| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 9.99995686494414, | |
| "global_step": 115910, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 5.172413793103448e-07, | |
| "loss": 2.1185, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00012931034482758618, | |
| "loss": 1.5115, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00025862068965517237, | |
| "loss": 1.2814, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003879310344827586, | |
| "loss": 1.2284, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0005172413793103447, | |
| "loss": 1.2049, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0005999990893103199, | |
| "loss": 1.1888, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0005999870030839344, | |
| "loss": 1.1754, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0005999608636610692, | |
| "loss": 1.1618, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0005999206722662517, | |
| "loss": 1.1522, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0005998664307822891, | |
| "loss": 1.1431, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0005997981417501787, | |
| "loss": 1.139, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0005997158083689897, | |
| "loss": 1.134, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0005996194344957133, | |
| "loss": 1.1267, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0005995090246450815, | |
| "loss": 1.1215, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0005993845839893562, | |
| "loss": 1.1198, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0005992461183580866, | |
| "loss": 1.1145, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0005990936342378362, | |
| "loss": 1.1097, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0005989271387718786, | |
| "loss": 1.1071, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0005987466397598635, | |
| "loss": 1.1049, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0005985521456574509, | |
| "loss": 1.1012, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0005983436655759147, | |
| "loss": 1.0981, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0005981212092817168, | |
| "loss": 1.0959, | |
| "step": 5250 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005978847871960486, | |
| "loss": 1.092, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005976344103943434, | |
| "loss": 1.0911, | |
| "step": 5750 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0005973700906057571, | |
| "loss": 1.0878, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0005970918402126193, | |
| "loss": 1.0856, | |
| "step": 6250 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.000596799672249853, | |
| "loss": 1.0818, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0005964936004043636, | |
| "loss": 1.0816, | |
| "step": 6750 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0005961736390143983, | |
| "loss": 1.0793, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0005958398030688742, | |
| "loss": 1.0774, | |
| "step": 7250 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0005954921082066758, | |
| "loss": 1.0738, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0005951305707159228, | |
| "loss": 1.0751, | |
| "step": 7750 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0005947552075332069, | |
| "loss": 1.0709, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0005943660362427983, | |
| "loss": 1.0707, | |
| "step": 8250 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0005939630750758225, | |
| "loss": 1.0682, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0005935463429094053, | |
| "loss": 1.0665, | |
| "step": 8750 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0005931158592657892, | |
| "loss": 1.0654, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0005926716443114187, | |
| "loss": 1.0635, | |
| "step": 9250 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.0005922137188559952, | |
| "loss": 1.0623, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 0.0005917421043515032, | |
| "loss": 1.0621, | |
| "step": 9750 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 0.0005912568228912035, | |
| "loss": 1.0601, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.0005907578972086006, | |
| "loss": 1.0587, | |
| "step": 10250 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.0005902453506763757, | |
| "loss": 1.057, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.0005897192073052928, | |
| "loss": 1.0566, | |
| "step": 10750 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 0.0005891794917430733, | |
| "loss": 1.0567, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.0005886262292732426, | |
| "loss": 1.0535, | |
| "step": 11250 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 0.0005880594458139437, | |
| "loss": 1.0534, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "eval_alliteration_score": 0.3529233565586186, | |
| "eval_harmonic_meter_score": 0.15390579451720435, | |
| "eval_harmonic_rhyme_score": 0.29888189056198133, | |
| "eval_meter_score": 0.37094404850825824, | |
| "eval_rhyme_score": 0.652105690320114, | |
| "eval_runtime": 1096.4036, | |
| "eval_samples_per_second": 2.463, | |
| "eval_steps_per_second": 0.308, | |
| "step": 11591 | |
| }, | |
| { | |
| "epoch": 1.01, | |
| "learning_rate": 0.0005874791679167249, | |
| "loss": 1.0499, | |
| "step": 11750 | |
| }, | |
| { | |
| "epoch": 1.04, | |
| "learning_rate": 0.0005868854227652948, | |
| "loss": 1.0418, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.0005862782381742493, | |
| "loss": 1.0432, | |
| "step": 12250 | |
| }, | |
| { | |
| "epoch": 1.08, | |
| "learning_rate": 0.0005856576425877683, | |
| "loss": 1.0423, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 1.1, | |
| "learning_rate": 0.0005850236650782842, | |
| "loss": 1.0433, | |
| "step": 12750 | |
| }, | |
| { | |
| "epoch": 1.12, | |
| "learning_rate": 0.0005843763353451184, | |
| "loss": 1.0416, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 1.14, | |
| "learning_rate": 0.0005837156837130911, | |
| "loss": 1.0425, | |
| "step": 13250 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.0005830417411311002, | |
| "loss": 1.0394, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.0005823545391706722, | |
| "loss": 1.0401, | |
| "step": 13750 | |
| }, | |
| { | |
| "epoch": 1.21, | |
| "learning_rate": 0.0005816541100244823, | |
| "loss": 1.0391, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 1.23, | |
| "learning_rate": 0.0005809404865048467, | |
| "loss": 1.0378, | |
| "step": 14250 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0005802137020421853, | |
| "loss": 1.0375, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 1.27, | |
| "learning_rate": 0.0005794737906834562, | |
