Training in progress, step 1000

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:96f29db797cf0416b61c8eaa483b098ffeab002b0817273e3e6040d753a2bd36
+oid sha256:9580ae263eb142fd985c53300cc6d17b67c0a0b350d846e9a141405514d598d7
 size 65309632

trainer_log.jsonl

@@ -49,3 +49,54 @@
 {"current_steps": 490, "total_steps": 1686, "loss": 1.5343, "accuracy": 0.512499988079071, "learning_rate": 4.028624486874608e-06, "epoch": 0.871304734385419, "percentage": 29.06, "elapsed_time": "1:20:19", "remaining_time": "3:16:03"}
 {"current_steps": 500, "total_steps": 1686, "loss": 1.6309, "accuracy": 0.4937500059604645, "learning_rate": 3.99149827895177e-06, "epoch": 0.8890864636585908, "percentage": 29.66, "elapsed_time": "1:22:01", "remaining_time": "3:14:33"}
 {"current_steps": 500, "total_steps": 1686, "eval_loss": 1.5815595388412476, "epoch": 0.8890864636585908, "percentage": 29.66, "elapsed_time": "1:25:25", "remaining_time": "3:22:36"}
+{"current_steps": 510, "total_steps": 1686, "loss": 1.6424, "accuracy": 0.550000011920929, "learning_rate": 3.953854170549114e-06, "epoch": 0.9068681929317626, "percentage": 30.25, "elapsed_time": "1:27:07", "remaining_time": "3:20:53"}
+{"current_steps": 520, "total_steps": 1686, "loss": 1.4923, "accuracy": 0.5249999761581421, "learning_rate": 3.91570523302051e-06, "epoch": 0.9246499222049345, "percentage": 30.84, "elapsed_time": "1:28:41", "remaining_time": "3:18:53"}
+{"current_steps": 530, "total_steps": 1686, "loss": 1.5466, "accuracy": 0.4625000059604645, "learning_rate": 3.8770647130141996e-06, "epoch": 0.9424316514781063, "percentage": 31.44, "elapsed_time": "1:30:23", "remaining_time": "3:17:08"}
+{"current_steps": 540, "total_steps": 1686, "loss": 1.661, "accuracy": 0.53125, "learning_rate": 3.837946027873086e-06, "epoch": 0.960213380751278, "percentage": 32.03, "elapsed_time": "1:32:00", "remaining_time": "3:15:14"}
+{"current_steps": 550, "total_steps": 1686, "loss": 1.6249, "accuracy": 0.4312500059604645, "learning_rate": 3.7983627609757713e-06, "epoch": 0.9779951100244498, "percentage": 32.62, "elapsed_time": "1:33:41", "remaining_time": "3:13:31"}
+{"current_steps": 560, "total_steps": 1686, "loss": 1.5469, "accuracy": 0.5062500238418579, "learning_rate": 3.758328657019924e-06, "epoch": 0.9957768392976217, "percentage": 33.21, "elapsed_time": "1:35:22", "remaining_time": "3:11:45"}
+{"current_steps": 570, "total_steps": 1686, "loss": 1.6007, "accuracy": 0.48124998807907104, "learning_rate": 3.717857617249642e-06, "epoch": 1.0135585685707935, "percentage": 33.81, "elapsed_time": "1:37:01", "remaining_time": "3:09:57"}
+{"current_steps": 580, "total_steps": 1686, "loss": 1.4702, "accuracy": 0.48750001192092896, "learning_rate": 3.6769636946284543e-06, "epoch": 1.0313402978439654, "percentage": 34.4, "elapsed_time": "1:38:38", "remaining_time": "3:08:05"}
+{"current_steps": 590, "total_steps": 1686, "loss": 1.6438, "accuracy": 0.4625000059604645, "learning_rate": 3.6356610889596355e-06, "epoch": 1.049122027117137, "percentage": 34.99, "elapsed_time": "1:40:19", "remaining_time": "3:06:22"}
+{"current_steps": 600, "total_steps": 1686, "loss": 1.5708, "accuracy": 0.5, "learning_rate": 3.593964141955541e-06, "epoch": 1.066903756390309, "percentage": 35.59, "elapsed_time": "1:41:53", "remaining_time": "3:04:25"}
