Upload trainer_log.jsonl with huggingface_hub
Browse files- trainer_log.jsonl +87 -0
trainer_log.jsonl
ADDED
|
@@ -0,0 +1,87 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"current_steps": 10, "total_steps": 10164, "loss": 4.4021, "learning_rate": 3.4414945919370703e-07, "epoch": 0.002950940612320177, "percentage": 0.1, "elapsed_time": "0:04:41", "remaining_time": "3 days, 7:25:38", "throughput": "0.00", "total_tokens": 0}
|
| 2 |
+
{"current_steps": 20, "total_steps": 10164, "loss": 4.0071, "learning_rate": 7.866273352999017e-07, "epoch": 0.005901881224640354, "percentage": 0.2, "elapsed_time": "0:09:22", "remaining_time": "3 days, 7:16:00", "throughput": "0.00", "total_tokens": 0}
|
| 3 |
+
{"current_steps": 10, "total_steps": 847, "loss": 3.4593, "learning_rate": 5.294117647058824e-06, "epoch": 0.011802891708468575, "percentage": 1.18, "elapsed_time": "0:02:04", "remaining_time": "2:54:06", "throughput": "0.00", "total_tokens": 0}
|
| 4 |
+
{"current_steps": 20, "total_steps": 847, "loss": 2.9512, "learning_rate": 1e-05, "epoch": 0.02360578341693715, "percentage": 2.36, "elapsed_time": "0:04:08", "remaining_time": "2:51:33", "throughput": "0.00", "total_tokens": 0}
|
| 5 |
+
{"current_steps": 30, "total_steps": 847, "loss": 2.2602, "learning_rate": 1.588235294117647e-05, "epoch": 0.03540867512540572, "percentage": 3.54, "elapsed_time": "0:06:13", "remaining_time": "2:49:21", "throughput": "0.00", "total_tokens": 0}
|
| 6 |
+
{"current_steps": 40, "total_steps": 847, "loss": 1.9542, "learning_rate": 2.1764705882352943e-05, "epoch": 0.0472115668338743, "percentage": 4.72, "elapsed_time": "0:08:17", "remaining_time": "2:47:21", "throughput": "0.00", "total_tokens": 0}
|
| 7 |
+
{"current_steps": 50, "total_steps": 847, "loss": 1.7491, "learning_rate": 2.7647058823529416e-05, "epoch": 0.05901445854234287, "percentage": 5.9, "elapsed_time": "0:10:22", "remaining_time": "2:45:16", "throughput": "0.00", "total_tokens": 0}
|
| 8 |
+
{"current_steps": 60, "total_steps": 847, "loss": 1.7127, "learning_rate": 3.352941176470588e-05, "epoch": 0.07081735025081144, "percentage": 7.08, "elapsed_time": "0:12:26", "remaining_time": "2:43:13", "throughput": "0.00", "total_tokens": 0}
|
| 9 |
+
{"current_steps": 70, "total_steps": 847, "loss": 1.5587, "learning_rate": 3.9411764705882356e-05, "epoch": 0.08262024195928003, "percentage": 8.26, "elapsed_time": "0:14:31", "remaining_time": "2:41:15", "throughput": "0.00", "total_tokens": 0}
|
| 10 |
+
{"current_steps": 80, "total_steps": 847, "loss": 1.6199, "learning_rate": 4.5294117647058826e-05, "epoch": 0.0944231336677486, "percentage": 9.45, "elapsed_time": "0:16:36", "remaining_time": "2:39:12", "throughput": "0.00", "total_tokens": 0}
|
| 11 |
+
{"current_steps": 90, "total_steps": 847, "loss": 1.5481, "learning_rate": 4.999915012051437e-05, "epoch": 0.10622602537621717, "percentage": 10.63, "elapsed_time": "0:18:39", "remaining_time": "2:37:00", "throughput": "0.00", "total_tokens": 0}
|
| 12 |
+
{"current_steps": 100, "total_steps": 847, "loss": 1.4648, "learning_rate": 4.996941040535653e-05, "epoch": 0.11802891708468574, "percentage": 11.81, "elapsed_time": "0:20:44", "remaining_time": "2:34:58", "throughput": "0.00", "total_tokens": 0}