| "loss": 1.0381, | |
| "step": 14750 | |
| }, | |
| { | |
| "epoch": 1.29, | |
| "learning_rate": 0.0005787207870905601, | |
| "loss": 1.0368, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 1.32, | |
| "learning_rate": 0.0005779547265387164, | |
| "loss": 1.0357, | |
| "step": 15250 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.0005771756449148116, | |
| "loss": 1.0336, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 1.36, | |
| "learning_rate": 0.0005763835787157171, | |
| "loss": 1.0354, | |
| "step": 15750 | |
| }, | |
| { | |
| "epoch": 1.38, | |
| "learning_rate": 0.0005755785650465805, | |
| "loss": 1.035, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0005747606416190861, | |
| "loss": 1.0331, | |
| "step": 16250 | |
| }, | |
| { | |
| "epoch": 1.42, | |
| "learning_rate": 0.0005739298467496894, | |
| "loss": 1.0329, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 1.45, | |
| "learning_rate": 0.0005730862193578216, | |
| "loss": 1.0316, | |
| "step": 16750 | |
| }, | |
| { | |
| "epoch": 1.47, | |
| "learning_rate": 0.0005722297989640667, | |
| "loss": 1.0333, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0005713606256883096, | |
| "loss": 1.0299, | |
| "step": 17250 | |
| }, | |
| { | |
| "epoch": 1.51, | |
| "learning_rate": 0.0005704787402478568, | |
| "loss": 1.0304, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 1.53, | |
| "learning_rate": 0.0005695841839555298, | |
| "loss": 1.0291, | |
| "step": 17750 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.000568676998717728, | |
| "loss": 1.0303, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 1.57, | |
| "learning_rate": 0.0005677572270324676, | |
| "loss": 1.0292, | |
| "step": 18250 | |
| }, | |
| { | |
| "epoch": 1.6, | |
| "learning_rate": 0.0005668249119873892, | |
| "loss": 1.0273, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 1.62, | |
| "learning_rate": 0.00056588009725774, | |
| "loss": 1.0277, | |
| "step": 18750 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 0.0005649228271043274, | |
| "loss": 1.0266, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 1.66, | |
| "learning_rate": 0.0005639531463714464, | |
| "loss": 1.0252, | |
| "step": 19250 | |
| }, | |
| { | |
| "epoch": 1.68, | |
| "learning_rate": 0.0005629711004847776, | |
| "loss": 1.0254, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 0.0005619767354492601, | |
| "loss": 1.0257, | |
| "step": 19750 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.0005609700978469361, | |
| "loss": 1.025, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 1.75, | |
| "learning_rate": 0.0005599512348347686, | |
| "loss": 1.0238, | |
| "step": 20250 | |
| }, | |
| { | |
| "epoch": 1.77, | |
| "learning_rate": 0.0005589201941424324, | |
| "loss": 1.0226, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.000557877024070078, | |
| "loss": 1.0232, | |
| "step": 20750 | |
| }, | |
| { | |
| "epoch": 1.81, | |
| "learning_rate": 0.0005568217734860691, | |
| "loss": 1.0233, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 0.0005557544918246936, | |
| "loss": 1.0206, | |
| "step": 21250 | |
| }, | |
| { | |
| "epoch": 1.85, | |
| "learning_rate": 0.0005546752290838471, | |
| "loss": 1.0211, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 1.88, | |
| "learning_rate": 0.0005535840358226909, | |
| "loss": 1.0221, | |
| "step": 21750 | |
| }, | |
| { | |
| "epoch": 1.9, | |
| "learning_rate": 0.0005524809631592841, | |
| "loss": 1.0207, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.0005513660627681884, | |
| "loss": 1.019, | |
| "step": 22250 | |
| }, | |
| { | |
| "epoch": 1.94, | |
| "learning_rate": 0.0005502393868780472, | |
| "loss": 1.0193, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 1.96, | |
| "learning_rate": 0.0005491009882691392, | |
| "loss": 1.019, | |
| "step": 22750 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.000547950920270906, | |
| "loss": 1.0172, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_alliteration_score": 0.42727272727272725, | |
| "eval_harmonic_meter_score": 0.20045849813764974, | |
| "eval_harmonic_rhyme_score": 0.5208998846302836, | |
| "eval_meter_score": 0.43620708634823174, | |
| "eval_rhyme_score": 0.8331009411550936, | |
| "eval_runtime": 1187.4498, | |
| "eval_samples_per_second": 2.274, | |
| "eval_steps_per_second": 0.285, | |
| "step": 23182 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0005467892367594536, | |
| "loss": 1.0168, | |
| "step": 23250 | |
| }, | |
| { | |
| "epoch": 2.03, | |
| "learning_rate": 0.0005456159921550282, | |
| "loss": 1.0074, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 2.05, | |
| "learning_rate": 0.0005444312414194676, | |
| "loss": 1.0084, | |
| "step": 23750 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.0005432350400536257, | |
| "loss": 1.0082, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 2.09, | |
| "learning_rate": 0.0005420274440947728, | |
| "loss": 1.0091, | |
| "step": 24250 | |
| }, | |
| { | |
| "epoch": 2.11, | |
| "learning_rate": 0.0005408085101139711, | |
| "loss": 1.0069, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 2.14, | |
| "learning_rate": 0.0005395782952134231, | |
| "loss": 1.0079, | |
| "step": 24750 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 0.0005383368570237981, | |
| "loss": 1.0084, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 2.18, | |
| "learning_rate": 0.0005370842537015316, | |
| "loss": 1.0084, | |
| "step": 25250 | |
| }, | |
| { | |
| "epoch": 2.2, | |
| "learning_rate": 0.0005358205439261011, | |
| "loss": 1.0084, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 0.0005345457868972771, | |
| "loss": 1.0078, | |
| "step": 25750 | |
| }, | |
| { | |
| "epoch": 2.24, | |
| "learning_rate": 0.0005332600423323502, | |
| "loss": 1.0082, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 2.26, | |
| "learning_rate": 0.0005319633704633331, | |
| "loss": 1.0064, | |
| "step": 26250 | |
| }, | |
| { | |
| "epoch": 2.29, | |
| "learning_rate": 0.0005306558320341392, | |
| "loss": 1.0069, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.0005293374882977369, | |
| "loss": 1.0067, | |
| "step": 26750 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.0005280084010132805, | |
| "loss": 1.0057, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 2.35, | |
| "learning_rate": 0.0005266686324432164, | |
| "loss": 1.0074, | |
| "step": 27250 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.000525318245350367, | |
| "loss": 1.0048, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 2.39, | |
| "learning_rate": 0.0005239573029949899, | |
| "loss": 1.0048, | |
| "step": 27750 | |
| }, | |
| { | |
| "epoch": 2.42, | |
| "learning_rate": 0.0005225858691318151, | |
