Training in progress, step 1500

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c2621805c06ea3b388ce3bb78df949ec9cf940808a50f3ccc5b0f0268a9e0aa7
+oid sha256:4bbe7236909192cae9c1ee2a93dc6e6cebabeb8942c08ef415e34472005388f6
 size 100059752

trainer_log.jsonl

@@ -100,3 +100,54 @@
 {"current_steps": 990, "total_steps": 1686, "loss": 1.3565, "accuracy": 0.550000011920929, "learning_rate": 1.8239381598343576e-06, "epoch": 1.76039119804401, "percentage": 58.72, "elapsed_time": "3:41:20", "remaining_time": "2:35:36"}
 {"current_steps": 1000, "total_steps": 1686, "loss": 1.428, "accuracy": 0.574999988079071, "learning_rate": 1.779207600392312e-06, "epoch": 1.7781729273171816, "percentage": 59.31, "elapsed_time": "3:43:37", "remaining_time": "2:33:24"}
 {"current_steps": 1000, "total_steps": 1686, "eval_loss": 1.3595393896102905, "epoch": 1.7781729273171816, "percentage": 59.31, "elapsed_time": "3:47:14", "remaining_time": "2:35:53"}
+{"current_steps": 1010, "total_steps": 1686, "loss": 1.3422, "accuracy": 0.512499988079071, "learning_rate": 1.7347273253353552e-06, "epoch": 1.7959546565903532, "percentage": 59.91, "elapsed_time": "3:49:28", "remaining_time": "2:33:35"}
+{"current_steps": 1020, "total_steps": 1686, "loss": 1.3849, "accuracy": 0.574999988079071, "learning_rate": 1.690512779774029e-06, "epoch": 1.8137363858635251, "percentage": 60.5, "elapsed_time": "3:51:43", "remaining_time": "2:31:18"}
+{"current_steps": 1030, "total_steps": 1686, "loss": 1.2779, "accuracy": 0.5625, "learning_rate": 1.6465793165482838e-06, "epoch": 1.831518115136697, "percentage": 61.09, "elapsed_time": "3:53:57", "remaining_time": "2:29:00"}
+{"current_steps": 1040, "total_steps": 1686, "loss": 1.2943, "accuracy": 0.574999988079071, "learning_rate": 1.6029421908964305e-06, "epoch": 1.849299844409869, "percentage": 61.68, "elapsed_time": "3:55:58", "remaining_time": "2:26:34"}
+{"current_steps": 1050, "total_steps": 1686, "loss": 1.3589, "accuracy": 0.550000011920929, "learning_rate": 1.559616555157985e-06, "epoch": 1.8670815736830408, "percentage": 62.28, "elapsed_time": "3:58:14", "remaining_time": "2:24:18"}
+{"current_steps": 1060, "total_steps": 1686, "loss": 1.4237, "accuracy": 0.5249999761581421, "learning_rate": 1.516617453512252e-06, "epoch": 1.8848633029562125, "percentage": 62.87, "elapsed_time": "4:00:28", "remaining_time": "2:22:00"}
+{"current_steps": 1070, "total_steps": 1686, "loss": 1.3067, "accuracy": 0.5249999761581421, "learning_rate": 1.473959816754449e-06, "epoch": 1.9026450322293842, "percentage": 63.46, "elapsed_time": "4:02:33", "remaining_time": "2:19:38"}
+{"current_steps": 1080, "total_steps": 1686, "loss": 1.3965, "accuracy": 0.543749988079071, "learning_rate": 1.4316584571112213e-06, "epoch": 1.920426761502556, "percentage": 64.06, "elapsed_time": "4:04:46", "remaining_time": "2:17:20"}
+{"current_steps": 1090, "total_steps": 1686, "loss": 1.3875, "accuracy": 0.5625, "learning_rate": 1.389728063097306e-06, "epoch": 1.938208490775728, "percentage": 64.65, "elapsed_time": "4:07:04", "remaining_time": "2:15:05"}
+{"current_steps": 1100, "total_steps": 1686, "loss": 1.3544, "accuracy": 0.612500011920929, "learning_rate": 1.348183194415179e-06, "epoch": 1.9559902200488999, "percentage": 65.24, "elapsed_time": "4:09:04", "remaining_time": "2:12:41"}