|
| 13 |
+
{"current_steps": 110, "total_steps": 847, "loss": 1.5273, "learning_rate": 4.989723448187131e-05, "epoch": 0.1298318087931543, "percentage": 12.99, "elapsed_time": "0:22:49", "remaining_time": "2:32:57", "throughput": "0.00", "total_tokens": 0}
|
| 14 |
+
{"current_steps": 120, "total_steps": 847, "loss": 1.5472, "learning_rate": 4.978274501505061e-05, "epoch": 0.14163470050162288, "percentage": 14.17, "elapsed_time": "0:24:53", "remaining_time": "2:30:46", "throughput": "0.00", "total_tokens": 0}
|
| 15 |
+
{"current_steps": 130, "total_steps": 847, "loss": 1.5428, "learning_rate": 4.962613658293158e-05, "epoch": 0.15343759221009148, "percentage": 15.35, "elapsed_time": "0:26:57", "remaining_time": "2:28:39", "throughput": "0.00", "total_tokens": 0}
|
| 16 |
+
{"current_steps": 140, "total_steps": 847, "loss": 1.3644, "learning_rate": 4.942767534590581e-05, "epoch": 0.16524048391856005, "percentage": 16.53, "elapsed_time": "0:29:00", "remaining_time": "2:26:29", "throughput": "0.00", "total_tokens": 0}
|
| 17 |
+
{"current_steps": 150, "total_steps": 847, "loss": 1.4653, "learning_rate": 4.918769859437232e-05, "epoch": 0.17704337562702863, "percentage": 17.71, "elapsed_time": "0:31:04", "remaining_time": "2:24:25", "throughput": "0.00", "total_tokens": 0}
|
| 18 |
+
{"current_steps": 160, "total_steps": 847, "loss": 1.5003, "learning_rate": 4.890661417550319e-05, "epoch": 0.1888462673354972, "percentage": 18.89, "elapsed_time": "0:33:08", "remaining_time": "2:22:19", "throughput": "0.00", "total_tokens": 0}
|
| 19 |
+
{"current_steps": 170, "total_steps": 847, "loss": 1.462, "learning_rate": 4.8584899800095864e-05, "epoch": 0.20064915904396577, "percentage": 20.07, "elapsed_time": "0:35:12", "remaining_time": "2:20:13", "throughput": "0.00", "total_tokens": 0}
|
| 20 |
+
{"current_steps": 180, "total_steps": 847, "loss": 1.3805, "learning_rate": 4.822310223069039e-05, "epoch": 0.21245205075243434, "percentage": 21.25, "elapsed_time": "0:37:16", "remaining_time": "2:18:07", "throughput": "0.00", "total_tokens": 0}
|
| 21 |
+
{"current_steps": 190, "total_steps": 847, "loss": 1.4678, "learning_rate": 4.782183635233124e-05, "epoch": 0.2242549424609029, "percentage": 22.43, "elapsed_time": "0:39:21", "remaining_time": "2:16:05", "throughput": "0.00", "total_tokens": 0}
|
| 22 |
+
{"current_steps": 200, "total_steps": 847, "loss": 1.378, "learning_rate": 4.738178412755306e-05, "epoch": 0.23605783416937148, "percentage": 23.61, "elapsed_time": "0:41:25", "remaining_time": "2:13:59", "throughput": "0.00", "total_tokens": 0}
|
| 23 |
+
{"current_steps": 210, "total_steps": 847, "loss": 1.4216, "learning_rate": 4.690369343736636e-05, "epoch": 0.24786072587784008, "percentage": 24.79, "elapsed_time": "0:43:29", "remaining_time": "2:11:55", "throughput": "0.00", "total_tokens": 0}
|
| 24 |
+
{"current_steps": 220, "total_steps": 847, "loss": 1.3915, "learning_rate": 4.6388376810212905e-05, "epoch": 0.2596636175863086, "percentage": 25.97, "elapsed_time": "0:45:33", "remaining_time": "2:09:50", "throughput": "0.00", "total_tokens": 0}
|
| 25 |
+
