Training in progress, step 1500

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48022570d812624f802c97d6342ea39098e5a3a123ff0bc6282c47d222a5b25a
+oid sha256:528189bd7166dd50eb58370cae6ca5ea4a9f5079366ab43c1e909f8a05f15e16
 size 65309632

trainer_log.jsonl

@@ -100,3 +100,54 @@
 {"current_steps": 990, "total_steps": 1770, "loss": 1.3763, "accuracy": 0.5625, "learning_rate": 2.0369703571452387e-06, "epoch": 1.676545300592718, "percentage": 55.93, "elapsed_time": "3:10:33", "remaining_time": "2:30:07"}
 {"current_steps": 1000, "total_steps": 1770, "loss": 1.4213, "accuracy": 0.4749999940395355, "learning_rate": 1.993438056859441e-06, "epoch": 1.6934801016088061, "percentage": 56.5, "elapsed_time": "3:13:51", "remaining_time": "2:29:16"}
 {"current_steps": 1000, "total_steps": 1770, "eval_loss": 1.509663701057434, "epoch": 1.6934801016088061, "percentage": 56.5, "elapsed_time": "3:21:17", "remaining_time": "2:34:59"}
+{"current_steps": 1010, "total_steps": 1770, "loss": 1.4302, "accuracy": 0.518750011920929, "learning_rate": 1.9500653531031917e-06, "epoch": 1.710414902624894, "percentage": 57.06, "elapsed_time": "3:24:41", "remaining_time": "2:34:01"}
+{"current_steps": 1020, "total_steps": 1770, "loss": 1.5008, "accuracy": 0.5249999761581421, "learning_rate": 1.9068659108055117e-06, "epoch": 1.7273497036409822, "percentage": 57.63, "elapsed_time": "3:28:20", "remaining_time": "2:33:11"}
+{"current_steps": 1030, "total_steps": 1770, "loss": 1.3634, "accuracy": 0.5687500238418579, "learning_rate": 1.863853340307962e-06, "epoch": 1.7442845046570703, "percentage": 58.19, "elapsed_time": "3:31:39", "remaining_time": "2:32:03"}
+{"current_steps": 1040, "total_steps": 1770, "loss": 1.547, "accuracy": 0.5625, "learning_rate": 1.8210411930766019e-06, "epoch": 1.7612193056731584, "percentage": 58.76, "elapsed_time": "3:35:03", "remaining_time": "2:30:56"}
+{"current_steps": 1050, "total_steps": 1770, "loss": 1.4567, "accuracy": 0.5562499761581421, "learning_rate": 1.7784429574324803e-06, "epoch": 1.7781541066892466, "percentage": 59.32, "elapsed_time": "3:38:25", "remaining_time": "2:29:46"}
+{"current_steps": 1060, "total_steps": 1770, "loss": 1.3747, "accuracy": 0.5062500238418579, "learning_rate": 1.7360720543020327e-06, "epoch": 1.7950889077053345, "percentage": 59.89, "elapsed_time": "3:41:45", "remaining_time": "2:28:32"}
+{"current_steps": 1070, "total_steps": 1770, "loss": 1.5049, "accuracy": 0.512499988079071, "learning_rate": 1.6939418329887042e-06, "epoch": 1.8120237087214224, "percentage": 60.45, "elapsed_time": "3:45:07", "remaining_time": "2:27:16"}
+{"current_steps": 1080, "total_steps": 1770, "loss": 1.519, "accuracy": 0.44999998807907104, "learning_rate": 1.6520655669671467e-06, "epoch": 1.8289585097375105, "percentage": 61.02, "elapsed_time": "3:48:21", "remaining_time": "2:25:53"}
+{"current_steps": 1090, "total_steps": 1770, "loss": 1.5548, "accuracy": 0.5062500238418579, "learning_rate": 1.610456449701294e-06, "epoch": 1.8458933107535986, "percentage": 61.58, "elapsed_time": "3:51:34", "remaining_time": "2:24:28"}
+{"current_steps": 1100, "total_steps": 1770, "loss": 1.474, "accuracy": 0.550000011920929, "learning_rate": 1.5691275904876545e-06, "epoch": 1.8628281117696868, "percentage": 62.15, "elapsed_time": "3:54:46", "remaining_time": "2:23:00"}