| "loss": 1.0061, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 0.0005212040080070575, | |
| "loss": 1.0045, | |
| "step": 28250 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0005198117843554081, | |
| "loss": 1.0039, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 2.48, | |
| "learning_rate": 0.0005184092633970007, | |
| "loss": 1.0032, | |
| "step": 28750 | |
| }, | |
| { | |
| "epoch": 2.5, | |
| "learning_rate": 0.0005169965108343571, | |
| "loss": 1.0026, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.000515573592849309, | |
| "loss": 1.0038, | |
| "step": 29250 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.0005141405760998976, | |
| "loss": 1.0011, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 2.57, | |
| "learning_rate": 0.0005126975277172511, | |
| "loss": 1.0014, | |
| "step": 29750 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.0005112445153024395, | |
| "loss": 1.0017, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 2.61, | |
| "learning_rate": 0.0005097816069233083, | |
| "loss": 1.0012, | |
| "step": 30250 | |
| }, | |
| { | |
| "epoch": 2.63, | |
| "learning_rate": 0.0005083088711112896, | |
| "loss": 0.9992, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.0005068263768581912, | |
| "loss": 1.0003, | |
| "step": 30750 | |
| }, | |
| { | |
| "epoch": 2.67, | |
| "learning_rate": 0.0005053341936129654, | |
| "loss": 1.0, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 2.7, | |
| "learning_rate": 0.0005038323912784549, | |
| "loss": 0.9995, | |
| "step": 31250 | |
| }, | |
| { | |
| "epoch": 2.72, | |
| "learning_rate": 0.0005023210402081186, | |
| "loss": 0.9998, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 0.0005008002112027358, | |
| "loss": 0.9994, | |
| "step": 31750 | |
| }, | |
| { | |
| "epoch": 2.76, | |
| "learning_rate": 0.000499269975507089, | |
| "loss": 0.9982, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 2.78, | |
| "learning_rate": 0.0004977304048066266, | |
| "loss": 0.9975, | |
| "step": 32250 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 0.0004961815712241053, | |
| "loss": 0.9977, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0004946235473162105, | |
| "loss": 0.9982, | |
| "step": 32750 | |
| }, | |
| { | |
| "epoch": 2.85, | |
| "learning_rate": 0.0004930564060701579, | |
| "loss": 0.9968, | |
| "step": 33000 | |
| }, | |
| { | |
| "epoch": 2.87, | |
| "learning_rate": 0.0004914802209002742, | |
| "loss": 0.9981, | |
| "step": 33250 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.0004898950656445579, | |
| "loss": 0.9971, | |
| "step": 33500 | |
| }, | |
| { | |
| "epoch": 2.91, | |
| "learning_rate": 0.0004883010145612202, | |
| "loss": 0.9956, | |
| "step": 33750 | |
| }, | |
| { | |
| "epoch": 2.93, | |
| "learning_rate": 0.00048669814232520656, | |
| "loss": 0.9953, | |
| "step": 34000 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.0004850865240246984, | |
| "loss": 0.9946, | |
| "step": 34250 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.000483466235157595, | |
| "loss": 0.9945, | |
| "step": 34500 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "learning_rate": 0.00048183735162797754, | |
| "loss": 0.9952, | |
| "step": 34750 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_alliteration_score": 0.4316546762589928, | |
| "eval_harmonic_meter_score": 0.21624795235944666, | |
| "eval_harmonic_rhyme_score": 0.5329191903361169, | |
| "eval_meter_score": 0.4587688559480128, | |
| "eval_rhyme_score": 0.8467749122975762, | |
| "eval_runtime": 1157.9301, | |
| "eval_samples_per_second": 2.332, | |
| "eval_steps_per_second": 0.292, | |
| "step": 34773 | |
| }, | |
| { | |
| "epoch": 3.02, | |
| "learning_rate": 0.00048019994974255294, | |
| "loss": 0.9837, | |
| "step": 35000 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.0004785541062070789, | |
| "loss": 0.9825, | |
| "step": 35250 | |
| }, | |
| { | |
| "epoch": 3.06, | |
| "learning_rate": 0.00047689989812277095, | |
| "loss": 0.9844, | |
| "step": 35500 | |
| }, | |
| { | |
| "epoch": 3.08, | |
| "learning_rate": 0.00047523740298269045, | |
| "loss": 0.9843, | |
| "step": 35750 | |
| }, | |
| { | |
| "epoch": 3.11, | |
| "learning_rate": 0.0004735666986681142, | |
| "loss": 0.9834, | |
| "step": 36000 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 0.00047188786344488643, | |
| "loss": 0.983, | |
| "step": 36250 | |
| }, | |
| { | |
| "epoch": 3.15, | |
| "learning_rate": 0.0004702009759597517, | |
| "loss": 0.9841, | |
| "step": 36500 | |
| }, | |
| { | |
| "epoch": 3.17, | |
| "learning_rate": 0.0004685061152366711, | |
| "loss": 0.9836, | |
| "step": 36750 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 0.00046680336067312034, | |
| "loss": 0.9848, | |
| "step": 37000 | |
| }, | |
| { | |
| "epoch": 3.21, | |
| "learning_rate": 0.0004650927920363698, | |
| "loss": 0.9833, | |
| "step": 37250 | |
| }, | |
| { | |
| "epoch": 3.24, | |
| "learning_rate": 0.00046337448945974846, | |
| "loss": 0.9819, | |
| "step": 37500 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 0.00046164853343888914, | |
| "loss": 0.9827, | |
| "step": 37750 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0004599150048279585, | |
| "loss": 0.9835, | |
| "step": 38000 | |
| }, | |
| { | |
| "epoch": 3.3, | |
| "learning_rate": 0.0004581739848358684, | |
| "loss": 0.9822, | |
| "step": 38250 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 0.0004564255550224727, | |
| "loss": 0.9823, | |
| "step": 38500 | |
| }, | |
| { | |
| "epoch": 3.34, | |
| "learning_rate": 0.0004546697972947452, | |
| "loss": 0.981, | |
| "step": 38750 | |
| }, | |
| { | |
| "epoch": 3.36, | |
| "learning_rate": 0.00045290679390294386, | |
| "loss": 0.9815, | |
| "step": 39000 | |
| }, | |
| { | |
| "epoch": 3.39, | |
| "learning_rate": 0.0004511366274367566, | |
| "loss": 0.9817, | |
| "step": 39250 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0004493593808214336, | |
| "loss": 0.9814, | |
| "step": 39500 | |
| }, | |
| { | |
| "epoch": 3.43, | |
| "learning_rate": 0.0004475751373139011, | |
| "loss": 0.981, | |
| "step": 39750 | |
| }, | |
| { | |
| "epoch": 3.45, | |
| "learning_rate": 0.00044578398049886233, | |
| "loss": 0.9797, | |
| "step": 40000 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.00044398599428488114, | |
| "loss": 0.981, | |
| "step": 40250 | |
| }, | |
| { | |
| "epoch": 3.49, | |
| "learning_rate": 0.00044218126290045183, | |
| "loss": 0.9814, | |
| "step": 40500 | |
| }, | |
| { | |
| "epoch": 3.52, | |
| "learning_rate": 0.00044036987089005275, | |
| "loss": 0.9796, | |
| "step": 40750 | |
| }, | |
| { | |
| "epoch": 3.54, | |
| "learning_rate": 0.0004385519031101863, | |
| "loss": 0.9789, | |
| "step": 41000 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 0.0004367274447254036, | |