+{"current_steps": 1110, "total_steps": 1686, "loss": 1.3195, "accuracy": 0.518750011920929, "learning_rate": 1.3070382768994015e-06, "epoch": 1.9737719493220716, "percentage": 65.84, "elapsed_time": "4:11:20", "remaining_time": "2:10:25"}
+{"current_steps": 1120, "total_steps": 1686, "loss": 1.3361, "accuracy": 0.543749988079071, "learning_rate": 1.2663075975074746e-06, "epoch": 1.9915536785952432, "percentage": 66.43, "elapsed_time": "4:13:39", "remaining_time": "2:08:11"}
+{"current_steps": 1130, "total_steps": 1686, "loss": 1.4548, "accuracy": 0.48750001192092896, "learning_rate": 1.2260052993589034e-06, "epoch": 2.009335407868415, "percentage": 67.02, "elapsed_time": "4:15:52", "remaining_time": "2:05:54"}
+{"current_steps": 1140, "total_steps": 1686, "loss": 1.2835, "accuracy": 0.6312500238418579, "learning_rate": 1.1861453768242099e-06, "epoch": 2.027117137141587, "percentage": 67.62, "elapsed_time": "4:18:03", "remaining_time": "2:03:35"}
+{"current_steps": 1150, "total_steps": 1686, "loss": 1.4101, "accuracy": 0.5249999761581421, "learning_rate": 1.1467416706655982e-06, "epoch": 2.044898866414759, "percentage": 68.21, "elapsed_time": "4:20:25", "remaining_time": "2:01:22"}
+{"current_steps": 1160, "total_steps": 1686, "loss": 1.3236, "accuracy": 0.59375, "learning_rate": 1.1078078632309559e-06, "epoch": 2.062680595687931, "percentage": 68.8, "elapsed_time": "4:22:33", "remaining_time": "1:59:03"}
+{"current_steps": 1170, "total_steps": 1686, "loss": 1.3642, "accuracy": 0.53125, "learning_rate": 1.0693574737028627e-06, "epoch": 2.0804623249611023, "percentage": 69.4, "elapsed_time": "4:24:48", "remaining_time": "1:56:47"}
+{"current_steps": 1180, "total_steps": 1686, "loss": 1.2441, "accuracy": 0.5874999761581421, "learning_rate": 1.0314038534042586e-06, "epoch": 2.098244054234274, "percentage": 69.99, "elapsed_time": "4:27:05", "remaining_time": "1:54:32"}
+{"current_steps": 1190, "total_steps": 1686, "loss": 1.3104, "accuracy": 0.4937500059604645, "learning_rate": 9.939601811623946e-07, "epoch": 2.116025783507446, "percentage": 70.58, "elapsed_time": "4:29:13", "remaining_time": "1:52:12"}
+{"current_steps": 1200, "total_steps": 1686, "loss": 1.3354, "accuracy": 0.6000000238418579, "learning_rate": 9.570394587326825e-07, "epoch": 2.133807512780618, "percentage": 71.17, "elapsed_time": "4:31:24", "remaining_time": "1:49:55"}
+{"current_steps": 1210, "total_steps": 1686, "loss": 1.3088, "accuracy": 0.59375, "learning_rate": 9.206545062840302e-07, "epoch": 2.15158924205379, "percentage": 71.77, "elapsed_time": "4:33:34", "remaining_time": "1:47:37"}
+{"current_steps": 1220, "total_steps": 1686, "loss": 1.2941, "accuracy": 0.518750011920929, "learning_rate": 8.848179579472285e-07, "epoch": 2.1693709713269618, "percentage": 72.36, "elapsed_time": "4:35:49", "remaining_time": "1:45:21"}
+{"current_steps": 1230, "total_steps": 1686, "loss": 1.2647, "accuracy": 0.6312500238418579, "learning_rate": 8.495422574279403e-07, "epoch": 2.1871527006001332, "percentage": 72.95, "elapsed_time": "4:37:53", "remaining_time": "1:43:01"}