+{"current_steps": 1110, "total_steps": 1770, "loss": 1.3829, "accuracy": 0.512499988079071, "learning_rate": 1.5280920103251235e-06, "epoch": 1.879762912785775, "percentage": 62.71, "elapsed_time": "3:58:03", "remaining_time": "2:21:32"}
+{"current_steps": 1120, "total_steps": 1770, "loss": 1.3984, "accuracy": 0.543749988079071, "learning_rate": 1.4873626378126015e-06, "epoch": 1.8966977138018628, "percentage": 63.28, "elapsed_time": "4:01:15", "remaining_time": "2:20:01"}
+{"current_steps": 1130, "total_steps": 1770, "loss": 1.4306, "accuracy": 0.48124998807907104, "learning_rate": 1.446952305075738e-06, "epoch": 1.913632514817951, "percentage": 63.84, "elapsed_time": "4:04:26", "remaining_time": "2:18:26"}
+{"current_steps": 1140, "total_steps": 1770, "loss": 1.5314, "accuracy": 0.5625, "learning_rate": 1.406873743724065e-06, "epoch": 1.9305673158340388, "percentage": 64.41, "elapsed_time": "4:07:50", "remaining_time": "2:16:57"}
+{"current_steps": 1150, "total_steps": 1770, "loss": 1.4094, "accuracy": 0.4749999940395355, "learning_rate": 1.3671395808397898e-06, "epoch": 1.947502116850127, "percentage": 64.97, "elapsed_time": "4:11:19", "remaining_time": "2:15:29"}
+{"current_steps": 1160, "total_steps": 1770, "loss": 1.4653, "accuracy": 0.48124998807907104, "learning_rate": 1.3277623349995418e-06, "epoch": 1.964436917866215, "percentage": 65.54, "elapsed_time": "4:14:21", "remaining_time": "2:13:45"}
+{"current_steps": 1170, "total_steps": 1770, "loss": 1.4772, "accuracy": 0.4437499940395355, "learning_rate": 1.2887544123302781e-06, "epoch": 1.9813717188823032, "percentage": 66.1, "elapsed_time": "4:17:28", "remaining_time": "2:12:02"}
+{"current_steps": 1180, "total_steps": 1770, "loss": 1.5002, "accuracy": 0.44999998807907104, "learning_rate": 1.2501281026006393e-06, "epoch": 1.9983065198983911, "percentage": 66.67, "elapsed_time": "4:20:31", "remaining_time": "2:10:15"}
+{"current_steps": 1190, "total_steps": 1770, "loss": 1.4511, "accuracy": 0.4749999940395355, "learning_rate": 1.2118955753489523e-06, "epoch": 2.015241320914479, "percentage": 67.23, "elapsed_time": "4:23:54", "remaining_time": "2:08:37"}
+{"current_steps": 1200, "total_steps": 1770, "loss": 1.4435, "accuracy": 0.518750011920929, "learning_rate": 1.1740688760491189e-06, "epoch": 2.032176121930567, "percentage": 67.8, "elapsed_time": "4:27:11", "remaining_time": "2:06:54"}
+{"current_steps": 1210, "total_steps": 1770, "loss": 1.4461, "accuracy": 0.5062500238418579, "learning_rate": 1.1366599223155847e-06, "epoch": 2.0491109229466553, "percentage": 68.36, "elapsed_time": "4:30:24", "remaining_time": "2:05:08"}
+{"current_steps": 1220, "total_steps": 1770, "loss": 1.4055, "accuracy": 0.581250011920929, "learning_rate": 1.0996805001486067e-06, "epoch": 2.0660457239627434, "percentage": 68.93, "elapsed_time": "4:33:39", "remaining_time": "2:03:22"}
+{"current_steps": 1230, "total_steps": 1770, "loss": 1.6016, "accuracy": 0.4437499940395355, "learning_rate": 1.0631422602209608e-06, "epoch": 2.0829805249788316, "percentage": 69.49, "elapsed_time": "4:37:03", "remaining_time": "2:01:38"}