{"current_steps": 230, "total_steps": 847, "loss": 1.4104, "learning_rate": 4.583671004105096e-05, "epoch": 0.2714665092947772, "percentage": 27.15, "elapsed_time": "0:47:37", "remaining_time": "2:07:45", "throughput": "0.00", "total_tokens": 0}
|
| 26 |
+
{"current_steps": 240, "total_steps": 847, "loss": 1.411, "learning_rate": 4.524963070291744e-05, "epoch": 0.28326940100324577, "percentage": 28.34, "elapsed_time": "0:49:42", "remaining_time": "2:05:42", "throughput": "0.00", "total_tokens": 0}
|
| 27 |
+
{"current_steps": 250, "total_steps": 847, "loss": 1.3628, "learning_rate": 4.4628136553496375e-05, "epoch": 0.29507229271171437, "percentage": 29.52, "elapsed_time": "0:51:46", "remaining_time": "2:03:38", "throughput": "0.00", "total_tokens": 0}
|
| 28 |
+
{"current_steps": 260, "total_steps": 847, "loss": 1.331, "learning_rate": 4.397328383940196e-05, "epoch": 0.30687518442018297, "percentage": 30.7, "elapsed_time": "0:53:50", "remaining_time": "2:01:33", "throughput": "0.00", "total_tokens": 0}
|
| 29 |
+
{"current_steps": 270, "total_steps": 847, "loss": 1.338, "learning_rate": 4.328618550105802e-05, "epoch": 0.3186780761286515, "percentage": 31.88, "elapsed_time": "0:55:54", "remaining_time": "1:59:29", "throughput": "0.00", "total_tokens": 0}
|
| 30 |
+
{"current_steps": 280, "total_steps": 847, "loss": 1.4131, "learning_rate": 4.256800928122475e-05, "epoch": 0.3304809678371201, "percentage": 33.06, "elapsed_time": "0:57:58", "remaining_time": "1:57:24", "throughput": "0.00", "total_tokens": 0}
|
| 31 |
+
{"current_steps": 290, "total_steps": 847, "loss": 1.3584, "learning_rate": 4.181997574038741e-05, "epoch": 0.34228385954558865, "percentage": 34.24, "elapsed_time": "1:00:02", "remaining_time": "1:55:20", "throughput": "0.00", "total_tokens": 0}
|
| 32 |
+
{"current_steps": 300, "total_steps": 847, "loss": 1.3541, "learning_rate": 4.104335618237972e-05, "epoch": 0.35408675125405725, "percentage": 35.42, "elapsed_time": "1:02:06", "remaining_time": "1:53:15", "throughput": "0.00", "total_tokens": 0}
|
| 33 |
+
{"current_steps": 310, "total_steps": 847, "loss": 1.359, "learning_rate": 4.0239470493767704e-05, "epoch": 0.3658896429625258, "percentage": 36.6, "elapsed_time": "1:04:11", "remaining_time": "1:51:12", "throughput": "0.00", "total_tokens": 0}
|
| 34 |
+
{"current_steps": 320, "total_steps": 847, "loss": 1.261, "learning_rate": 3.940968490066559e-05, "epoch": 0.3776925346709944, "percentage": 37.78, "elapsed_time": "1:06:15", "remaining_time": "1:49:07", "throughput": "0.00", "total_tokens": 0}
|
| 35 |
+
{"current_steps": 330, "total_steps": 847, "loss": 1.2903, "learning_rate": 3.855540964679658e-05, "epoch": 0.389495426379463, "percentage": 38.96, "elapsed_time": "1:08:20", "remaining_time": "1:47:03", "throughput": "0.00", "total_tokens": 0}
|
| 36 |
+
{"current_steps": 340, "total_steps": 847, "loss": 1.3593, "learning_rate": 3.767809659674433e-05, "epoch": 0.40129831808793154, "percentage": 40.14, "elapsed_time": "1:10:24", "remaining_time": "1:44:58", "throughput": "0.00", "total_tokens": 0}
|
| 37 |
+
{"current_steps": 350, "total_steps": 847, "loss": 1.3608, "learning_rate": 3.677923676846864e-05, "epoch": 0.41310120979640014, "percentage": 41.32, "elapsed_time": "1:12:28", "remaining_time": "1:42:55", "throughput": "0.00", "total_tokens": 0}
