Training in progress, step 1000

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:4f539e713d8ea0369fb2b9305e870081b4aa921e82a71bcef544e8bd0f7af069
+oid sha256:5b2b32a63befd25f17197326aaace88a9194875999714a10643ff931bbfcd325
 size 83945296

trainer_log.jsonl

@@ -49,3 +49,54 @@
 {"current_steps": 490, "total_steps": 1770, "loss": 0.8576, "accuracy": 0.6000000238418579, "learning_rate": 4.11287244177176e-06, "epoch": 0.8298052497883149, "percentage": 27.68, "elapsed_time": "1:18:24", "remaining_time": "3:24:49"}
 {"current_steps": 500, "total_steps": 1770, "loss": 0.8758, "accuracy": 0.637499988079071, "learning_rate": 4.078714909616215e-06, "epoch": 0.8467400508044031, "percentage": 28.25, "elapsed_time": "1:20:00", "remaining_time": "3:23:13"}
 {"current_steps": 500, "total_steps": 1770, "eval_loss": 0.8691067099571228, "epoch": 0.8467400508044031, "percentage": 28.25, "elapsed_time": "1:23:14", "remaining_time": "3:31:26"}
+{"current_steps": 510, "total_steps": 1770, "loss": 0.9067, "accuracy": 0.550000011920929, "learning_rate": 4.044059990272125e-06, "epoch": 0.8636748518204911, "percentage": 28.81, "elapsed_time": "1:24:55", "remaining_time": "3:29:47"}
+{"current_steps": 520, "total_steps": 1770, "loss": 0.9597, "accuracy": 0.574999988079071, "learning_rate": 4.0089186020584345e-06, "epoch": 0.8806096528365792, "percentage": 29.38, "elapsed_time": "1:26:31", "remaining_time": "3:28:00"}
+{"current_steps": 530, "total_steps": 1770, "loss": 0.871, "accuracy": 0.581250011920929, "learning_rate": 3.973301816560124e-06, "epoch": 0.8975444538526672, "percentage": 29.94, "elapsed_time": "1:28:04", "remaining_time": "3:26:04"}
+{"current_steps": 540, "total_steps": 1770, "loss": 0.8891, "accuracy": 0.5249999761581421, "learning_rate": 3.937220855140021e-06, "epoch": 0.9144792548687553, "percentage": 30.51, "elapsed_time": "1:29:38", "remaining_time": "3:24:10"}
+{"current_steps": 550, "total_steps": 1770, "loss": 0.8817, "accuracy": 0.4937500059604645, "learning_rate": 3.900687085403418e-06, "epoch": 0.9314140558848434, "percentage": 31.07, "elapsed_time": "1:31:12", "remaining_time": "3:22:18"}
+{"current_steps": 560, "total_steps": 1770, "loss": 0.8764, "accuracy": 0.5625, "learning_rate": 3.863712017616614e-06, "epoch": 0.9483488569009314, "percentage": 31.64, "elapsed_time": "1:32:51", "remaining_time": "3:20:39"}
+{"current_steps": 570, "total_steps": 1770, "loss": 0.8764, "accuracy": 0.5625, "learning_rate": 3.826307301080504e-06, "epoch": 0.9652836579170194, "percentage": 32.2, "elapsed_time": "1:34:22", "remaining_time": "3:18:41"}
+{"current_steps": 580, "total_steps": 1770, "loss": 0.8812, "accuracy": 0.48750001192092896, "learning_rate": 3.7884847204603775e-06, "epoch": 0.9822184589331076, "percentage": 32.77, "elapsed_time": "1:36:03", "remaining_time": "3:17:05"}
+{"current_steps": 590, "total_steps": 1770, "loss": 1.0009, "accuracy": 0.5375000238418579, "learning_rate": 3.750256192073058e-06, "epoch": 0.9991532599491956, "percentage": 33.33, "elapsed_time": "1:37:45", "remaining_time": "3:15:30"}
+{"current_steps": 600, "total_steps": 1770, "loss": 0.8166, "accuracy": 0.5375000238418579, "learning_rate": 3.7116337601325715e-06, "epoch": 1.0160880609652836, "percentage": 33.9, "elapsed_time": "1:39:22", "remaining_time": "3:13:46"}