+{"current_steps": 1240, "total_steps": 1770, "loss": 1.5034, "accuracy": 0.4749999940395355, "learning_rate": 1.027056714207319e-06, "epoch": 2.0999153259949197, "percentage": 70.06, "elapsed_time": "4:40:35", "remaining_time": "1:59:55"}
+{"current_steps": 1250, "total_steps": 1770, "loss": 1.3887, "accuracy": 0.518750011920929, "learning_rate": 9.914352311573838e-07, "epoch": 2.116850127011008, "percentage": 70.62, "elapsed_time": "4:43:48", "remaining_time": "1:58:03"}
+{"current_steps": 1260, "total_steps": 1770, "loss": 1.4097, "accuracy": 0.48750001192092896, "learning_rate": 9.562890339139877e-07, "epoch": 2.1337849280270955, "percentage": 71.19, "elapsed_time": "4:47:00", "remaining_time": "1:56:10"}
+{"current_steps": 1270, "total_steps": 1770, "loss": 1.4659, "accuracy": 0.5062500238418579, "learning_rate": 9.216291955772374e-07, "epoch": 2.1507197290431836, "percentage": 71.75, "elapsed_time": "4:50:09", "remaining_time": "1:54:14"}
+{"current_steps": 1280, "total_steps": 1770, "loss": 1.4341, "accuracy": 0.4625000059604645, "learning_rate": 8.874666360158457e-07, "epoch": 2.167654530059272, "percentage": 72.32, "elapsed_time": "4:53:33", "remaining_time": "1:52:22"}
+{"current_steps": 1290, "total_steps": 1770, "loss": 1.3574, "accuracy": 0.4937500059604645, "learning_rate": 8.538121184267315e-07, "epoch": 2.18458933107536, "percentage": 72.88, "elapsed_time": "4:56:43", "remaining_time": "1:50:24"}
+{"current_steps": 1300, "total_steps": 1770, "loss": 1.4865, "accuracy": 0.45625001192092896, "learning_rate": 8.206762459439907e-07, "epoch": 2.201524132091448, "percentage": 73.45, "elapsed_time": "5:00:00", "remaining_time": "1:48:27"}
+{"current_steps": 1310, "total_steps": 1770, "loss": 1.506, "accuracy": 0.48750001192092896, "learning_rate": 7.880694582982898e-07, "epoch": 2.218458933107536, "percentage": 74.01, "elapsed_time": "5:03:19", "remaining_time": "1:46:30"}
+{"current_steps": 1320, "total_steps": 1770, "loss": 1.4741, "accuracy": 0.4437499940395355, "learning_rate": 7.560020285277401e-07, "epoch": 2.235393734123624, "percentage": 74.58, "elapsed_time": "5:06:28", "remaining_time": "1:44:28"}
+{"current_steps": 1330, "total_steps": 1770, "loss": 1.5981, "accuracy": 0.45625001192092896, "learning_rate": 7.244840597412956e-07, "epoch": 2.252328535139712, "percentage": 75.14, "elapsed_time": "5:09:46", "remaining_time": "1:42:28"}
+{"current_steps": 1340, "total_steps": 1770, "loss": 1.4773, "accuracy": 0.48750001192092896, "learning_rate": 6.935254819356796e-07, "epoch": 2.2692633361558, "percentage": 75.71, "elapsed_time": "5:12:59", "remaining_time": "1:40:26"}
+{"current_steps": 1350, "total_steps": 1770, "loss": 1.3662, "accuracy": 0.512499988079071, "learning_rate": 6.631360488668662e-07, "epoch": 2.2861981371718882, "percentage": 76.27, "elapsed_time": "5:16:13", "remaining_time": "1:38:22"}
+{"current_steps": 1360, "total_steps": 1770, "loss": 1.5184, "accuracy": 0.48750001192092896, "learning_rate": 6.333253349770672e-07, "epoch": 2.3031329381879764, "percentage": 76.84, "elapsed_time": "5:19:28", "remaining_time": "1:36:18"}
+{"current_steps": 1370, "total_steps": 1770, "loss": 1.4568, "accuracy": 0.518750011920929, "learning_rate": 6.041027323782364e-07, "epoch": 2.3200677392040645, "percentage": 77.4, "elapsed_time": "5:22:51", "remaining_time": "1:34:15"}
