:37", "remaining_time": "3:37:06"}
+{"current_steps": 530, "total_steps": 1854, "loss": 1.4967, "accuracy": 0.5375000238418579, "learning_rate": 4.058014644460991e-06, "epoch": 0.8567387350979996, "percentage": 28.59, "elapsed_time": "1:26:00", "remaining_time": "3:34:52"}
+{"current_steps": 540, "total_steps": 1854, "loss": 1.4748, "accuracy": 0.4749999940395355, "learning_rate": 4.024660931092939e-06, "epoch": 0.8729036168923014, "percentage": 29.13, "elapsed_time": "1:27:36", "remaining_time": "3:33:10"}
+{"current_steps": 550, "total_steps": 1854, "loss": 1.5359, "accuracy": 0.550000011920929, "learning_rate": 3.990869403351272e-06, "epoch": 0.8890684986866033, "percentage": 29.67, "elapsed_time": "1:29:11", "remaining_time": "3:31:28"}
+{"current_steps": 560, "total_steps": 1854, "loss": 1.5325, "accuracy": 0.42500001192092896, "learning_rate": 3.956649764650206e-06, "epoch": 0.9052333804809052, "percentage": 30.2, "elapsed_time": "1:30:54", "remaining_time": "3:30:03"}
+{"current_steps": 570, "total_steps": 1854, "loss": 1.4679, "accuracy": 0.5249999761581421, "learning_rate": 3.92201184133826e-06, "epoch": 0.9213982622752072, "percentage": 30.74, "elapsed_time": "1:32:24", "remaining_time": "3:28:09"}
+{"current_steps": 580, "total_steps": 1854, "loss": 1.4501, "accuracy": 0.5249999761581421, "learning_rate": 3.886965579876572e-06, "epoch": 0.937563144069509, "percentage": 31.28, "elapsed_time": "1:33:56", "remaining_time": "3:26:21"}
+{"current_steps": 590, "total_steps": 1854, "loss": 1.4998, "accuracy": 0.45625001192092896, "learning_rate": 3.851521043982716e-06, "epoch": 0.9537280258638109, "percentage": 31.82, "elapsed_time": "1:35:31", "remaining_time": "3:24:39"}
+{"current_steps": 600, "total_steps": 1854, "loss": 1.5191, "accuracy": 0.5, "learning_rate": 3.81568841174086e-06, "epoch": 0.9698929076581128, "percentage": 32.36, "elapsed_time": "1:37:12", "remaining_time": "3:23:09"}
+{"current_steps": 610, "total_steps": 1854, "loss": 1.4561, "accuracy": 0.44999998807907104, "learning_rate": 3.7794779726790664e-06, "epoch": 0.9860577894524146, "percentage": 32.9, "elapsed_time": "1:38:46", "remaining_time": "3:21:25"}
+{"current_steps": 620, "total_steps": 1854, "loss": 1.4924, "accuracy": 0.53125, "learning_rate": 3.7429001248146096e-06, "epoch": 1.0022226712467166, "percentage": 33.44, "elapsed_time": "1:40:25", "remaining_time": "3:19:51"}
+{"current_steps": 630, "total_steps": 1854, "loss": 1.4861, "accuracy": 0.5, "learning_rate": 3.7059653716681227e-06, "epoch": 1.0183875530410185, "percentage": 33.98, "elapsed_time": "1:41:56", "remaining_time": "3:18:04"}
+{"current_steps": 640, "total_steps": 1854, "loss": 1.4348, "accuracy": 0.5562499761581421, "learning_rate": 3.668684319247463e-06, "epoch": 1.0345524348353203, "percentage": 34.52, "elapsed_time": "1:43:26", "remaining_time": "3:16:12"}
+{"current_steps": 650, "total_steps": 1854, "loss": 1.3979, "accuracy": 0.5062500238418579, "learning_rate": 3.6310676730021373e-06, "epoch": 1.0507173166296222, "percentage": 35.06, "elapsed_time": "1:44:56", "remaining_time": "3:14:23"}
+{"current_steps": 660, "total_steps": 1854, "loss": 1.4976, "accuracy": 0.4749999940395355, "learning_rate": 3.593126234749178e-06, "epoch": 1.066882198423924, "percentage": 35.6, "elapsed_time": "1:46:32", "remaining_time": "3:12:44"}
+{"current_steps": 670, "total_steps": 1854, "loss": 1.4767, "accuracy": 0.5625, "learning_rate": 3.554870899571343e-06, "epoch": 1.083047080218226, "percentage": 36.14, "elapsed_time": "1:48:07", "remaining_time": "3:11:04"}
+{"current_steps": 680, "total_steps": 1854, "loss": 1.4506, "accuracy": 0.46875, "learning_rate": 3.5163126526885373e-06, "epoch": 1.0992119620125278, "percentage": 36.68, "elapsed_time": "1:49:42", "remaining_time": "3:09:24"}