| "loss": 0.9791, | |
| "step": 41250 | |
| }, | |
| { | |
| "epoch": 3.58, | |
| "learning_rate": 0.0004348965812043143, | |
| "loss": 0.9782, | |
| "step": 41500 | |
| }, | |
| { | |
| "epoch": 3.6, | |
| "learning_rate": 0.00043305939831558366, | |
| "loss": 0.9779, | |
| "step": 41750 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0004312159821239138, | |
| "loss": 0.9787, | |
| "step": 42000 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 0.0004293664189860124, | |
| "loss": 0.9767, | |
| "step": 42250 | |
| }, | |
| { | |
| "epoch": 3.67, | |
| "learning_rate": 0.0004275107955465473, | |
| "loss": 0.9765, | |
| "step": 42500 | |
| }, | |
| { | |
| "epoch": 3.69, | |
| "learning_rate": 0.000425649198734087, | |
| "loss": 0.9756, | |
| "step": 42750 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 0.0004237817157570289, | |
| "loss": 0.9763, | |
| "step": 43000 | |
| }, | |
| { | |
| "epoch": 3.73, | |
| "learning_rate": 0.00042190843409951374, | |
| "loss": 0.9757, | |
| "step": 43250 | |
| }, | |
| { | |
| "epoch": 3.75, | |
| "learning_rate": 0.00042002944151732724, | |
| "loss": 0.9756, | |
| "step": 43500 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.00041814482603378945, | |
| "loss": 0.9735, | |
| "step": 43750 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.00041625467593563083, | |
| "loss": 0.9752, | |
| "step": 44000 | |
| }, | |
| { | |
| "epoch": 3.82, | |
| "learning_rate": 0.00041435907976885645, | |
| "loss": 0.9719, | |
| "step": 44250 | |
| }, | |
| { | |
| "epoch": 3.84, | |
| "learning_rate": 0.000412458126334598, | |
| "loss": 0.9736, | |
| "step": 44500 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.000410551904684954, | |
| "loss": 0.9745, | |
| "step": 44750 | |
| }, | |
| { | |
| "epoch": 3.88, | |
| "learning_rate": 0.0004086405041188175, | |
| "loss": 0.9739, | |
| "step": 45000 | |
| }, | |
| { | |
| "epoch": 3.9, | |
| "learning_rate": 0.0004067240141776935, | |
| "loss": 0.9714, | |
| "step": 45250 | |
| }, | |
| { | |
| "epoch": 3.93, | |
| "learning_rate": 0.0004048025246415039, | |
| "loss": 0.973, | |
| "step": 45500 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.0004028761255243818, | |
| "loss": 0.9724, | |
| "step": 45750 | |
| }, | |
| { | |
| "epoch": 3.97, | |
| "learning_rate": 0.00040094490707045426, | |
| "loss": 0.9722, | |
| "step": 46000 | |
| }, | |
| { | |
| "epoch": 3.99, | |
| "learning_rate": 0.0003990089597496154, | |
| "loss": 0.9716, | |
| "step": 46250 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_alliteration_score": 0.43804537521815007, | |
| "eval_harmonic_meter_score": 0.21317668443125748, | |
| "eval_harmonic_rhyme_score": 0.5018255179872321, | |
| "eval_meter_score": 0.45256004654409726, | |
| "eval_rhyme_score": 0.8339540595241286, | |
| "eval_runtime": 1179.1819, | |
| "eval_samples_per_second": 2.29, | |
| "eval_steps_per_second": 0.287, | |
| "step": 46364 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 0.0003970683742532878, | |
| "loss": 0.9648, | |
| "step": 46500 | |
| }, | |
| { | |
| "epoch": 4.03, | |
| "learning_rate": 0.00039512324149017415, | |
| "loss": 0.9568, | |
| "step": 46750 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 0.00039317365258199856, | |
| "loss": 0.9576, | |
| "step": 47000 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 0.0003912196988592375, | |
| "loss": 0.9586, | |
| "step": 47250 | |
| }, | |
| { | |
| "epoch": 4.1, | |
| "learning_rate": 0.0003892614718568419, | |
| "loss": 0.9585, | |
| "step": 47500 | |
| }, | |
| { | |
| "epoch": 4.12, | |
| "learning_rate": 0.0003872990633099488, | |
| "loss": 0.9601, | |
| "step": 47750 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 0.00038533256514958366, | |
| "loss": 0.9588, | |
| "step": 48000 | |
| }, | |
| { | |
| "epoch": 4.16, | |
| "learning_rate": 0.0003833620694983544, | |
| "loss": 0.9577, | |
| "step": 48250 | |
| }, | |
| { | |
| "epoch": 4.18, | |
| "learning_rate": 0.0003813876686661354, | |
| "loss": 0.9575, | |
| "step": 48500 | |
| }, | |
| { | |
| "epoch": 4.21, | |
| "learning_rate": 0.000379409455145743, | |
| "loss": 0.9585, | |
| "step": 48750 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 0.0003774275216086031, | |
| "loss": 0.959, | |
| "step": 49000 | |
| }, | |
| { | |
| "epoch": 4.25, | |
| "learning_rate": 0.0003754419609004093, | |
| "loss": 0.9574, | |
| "step": 49250 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 0.0003734528660367739, | |
| "loss": 0.9575, | |
| "step": 49500 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 0.0003714603301988702, | |
| "loss": 0.9561, | |
| "step": 49750 | |
| }, | |
| { | |
| "epoch": 4.31, | |
| "learning_rate": 0.00036946444672906754, | |
| "loss": 0.9581, | |
| "step": 50000 | |
| }, | |
| { | |
| "epoch": 4.34, | |
| "learning_rate": 0.00036746530912655837, | |
| "loss": 0.9569, | |
| "step": 50250 | |
| }, | |
| { | |
| "epoch": 4.36, | |
| "learning_rate": 0.0003654630110429785, | |
| "loss": 0.956, | |
| "step": 50500 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 0.0003634576462780193, | |
| "loss": 0.9559, | |
| "step": 50750 | |
| }, | |
| { | |
| "epoch": 4.4, | |
| "learning_rate": 0.0003614493087750344, | |
| "loss": 0.9559, | |
| "step": 51000 | |
| }, | |
| { | |
| "epoch": 4.42, | |
| "learning_rate": 0.00035943809261663816, | |
| "loss": 0.9537, | |
| "step": 51250 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 0.0003574240920202984, | |
| "loss": 0.9524, | |
| "step": 51500 | |
| }, | |
| { | |
| "epoch": 4.46, | |
| "learning_rate": 0.000355407401333923, | |
| "loss": 0.9553, | |
| "step": 51750 | |
| }, | |
| { | |
| "epoch": 4.49, | |
| "learning_rate": 0.0003533881150314395, | |
| "loss": 0.9539, | |
| "step": 52000 | |
| }, | |
| { | |
| "epoch": 4.51, | |
| "learning_rate": 0.00035136632770837013, | |
| "loss": 0.9529, | |
| "step": 52250 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 0.0003493421340773996, | |
| "loss": 0.9533, | |
| "step": 52500 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 0.000347315628963939, | |
| "loss": 0.9522, | |
| "step": 52750 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 0.00034528690730168256, | |
| "loss": 0.9525, | |
| "step": 53000 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 0.0003432560641281616, | |
| "loss": 0.9522, | |
| "step": 53250 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 0.00034122319458029137, | |
| "loss": 0.9513, | |
| "step": 53500 | |
| }, | |
| { | |
| "epoch": 4.64, | |
| "learning_rate": 0.0003391883938899151, | |
| "loss": 0.9512, | |
| "step": 53750 | |
| }, | |
| { | |
| "epoch": 4.66, | |
| "learning_rate": 0.00033715175737934175, | |
| "loss": 0.9495, | |
