Training in progress, step 1000

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:ce648971f97a29fd8714d80f03599bddf13945c172979834e54cd0326983ef71
+oid sha256:10a19d704e4d53853df933e1353ec1b34230b2ff53b0cdbff6826ddb4012ddb3
 size 83945296

trainer_log.jsonl

@@ -49,3 +49,54 @@
 {"current_steps": 490, "total_steps": 1854, "loss": 0.8348, "accuracy": 0.550000011920929, "learning_rate": 4.186861743633911e-06, "epoch": 0.7920792079207921, "percentage": 26.43, "elapsed_time": "2:24:24", "remaining_time": "6:42:00"}
 {"current_steps": 500, "total_steps": 1854, "loss": 0.8803, "accuracy": 0.5375000238418579, "learning_rate": 4.155353766456497e-06, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "2:27:31", "remaining_time": "6:39:30"}
 {"current_steps": 500, "total_steps": 1854, "eval_loss": 0.8619003891944885, "epoch": 0.808244089715094, "percentage": 26.97, "elapsed_time": "2:33:41", "remaining_time": "6:56:10"}
+{"current_steps": 510, "total_steps": 1854, "loss": 0.8606, "accuracy": 0.550000011920929, "learning_rate": 4.123370445773134e-06, "epoch": 0.8244089715093958, "percentage": 27.51, "elapsed_time": "2:36:45", "remaining_time": "6:53:05"}
+{"current_steps": 520, "total_steps": 1854, "loss": 0.8752, "accuracy": 0.6187499761581421, "learning_rate": 4.090920965761906e-06, "epoch": 0.8405738533036977, "percentage": 28.05, "elapsed_time": "2:39:34", "remaining_time": "6:49:23"}
+{"current_steps": 530, "total_steps": 1854, "loss": 0.8179, "accuracy": 0.6000000238418579, "learning_rate": 4.058014644460991e-06, "epoch": 0.8567387350979996, "percentage": 28.59, "elapsed_time": "2:42:16", "remaining_time": "6:45:21"}
+{"current_steps": 540, "total_steps": 1854, "loss": 0.8497, "accuracy": 0.5874999761581421, "learning_rate": 4.024660931092939e-06, "epoch": 0.8729036168923014, "percentage": 29.13, "elapsed_time": "2:45:17", "remaining_time": "6:42:13"}
+{"current_steps": 550, "total_steps": 1854, "loss": 0.8507, "accuracy": 0.675000011920929, "learning_rate": 3.990869403351272e-06, "epoch": 0.8890684986866033, "percentage": 29.67, "elapsed_time": "2:48:20", "remaining_time": "6:39:08"}
+{"current_steps": 560, "total_steps": 1854, "loss": 0.864, "accuracy": 0.4625000059604645, "learning_rate": 3.956649764650206e-06, "epoch": 0.9052333804809052, "percentage": 30.2, "elapsed_time": "2:51:33", "remaining_time": "6:36:24"}
+{"current_steps": 570, "total_steps": 1854, "loss": 0.8159, "accuracy": 0.637499988079071, "learning_rate": 3.92201184133826e-06, "epoch": 0.9213982622752072, "percentage": 30.74, "elapsed_time": "2:54:25", "remaining_time": "6:32:54"}
+{"current_steps": 580, "total_steps": 1854, "loss": 0.8214, "accuracy": 0.53125, "learning_rate": 3.886965579876572e-06, "epoch": 0.937563144069509, "percentage": 31.28, "elapsed_time": "2:57:19", "remaining_time": "6:29:30"}
+{"current_steps": 590, "total_steps": 1854, "loss": 0.8603, "accuracy": 0.5625, "learning_rate": 3.851521043982716e-06, "epoch": 0.9537280258638109, "percentage": 31.82, "elapsed_time": "3:00:21", "remaining_time": "6:26:23"}
+{"current_steps": 600, "total_steps": 1854, "loss": 0.8172, "accuracy": 0.6312500238418579, "learning_rate": 3.81568841174086e-06, "epoch": 0.9698929076581128, "percentage": 32.36, "elapsed_time": "3:03:33", "remaining_time": "6:23:37"}