+{"current_steps": 690, "total_steps": 1854, "loss": 1.4783, "accuracy": 0.518750011920929, "learning_rate": 3.4774625663033484e-06, "epoch": 1.1153768438068297, "percentage": 37.22, "elapsed_time": "1:51:14", "remaining_time": "3:07:39"}
+{"current_steps": 700, "total_steps": 1854, "loss": 1.4108, "accuracy": 0.4625000059604645, "learning_rate": 3.4383317964216067e-06, "epoch": 1.1315417256011315, "percentage": 37.76, "elapsed_time": "1:52:50", "remaining_time": "3:06:01"}
+{"current_steps": 710, "total_steps": 1854, "loss": 1.5203, "accuracy": 0.5375000238418579, "learning_rate": 3.398931579648877e-06, "epoch": 1.1477066073954334, "percentage": 38.3, "elapsed_time": "1:54:31", "remaining_time": "3:04:31"}
+{"current_steps": 720, "total_steps": 1854, "loss": 1.421, "accuracy": 0.4625000059604645, "learning_rate": 3.359273229963813e-06, "epoch": 1.1638714891897353, "percentage": 38.83, "elapsed_time": "1:56:05", "remaining_time": "3:02:51"}
+{"current_steps": 730, "total_steps": 1854, "loss": 1.4538, "accuracy": 0.45625001192092896, "learning_rate": 3.319368135469285e-06, "epoch": 1.1800363709840371, "percentage": 39.37, "elapsed_time": "1:57:34", "remaining_time": "3:01:01"}
+{"current_steps": 740, "total_steps": 1854, "loss": 1.3866, "accuracy": 0.612500011920929, "learning_rate": 3.279227755122228e-06, "epoch": 1.196201252778339, "percentage": 39.91, "elapsed_time": "1:59:10", "remaining_time": "2:59:24"}
+{"current_steps": 750, "total_steps": 1854, "loss": 1.502, "accuracy": 0.48750001192092896, "learning_rate": 3.2388636154431417e-06, "epoch": 1.2123661345726409, "percentage": 40.45, "elapsed_time": "2:00:50", "remaining_time": "2:57:52"}
+{"current_steps": 760, "total_steps": 1854, "loss": 1.4829, "accuracy": 0.4749999940395355, "learning_rate": 3.198287307206192e-06, "epoch": 1.2285310163669427, "percentage": 40.99, "elapsed_time": "2:02:23", "remaining_time": "2:56:11"}
+{"current_steps": 770, "total_steps": 1854, "loss": 1.4004, "accuracy": 0.5249999761581421, "learning_rate": 3.157510482110856e-06, "epoch": 1.2446958981612446, "percentage": 41.53, "elapsed_time": "2:03:59", "remaining_time": "2:54:33"}
+{"current_steps": 780, "total_steps": 1854, "loss": 1.588, "accuracy": 0.4749999940395355, "learning_rate": 3.116544849436077e-06, "epoch": 1.2608607799555465, "percentage": 42.07, "elapsed_time": "2:05:36", "remaining_time": "2:52:57"}
+{"current_steps": 790, "total_steps": 1854, "loss": 1.4162, "accuracy": 0.550000011920929, "learning_rate": 3.0754021726778848e-06, "epoch": 1.2770256617498483, "percentage": 42.61, "elapsed_time": "2:07:11", "remaining_time": "2:51:18"}
+{"current_steps": 800, "total_steps": 1854, "loss": 1.5077, "accuracy": 0.4375, "learning_rate": 3.0340942661714463e-06, "epoch": 1.2931905435441502, "percentage": 43.15, "elapsed_time": "2:08:48", "remaining_time": "2:49:41"}
+{"current_steps": 810, "total_steps": 1854, "loss": 1.4207, "accuracy": 0.550000011920929, "learning_rate": 2.992632991698512e-06, "epoch": 1.3093554253384523, "percentage": 43.69, "elapsed_time": "2:10:19", "remaining_time": "2:47:59"}
+{"current_steps": 820, "total_steps": 1854, "loss": 1.4155, "accuracy": 0.48124998807907104, "learning_rate": 2.9510302550812537e-06, "epoch": 1.3255203071327541, "percentage": 44.23, "elapsed_time": "2:11:52", "remaining_time": "2:46:17"}
+{"current_steps": 830, "total_steps": 1854, "loss": 1.3858, "accuracy": 0.48124998807907104, "learning_rate": 2.9092980027634325e-06, "epoch": 1.341685188927056, "percentage": 44.77, "elapsed_time": "2:13:27", "remaining_time": "2:44:38"}
+{"current_steps": 840, "total_steps": 1854, "loss": 1.4355, "accuracy": 0.512499988079071, "learning_rate": 2.867448218379927e-06, "epoch": 1.3578500707213579, "percentage": 45.31, "elapsed_time": "2:14:57", "remaining_time": "2:42:55"}