+{"current_steps": 610, "total_steps": 1770, "loss": 0.8117, "accuracy": 0.53125, "learning_rate": 3.6726295929555154e-06, "epoch": 1.0330228619813717, "percentage": 34.46, "elapsed_time": "1:40:57", "remaining_time": "3:11:58"}
+{"current_steps": 620, "total_steps": 1770, "loss": 0.843, "accuracy": 0.59375, "learning_rate": 3.6332559791273307e-06, "epoch": 1.0499576629974599, "percentage": 35.03, "elapsed_time": "1:42:34", "remaining_time": "3:10:14"}
+{"current_steps": 630, "total_steps": 1770, "loss": 0.8715, "accuracy": 0.59375, "learning_rate": 3.593525323630681e-06, "epoch": 1.0668924640135478, "percentage": 35.59, "elapsed_time": "1:44:12", "remaining_time": "3:08:34"}
+{"current_steps": 640, "total_steps": 1770, "loss": 0.8548, "accuracy": 0.6187499761581421, "learning_rate": 3.5534501439371615e-06, "epoch": 1.083827265029636, "percentage": 36.16, "elapsed_time": "1:45:50", "remaining_time": "3:06:52"}
+{"current_steps": 650, "total_steps": 1770, "loss": 0.864, "accuracy": 0.6187499761581421, "learning_rate": 3.5130430660635633e-06, "epoch": 1.100762066045724, "percentage": 36.72, "elapsed_time": "1:47:26", "remaining_time": "3:05:07"}
+{"current_steps": 660, "total_steps": 1770, "loss": 0.8483, "accuracy": 0.5375000238418579, "learning_rate": 3.4723168205939444e-06, "epoch": 1.117696867061812, "percentage": 37.29, "elapsed_time": "1:48:57", "remaining_time": "3:03:15"}
+{"current_steps": 670, "total_steps": 1770, "loss": 0.9059, "accuracy": 0.5375000238418579, "learning_rate": 3.431284238668754e-06, "epoch": 1.1346316680779, "percentage": 37.85, "elapsed_time": "1:50:32", "remaining_time": "3:01:28"}
+{"current_steps": 680, "total_steps": 1770, "loss": 0.9091, "accuracy": 0.5249999761581421, "learning_rate": 3.389958247942274e-06, "epoch": 1.1515664690939882, "percentage": 38.42, "elapsed_time": "1:52:07", "remaining_time": "2:59:44"}
+{"current_steps": 690, "total_steps": 1770, "loss": 0.9066, "accuracy": 0.5625, "learning_rate": 3.3483518685096588e-06, "epoch": 1.168501270110076, "percentage": 38.98, "elapsed_time": "1:53:44", "remaining_time": "2:58:01"}
+{"current_steps": 700, "total_steps": 1770, "loss": 0.8451, "accuracy": 0.550000011920929, "learning_rate": 3.306478208804839e-06, "epoch": 1.1854360711261642, "percentage": 39.55, "elapsed_time": "1:55:14", "remaining_time": "2:56:09"}
+{"current_steps": 710, "total_steps": 1770, "loss": 0.7814, "accuracy": 0.606249988079071, "learning_rate": 3.264350461470608e-06, "epoch": 1.2023708721422524, "percentage": 40.11, "elapsed_time": "1:56:46", "remaining_time": "2:54:20"}
+{"current_steps": 720, "total_steps": 1770, "loss": 0.7828, "accuracy": 0.625, "learning_rate": 3.2219818992021685e-06, "epoch": 1.2193056731583405, "percentage": 40.68, "elapsed_time": "1:58:27", "remaining_time": "2:52:44"}
+{"current_steps": 730, "total_steps": 1770, "loss": 0.7733, "accuracy": 0.5874999761581421, "learning_rate": 3.1793858705654595e-06, "epoch": 1.2362404741744284, "percentage": 41.24, "elapsed_time": "2:00:05", "remaining_time": "2:51:06"}