|
| 38 |
+
{"current_steps": 360, "total_steps": 847, "loss": 1.2742, "learning_rate": 3.586035779927896e-05, "epoch": 0.4249041015048687, "percentage": 42.5, "elapsed_time": "1:14:33", "remaining_time": "1:40:51", "throughput": "0.00", "total_tokens": 0}
|
| 39 |
+
{"current_steps": 370, "total_steps": 847, "loss": 1.3217, "learning_rate": 3.492302134957218e-05, "epoch": 0.4367069932133373, "percentage": 43.68, "elapsed_time": "1:16:37", "remaining_time": "1:38:46", "throughput": "0.00", "total_tokens": 0}
|
| 40 |
+
{"current_steps": 380, "total_steps": 847, "loss": 1.2824, "learning_rate": 3.396882044874736e-05, "epoch": 0.4485098849218058, "percentage": 44.86, "elapsed_time": "1:18:41", "remaining_time": "1:36:42", "throughput": "0.00", "total_tokens": 0}
|
| 41 |
+
{"current_steps": 390, "total_steps": 847, "loss": 1.344, "learning_rate": 3.2999376787807864e-05, "epoch": 0.4603127766302744, "percentage": 46.04, "elapsed_time": "1:20:45", "remaining_time": "1:34:38", "throughput": "0.00", "total_tokens": 0}
|
| 42 |
+
{"current_steps": 400, "total_steps": 847, "loss": 1.3372, "learning_rate": 3.201633796325233e-05, "epoch": 0.47211566833874297, "percentage": 47.23, "elapsed_time": "1:22:49", "remaining_time": "1:32:33", "throughput": "0.00", "total_tokens": 0}
|
| 43 |
+
{"current_steps": 410, "total_steps": 847, "loss": 1.33, "learning_rate": 3.1021374676938584e-05, "epoch": 0.48391856004721157, "percentage": 48.41, "elapsed_time": "1:24:54", "remaining_time": "1:30:29", "throughput": "0.00", "total_tokens": 0}
|
| 44 |
+
{"current_steps": 420, "total_steps": 847, "loss": 1.2919, "learning_rate": 3.0016177896679255e-05, "epoch": 0.49572145175568016, "percentage": 49.59, "elapsed_time": "1:26:59", "remaining_time": "1:28:26", "throughput": "0.00", "total_tokens": 0}
|
| 45 |
+
{"current_steps": 430, "total_steps": 847, "loss": 1.2649, "learning_rate": 2.9002455982394944e-05, "epoch": 0.5075243434641488, "percentage": 50.77, "elapsed_time": "1:29:03", "remaining_time": "1:26:21", "throughput": "0.00", "total_tokens": 0}
|
| 46 |
+
{"current_steps": 440, "total_steps": 847, "loss": 1.3047, "learning_rate": 2.798193178270889e-05, "epoch": 0.5193272351726173, "percentage": 51.95, "elapsed_time": "1:31:07", "remaining_time": "1:24:17", "throughput": "0.00", "total_tokens": 0}
|
| 47 |
+
{"current_steps": 450, "total_steps": 847, "loss": 1.2862, "learning_rate": 2.695633970691786e-05, "epoch": 0.5311301268810859, "percentage": 53.13, "elapsed_time": "1:33:12", "remaining_time": "1:22:13", "throughput": "0.00", "total_tokens": 0}
|
| 48 |
+
{"current_steps": 460, "total_steps": 847, "loss": 1.2843, "learning_rate": 2.592742277731513e-05, "epoch": 0.5429330185895545, "percentage": 54.31, "elapsed_time": "1:35:16", "remaining_time": "1:20:09", "throughput": "0.00", "total_tokens": 0}
|
| 49 |
+
{"current_steps": 470, "total_steps": 847, "loss": 1.2795, "learning_rate": 2.489692966687566e-05, "epoch": 0.554735910298023, "percentage": 55.49, "elapsed_time": "1:37:20", "remaining_time": "1:18:04", "throughput": "0.00", "total_tokens": 0}
|
| 50 |
+