+{"current_steps": 610, "total_steps": 1854, "loss": 0.798, "accuracy": 0.5687500238418579, "learning_rate": 3.7794779726790664e-06, "epoch": 0.9860577894524146, "percentage": 32.9, "elapsed_time": "3:06:33", "remaining_time": "6:20:26"}
+{"current_steps": 620, "total_steps": 1854, "loss": 0.8118, "accuracy": 0.625, "learning_rate": 3.7429001248146096e-06, "epoch": 1.0022226712467166, "percentage": 33.44, "elapsed_time": "3:09:42", "remaining_time": "6:17:34"}
+{"current_steps": 630, "total_steps": 1854, "loss": 0.8821, "accuracy": 0.5562499761581421, "learning_rate": 3.7059653716681227e-06, "epoch": 1.0183875530410185, "percentage": 33.98, "elapsed_time": "3:12:37", "remaining_time": "6:14:15"}
+{"current_steps": 640, "total_steps": 1854, "loss": 0.7692, "accuracy": 0.612500011920929, "learning_rate": 3.668684319247463e-06, "epoch": 1.0345524348353203, "percentage": 34.52, "elapsed_time": "3:15:27", "remaining_time": "6:10:45"}
+{"current_steps": 650, "total_steps": 1854, "loss": 0.8402, "accuracy": 0.5687500238418579, "learning_rate": 3.6310676730021373e-06, "epoch": 1.0507173166296222, "percentage": 35.06, "elapsed_time": "3:18:21", "remaining_time": "6:07:24"}
+{"current_steps": 660, "total_steps": 1854, "loss": 0.8843, "accuracy": 0.5562499761581421, "learning_rate": 3.593126234749178e-06, "epoch": 1.066882198423924, "percentage": 35.6, "elapsed_time": "3:21:23", "remaining_time": "6:04:20"}
+{"current_steps": 670, "total_steps": 1854, "loss": 0.8313, "accuracy": 0.6187499761581421, "learning_rate": 3.554870899571343e-06, "epoch": 1.083047080218226, "percentage": 36.14, "elapsed_time": "3:24:27", "remaining_time": "6:01:18"}
+{"current_steps": 680, "total_steps": 1854, "loss": 0.8024, "accuracy": 0.6187499761581421, "learning_rate": 3.5163126526885373e-06, "epoch": 1.0992119620125278, "percentage": 36.68, "elapsed_time": "3:27:31", "remaining_time": "5:58:17"}
+{"current_steps": 690, "total_steps": 1854, "loss": 0.8092, "accuracy": 0.5687500238418579, "learning_rate": 3.4774625663033484e-06, "epoch": 1.1153768438068297, "percentage": 37.22, "elapsed_time": "3:30:28", "remaining_time": "5:55:03"}
+{"current_steps": 700, "total_steps": 1854, "loss": 0.7881, "accuracy": 0.5625, "learning_rate": 3.4383317964216067e-06, "epoch": 1.1315417256011315, "percentage": 37.76, "elapsed_time": "3:33:29", "remaining_time": "5:51:56"}
+{"current_steps": 710, "total_steps": 1854, "loss": 0.8534, "accuracy": 0.6187499761581421, "learning_rate": 3.398931579648877e-06, "epoch": 1.1477066073954334, "percentage": 38.3, "elapsed_time": "3:36:37", "remaining_time": "5:49:02"}
+{"current_steps": 720, "total_steps": 1854, "loss": 0.7957, "accuracy": 0.5375000238418579, "learning_rate": 3.359273229963813e-06, "epoch": 1.1638714891897353, "percentage": 38.83, "elapsed_time": "3:39:40", "remaining_time": "5:45:58"}
+{"current_steps": 730, "total_steps": 1854, "loss": 0.837, "accuracy": 0.574999988079071, "learning_rate": 3.319368135469285e-06, "epoch": 1.1800363709840371, "percentage": 39.37, "elapsed_time": "3:42:31", "remaining_time": "5:42:38"}