| "step": 54000 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 0.0003351133804568817, | |
| "loss": 0.9519, | |
| "step": 54250 | |
| }, | |
| { | |
| "epoch": 4.7, | |
| "learning_rate": 0.00033307335861237653, | |
| "loss": 0.9502, | |
| "step": 54500 | |
| }, | |
| { | |
| "epoch": 4.72, | |
| "learning_rate": 0.00033103178741272587, | |
| "loss": 0.9481, | |
| "step": 54750 | |
| }, | |
| { | |
| "epoch": 4.75, | |
| "learning_rate": 0.0003289887624974103, | |
| "loss": 0.9505, | |
| "step": 55000 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 0.0003269443795740114, | |
| "loss": 0.9481, | |
| "step": 55250 | |
| }, | |
| { | |
| "epoch": 4.79, | |
| "learning_rate": 0.00032489873441372786, | |
| "loss": 0.9488, | |
| "step": 55500 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 0.0003228519228468893, | |
| "loss": 0.9471, | |
| "step": 55750 | |
| }, | |
| { | |
| "epoch": 4.83, | |
| "learning_rate": 0.00032080404075846686, | |
| "loss": 0.9474, | |
| "step": 56000 | |
| }, | |
| { | |
| "epoch": 4.85, | |
| "learning_rate": 0.0003187551840835811, | |
| "loss": 0.9475, | |
| "step": 56250 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 0.0003167054488030084, | |
| "loss": 0.9469, | |
| "step": 56500 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 0.000314654930938684, | |
| "loss": 0.9456, | |
| "step": 56750 | |
| }, | |
| { | |
| "epoch": 4.92, | |
| "learning_rate": 0.00031260372654920427, | |
| "loss": 0.9449, | |
| "step": 57000 | |
| }, | |
| { | |
| "epoch": 4.94, | |
| "learning_rate": 0.00031055193172532656, | |
| "loss": 0.9451, | |
| "step": 57250 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 0.00030849964258546764, | |
| "loss": 0.9445, | |
| "step": 57500 | |
| }, | |
| { | |
| "epoch": 4.98, | |
| "learning_rate": 0.0003064469552712011, | |
| "loss": 0.9436, | |
| "step": 57750 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_alliteration_score": 0.448066935949221, | |
| "eval_harmonic_meter_score": 0.21530994103058235, | |
| "eval_harmonic_rhyme_score": 0.5565157424339893, | |
| "eval_meter_score": 0.4616261926927596, | |
| "eval_rhyme_score": 0.8542501571432405, | |
| "eval_runtime": 1173.576, | |
| "eval_samples_per_second": 2.301, | |
| "eval_steps_per_second": 0.288, | |
| "step": 57955 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "learning_rate": 0.00030439396594275327, | |
| "loss": 0.9417, | |
| "step": 58000 | |
| }, | |
| { | |
| "epoch": 5.03, | |
| "learning_rate": 0.00030234077077449876, | |
| "loss": 0.9273, | |
| "step": 58250 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 0.0003002874659504549, | |
| "loss": 0.9289, | |
| "step": 58500 | |
| }, | |
| { | |
| "epoch": 5.07, | |
| "learning_rate": 0.00029823414765977604, | |
| "loss": 0.9294, | |
| "step": 58750 | |
| }, | |
| { | |
| "epoch": 5.09, | |
| "learning_rate": 0.00029618091209224727, | |
| "loss": 0.9298, | |
| "step": 59000 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 0.0002941278554337787, | |
| "loss": 0.9286, | |
| "step": 59250 | |
| }, | |
| { | |
| "epoch": 5.13, | |
| "learning_rate": 0.00029207507386189873, | |
| "loss": 0.9288, | |
| "step": 59500 | |
| }, | |
| { | |
| "epoch": 5.15, | |
| "learning_rate": 0.0002900226635412495, | |
| "loss": 0.929, | |
| "step": 59750 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 0.0002879707206190815, | |
| "loss": 0.9284, | |
| "step": 60000 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 0.0002859193412207492, | |
| "loss": 0.9289, | |
| "step": 60250 | |
| }, | |
| { | |
| "epoch": 5.22, | |
| "learning_rate": 0.0002838686214452084, | |
| "loss": 0.9272, | |
| "step": 60500 | |
| }, | |
| { | |
| "epoch": 5.24, | |
| "learning_rate": 0.00028181865736051414, | |
| "loss": 0.9271, | |
| "step": 60750 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 0.0002797695449993205, | |
| "loss": 0.9268, | |
| "step": 61000 | |
| }, | |
| { | |
| "epoch": 5.28, | |
| "learning_rate": 0.0002777213803543814, | |
| "loss": 0.928, | |
| "step": 61250 | |
| }, | |
| { | |
| "epoch": 5.31, | |
| "learning_rate": 0.00027567425937405436, | |
| "loss": 0.9263, | |
| "step": 61500 | |
| }, | |
| { | |
| "epoch": 5.33, | |
| "learning_rate": 0.00027362827795780515, | |
| "loss": 0.9264, | |
| "step": 61750 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 0.0002715835319517157, | |
| "loss": 0.9259, | |
| "step": 62000 | |
| }, | |
| { | |
| "epoch": 5.37, | |
| "learning_rate": 0.0002695401171439936, | |
| "loss": 0.9253, | |
| "step": 62250 | |
| }, | |
| { | |
| "epoch": 5.39, | |
| "learning_rate": 0.0002674981292604852, | |
| "loss": 0.9244, | |
| "step": 62500 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 0.00026545766396019144, | |
| "loss": 0.9247, | |
| "step": 62750 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 0.0002634188168307859, | |
| "loss": 0.9255, | |
| "step": 63000 | |
| }, | |
| { | |
| "epoch": 5.46, | |
| "learning_rate": 0.00026138168338413746, | |
| "loss": 0.9238, | |
| "step": 63250 | |
| }, | |
| { | |
| "epoch": 5.48, | |
| "learning_rate": 0.00025934635905183577, | |
| "loss": 0.9228, | |
| "step": 63500 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 0.00025731293918072076, | |
| "loss": 0.9244, | |
| "step": 63750 | |
| }, | |
| { | |
| "epoch": 5.52, | |
| "learning_rate": 0.00025528151902841584, | |
| "loss": 0.9229, | |
| "step": 64000 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 0.00025325219375886554, | |
| "loss": 0.9231, | |
| "step": 64250 | |
| }, | |
| { | |
| "epoch": 5.56, | |
| "learning_rate": 0.0002512250584378775, | |
| "loss": 0.9222, | |
| "step": 64500 | |
| }, | |
| { | |
| "epoch": 5.59, | |
| "learning_rate": 0.00024920020802866925, | |
| "loss": 0.9221, | |
| "step": 64750 | |
| }, | |
| { | |
| "epoch": 5.61, | |
| "learning_rate": 0.00024717773738741866, | |
| "loss": 0.9218, | |
| "step": 65000 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 0.0002451577412588216, | |
| "loss": 0.9193, | |
| "step": 65250 | |
| }, | |
| { | |
| "epoch": 5.65, | |
| "learning_rate": 0.00024314031427165252, | |
| "loss": 0.9192, | |
| "step": 65500 | |
| }, | |
| { | |
| "epoch": 5.67, | |
| "learning_rate": 0.00024112555093433206, | |
| "loss": 0.9185, | |
| "step": 65750 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 0.0002391135456304992, | |
| "loss": 0.9183, | |
| "step": 66000 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 0.0002371043926145902, | |
| "loss": 0.9181, | |
| "step": 66250 | |
| }, | |
| { | |
| "epoch": 5.74, | |
| "learning_rate": 0.00023509818600742305, | |
| "loss": 0.9167, | |
| "step": 66500 | |
| }, | |
| { | |
| "epoch": 5.76, | |
| "learning_rate": 0.0002330950197917882, | |
| "loss": 0.9175, | |
| "step": 66750 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 0.00023109498780804545, | |