+{"current_steps": 1240, "total_steps": 1686, "loss": 1.3784, "accuracy": 0.5687500238418579, "learning_rate": 8.148396536858063e-07, "epoch": 2.204934429873305, "percentage": 73.55, "elapsed_time": "4:40:04", "remaining_time": "1:40:44"}
+{"current_steps": 1250, "total_steps": 1686, "loss": 1.2915, "accuracy": 0.518750011920929, "learning_rate": 7.807221966811815e-07, "epoch": 2.222716159146477, "percentage": 74.14, "elapsed_time": "4:42:21", "remaining_time": "1:38:29"}
+{"current_steps": 1260, "total_steps": 1686, "loss": 1.2993, "accuracy": 0.48750001192092896, "learning_rate": 7.47201733190962e-07, "epoch": 2.240497888419649, "percentage": 74.73, "elapsed_time": "4:44:35", "remaining_time": "1:36:13"}
+{"current_steps": 1270, "total_steps": 1686, "loss": 1.3351, "accuracy": 0.574999988079071, "learning_rate": 7.142899026949721e-07, "epoch": 2.258279617692821, "percentage": 75.33, "elapsed_time": "4:46:50", "remaining_time": "1:33:57"}
+{"current_steps": 1280, "total_steps": 1686, "loss": 1.2691, "accuracy": 0.543749988079071, "learning_rate": 6.819981333343273e-07, "epoch": 2.2760613469659923, "percentage": 75.92, "elapsed_time": "4:48:56", "remaining_time": "1:31:38"}
+{"current_steps": 1290, "total_steps": 1686, "loss": 1.423, "accuracy": 0.4749999940395355, "learning_rate": 6.503376379431839e-07, "epoch": 2.293843076239164, "percentage": 76.51, "elapsed_time": "4:51:09", "remaining_time": "1:29:22"}
+{"current_steps": 1300, "total_steps": 1686, "loss": 1.3081, "accuracy": 0.6187499761581421, "learning_rate": 6.193194101552502e-07, "epoch": 2.311624805512336, "percentage": 77.11, "elapsed_time": "4:53:27", "remaining_time": "1:27:08"}
+{"current_steps": 1310, "total_steps": 1686, "loss": 1.303, "accuracy": 0.59375, "learning_rate": 5.889542205864083e-07, "epoch": 2.329406534785508, "percentage": 77.7, "elapsed_time": "4:55:31", "remaining_time": "1:24:49"}
+{"current_steps": 1320, "total_steps": 1686, "loss": 1.3841, "accuracy": 0.48124998807907104, "learning_rate": 5.592526130947862e-07, "epoch": 2.34718826405868, "percentage": 78.29, "elapsed_time": "4:57:49", "remaining_time": "1:22:34"}
+{"current_steps": 1330, "total_steps": 1686, "loss": 1.3008, "accuracy": 0.550000011920929, "learning_rate": 5.302249011195507e-07, "epoch": 2.3649699933318518, "percentage": 78.88, "elapsed_time": "4:59:50", "remaining_time": "1:20:15"}
+{"current_steps": 1340, "total_steps": 1686, "loss": 1.3833, "accuracy": 0.59375, "learning_rate": 5.018811640997307e-07, "epoch": 2.382751722605023, "percentage": 79.48, "elapsed_time": "5:02:11", "remaining_time": "1:18:01"}
+{"current_steps": 1350, "total_steps": 1686, "loss": 1.3209, "accuracy": 0.5625, "learning_rate": 4.7423124397427105e-07, "epoch": 2.400533451878195, "percentage": 80.07, "elapsed_time": "5:04:10", "remaining_time": "1:15:42"}
+{"current_steps": 1360, "total_steps": 1686, "loss": 1.3842, "accuracy": 0.606249988079071, "learning_rate": 4.472847417645787e-07, "epoch": 2.418315181151367, "percentage": 80.66, "elapsed_time": "5:06:17", "remaining_time": "1:13:25"}
+{"current_steps": 1370, "total_steps": 1686, "loss": 1.3596, "accuracy": 0.543749988079071, "learning_rate": 4.210510142406993e-07, "epoch": 2.436096910424539, "percentage": 81.26, "elapsed_time": "5:08:24", "remaining_time": "1:11:08"}