+{"current_steps": 740, "total_steps": 1854, "loss": 0.778, "accuracy": 0.668749988079071, "learning_rate": 3.279227755122228e-06, "epoch": 1.196201252778339, "percentage": 39.91, "elapsed_time": "3:45:35", "remaining_time": "5:39:36"}
+{"current_steps": 750, "total_steps": 1854, "loss": 0.846, "accuracy": 0.581250011920929, "learning_rate": 3.2388636154431417e-06, "epoch": 1.2123661345726409, "percentage": 40.45, "elapsed_time": "3:48:42", "remaining_time": "5:36:38"}
+{"current_steps": 760, "total_steps": 1854, "loss": 0.7952, "accuracy": 0.5874999761581421, "learning_rate": 3.198287307206192e-06, "epoch": 1.2285310163669427, "percentage": 40.99, "elapsed_time": "3:51:39", "remaining_time": "5:33:28"}
+{"current_steps": 770, "total_steps": 1854, "loss": 0.8251, "accuracy": 0.5625, "learning_rate": 3.157510482110856e-06, "epoch": 1.2446958981612446, "percentage": 41.53, "elapsed_time": "3:54:42", "remaining_time": "5:30:25"}
+{"current_steps": 780, "total_steps": 1854, "loss": 0.8661, "accuracy": 0.5625, "learning_rate": 3.116544849436077e-06, "epoch": 1.2608607799555465, "percentage": 42.07, "elapsed_time": "3:57:46", "remaining_time": "5:27:23"}
+{"current_steps": 790, "total_steps": 1854, "loss": 0.7785, "accuracy": 0.6625000238418579, "learning_rate": 3.0754021726778848e-06, "epoch": 1.2770256617498483, "percentage": 42.61, "elapsed_time": "4:00:46", "remaining_time": "5:24:17"}
+{"current_steps": 800, "total_steps": 1854, "loss": 0.8595, "accuracy": 0.5375000238418579, "learning_rate": 3.0340942661714463e-06, "epoch": 1.2931905435441502, "percentage": 43.15, "elapsed_time": "4:03:52", "remaining_time": "5:21:18"}
+{"current_steps": 810, "total_steps": 1854, "loss": 0.8377, "accuracy": 0.625, "learning_rate": 2.992632991698512e-06, "epoch": 1.3093554253384523, "percentage": 43.69, "elapsed_time": "4:06:48", "remaining_time": "5:18:06"}
+{"current_steps": 820, "total_steps": 1854, "loss": 0.7382, "accuracy": 0.581250011920929, "learning_rate": 2.9510302550812537e-06, "epoch": 1.3255203071327541, "percentage": 44.23, "elapsed_time": "4:09:43", "remaining_time": "5:14:54"}
+{"current_steps": 830, "total_steps": 1854, "loss": 0.773, "accuracy": 0.581250011920929, "learning_rate": 2.9092980027634325e-06, "epoch": 1.341685188927056, "percentage": 44.77, "elapsed_time": "4:12:46", "remaining_time": "5:11:50"}
+{"current_steps": 840, "total_steps": 1854, "loss": 0.8783, "accuracy": 0.574999988079071, "learning_rate": 2.867448218379927e-06, "epoch": 1.3578500707213579, "percentage": 45.31, "elapsed_time": "4:15:41", "remaining_time": "5:08:38"}
+{"current_steps": 850, "total_steps": 1854, "loss": 0.9094, "accuracy": 0.5874999761581421, "learning_rate": 2.825492919315559e-06, "epoch": 1.3740149525156597, "percentage": 45.85, "elapsed_time": "4:18:40", "remaining_time": "5:05:31"}
+{"current_steps": 860, "total_steps": 1854, "loss": 0.8093, "accuracy": 0.59375, "learning_rate": 2.7834441532542482e-06, "epoch": 1.3901798343099616, "percentage": 46.39, "elapsed_time": "4:21:43", "remaining_time": "5:02:30"}
+{"current_steps": 870, "total_steps": 1854, "loss": 0.8335, "accuracy": 0.574999988079071, "learning_rate": 2.74131399471945e-06, "epoch": 1.4063447161042635, "percentage": 46.93, "elapsed_time": "4:24:41", "remaining_time": "4:59:22"}