+{"current_steps": 740, "total_steps": 1770, "loss": 0.8832, "accuracy": 0.4937500059604645, "learning_rate": 3.1365757957915787e-06, "epoch": 1.2531752751905165, "percentage": 41.81, "elapsed_time": "2:01:40", "remaining_time": "2:49:21"}
+{"current_steps": 750, "total_steps": 1770, "loss": 0.9282, "accuracy": 0.543749988079071, "learning_rate": 3.093565162548633e-06, "epoch": 1.2701100762066047, "percentage": 42.37, "elapsed_time": "2:03:16", "remaining_time": "2:47:39"}
+{"current_steps": 760, "total_steps": 1770, "loss": 0.8093, "accuracy": 0.625, "learning_rate": 3.0503675216923294e-06, "epoch": 1.2870448772226926, "percentage": 42.94, "elapsed_time": "2:04:53", "remaining_time": "2:45:58"}
+{"current_steps": 770, "total_steps": 1770, "loss": 0.844, "accuracy": 0.5562499761581421, "learning_rate": 3.0069964829966748e-06, "epoch": 1.3039796782387807, "percentage": 43.5, "elapsed_time": "2:06:32", "remaining_time": "2:44:20"}
+{"current_steps": 780, "total_steps": 1770, "loss": 0.8183, "accuracy": 0.6000000238418579, "learning_rate": 2.963465710866094e-06, "epoch": 1.3209144792548688, "percentage": 44.07, "elapsed_time": "2:08:13", "remaining_time": "2:42:45"}
+{"current_steps": 790, "total_steps": 1770, "loss": 0.8826, "accuracy": 0.5562499761581421, "learning_rate": 2.919788920030357e-06, "epoch": 1.337849280270957, "percentage": 44.63, "elapsed_time": "2:09:55", "remaining_time": "2:41:10"}
+{"current_steps": 800, "total_steps": 1770, "loss": 0.8173, "accuracy": 0.6000000238418579, "learning_rate": 2.8759798712236303e-06, "epoch": 1.3547840812870449, "percentage": 45.2, "elapsed_time": "2:11:34", "remaining_time": "2:39:31"}
+{"current_steps": 810, "total_steps": 1770, "loss": 0.8841, "accuracy": 0.5375000238418579, "learning_rate": 2.8320523668490507e-06, "epoch": 1.371718882303133, "percentage": 45.76, "elapsed_time": "2:13:07", "remaining_time": "2:37:47"}
+{"current_steps": 820, "total_steps": 1770, "loss": 0.8735, "accuracy": 0.543749988079071, "learning_rate": 2.7880202466301597e-06, "epoch": 1.388653683319221, "percentage": 46.33, "elapsed_time": "2:14:45", "remaining_time": "2:36:07"}
+{"current_steps": 830, "total_steps": 1770, "loss": 0.8235, "accuracy": 0.5874999761581421, "learning_rate": 2.7438973832505854e-06, "epoch": 1.405588484335309, "percentage": 46.89, "elapsed_time": "2:16:18", "remaining_time": "2:34:22"}
+{"current_steps": 840, "total_steps": 1770, "loss": 0.8831, "accuracy": 0.512499988079071, "learning_rate": 2.699697677983341e-06, "epoch": 1.4225232853513972, "percentage": 47.46, "elapsed_time": "2:17:53", "remaining_time": "2:32:40"}
+{"current_steps": 850, "total_steps": 1770, "loss": 0.9054, "accuracy": 0.53125, "learning_rate": 2.6554350563111115e-06, "epoch": 1.4394580863674853, "percentage": 48.02, "elapsed_time": "2:19:30", "remaining_time": "2:31:00"}
+{"current_steps": 860, "total_steps": 1770, "loss": 0.7775, "accuracy": 0.5625, "learning_rate": 2.611123463538913e-06, "epoch": 1.4563928873835732, "percentage": 48.59, "elapsed_time": "2:21:08", "remaining_time": "2:29:21"}
+{"current_steps": 870, "total_steps": 1770, "loss": 0.898, "accuracy": 0.612500011920929, "learning_rate": 2.566776860400514e-06, "epoch": 1.4733276883996613, "percentage": 49.15, "elapsed_time": "2:22:46", "remaining_time": "2:27:41"}