+{"current_steps": 1380, "total_steps": 1770, "loss": 1.4726, "accuracy": 0.46875, "learning_rate": 5.754774478929969e-07, "epoch": 2.337002540220152, "percentage": 77.97, "elapsed_time": "5:26:18", "remaining_time": "1:32:13"}
+{"current_steps": 1390, "total_steps": 1770, "loss": 1.3692, "accuracy": 0.5249999761581421, "learning_rate": 5.474585001539634e-07, "epoch": 2.3539373412362403, "percentage": 78.53, "elapsed_time": "5:29:31", "remaining_time": "1:30:05"}
+{"current_steps": 1400, "total_steps": 1770, "loss": 1.4923, "accuracy": 0.518750011920929, "learning_rate": 5.200547167623424e-07, "epoch": 2.3708721422523285, "percentage": 79.1, "elapsed_time": "5:32:48", "remaining_time": "1:27:57"}
+{"current_steps": 1410, "total_steps": 1770, "loss": 1.4755, "accuracy": 0.48750001192092896, "learning_rate": 4.932747315067271e-07, "epoch": 2.3878069432684166, "percentage": 79.66, "elapsed_time": "5:36:14", "remaining_time": "1:25:51"}
+{"current_steps": 1420, "total_steps": 1770, "loss": 1.5183, "accuracy": 0.5375000238418579, "learning_rate": 4.6712698164294553e-07, "epoch": 2.4047417442845047, "percentage": 80.23, "elapsed_time": "5:39:37", "remaining_time": "1:23:42"}
+{"current_steps": 1430, "total_steps": 1770, "loss": 1.4439, "accuracy": 0.53125, "learning_rate": 4.41619705235842e-07, "epoch": 2.421676545300593, "percentage": 80.79, "elapsed_time": "5:43:03", "remaining_time": "1:21:34"}
+{"current_steps": 1440, "total_steps": 1770, "loss": 1.4478, "accuracy": 0.512499988079071, "learning_rate": 4.167609385637961e-07, "epoch": 2.438611346316681, "percentage": 81.36, "elapsed_time": "5:46:20", "remaining_time": "1:19:22"}
+{"current_steps": 1450, "total_steps": 1770, "loss": 1.4184, "accuracy": 0.4625000059604645, "learning_rate": 3.9255851358683567e-07, "epoch": 2.4555461473327687, "percentage": 81.92, "elapsed_time": "5:49:44", "remaining_time": "1:17:10"}
+{"current_steps": 1460, "total_steps": 1770, "loss": 1.44, "accuracy": 0.4937500059604645, "learning_rate": 3.690200554791082e-07, "epoch": 2.472480948348857, "percentage": 82.49, "elapsed_time": "5:52:58", "remaining_time": "1:14:56"}
+{"current_steps": 1470, "total_steps": 1770, "loss": 1.4382, "accuracy": 0.5062500238418579, "learning_rate": 3.461529802265079e-07, "epoch": 2.489415749364945, "percentage": 83.05, "elapsed_time": "5:56:18", "remaining_time": "1:12:42"}
+{"current_steps": 1480, "total_steps": 1770, "loss": 1.5066, "accuracy": 0.4749999940395355, "learning_rate": 3.2396449229020883e-07, "epoch": 2.506350550381033, "percentage": 83.62, "elapsed_time": "5:59:36", "remaining_time": "1:10:27"}
+{"current_steps": 1490, "total_steps": 1770, "loss": 1.4407, "accuracy": 0.5, "learning_rate": 3.024615823368371e-07, "epoch": 2.523285351397121, "percentage": 84.18, "elapsed_time": "6:02:32", "remaining_time": "1:08:07"}
+{"current_steps": 1500, "total_steps": 1770, "loss": 1.4234, "accuracy": 0.512499988079071, "learning_rate": 2.8165102503600716e-07, "epoch": 2.5402201524132093, "percentage": 84.75, "elapsed_time": "6:05:53", "remaining_time": "1:05:51"}
+{"current_steps": 1500, "total_steps": 1770, "eval_loss": 1.4967381954193115, "epoch": 2.5402201524132093, "percentage": 84.75, "elapsed_time": "6:13:19", "remaining_time": "1:07:11"}