Training in progress, step 1000

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:d6bdaa76d182665e5fc9cbfe55d2c39f93c81cd2702cbee48dc959bc747e4e8e
+oid sha256:9d30982d84fdb6a03f886df94d2d159ef59282a19ce192bb4d7698d78fa1ce71
 size 39256456

trainer_log.jsonl

@@ -49,3 +49,54 @@
 {"current_steps": 490, "total_steps": 1854, "loss": 1.3277, "accuracy": 0.48750001192092896, "learning_rate": 4.186861743633911e-06, "epoch": 0.7920792079207921, "percentage": 26.43, "elapsed_time": "0:36:24", "remaining_time": "1:41:20"}
 {"current_steps": 500, "total_steps": 1854, "loss": 1.4057, "accuracy": 0.5, "learning_rate": 4.155353766456497e-06, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "0:37:11", "remaining_time": "1:40:41"}
 {"current_steps": 500, "total_steps": 1854, "eval_loss": 1.38706636428833, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "0:38:33", "remaining_time": "1:44:24"}
+{"current_steps": 510, "total_steps": 1854, "loss": 1.3519, "accuracy": 0.53125, "learning_rate": 4.123370445773134e-06, "epoch": 0.8244089715093958, "percentage": 27.51, "elapsed_time": "0:39:29", "remaining_time": "1:44:05"}
+{"current_steps": 520, "total_steps": 1854, "loss": 1.3752, "accuracy": 0.5687500238418579, "learning_rate": 4.090920965761906e-06, "epoch": 0.8405738533036977, "percentage": 28.05, "elapsed_time": "0:40:11", "remaining_time": "1:43:06"}
+{"current_steps": 530, "total_steps": 1854, "loss": 1.3819, "accuracy": 0.5562499761581421, "learning_rate": 4.058014644460991e-06, "epoch": 0.8567387350979996, "percentage": 28.59, "elapsed_time": "0:40:51", "remaining_time": "1:42:02"}
+{"current_steps": 540, "total_steps": 1854, "loss": 1.4012, "accuracy": 0.512499988079071, "learning_rate": 4.024660931092939e-06, "epoch": 0.8729036168923014, "percentage": 29.13, "elapsed_time": "0:41:36", "remaining_time": "1:41:15"}
+{"current_steps": 550, "total_steps": 1854, "loss": 1.3887, "accuracy": 0.59375, "learning_rate": 3.990869403351272e-06, "epoch": 0.8890684986866033, "percentage": 29.67, "elapsed_time": "0:42:22", "remaining_time": "1:40:26"}
+{"current_steps": 560, "total_steps": 1854, "loss": 1.3633, "accuracy": 0.45625001192092896, "learning_rate": 3.956649764650206e-06, "epoch": 0.9052333804809052, "percentage": 30.2, "elapsed_time": "0:43:10", "remaining_time": "1:39:46"}
+{"current_steps": 570, "total_steps": 1854, "loss": 1.3613, "accuracy": 0.581250011920929, "learning_rate": 3.92201184133826e-06, "epoch": 0.9213982622752072, "percentage": 30.74, "elapsed_time": "0:43:53", "remaining_time": "1:38:53"}
+{"current_steps": 580, "total_steps": 1854, "loss": 1.3741, "accuracy": 0.5, "learning_rate": 3.886965579876572e-06, "epoch": 0.937563144069509, "percentage": 31.28, "elapsed_time": "0:44:37", "remaining_time": "1:38:01"}
+{"current_steps": 590, "total_steps": 1854, "loss": 1.3862, "accuracy": 0.48750001192092896, "learning_rate": 3.851521043982716e-06, "epoch": 0.9537280258638109, "percentage": 31.82, "elapsed_time": "0:45:22", "remaining_time": "1:37:12"}
+{"current_steps": 600, "total_steps": 1854, "loss": 1.4338, "accuracy": 0.574999988079071, "learning_rate": 3.81568841174086e-06, "epoch": 0.9698929076581128, "percentage": 32.36, "elapsed_time": "0:46:10", "remaining_time": "1:36:30"}