+{"current_steps": 610, "total_steps": 1686, "loss": 1.6005, "accuracy": 0.5062500238418579, "learning_rate": 3.5518873322576573e-06, "epoch": 1.0846854856634809, "percentage": 36.18, "elapsed_time": "1:43:30", "remaining_time": "3:02:34"}
+{"current_steps": 620, "total_steps": 1686, "loss": 1.5353, "accuracy": 0.4749999940395355, "learning_rate": 3.5094452704091143e-06, "epoch": 1.1024672149366526, "percentage": 36.77, "elapsed_time": "1:45:12", "remaining_time": "3:00:54"}
+{"current_steps": 630, "total_steps": 1686, "loss": 1.6576, "accuracy": 0.4312500059604645, "learning_rate": 3.46665269378139e-06, "epoch": 1.1202489442098245, "percentage": 37.37, "elapsed_time": "1:46:54", "remaining_time": "2:59:12"}
+{"current_steps": 640, "total_steps": 1686, "loss": 1.6419, "accuracy": 0.512499988079071, "learning_rate": 3.4235244614569794e-06, "epoch": 1.1380306734829961, "percentage": 37.96, "elapsed_time": "1:49:09", "remaining_time": "2:58:23"}
+{"current_steps": 650, "total_steps": 1686, "loss": 1.4832, "accuracy": 0.550000011920929, "learning_rate": 3.3800755490698008e-06, "epoch": 1.155812402756168, "percentage": 38.55, "elapsed_time": "1:50:43", "remaining_time": "2:56:29"}
+{"current_steps": 660, "total_steps": 1686, "loss": 1.6292, "accuracy": 0.5062500238418579, "learning_rate": 3.3363210436051287e-06, "epoch": 1.17359413202934, "percentage": 39.15, "elapsed_time": "1:52:59", "remaining_time": "2:55:39"}
+{"current_steps": 670, "total_steps": 1686, "loss": 1.5589, "accuracy": 0.41874998807907104, "learning_rate": 3.292276138160867e-06, "epoch": 1.1913758613025116, "percentage": 39.74, "elapsed_time": "1:54:35", "remaining_time": "2:53:46"}
+{"current_steps": 680, "total_steps": 1686, "loss": 1.6131, "accuracy": 0.4937500059604645, "learning_rate": 3.2479561266719694e-06, "epoch": 1.2091575905756835, "percentage": 40.33, "elapsed_time": "1:56:14", "remaining_time": "2:51:57"}
+{"current_steps": 690, "total_steps": 1686, "loss": 1.5038, "accuracy": 0.543749988079071, "learning_rate": 3.2033763985998533e-06, "epoch": 1.2269393198488552, "percentage": 40.93, "elapsed_time": "1:57:52", "remaining_time": "2:50:08"}
+{"current_steps": 700, "total_steps": 1686, "loss": 1.4937, "accuracy": 0.512499988079071, "learning_rate": 3.1585524335886335e-06, "epoch": 1.244721049122027, "percentage": 41.52, "elapsed_time": "1:59:28", "remaining_time": "2:48:17"}
+{"current_steps": 710, "total_steps": 1686, "loss": 1.4906, "accuracy": 0.512499988079071, "learning_rate": 3.1134997960900536e-06, "epoch": 1.262502778395199, "percentage": 42.11, "elapsed_time": "2:01:03", "remaining_time": "2:46:24"}
+{"current_steps": 720, "total_steps": 1686, "loss": 1.5498, "accuracy": 0.4000000059604645, "learning_rate": 3.0682341299589583e-06, "epoch": 1.2802845076683709, "percentage": 42.7, "elapsed_time": "2:02:46", "remaining_time": "2:44:43"}
+{"current_steps": 730, "total_steps": 1686, "loss": 1.5509, "accuracy": 0.543749988079071, "learning_rate": 3.022771153021201e-06, "epoch": 1.2980662369415426, "percentage": 43.3, "elapsed_time": "2:04:23", "remaining_time": "2:42:53"}