| "loss": 0.9162, | |
| "step": 67000 | |
| }, | |
| { | |
| "epoch": 5.8, | |
| "learning_rate": 0.000229098183749729, | |
| "loss": 0.9156, | |
| "step": 67250 | |
| }, | |
| { | |
| "epoch": 5.82, | |
| "learning_rate": 0.00022710470115915718, | |
| "loss": 0.9166, | |
| "step": 67500 | |
| }, | |
| { | |
| "epoch": 5.85, | |
| "learning_rate": 0.00022511463342305096, | |
| "loss": 0.9145, | |
| "step": 67750 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 0.0002231280737681587, | |
| "loss": 0.9136, | |
| "step": 68000 | |
| }, | |
| { | |
| "epoch": 5.89, | |
| "learning_rate": 0.0002211451152568894, | |
| "loss": 0.9133, | |
| "step": 68250 | |
| }, | |
| { | |
| "epoch": 5.91, | |
| "learning_rate": 0.00021916585078295267, | |
| "loss": 0.9126, | |
| "step": 68500 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 0.00021719037306700746, | |
| "loss": 0.9136, | |
| "step": 68750 | |
| }, | |
| { | |
| "epoch": 5.95, | |
| "learning_rate": 0.00021521877465231773, | |
| "loss": 0.9117, | |
| "step": 69000 | |
| }, | |
| { | |
| "epoch": 5.97, | |
| "learning_rate": 0.00021325114790041793, | |
| "loss": 0.9118, | |
| "step": 69250 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "learning_rate": 0.0002112875849867861, | |
| "loss": 0.9096, | |
| "step": 69500 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_alliteration_score": 0.467425320056899, | |
| "eval_harmonic_meter_score": 0.205647038467351, | |
| "eval_harmonic_rhyme_score": 0.5379127523573549, | |
| "eval_meter_score": 0.44894757325407986, | |
| "eval_rhyme_score": 0.8296400338920163, | |
| "eval_runtime": 1108.8436, | |
| "eval_samples_per_second": 2.435, | |
| "eval_steps_per_second": 0.305, | |
| "step": 69546 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 0.00020932817789652498, | |
| "loss": 0.8977, | |
| "step": 69750 | |
| }, | |
| { | |
| "epoch": 6.04, | |
| "learning_rate": 0.0002073730184200542, | |
| "loss": 0.8935, | |
| "step": 70000 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 0.0002054221981488093, | |
| "loss": 0.893, | |
| "step": 70250 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 0.00020347580847095156, | |
| "loss": 0.8944, | |
| "step": 70500 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 0.00020153394056708618, | |
| "loss": 0.8963, | |
| "step": 70750 | |
| }, | |
| { | |
| "epoch": 6.13, | |
| "learning_rate": 0.0001995966854059916, | |
| "loss": 0.8953, | |
| "step": 71000 | |
| }, | |
| { | |
| "epoch": 6.15, | |
| "learning_rate": 0.0001976641337403576, | |
| "loss": 0.8938, | |
| "step": 71250 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 0.0001957363761025339, | |
| "loss": 0.8928, | |
| "step": 71500 | |
| }, | |
| { | |
| "epoch": 6.19, | |
| "learning_rate": 0.0001938135028002889, | |
| "loss": 0.8931, | |
| "step": 71750 | |
| }, | |
| { | |
| "epoch": 6.21, | |
| "learning_rate": 0.00019189560391257962, | |
| "loss": 0.8941, | |
| "step": 72000 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 0.00018998276928533143, | |
| "loss": 0.8933, | |
| "step": 72250 | |
| }, | |
| { | |
| "epoch": 6.25, | |
| "learning_rate": 0.00018807508852722944, | |
| "loss": 0.8923, | |
| "step": 72500 | |
| }, | |
| { | |
| "epoch": 6.28, | |
| "learning_rate": 0.00018617265100552005, | |
| "loss": 0.8914, | |
| "step": 72750 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 0.00018427554584182545, | |
| "loss": 0.8903, | |
| "step": 73000 | |
| }, | |
| { | |
| "epoch": 6.32, | |
| "learning_rate": 0.0001823838619079677, | |
| "loss": 0.8923, | |
| "step": 73250 | |
| }, | |
| { | |
| "epoch": 6.34, | |
| "learning_rate": 0.00018049768782180627, | |
| "loss": 0.8893, | |
| "step": 73500 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 0.0001786171119430857, | |
| "loss": 0.8913, | |
| "step": 73750 | |
| }, | |
| { | |
| "epoch": 6.38, | |
| "learning_rate": 0.0001767422223692971, | |
| "loss": 0.8896, | |
| "step": 74000 | |
| }, | |
| { | |
| "epoch": 6.41, | |
| "learning_rate": 0.00017487310693155088, | |
| "loss": 0.8891, | |
| "step": 74250 | |
| }, | |
| { | |
| "epoch": 6.43, | |
| "learning_rate": 0.00017300985319046187, | |
| "loss": 0.8883, | |
| "step": 74500 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 0.00017115254843204803, | |
| "loss": 0.8891, | |
| "step": 74750 | |
| }, | |
| { | |
| "epoch": 6.47, | |
| "learning_rate": 0.0001693012796636411, | |
| "loss": 0.8876, | |
| "step": 75000 | |
| }, | |
| { | |
| "epoch": 6.49, | |
| "learning_rate": 0.0001674561336098108, | |
| "loss": 0.8879, | |
| "step": 75250 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 0.0001656171967083019, | |
| "loss": 0.8867, | |
| "step": 75500 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 0.00016378455510598517, | |
| "loss": 0.8857, | |
| "step": 75750 | |
| }, | |
| { | |
| "epoch": 6.56, | |
| "learning_rate": 0.00016195829465482172, | |
| "loss": 0.8843, | |
| "step": 76000 | |
| }, | |
| { | |
| "epoch": 6.58, | |
| "learning_rate": 0.00016013850090784126, | |
| "loss": 0.8848, | |
| "step": 76250 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 0.00015832525911513371, | |
| "loss": 0.8835, | |
| "step": 76500 | |
| }, | |
| { | |
| "epoch": 6.62, | |
| "learning_rate": 0.00015651865421985648, | |
| "loss": 0.8837, | |
| "step": 76750 | |
| }, | |
| { | |
| "epoch": 6.64, | |
| "learning_rate": 0.00015471877085425465, | |
| "loss": 0.8843, | |
| "step": 77000 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 0.0001529256933356964, | |
| "loss": 0.8826, | |
| "step": 77250 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 0.00015113950566272296, | |
| "loss": 0.8817, | |
| "step": 77500 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 0.00014936029151111374, | |
| "loss": 0.8817, | |
| "step": 77750 | |
| }, | |
| { | |
| "epoch": 6.73, | |
| "learning_rate": 0.00014758813422996664, | |
| "loss": 0.8811, | |
| "step": 78000 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 0.00014582311683779313, | |
| "loss": 0.8808, | |
| "step": 78250 | |
| }, | |
| { | |
| "epoch": 6.77, | |
| "learning_rate": 0.00014406532201862937, | |
| "loss": 0.8793, | |
| "step": 78500 | |
| }, | |
| { | |
| "epoch": 6.79, | |
| "learning_rate": 0.00014231483211816264, | |
| "loss": 0.8782, | |
| "step": 78750 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 0.00014057172913987414, | |
| "loss": 0.879, | |
| "step": 79000 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 0.00013883609474119667, | |
| "loss": 0.878, | |
| "step": 79250 | |
| }, | |
| { | |
| "epoch": 6.86, | |
| "learning_rate": 0.00013710801022969032, | |
| "loss": 0.8784, | |
| "step": 79500 | |
| }, | |
| { | |
| "epoch": 6.88, | |
| "learning_rate": 0.00013538755655923278, | |
| "loss": 0.8762, | |