+{"current_steps": 610, "total_steps": 1854, "loss": 1.2907, "accuracy": 0.518750011920929, "learning_rate": 3.7794779726790664e-06, "epoch": 0.9860577894524146, "percentage": 32.9, "elapsed_time": "0:46:55", "remaining_time": "1:35:41"}
+{"current_steps": 620, "total_steps": 1854, "loss": 1.3864, "accuracy": 0.6000000238418579, "learning_rate": 3.7429001248146096e-06, "epoch": 1.0022226712467166, "percentage": 33.44, "elapsed_time": "0:47:41", "remaining_time": "1:34:55"}
+{"current_steps": 630, "total_steps": 1854, "loss": 1.3899, "accuracy": 0.5249999761581421, "learning_rate": 3.7059653716681227e-06, "epoch": 1.0183875530410185, "percentage": 33.98, "elapsed_time": "0:48:25", "remaining_time": "1:34:05"}
+{"current_steps": 640, "total_steps": 1854, "loss": 1.298, "accuracy": 0.5874999761581421, "learning_rate": 3.668684319247463e-06, "epoch": 1.0345524348353203, "percentage": 34.52, "elapsed_time": "0:49:08", "remaining_time": "1:33:12"}
+{"current_steps": 650, "total_steps": 1854, "loss": 1.3324, "accuracy": 0.518750011920929, "learning_rate": 3.6310676730021373e-06, "epoch": 1.0507173166296222, "percentage": 35.06, "elapsed_time": "0:49:51", "remaining_time": "1:32:21"}
+{"current_steps": 660, "total_steps": 1854, "loss": 1.375, "accuracy": 0.518750011920929, "learning_rate": 3.593126234749178e-06, "epoch": 1.066882198423924, "percentage": 35.6, "elapsed_time": "0:50:36", "remaining_time": "1:31:33"}
+{"current_steps": 670, "total_steps": 1854, "loss": 1.3498, "accuracy": 0.5625, "learning_rate": 3.554870899571343e-06, "epoch": 1.083047080218226, "percentage": 36.14, "elapsed_time": "0:51:22", "remaining_time": "1:30:47"}
+{"current_steps": 680, "total_steps": 1854, "loss": 1.3095, "accuracy": 0.5249999761581421, "learning_rate": 3.5163126526885373e-06, "epoch": 1.0992119620125278, "percentage": 36.68, "elapsed_time": "0:52:07", "remaining_time": "1:29:59"}
+{"current_steps": 690, "total_steps": 1854, "loss": 1.3002, "accuracy": 0.5562499761581421, "learning_rate": 3.4774625663033484e-06, "epoch": 1.1153768438068297, "percentage": 37.22, "elapsed_time": "0:52:51", "remaining_time": "1:29:10"}
+{"current_steps": 700, "total_steps": 1854, "loss": 1.3116, "accuracy": 0.48124998807907104, "learning_rate": 3.4383317964216067e-06, "epoch": 1.1315417256011315, "percentage": 37.76, "elapsed_time": "0:53:37", "remaining_time": "1:28:23"}
+{"current_steps": 710, "total_steps": 1854, "loss": 1.3932, "accuracy": 0.5562499761581421, "learning_rate": 3.398931579648877e-06, "epoch": 1.1477066073954334, "percentage": 38.3, "elapsed_time": "0:54:25", "remaining_time": "1:27:40"}
+{"current_steps": 720, "total_steps": 1854, "loss": 1.315, "accuracy": 0.5249999761581421, "learning_rate": 3.359273229963813e-06, "epoch": 1.1638714891897353, "percentage": 38.83, "elapsed_time": "0:55:09", "remaining_time": "1:26:53"}
+{"current_steps": 730, "total_steps": 1854, "loss": 1.3682, "accuracy": 0.5249999761581421, "learning_rate": 3.319368135469285e-06, "epoch": 1.1800363709840371, "percentage": 39.37, "elapsed_time": "0:55:52", "remaining_time": "1:26:01"}