+{"current_steps": 740, "total_steps": 1686, "loss": 1.5376, "accuracy": 0.48750001192092896, "learning_rate": 2.9771266516158625e-06, "epoch": 1.3158479662147144, "percentage": 43.89, "elapsed_time": "2:06:00", "remaining_time": "2:41:04"}
+{"current_steps": 750, "total_steps": 1686, "loss": 1.5505, "accuracy": 0.4937500059604645, "learning_rate": 2.9313164751136802e-06, "epoch": 1.3336296954878861, "percentage": 44.48, "elapsed_time": "2:07:36", "remaining_time": "2:39:15"}
+{"current_steps": 760, "total_steps": 1686, "loss": 1.5409, "accuracy": 0.4375, "learning_rate": 2.8853565304135956e-06, "epoch": 1.351411424761058, "percentage": 45.08, "elapsed_time": "2:09:12", "remaining_time": "2:37:26"}
+{"current_steps": 770, "total_steps": 1686, "loss": 1.5416, "accuracy": 0.5249999761581421, "learning_rate": 2.839262776419313e-06, "epoch": 1.36919315403423, "percentage": 45.67, "elapsed_time": "2:11:12", "remaining_time": "2:36:04"}
+{"current_steps": 780, "total_steps": 1686, "loss": 1.5454, "accuracy": 0.42500001192092896, "learning_rate": 2.793051218497817e-06, "epoch": 1.3869748833074016, "percentage": 46.26, "elapsed_time": "2:12:51", "remaining_time": "2:34:19"}
+{"current_steps": 790, "total_steps": 1686, "loss": 1.517, "accuracy": 0.53125, "learning_rate": 2.7467379029217437e-06, "epoch": 1.4047566125805735, "percentage": 46.86, "elapsed_time": "2:14:31", "remaining_time": "2:32:34"}
+{"current_steps": 800, "total_steps": 1686, "loss": 1.6221, "accuracy": 0.45625001192092896, "learning_rate": 2.7003389112975546e-06, "epoch": 1.4225383418537452, "percentage": 47.45, "elapsed_time": "2:16:34", "remaining_time": "2:31:15"}
+{"current_steps": 810, "total_steps": 1686, "loss": 1.4137, "accuracy": 0.5062500238418579, "learning_rate": 2.653870354981437e-06, "epoch": 1.440320071126917, "percentage": 48.04, "elapsed_time": "2:18:08", "remaining_time": "2:29:23"}
+{"current_steps": 820, "total_steps": 1686, "loss": 1.5337, "accuracy": 0.518750011920929, "learning_rate": 2.6073483694848777e-06, "epoch": 1.458101800400089, "percentage": 48.64, "elapsed_time": "2:19:51", "remaining_time": "2:27:42"}
+{"current_steps": 830, "total_steps": 1686, "loss": 1.5064, "accuracy": 0.4937500059604645, "learning_rate": 2.560789108871847e-06, "epoch": 1.4758835296732609, "percentage": 49.23, "elapsed_time": "2:21:43", "remaining_time": "2:26:09"}
+{"current_steps": 840, "total_steps": 1686, "loss": 1.5523, "accuracy": 0.5062500238418579, "learning_rate": 2.514208740149544e-06, "epoch": 1.4936652589464325, "percentage": 49.82, "elapsed_time": "2:23:24", "remaining_time": "2:24:26"}
+{"current_steps": 850, "total_steps": 1686, "loss": 1.5463, "accuracy": 0.53125, "learning_rate": 2.46762343765464e-06, "epoch": 1.5114469882196042, "percentage": 50.42, "elapsed_time": "2:25:11", "remaining_time": "2:22:48"}
+{"current_steps": 860, "total_steps": 1686, "loss": 1.6351, "accuracy": 0.4625000059604645, "learning_rate": 2.4210493774369903e-06, "epoch": 1.5292287174927761, "percentage": 51.01, "elapsed_time": "2:26:49", "remaining_time": "2:21:01"}
+{"current_steps": 870, "total_steps": 1686, "loss": 1.5247, "accuracy": 0.5, "learning_rate": 2.374502731642732e-06, "epoch": 1.547010446765948, "percentage": 51.6, "elapsed_time": "2:28:27", "remaining_time": "2:19:14"}