+{"current_steps": 880, "total_steps": 1854, "loss": 0.8403, "accuracy": 0.5874999761581421, "learning_rate": 2.6991145416068947e-06, "epoch": 1.4225095978985653, "percentage": 47.46, "elapsed_time": "4:27:50", "remaining_time": "4:56:26"}
+{"current_steps": 890, "total_steps": 1854, "loss": 0.7832, "accuracy": 0.637499988079071, "learning_rate": 2.6568579117106143e-06, "epoch": 1.4386744796928672, "percentage": 48.0, "elapsed_time": "4:30:52", "remaining_time": "4:53:24"}
+{"current_steps": 900, "total_steps": 1854, "loss": 0.8259, "accuracy": 0.550000011920929, "learning_rate": 2.6145562392432544e-06, "epoch": 1.454839361487169, "percentage": 48.54, "elapsed_time": "4:33:49", "remaining_time": "4:50:15"}
+{"current_steps": 910, "total_steps": 1854, "loss": 0.7723, "accuracy": 0.59375, "learning_rate": 2.5722216713516682e-06, "epoch": 1.471004243281471, "percentage": 49.08, "elapsed_time": "4:36:41", "remaining_time": "4:47:02"}
+{"current_steps": 920, "total_steps": 1854, "loss": 0.8097, "accuracy": 0.543749988079071, "learning_rate": 2.5298663646288064e-06, "epoch": 1.4871691250757728, "percentage": 49.62, "elapsed_time": "4:39:45", "remaining_time": "4:44:01"}
+{"current_steps": 930, "total_steps": 1854, "loss": 0.8507, "accuracy": 0.65625, "learning_rate": 2.487502481622879e-06, "epoch": 1.503334006870075, "percentage": 50.16, "elapsed_time": "4:42:35", "remaining_time": "4:40:46"}
+{"current_steps": 940, "total_steps": 1854, "loss": 0.8309, "accuracy": 0.550000011920929, "learning_rate": 2.4451421873448253e-06, "epoch": 1.5194988886643768, "percentage": 50.7, "elapsed_time": "4:45:34", "remaining_time": "4:37:40"}
+{"current_steps": 950, "total_steps": 1854, "loss": 0.8488, "accuracy": 0.5687500238418579, "learning_rate": 2.40279764577506e-06, "epoch": 1.5356637704586786, "percentage": 51.24, "elapsed_time": "4:48:35", "remaining_time": "4:34:36"}
+{"current_steps": 960, "total_steps": 1854, "loss": 0.756, "accuracy": 0.6625000238418579, "learning_rate": 2.3604810163705242e-06, "epoch": 1.5518286522529805, "percentage": 51.78, "elapsed_time": "4:51:36", "remaining_time": "4:31:33"}
+{"current_steps": 970, "total_steps": 1854, "loss": 0.7224, "accuracy": 0.59375, "learning_rate": 2.3182044505730364e-06, "epoch": 1.5679935340472824, "percentage": 52.32, "elapsed_time": "4:54:31", "remaining_time": "4:28:24"}
+{"current_steps": 980, "total_steps": 1854, "loss": 0.7869, "accuracy": 0.574999988079071, "learning_rate": 2.275980088319941e-06, "epoch": 1.5841584158415842, "percentage": 52.86, "elapsed_time": "4:57:31", "remaining_time": "4:25:21"}
+{"current_steps": 990, "total_steps": 1854, "loss": 0.7727, "accuracy": 0.543749988079071, "learning_rate": 2.2338200545580577e-06, "epoch": 1.600323297635886, "percentage": 53.4, "elapsed_time": "5:00:24", "remaining_time": "4:22:10"}
+{"current_steps": 1000, "total_steps": 1854, "loss": 0.7341, "accuracy": 0.6312500238418579, "learning_rate": 2.191736455761947e-06, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "5:03:14", "remaining_time": "4:18:57"}
+{"current_steps": 1000, "total_steps": 1854, "eval_loss": 0.8449718356132507, "epoch": 1.616488179430188, "percentage": 53.94, "elapsed_time": "5:09:21", "remaining_time": "4:24:11"}