+{"current_steps": 1380, "total_steps": 1686, "loss": 1.3585, "accuracy": 0.5249999761581421, "learning_rate": 3.9553917067232966e-07, "epoch": 2.4538786396977104, "percentage": 81.85, "elapsed_time": "5:10:35", "remaining_time": "1:08:52"}
+{"current_steps": 1390, "total_steps": 1686, "loss": 1.3025, "accuracy": 0.5062500238418579, "learning_rate": 3.707580696657509e-07, "epoch": 2.4716603689708823, "percentage": 82.44, "elapsed_time": "5:12:42", "remaining_time": "1:06:35"}
+{"current_steps": 1400, "total_steps": 1686, "loss": 1.3758, "accuracy": 0.512499988079071, "learning_rate": 3.4671631608781815e-07, "epoch": 2.489442098244054, "percentage": 83.04, "elapsed_time": "5:14:45", "remaining_time": "1:04:18"}
+{"current_steps": 1410, "total_steps": 1686, "loss": 1.3485, "accuracy": 0.543749988079071, "learning_rate": 3.234222580780405e-07, "epoch": 2.507223827517226, "percentage": 83.63, "elapsed_time": "5:16:47", "remaining_time": "1:02:00"}
+{"current_steps": 1420, "total_steps": 1686, "loss": 1.3364, "accuracy": 0.550000011920929, "learning_rate": 3.0088398414982375e-07, "epoch": 2.525005556790398, "percentage": 84.22, "elapsed_time": "5:18:55", "remaining_time": "0:59:44"}
+{"current_steps": 1430, "total_steps": 1686, "loss": 1.3164, "accuracy": 0.6000000238418579, "learning_rate": 2.7910932038184487e-07, "epoch": 2.54278728606357, "percentage": 84.82, "elapsed_time": "5:21:00", "remaining_time": "0:57:27"}
+{"current_steps": 1440, "total_steps": 1686, "loss": 1.2763, "accuracy": 0.550000011920929, "learning_rate": 2.5810582770057325e-07, "epoch": 2.5605690153367417, "percentage": 85.41, "elapsed_time": "5:23:16", "remaining_time": "0:55:13"}
+{"current_steps": 1450, "total_steps": 1686, "loss": 1.3706, "accuracy": 0.5249999761581421, "learning_rate": 2.3788079925484402e-07, "epoch": 2.578350744609913, "percentage": 86.0, "elapsed_time": "5:25:34", "remaining_time": "0:52:59"}
+{"current_steps": 1460, "total_steps": 1686, "loss": 1.284, "accuracy": 0.6000000238418579, "learning_rate": 2.1844125788342661e-07, "epoch": 2.596132473883085, "percentage": 86.6, "elapsed_time": "5:27:40", "remaining_time": "0:50:43"}
+{"current_steps": 1470, "total_steps": 1686, "loss": 1.2838, "accuracy": 0.543749988079071, "learning_rate": 1.9979395367644428e-07, "epoch": 2.613914203156257, "percentage": 87.19, "elapsed_time": "5:29:55", "remaining_time": "0:48:28"}
+{"current_steps": 1480, "total_steps": 1686, "loss": 1.3463, "accuracy": 0.512499988079071, "learning_rate": 1.81945361631512e-07, "epoch": 2.631695932429429, "percentage": 87.78, "elapsed_time": "5:32:04", "remaining_time": "0:46:13"}
+{"current_steps": 1490, "total_steps": 1686, "loss": 1.289, "accuracy": 0.59375, "learning_rate": 1.6490167940538343e-07, "epoch": 2.6494776617026004, "percentage": 88.37, "elapsed_time": "5:34:21", "remaining_time": "0:43:58"}
+{"current_steps": 1500, "total_steps": 1686, "loss": 1.3302, "accuracy": 0.512499988079071, "learning_rate": 1.4866882516191339e-07, "epoch": 2.6672593909757722, "percentage": 88.97, "elapsed_time": "5:36:38", "remaining_time": "0:41:44"}
+{"current_steps": 1500, "total_steps": 1686, "eval_loss": 1.347064733505249, "epoch": 2.6672593909757722, "percentage": 88.97, "elapsed_time": "5:40:15", "remaining_time": "0:42:11"}