{"current_steps": 480, "total_steps": 847, "loss": 1.1897, "learning_rate": 2.386661172733762e-05, "epoch": 0.5665388020064915, "percentage": 56.67, "elapsed_time": "1:39:24", "remaining_time": "1:16:00", "throughput": "0.00", "total_tokens": 0}
|
| 51 |
+
{"current_steps": 490, "total_steps": 847, "loss": 1.3352, "learning_rate": 2.2838220012731365e-05, "epoch": 0.5783416937149601, "percentage": 57.85, "elapsed_time": "1:41:28", "remaining_time": "1:13:55", "throughput": "0.00", "total_tokens": 0}
|
| 52 |
+
{"current_steps": 500, "total_steps": 847, "loss": 1.2552, "learning_rate": 2.1813502303414306e-05, "epoch": 0.5901445854234287, "percentage": 59.03, "elapsed_time": "1:43:32", "remaining_time": "1:11:51", "throughput": "0.00", "total_tokens": 0}
|
| 53 |
+
{"current_steps": 510, "total_steps": 847, "loss": 1.2573, "learning_rate": 2.0794200135669584e-05, "epoch": 0.6019474771318973, "percentage": 60.21, "elapsed_time": "1:45:37", "remaining_time": "1:09:47", "throughput": "0.00", "total_tokens": 0}
|
| 54 |
+
{"current_steps": 520, "total_steps": 847, "loss": 1.2564, "learning_rate": 1.9782045841916625e-05, "epoch": 0.6137503688403659, "percentage": 61.39, "elapsed_time": "1:47:41", "remaining_time": "1:07:43", "throughput": "0.00", "total_tokens": 0}
|
| 55 |
+
{"current_steps": 530, "total_steps": 847, "loss": 1.1512, "learning_rate": 1.877875960656394e-05, "epoch": 0.6255532605488344, "percentage": 62.57, "elapsed_time": "1:49:44", "remaining_time": "1:05:38", "throughput": "0.00", "total_tokens": 0}
|
| 56 |
+
{"current_steps": 540, "total_steps": 847, "loss": 1.2434, "learning_rate": 1.7786046542507843e-05, "epoch": 0.637356152257303, "percentage": 63.75, "elapsed_time": "1:51:49", "remaining_time": "1:03:34", "throughput": "0.00", "total_tokens": 0}
|
| 57 |
+
{"current_steps": 550, "total_steps": 847, "loss": 1.325, "learning_rate": 1.680559379324558e-05, "epoch": 0.6491590439657716, "percentage": 64.94, "elapsed_time": "1:53:52", "remaining_time": "1:01:29", "throughput": "0.00", "total_tokens": 0}
|
| 58 |
+
{"current_steps": 560, "total_steps": 847, "loss": 1.2197, "learning_rate": 1.583906766552799e-05, "epoch": 0.6609619356742402, "percentage": 66.12, "elapsed_time": "1:55:57", "remaining_time": "0:59:25", "throughput": "0.00", "total_tokens": 0}
|
| 59 |
+
{"current_steps": 570, "total_steps": 847, "loss": 1.2821, "learning_rate": 1.4888110797424782e-05, "epoch": 0.6727648273827088, "percentage": 67.3, "elapsed_time": "1:58:01", "remaining_time": "0:57:21", "throughput": "0.00", "total_tokens": 0}
|
| 60 |
+
{"current_steps": 580, "total_steps": 847, "loss": 1.239, "learning_rate": 1.3954339366615334e-05, "epoch": 0.6845677190911773, "percentage": 68.48, "elapsed_time": "2:00:06", "remaining_time": "0:55:17", "throughput": "0.00", "total_tokens": 0}
|
| 61 |
+
{"current_steps": 590, "total_steps": 847, "loss": 1.2403, "learning_rate": 1.303934034364983e-05, "epoch": 0.6963706107996459, "percentage": 69.66, "elapsed_time": "2:02:11", "remaining_time": "0:53:13", "throughput": "0.00", "total_tokens": 0}
|
| 62 |
+
{"current_steps": 600, "total_steps": 847, "loss": 1.2466, "learning_rate": 1.21446687948485e-05, "epoch": 0.7081735025081145, "percentage": 70.84, "elapsed_time": "2:04:15", "remaining_time": "0:51:09", "throughput": "0.00", "total_tokens": 0}