+{"current_steps": 880, "total_steps": 1770, "loss": 0.8512, "accuracy": 0.512499988079071, "learning_rate": 2.522409218659989e-06, "epoch": 1.4902624894157492, "percentage": 49.72, "elapsed_time": "2:24:22", "remaining_time": "2:26:00"}
+{"current_steps": 890, "total_steps": 1770, "loss": 0.8459, "accuracy": 0.6499999761581421, "learning_rate": 2.4780345167097976e-06, "epoch": 1.5071972904318374, "percentage": 50.28, "elapsed_time": "2:26:02", "remaining_time": "2:24:24"}
+{"current_steps": 900, "total_steps": 1770, "loss": 0.8792, "accuracy": 0.6187499761581421, "learning_rate": 2.4336667351667747e-06, "epoch": 1.5241320914479255, "percentage": 50.85, "elapsed_time": "2:27:36", "remaining_time": "2:22:41"}
+{"current_steps": 910, "total_steps": 1770, "loss": 0.8478, "accuracy": 0.5874999761581421, "learning_rate": 2.3893198524674264e-06, "epoch": 1.5410668924640136, "percentage": 51.41, "elapsed_time": "2:29:15", "remaining_time": "2:21:03"}
+{"current_steps": 920, "total_steps": 1770, "loss": 0.8759, "accuracy": 0.4937500059604645, "learning_rate": 2.345007840463904e-06, "epoch": 1.5580016934801018, "percentage": 51.98, "elapsed_time": "2:30:57", "remaining_time": "2:19:28"}
+{"current_steps": 930, "total_steps": 1770, "loss": 0.8639, "accuracy": 0.5687500238418579, "learning_rate": 2.3007446600220572e-06, "epoch": 1.5749364944961897, "percentage": 52.54, "elapsed_time": "2:32:37", "remaining_time": "2:17:51"}
+{"current_steps": 940, "total_steps": 1770, "loss": 0.862, "accuracy": 0.4375, "learning_rate": 2.2565442566229507e-06, "epoch": 1.5918712955122776, "percentage": 53.11, "elapsed_time": "2:34:16", "remaining_time": "2:16:13"}
+{"current_steps": 950, "total_steps": 1770, "loss": 0.8399, "accuracy": 0.6499999761581421, "learning_rate": 2.2124205559692195e-06, "epoch": 1.6088060965283657, "percentage": 53.67, "elapsed_time": "2:35:54", "remaining_time": "2:14:34"}
+{"current_steps": 960, "total_steps": 1770, "loss": 0.8565, "accuracy": 0.6312500238418579, "learning_rate": 2.168387459597666e-06, "epoch": 1.6257408975444538, "percentage": 54.24, "elapsed_time": "2:37:36", "remaining_time": "2:12:58"}
+{"current_steps": 970, "total_steps": 1770, "loss": 0.8408, "accuracy": 0.581250011920929, "learning_rate": 2.1244588404994648e-06, "epoch": 1.642675698560542, "percentage": 54.8, "elapsed_time": "2:39:12", "remaining_time": "2:11:18"}
+{"current_steps": 980, "total_steps": 1770, "loss": 0.8786, "accuracy": 0.53125, "learning_rate": 2.08064853874936e-06, "epoch": 1.65961049957663, "percentage": 55.37, "elapsed_time": "2:40:51", "remaining_time": "2:09:40"}
+{"current_steps": 990, "total_steps": 1770, "loss": 0.7743, "accuracy": 0.6499999761581421, "learning_rate": 2.0369703571452387e-06, "epoch": 1.676545300592718, "percentage": 55.93, "elapsed_time": "2:42:30", "remaining_time": "2:08:02"}
+{"current_steps": 1000, "total_steps": 1770, "loss": 0.8098, "accuracy": 0.5687500238418579, "learning_rate": 1.993438056859441e-06, "epoch": 1.6934801016088061, "percentage": 56.5, "elapsed_time": "2:44:07", "remaining_time": "2:06:22"}
+{"current_steps": 1000, "total_steps": 1770, "eval_loss": 0.8548597693443298, "epoch": 1.6934801016088061, "percentage": 56.5, "elapsed_time": "2:47:22", "remaining_time": "2:08:52"}