| "step": 79750 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 0.00013367481432622743, | |
| "loss": 0.8747, | |
| "step": 80000 | |
| }, | |
| { | |
| "epoch": 6.92, | |
| "learning_rate": 0.00013196986376582716, | |
| "loss": 0.8744, | |
| "step": 80250 | |
| }, | |
| { | |
| "epoch": 6.95, | |
| "learning_rate": 0.0001302727847481763, | |
| "loss": 0.875, | |
| "step": 80500 | |
| }, | |
| { | |
| "epoch": 6.97, | |
| "learning_rate": 0.00012858365677466868, | |
| "loss": 0.8746, | |
| "step": 80750 | |
| }, | |
| { | |
| "epoch": 6.99, | |
| "learning_rate": 0.00012690255897422343, | |
| "loss": 0.8735, | |
| "step": 81000 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_alliteration_score": 0.4531559581092556, | |
| "eval_harmonic_meter_score": 0.2045466700855701, | |
| "eval_harmonic_rhyme_score": 0.5273458131288548, | |
| "eval_meter_score": 0.46399703611329074, | |
| "eval_rhyme_score": 0.8458613205511866, | |
| "eval_runtime": 1177.9262, | |
| "eval_samples_per_second": 2.292, | |
| "eval_steps_per_second": 0.287, | |
| "step": 81137 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 0.00012522957009957795, | |
| "loss": 0.8649, | |
| "step": 81250 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 0.00012356476852359884, | |
| "loss": 0.8542, | |
| "step": 81500 | |
| }, | |
| { | |
| "epoch": 7.05, | |
| "learning_rate": 0.00012190823223561035, | |
| "loss": 0.8522, | |
| "step": 81750 | |
| }, | |
| { | |
| "epoch": 7.07, | |
| "learning_rate": 0.00012026003883774089, | |
| "loss": 0.8536, | |
| "step": 82000 | |
| }, | |
| { | |
| "epoch": 7.1, | |
| "learning_rate": 0.00011862026554128754, | |
| "loss": 0.8549, | |
| "step": 82250 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 0.00011698898916309941, | |
| "loss": 0.855, | |
| "step": 82500 | |
| }, | |
| { | |
| "epoch": 7.14, | |
| "learning_rate": 0.00011536628612197879, | |
| "loss": 0.8553, | |
| "step": 82750 | |
| }, | |
| { | |
| "epoch": 7.16, | |
| "learning_rate": 0.00011375223243510097, | |
| "loss": 0.8548, | |
| "step": 83000 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 0.00011214690371445381, | |
| "loss": 0.8542, | |
| "step": 83250 | |
| }, | |
| { | |
| "epoch": 7.2, | |
| "learning_rate": 0.00011055037516329507, | |
| "loss": 0.8551, | |
| "step": 83500 | |
| }, | |
| { | |
| "epoch": 7.23, | |
| "learning_rate": 0.00010896272157262967, | |
| "loss": 0.8546, | |
| "step": 83750 | |
| }, | |
| { | |
| "epoch": 7.25, | |
| "learning_rate": 0.00010738401731770592, | |
| "loss": 0.853, | |
| "step": 84000 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 0.00010581433635453141, | |
| "loss": 0.8537, | |
| "step": 84250 | |
| }, | |
| { | |
| "epoch": 7.29, | |
| "learning_rate": 0.00010425375221640848, | |
| "loss": 0.8523, | |
| "step": 84500 | |
| }, | |
| { | |
| "epoch": 7.31, | |
| "learning_rate": 0.00010270233801048948, | |
| "loss": 0.8521, | |
| "step": 84750 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 0.0001011601664143517, | |
| "loss": 0.8522, | |
| "step": 85000 | |
| }, | |
| { | |
| "epoch": 7.35, | |
| "learning_rate": 9.962730967259322e-05, | |
| "loss": 0.8531, | |
| "step": 85250 | |
| }, | |
| { | |
| "epoch": 7.38, | |
| "learning_rate": 9.810383959344827e-05, | |
| "loss": 0.8509, | |
| "step": 85500 | |
| }, | |
| { | |
| "epoch": 7.4, | |
| "learning_rate": 9.65898275454233e-05, | |
| "loss": 0.8509, | |
| "step": 85750 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 9.508534445395351e-05, | |
| "loss": 0.8492, | |
| "step": 86000 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 9.359046079808056e-05, | |
| "loss": 0.8488, | |
| "step": 86250 | |
| }, | |
| { | |
| "epoch": 7.46, | |
| "learning_rate": 9.210524660715084e-05, | |
| "loss": 0.8498, | |
| "step": 86500 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 9.062977145753468e-05, | |
| "loss": 0.8501, | |
| "step": 86750 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 8.916410446936725e-05, | |
| "loss": 0.8499, | |
| "step": 87000 | |
| }, | |
| { | |
| "epoch": 7.53, | |
| "learning_rate": 8.770831430331029e-05, | |
| "loss": 0.8486, | |
| "step": 87250 | |
| }, | |
| { | |
| "epoch": 7.55, | |
| "learning_rate": 8.626246915733591e-05, | |
| "loss": 0.8475, | |
| "step": 87500 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 8.482663676353128e-05, | |
| "loss": 0.8469, | |
| "step": 87750 | |
| }, | |
| { | |
| "epoch": 7.59, | |
| "learning_rate": 8.340088438492646e-05, | |
| "loss": 0.8465, | |
| "step": 88000 | |
| }, | |
| { | |
| "epoch": 7.61, | |
| "learning_rate": 8.198527881234261e-05, | |
| "loss": 0.8454, | |
| "step": 88250 | |
| }, | |
| { | |
| "epoch": 7.64, | |
| "learning_rate": 8.057988636126368e-05, | |
| "loss": 0.8455, | |
| "step": 88500 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 7.918477286872927e-05, | |
| "loss": 0.8463, | |
| "step": 88750 | |
| }, | |
| { | |
| "epoch": 7.68, | |
| "learning_rate": 7.780000369025093e-05, | |
| "loss": 0.8439, | |
| "step": 89000 | |
| }, | |
| { | |
| "epoch": 7.7, | |
| "learning_rate": 7.642564369675017e-05, | |
| "loss": 0.8437, | |
| "step": 89250 | |
| }, | |
| { | |
| "epoch": 7.72, | |
| "learning_rate": 7.506175727151986e-05, | |
| "loss": 0.8444, | |
| "step": 89500 | |
| }, | |
| { | |
| "epoch": 7.74, | |
| "learning_rate": 7.370840830720743e-05, | |
| "loss": 0.8433, | |
| "step": 89750 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 7.236566020282263e-05, | |
| "loss": 0.8423, | |
| "step": 90000 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 7.103357586076716e-05, | |
| "loss": 0.8417, | |
| "step": 90250 | |
| }, | |
| { | |
| "epoch": 7.81, | |
| "learning_rate": 6.971221768388777e-05, | |
| "loss": 0.8428, | |
| "step": 90500 | |
| }, | |
| { | |
| "epoch": 7.83, | |
| "learning_rate": 6.840164757255298e-05, | |
| "loss": 0.8406, | |
| "step": 90750 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 6.710192692175355e-05, | |
| "loss": 0.84, | |
| "step": 91000 | |
| }, | |
| { | |
| "epoch": 7.87, | |
| "learning_rate": 6.581311661822627e-05, | |
| "loss": 0.8402, | |
| "step": 91250 | |
| }, | |
| { | |
| "epoch": 7.89, | |
| "learning_rate": 6.453527703760158e-05, | |
| "loss": 0.838, | |
| "step": 91500 | |
| }, | |
| { | |
| "epoch": 7.92, | |
| "learning_rate": 6.326846804157504e-05, | |
| "loss": 0.838, | |
| "step": 91750 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 6.201274897510344e-05, | |
| "loss": 0.8381, | |
| "step": 92000 | |
| }, | |
| { | |
| "epoch": 7.96, | |
| "learning_rate": 6.076817866362461e-05, | |
| "loss": 0.837, | |
| "step": 92250 | |
| }, | |
| { | |
| "epoch": 7.98, | |
| "learning_rate": 5.953481541030134e-05, | |
| "loss": 0.8385, | |
| "step": 92500 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_alliteration_score": 0.4492298916143754, | |