+{"current_steps": 740, "total_steps": 1854, "loss": 1.3163, "accuracy": 0.612500011920929, "learning_rate": 3.279227755122228e-06, "epoch": 1.196201252778339, "percentage": 39.91, "elapsed_time": "0:56:37", "remaining_time": "1:25:15"}
+{"current_steps": 750, "total_steps": 1854, "loss": 1.4216, "accuracy": 0.5625, "learning_rate": 3.2388636154431417e-06, "epoch": 1.2123661345726409, "percentage": 40.45, "elapsed_time": "0:57:25", "remaining_time": "1:24:31"}
+{"current_steps": 760, "total_steps": 1854, "loss": 1.3241, "accuracy": 0.5375000238418579, "learning_rate": 3.198287307206192e-06, "epoch": 1.2285310163669427, "percentage": 40.99, "elapsed_time": "0:58:09", "remaining_time": "1:23:42"}
+{"current_steps": 770, "total_steps": 1854, "loss": 1.337, "accuracy": 0.53125, "learning_rate": 3.157510482110856e-06, "epoch": 1.2446958981612446, "percentage": 41.53, "elapsed_time": "0:58:54", "remaining_time": "1:22:55"}
+{"current_steps": 780, "total_steps": 1854, "loss": 1.4243, "accuracy": 0.512499988079071, "learning_rate": 3.116544849436077e-06, "epoch": 1.2608607799555465, "percentage": 42.07, "elapsed_time": "0:59:40", "remaining_time": "1:22:10"}
+{"current_steps": 790, "total_steps": 1854, "loss": 1.2508, "accuracy": 0.5625, "learning_rate": 3.0754021726778848e-06, "epoch": 1.2770256617498483, "percentage": 42.61, "elapsed_time": "1:00:25", "remaining_time": "1:21:23"}
+{"current_steps": 800, "total_steps": 1854, "loss": 1.4051, "accuracy": 0.53125, "learning_rate": 3.0340942661714463e-06, "epoch": 1.2931905435441502, "percentage": 43.15, "elapsed_time": "1:01:11", "remaining_time": "1:20:37"}
+{"current_steps": 810, "total_steps": 1854, "loss": 1.3541, "accuracy": 0.6000000238418579, "learning_rate": 2.992632991698512e-06, "epoch": 1.3093554253384523, "percentage": 43.69, "elapsed_time": "1:01:55", "remaining_time": "1:19:48"}
+{"current_steps": 820, "total_steps": 1854, "loss": 1.2925, "accuracy": 0.543749988079071, "learning_rate": 2.9510302550812537e-06, "epoch": 1.3255203071327541, "percentage": 44.23, "elapsed_time": "1:02:38", "remaining_time": "1:18:59"}
+{"current_steps": 830, "total_steps": 1854, "loss": 1.2381, "accuracy": 0.5687500238418579, "learning_rate": 2.9092980027634325e-06, "epoch": 1.341685188927056, "percentage": 44.77, "elapsed_time": "1:03:23", "remaining_time": "1:18:12"}
+{"current_steps": 840, "total_steps": 1854, "loss": 1.359, "accuracy": 0.550000011920929, "learning_rate": 2.867448218379927e-06, "epoch": 1.3578500707213579, "percentage": 45.31, "elapsed_time": "1:04:06", "remaining_time": "1:17:23"}
+{"current_steps": 850, "total_steps": 1854, "loss": 1.4234, "accuracy": 0.518750011920929, "learning_rate": 2.825492919315559e-06, "epoch": 1.3740149525156597, "percentage": 45.85, "elapsed_time": "1:04:51", "remaining_time": "1:16:36"}
+{"current_steps": 860, "total_steps": 1854, "loss": 1.3028, "accuracy": 0.581250011920929, "learning_rate": 2.7834441532542482e-06, "epoch": 1.3901798343099616, "percentage": 46.39, "elapsed_time": "1:05:37", "remaining_time": "1:15:51"}
+{"current_steps": 870, "total_steps": 1854, "loss": 1.3586, "accuracy": 0.5, "learning_rate": 2.74131399471945e-06, "epoch": 1.4063447161042635, "percentage": 46.93, "elapsed_time": "1:06:20", "remaining_time": "1:15:02"}