+{"current_steps": 880, "total_steps": 1686, "loss": 1.5356, "accuracy": 0.5249999761581421, "learning_rate": 2.3279996628987556e-06, "epoch": 1.56479217603912, "percentage": 52.19, "elapsed_time": "2:30:00", "remaining_time": "2:17:23"}
+{"current_steps": 890, "total_steps": 1686, "loss": 1.5144, "accuracy": 0.4375, "learning_rate": 2.281556318700474e-06, "epoch": 1.5825739053122916, "percentage": 52.79, "elapsed_time": "2:32:00", "remaining_time": "2:15:57"}
+{"current_steps": 900, "total_steps": 1686, "loss": 1.5647, "accuracy": 0.518750011920929, "learning_rate": 2.2351888258048408e-06, "epoch": 1.6003556345854635, "percentage": 53.38, "elapsed_time": "2:35:01", "remaining_time": "2:15:23"}
+{"current_steps": 910, "total_steps": 1686, "loss": 1.6264, "accuracy": 0.4749999940395355, "learning_rate": 2.188913284630584e-06, "epoch": 1.6181373638586352, "percentage": 53.97, "elapsed_time": "2:38:10", "remaining_time": "2:14:53"}
+{"current_steps": 920, "total_steps": 1686, "loss": 1.609, "accuracy": 0.45625001192092896, "learning_rate": 2.1427457636675652e-06, "epoch": 1.635919093131807, "percentage": 54.57, "elapsed_time": "2:41:07", "remaining_time": "2:14:09"}
+{"current_steps": 930, "total_steps": 1686, "loss": 1.4753, "accuracy": 0.5375000238418579, "learning_rate": 2.096702293897247e-06, "epoch": 1.653700822404979, "percentage": 55.16, "elapsed_time": "2:44:11", "remaining_time": "2:13:28"}
+{"current_steps": 940, "total_steps": 1686, "loss": 1.5523, "accuracy": 0.5, "learning_rate": 2.0507988632261672e-06, "epoch": 1.6714825516781509, "percentage": 55.75, "elapsed_time": "2:47:30", "remaining_time": "2:12:56"}
+{"current_steps": 950, "total_steps": 1686, "loss": 1.6259, "accuracy": 0.48750001192092896, "learning_rate": 2.005051410934382e-06, "epoch": 1.6892642809513225, "percentage": 56.35, "elapsed_time": "2:50:35", "remaining_time": "2:12:09"}
+{"current_steps": 960, "total_steps": 1686, "loss": 1.453, "accuracy": 0.5687500238418579, "learning_rate": 1.9594758221407843e-06, "epoch": 1.7070460102244942, "percentage": 56.94, "elapsed_time": "2:53:43", "remaining_time": "2:11:22"}
+{"current_steps": 970, "total_steps": 1686, "loss": 1.4772, "accuracy": 0.48750001192092896, "learning_rate": 1.9140879222872408e-06, "epoch": 1.724827739497666, "percentage": 57.53, "elapsed_time": "2:56:53", "remaining_time": "2:10:34"}
+{"current_steps": 980, "total_steps": 1686, "loss": 1.5681, "accuracy": 0.45625001192092896, "learning_rate": 1.8689034716434346e-06, "epoch": 1.742609468770838, "percentage": 58.13, "elapsed_time": "2:59:55", "remaining_time": "2:09:37"}
+{"current_steps": 990, "total_steps": 1686, "loss": 1.4872, "accuracy": 0.44999998807907104, "learning_rate": 1.8239381598343576e-06, "epoch": 1.76039119804401, "percentage": 58.72, "elapsed_time": "3:02:59", "remaining_time": "2:08:39"}
+{"current_steps": 1000, "total_steps": 1686, "loss": 1.5401, "accuracy": 0.4749999940395355, "learning_rate": 1.779207600392312e-06, "epoch": 1.7781729273171816, "percentage": 59.31, "elapsed_time": "3:06:18", "remaining_time": "2:07:48"}
+{"current_steps": 1000, "total_steps": 1686, "eval_loss": 1.5269325971603394, "epoch": 1.7781729273171816, "percentage": 59.31, "elapsed_time": "3:13:06", "remaining_time": "2:12:28"}