|
| 63 |
+
{"current_steps": 610, "total_steps": 847, "loss": 1.1662, "learning_rate": 1.1271845239423196e-05, "epoch": 0.7199763942165831, "percentage": 72.02, "elapsed_time": "2:06:19", "remaining_time": "0:49:04", "throughput": "0.00", "total_tokens": 0}
|
| 64 |
+
{"current_steps": 620, "total_steps": 847, "loss": 1.3161, "learning_rate": 1.0422353065312573e-05, "epoch": 0.7317792859250516, "percentage": 73.2, "elapsed_time": "2:08:23", "remaining_time": "0:47:00", "throughput": "0.00", "total_tokens": 0}
|
| 65 |
+
{"current_steps": 630, "total_steps": 847, "loss": 1.2608, "learning_rate": 9.59763600812305e-06, "epoch": 0.7435821776335202, "percentage": 74.38, "elapsed_time": "2:10:27", "remaining_time": "0:44:56", "throughput": "0.00", "total_tokens": 0}
|
| 66 |
+
{"current_steps": 640, "total_steps": 847, "loss": 1.1507, "learning_rate": 8.79909569745987e-06, "epoch": 0.7553850693419888, "percentage": 75.56, "elapsed_time": "2:12:31", "remaining_time": "0:42:51", "throughput": "0.00", "total_tokens": 0}
|
| 67 |
+
{"current_steps": 650, "total_steps": 847, "loss": 1.3008, "learning_rate": 8.028089274818624e-06, "epoch": 0.7671879610504574, "percentage": 76.74, "elapsed_time": "2:14:36", "remaining_time": "0:40:47", "throughput": "0.00", "total_tokens": 0}
|
| 68 |
+
{"current_steps": 660, "total_steps": 847, "loss": 1.272, "learning_rate": 7.285927087085423e-06, "epoch": 0.778990852758926, "percentage": 77.92, "elapsed_time": "2:16:39", "remaining_time": "0:38:43", "throughput": "0.00", "total_tokens": 0}
|
| 69 |
+
{"current_steps": 670, "total_steps": 847, "loss": 1.1615, "learning_rate": 6.5738704595659065e-06, "epoch": 0.7907937444673945, "percentage": 79.1, "elapsed_time": "2:18:44", "remaining_time": "0:36:39", "throughput": "0.00", "total_tokens": 0}
|
| 70 |
+
{"current_steps": 680, "total_steps": 847, "loss": 1.1878, "learning_rate": 5.893129552327781e-06, "epoch": 0.8025966361758631, "percentage": 80.28, "elapsed_time": "2:20:48", "remaining_time": "0:34:34", "throughput": "0.00", "total_tokens": 0}
|
| 71 |
+
{"current_steps": 690, "total_steps": 847, "loss": 1.2436, "learning_rate": 5.244861303500026e-06, "epoch": 0.8143995278843317, "percentage": 81.46, "elapsed_time": "2:22:52", "remaining_time": "0:32:30", "throughput": "0.00", "total_tokens": 0}
|
| 72 |
+
{"current_steps": 700, "total_steps": 847, "loss": 1.0838, "learning_rate": 4.630167463024393e-06, "epoch": 0.8262024195928003, "percentage": 82.64, "elapsed_time": "2:24:55", "remaining_time": "0:30:26", "throughput": "0.00", "total_tokens": 0}
|
| 73 |
+
{"current_steps": 710, "total_steps": 847, "loss": 1.1495, "learning_rate": 4.050092720200638e-06, "epoch": 0.8380053113012688, "percentage": 83.83, "elapsed_time": "2:26:59", "remaining_time": "0:28:21", "throughput": "0.00", "total_tokens": 0}
|
| 74 |
+
{"current_steps": 720, "total_steps": 847, "loss": 1.234, "learning_rate": 3.5056229282080077e-06, "epoch": 0.8498082030097374, "percentage": 85.01, "elapsed_time": "2:29:04", "remaining_time": "0:26:17", "throughput": "0.00", "total_tokens": 0}
|
| 75 |
+