+{"current_steps": 850, "total_steps": 1854, "loss": 1.4789, "accuracy": 0.5062500238418579, "learning_rate": 2.825492919315559e-06, "epoch": 1.3740149525156597, "percentage": 45.85, "elapsed_time": "2:16:31", "remaining_time": "2:41:16"}
+{"current_steps": 860, "total_steps": 1854, "loss": 1.4297, "accuracy": 0.581250011920929, "learning_rate": 2.7834441532542482e-06, "epoch": 1.3901798343099616, "percentage": 46.39, "elapsed_time": "2:18:08", "remaining_time": "2:39:39"}
+{"current_steps": 870, "total_steps": 1854, "loss": 1.4755, "accuracy": 0.5, "learning_rate": 2.74131399471945e-06, "epoch": 1.4063447161042635, "percentage": 46.93, "elapsed_time": "2:19:40", "remaining_time": "2:37:59"}
+{"current_steps": 880, "total_steps": 1854, "loss": 1.4645, "accuracy": 0.4625000059604645, "learning_rate": 2.6991145416068947e-06, "epoch": 1.4225095978985653, "percentage": 47.46, "elapsed_time": "2:21:19", "remaining_time": "2:36:25"}
+{"current_steps": 890, "total_steps": 1854, "loss": 1.3933, "accuracy": 0.550000011920929, "learning_rate": 2.6568579117106143e-06, "epoch": 1.4386744796928672, "percentage": 48.0, "elapsed_time": "2:22:56", "remaining_time": "2:34:49"}
+{"current_steps": 900, "total_steps": 1854, "loss": 1.4466, "accuracy": 0.5, "learning_rate": 2.6145562392432544e-06, "epoch": 1.454839361487169, "percentage": 48.54, "elapsed_time": "2:24:28", "remaining_time": "2:33:08"}
+{"current_steps": 910, "total_steps": 1854, "loss": 1.3653, "accuracy": 0.48124998807907104, "learning_rate": 2.5722216713516682e-06, "epoch": 1.471004243281471, "percentage": 49.08, "elapsed_time": "2:25:56", "remaining_time": "2:31:23"}
+{"current_steps": 920, "total_steps": 1854, "loss": 1.4084, "accuracy": 0.518750011920929, "learning_rate": 2.5298663646288064e-06, "epoch": 1.4871691250757728, "percentage": 49.62, "elapsed_time": "2:27:35", "remaining_time": "2:29:50"}
+{"current_steps": 930, "total_steps": 1854, "loss": 1.3983, "accuracy": 0.4937500059604645, "learning_rate": 2.487502481622879e-06, "epoch": 1.503334006870075, "percentage": 50.16, "elapsed_time": "2:29:04", "remaining_time": "2:28:07"}
+{"current_steps": 940, "total_steps": 1854, "loss": 1.508, "accuracy": 0.44999998807907104, "learning_rate": 2.4451421873448253e-06, "epoch": 1.5194988886643768, "percentage": 50.7, "elapsed_time": "2:30:38", "remaining_time": "2:26:28"}
+{"current_steps": 950, "total_steps": 1854, "loss": 1.48, "accuracy": 0.44999998807907104, "learning_rate": 2.40279764577506e-06, "epoch": 1.5356637704586786, "percentage": 51.24, "elapsed_time": "2:32:14", "remaining_time": "2:24:52"}
+{"current_steps": 960, "total_steps": 1854, "loss": 1.377, "accuracy": 0.512499988079071, "learning_rate": 2.3604810163705242e-06, "epoch": 1.5518286522529805, "percentage": 51.78, "elapsed_time": "2:33:49", "remaining_time": "2:23:14"}
+{"current_steps": 970, "total_steps": 1854, "loss": 1.3349, "accuracy": 0.5062500238418579, "learning_rate": 2.3182044505730364e-06, "epoch": 1.5679935340472824, "percentage": 52.32, "elapsed_time": "2:35:22", "remaining_time": "2:21:35"}
+{"current_steps": 980, "total_steps": 1854, "loss": 1.3406, "accuracy": 0.5062500238418579, "learning_rate": 2.275980088319941e-06, "epoch": 1.5841584158415842, "percentage": 52.86, "elapsed_time": "2:36:56", "remaining_time": "2:19:57"}
+{"current_steps": 990, "total_steps": 1854, "loss": 1.3345, "accuracy": 0.5, "learning_rate": 2.2338200545580577e-06, "epoch": 1.600323297635886, "percentage": 53.4, "elapsed_time": "2:38:27", "remaining_time": "2:18:17"}
+{"current_steps": 1000, "total_steps": 1854, "loss": 1.3401, "accuracy": 0.4437499940395355, "learning_rate": 2.191736455761947e-06, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "2:39:58", "remaining_time": "2:16:36"}
+{"current_steps": 1000, "total_steps": 1854, "eval_loss": 1.4635207653045654, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "2:43:26", "remaining_time": "2:19:34"}