+{"current_steps": 880, "total_steps": 1854, "loss": 1.3658, "accuracy": 0.53125, "learning_rate": 2.6991145416068947e-06, "epoch": 1.4225095978985653, "percentage": 47.46, "elapsed_time": "1:07:07", "remaining_time": "1:14:18"}
+{"current_steps": 890, "total_steps": 1854, "loss": 1.2749, "accuracy": 0.6000000238418579, "learning_rate": 2.6568579117106143e-06, "epoch": 1.4386744796928672, "percentage": 48.0, "elapsed_time": "1:07:52", "remaining_time": "1:13:31"}
+{"current_steps": 900, "total_steps": 1854, "loss": 1.3382, "accuracy": 0.5062500238418579, "learning_rate": 2.6145562392432544e-06, "epoch": 1.454839361487169, "percentage": 48.54, "elapsed_time": "1:08:36", "remaining_time": "1:12:42"}
+{"current_steps": 910, "total_steps": 1854, "loss": 1.284, "accuracy": 0.59375, "learning_rate": 2.5722216713516682e-06, "epoch": 1.471004243281471, "percentage": 49.08, "elapsed_time": "1:09:18", "remaining_time": "1:11:53"}
+{"current_steps": 920, "total_steps": 1854, "loss": 1.2822, "accuracy": 0.5249999761581421, "learning_rate": 2.5298663646288064e-06, "epoch": 1.4871691250757728, "percentage": 49.62, "elapsed_time": "1:10:29", "remaining_time": "1:11:33"}
+{"current_steps": 930, "total_steps": 1854, "loss": 1.3892, "accuracy": 0.5249999761581421, "learning_rate": 2.487502481622879e-06, "epoch": 1.503334006870075, "percentage": 50.16, "elapsed_time": "1:11:12", "remaining_time": "1:10:44"}
+{"current_steps": 940, "total_steps": 1854, "loss": 1.3282, "accuracy": 0.4937500059604645, "learning_rate": 2.4451421873448253e-06, "epoch": 1.5194988886643768, "percentage": 50.7, "elapsed_time": "1:11:57", "remaining_time": "1:09:57"}
+{"current_steps": 950, "total_steps": 1854, "loss": 1.4328, "accuracy": 0.5062500238418579, "learning_rate": 2.40279764577506e-06, "epoch": 1.5356637704586786, "percentage": 51.24, "elapsed_time": "1:12:42", "remaining_time": "1:09:11"}
+{"current_steps": 960, "total_steps": 1854, "loss": 1.2808, "accuracy": 0.5625, "learning_rate": 2.3604810163705242e-06, "epoch": 1.5518286522529805, "percentage": 51.78, "elapsed_time": "1:13:28", "remaining_time": "1:08:24"}
+{"current_steps": 970, "total_steps": 1854, "loss": 1.2392, "accuracy": 0.512499988079071, "learning_rate": 2.3182044505730364e-06, "epoch": 1.5679935340472824, "percentage": 52.32, "elapsed_time": "1:14:12", "remaining_time": "1:07:37"}
+{"current_steps": 980, "total_steps": 1854, "loss": 1.2758, "accuracy": 0.5375000238418579, "learning_rate": 2.275980088319941e-06, "epoch": 1.5841584158415842, "percentage": 52.86, "elapsed_time": "1:14:57", "remaining_time": "1:06:50"}
+{"current_steps": 990, "total_steps": 1854, "loss": 1.2503, "accuracy": 0.5375000238418579, "learning_rate": 2.2338200545580577e-06, "epoch": 1.600323297635886, "percentage": 53.4, "elapsed_time": "1:15:40", "remaining_time": "1:06:02"}
+{"current_steps": 1000, "total_steps": 1854, "loss": 1.2196, "accuracy": 0.5562499761581421, "learning_rate": 2.191736455761947e-06, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "1:16:23", "remaining_time": "1:05:14"}
+{"current_steps": 1000, "total_steps": 1854, "eval_loss": 1.3396929502487183, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "1:19:02", "remaining_time": "1:07:29"}