{"current_steps": 730, "total_steps": 847, "loss": 1.1803, "learning_rate": 2.997683428620296e-06, "epoch": 0.861611094718206, "percentage": 86.19, "elapsed_time": "2:31:08", "remaining_time": "0:24:13", "throughput": "0.00", "total_tokens": 0}
|
| 76 |
+
{"current_steps": 740, "total_steps": 847, "loss": 1.2197, "learning_rate": 2.527137478762037e-06, "epoch": 0.8734139864266746, "percentage": 87.37, "elapsed_time": "2:33:12", "remaining_time": "0:22:09", "throughput": "0.00", "total_tokens": 0}
|
| 77 |
+
{"current_steps": 750, "total_steps": 847, "loss": 1.2354, "learning_rate": 2.094784784578707e-06, "epoch": 0.8852168781351432, "percentage": 88.55, "elapsed_time": "2:35:17", "remaining_time": "0:20:05", "throughput": "0.00", "total_tokens": 0}
|
| 78 |
+
{"current_steps": 760, "total_steps": 847, "loss": 1.1835, "learning_rate": 1.7013601415141383e-06, "epoch": 0.8970197698436116, "percentage": 89.73, "elapsed_time": "2:37:22", "remaining_time": "0:18:00", "throughput": "0.00", "total_tokens": 0}
|
| 79 |
+
{"current_steps": 770, "total_steps": 847, "loss": 1.1651, "learning_rate": 1.3475321857052386e-06, "epoch": 0.9088226615520802, "percentage": 90.91, "elapsed_time": "2:39:26", "remaining_time": "0:15:56", "throughput": "0.00", "total_tokens": 0}
|
| 80 |
+
{"current_steps": 780, "total_steps": 847, "loss": 1.1662, "learning_rate": 1.03390225761624e-06, "epoch": 0.9206255532605488, "percentage": 92.09, "elapsed_time": "2:41:30", "remaining_time": "0:13:52", "throughput": "0.00", "total_tokens": 0}
|
| 81 |
+
{"current_steps": 790, "total_steps": 847, "loss": 1.1798, "learning_rate": 7.610033800438344e-07, "epoch": 0.9324284449690174, "percentage": 93.27, "elapsed_time": "2:43:34", "remaining_time": "0:11:48", "throughput": "0.00", "total_tokens": 0}
|
| 82 |
+
{"current_steps": 800, "total_steps": 847, "loss": 1.2053, "learning_rate": 5.292993522301005e-07, "epoch": 0.9442313366774859, "percentage": 94.45, "elapsed_time": "2:45:39", "remaining_time": "0:09:43", "throughput": "0.00", "total_tokens": 0}
|
| 83 |
+
{"current_steps": 810, "total_steps": 847, "loss": 1.2049, "learning_rate": 3.3918396162275214e-07, "epoch": 0.9560342283859545, "percentage": 95.63, "elapsed_time": "2:47:43", "remaining_time": "0:07:39", "throughput": "0.00", "total_tokens": 0}
|
| 84 |
+
{"current_steps": 820, "total_steps": 847, "loss": 1.2097, "learning_rate": 1.9098031462242705e-07, "epoch": 0.9678371200944231, "percentage": 96.81, "elapsed_time": "2:49:47", "remaining_time": "0:05:35", "throughput": "0.00", "total_tokens": 0}
|
| 85 |
+
{"current_steps": 830, "total_steps": 847, "loss": 1.2085, "learning_rate": 8.494028745434368e-08, "epoch": 0.9796400118028917, "percentage": 97.99, "elapsed_time": "2:51:51", "remaining_time": "0:03:31", "throughput": "0.00", "total_tokens": 0}
|
| 86 |
+
{"current_steps": 840, "total_steps": 847, "loss": 1.1854, "learning_rate": 2.124409809766692e-08, "epoch": 0.9914429035113603, "percentage": 99.17, "elapsed_time": "2:53:55", "remaining_time": "0:01:26", "throughput": "0.00", "total_tokens": 0}
|
| 87 |
+
{"current_steps": 847, "total_steps": 847, "epoch": 0.9997049277072882, "percentage": 100.0, "elapsed_time": "2:55:24", "remaining_time": "0:00:00", "throughput": "0.00", "total_tokens": 0}
|