| "eval_harmonic_meter_score": 0.22527704736583123, | |
| "eval_harmonic_rhyme_score": 0.55956489384148, | |
| "eval_meter_score": 0.4760072853422334, | |
| "eval_rhyme_score": 0.8549378831695242, | |
| "eval_runtime": 1170.1941, | |
| "eval_samples_per_second": 2.307, | |
| "eval_steps_per_second": 0.289, | |
| "step": 92728 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 5.831271699329055e-05, | |
| "loss": 0.8367, | |
| "step": 92750 | |
| }, | |
| { | |
| "epoch": 8.02, | |
| "learning_rate": 5.7101940663036404e-05, | |
| "loss": 0.8191, | |
| "step": 93000 | |
| }, | |
| { | |
| "epoch": 8.05, | |
| "learning_rate": 5.590254313958858e-05, | |
| "loss": 0.8209, | |
| "step": 93250 | |
| }, | |
| { | |
| "epoch": 8.07, | |
| "learning_rate": 5.471458060994458e-05, | |
| "loss": 0.819, | |
| "step": 93500 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 5.353810872541822e-05, | |
| "loss": 0.8203, | |
| "step": 93750 | |
| }, | |
| { | |
| "epoch": 8.11, | |
| "learning_rate": 5.237318259903237e-05, | |
| "loss": 0.8185, | |
| "step": 94000 | |
| }, | |
| { | |
| "epoch": 8.13, | |
| "learning_rate": 5.121985680293711e-05, | |
| "loss": 0.8197, | |
| "step": 94250 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 5.007818536585306e-05, | |
| "loss": 0.8193, | |
| "step": 94500 | |
| }, | |
| { | |
| "epoch": 8.17, | |
| "learning_rate": 4.8948221770540686e-05, | |
| "loss": 0.8189, | |
| "step": 94750 | |
| }, | |
| { | |
| "epoch": 8.2, | |
| "learning_rate": 4.7830018951294724e-05, | |
| "loss": 0.8181, | |
| "step": 95000 | |
| }, | |
| { | |
| "epoch": 8.22, | |
| "learning_rate": 4.6723629291464336e-05, | |
| "loss": 0.8196, | |
| "step": 95250 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 4.5629104620999126e-05, | |
| "loss": 0.8184, | |
| "step": 95500 | |
| }, | |
| { | |
| "epoch": 8.26, | |
| "learning_rate": 4.4546496214021265e-05, | |
| "loss": 0.8189, | |
| "step": 95750 | |
| }, | |
| { | |
| "epoch": 8.28, | |
| "learning_rate": 4.347585478642348e-05, | |
| "loss": 0.817, | |
| "step": 96000 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 4.241723049349309e-05, | |
| "loss": 0.8205, | |
| "step": 96250 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 4.1370672927562575e-05, | |
| "loss": 0.8183, | |
| "step": 96500 | |
| }, | |
| { | |
| "epoch": 8.35, | |
| "learning_rate": 4.033623111568628e-05, | |
| "loss": 0.8176, | |
| "step": 96750 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 3.931395351734373e-05, | |
| "loss": 0.816, | |
| "step": 97000 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 3.830388802216939e-05, | |
| "loss": 0.8177, | |
| "step": 97250 | |
| }, | |
| { | |
| "epoch": 8.41, | |
| "learning_rate": 3.730608194770944e-05, | |
| "loss": 0.8171, | |
| "step": 97500 | |
| }, | |
| { | |
| "epoch": 8.43, | |
| "learning_rate": 3.6320582037205024e-05, | |
| "loss": 0.8153, | |
| "step": 97750 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 3.53474344574025e-05, | |
| "loss": 0.8172, | |
| "step": 98000 | |
| }, | |
| { | |
| "epoch": 8.48, | |
| "learning_rate": 3.438668479639057e-05, | |
| "loss": 0.8164, | |
| "step": 98250 | |
| }, | |
| { | |
| "epoch": 8.5, | |
| "learning_rate": 3.3438378061465e-05, | |
| "loss": 0.8147, | |
| "step": 98500 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 3.250255867701994e-05, | |
| "loss": 0.8145, | |
| "step": 98750 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 3.157927048246697e-05, | |
| "loss": 0.8158, | |
| "step": 99000 | |
| }, | |
| { | |
| "epoch": 8.56, | |
| "learning_rate": 3.0668556730181194e-05, | |
| "loss": 0.8133, | |
| "step": 99250 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 2.97704600834753e-05, | |
| "loss": 0.814, | |
| "step": 99500 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 2.8885022614600705e-05, | |
| "loss": 0.8133, | |
| "step": 99750 | |
| }, | |
| { | |
| "epoch": 8.63, | |
| "learning_rate": 2.8012285802776934e-05, | |
| "loss": 0.8146, | |
| "step": 100000 | |
| }, | |
| { | |
| "epoch": 8.65, | |
| "learning_rate": 2.7152290532248077e-05, | |
| "loss": 0.8121, | |
| "step": 100250 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 2.6305077090367855e-05, | |
| "loss": 0.8137, | |
| "step": 100500 | |
| }, | |
| { | |
| "epoch": 8.69, | |
| "learning_rate": 2.5470685165712322e-05, | |
| "loss": 0.8126, | |
| "step": 100750 | |
| }, | |
| { | |
| "epoch": 8.71, | |
| "learning_rate": 2.4649153846220405e-05, | |
| "loss": 0.8137, | |
| "step": 101000 | |
| }, | |
| { | |
| "epoch": 8.74, | |
| "learning_rate": 2.3840521617362807e-05, | |
| "loss": 0.8121, | |
| "step": 101250 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 2.304482636033934e-05, | |
| "loss": 0.8136, | |
| "step": 101500 | |
| }, | |
| { | |
| "epoch": 8.78, | |
| "learning_rate": 2.2262105350304206e-05, | |
| "loss": 0.8119, | |
| "step": 101750 | |
| }, | |
| { | |
| "epoch": 8.8, | |
| "learning_rate": 2.149239525461971e-05, | |
| "loss": 0.8121, | |
| "step": 102000 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 2.0735732131138694e-05, | |
| "loss": 0.8122, | |
| "step": 102250 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 1.9992151426515303e-05, | |
| "loss": 0.81, | |
| "step": 102500 | |
| }, | |
| { | |
| "epoch": 8.86, | |
| "learning_rate": 1.9261687974544515e-05, | |
| "loss": 0.8127, | |
| "step": 102750 | |
| }, | |
| { | |
| "epoch": 8.89, | |
| "learning_rate": 1.8544375994530026e-05, | |
| "loss": 0.8113, | |
| "step": 103000 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 1.7840249089681714e-05, | |
| "loss": 0.8091, | |
| "step": 103250 | |
| }, | |
| { | |
| "epoch": 8.93, | |
| "learning_rate": 1.7149340245541043e-05, | |
| "loss": 0.8137, | |
| "step": 103500 | |
| }, | |
| { | |
| "epoch": 8.95, | |
| "learning_rate": 1.64716818284361e-05, | |
| "loss": 0.8096, | |
| "step": 103750 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 1.5807305583965002e-05, | |
| "loss": 0.8089, | |
| "step": 104000 | |
| }, | |
| { | |
| "epoch": 8.99, | |
| "learning_rate": 1.5156242635509187e-05, | |
| "loss": 0.8101, | |
| "step": 104250 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_alliteration_score": 0.45457142857142857, | |
| "eval_harmonic_meter_score": 0.22300740426243393, | |
| "eval_harmonic_rhyme_score": 0.5252016141180921, | |
| "eval_meter_score": 0.47249164549890543, | |
| "eval_rhyme_score": 0.8439913924542158, | |
| "eval_runtime": 1174.1795, | |
| "eval_samples_per_second": 2.299, | |
| "eval_steps_per_second": 0.288, | |
| "step": 104319 | |
| }, | |
| { | |
| "epoch": 9.02, | |
| "learning_rate": 1.451852348277508e-05, | |
| "loss": 0.8024, | |
| "step": 104500 | |
| }, | |
| { | |
| "epoch": 9.04, | |
| "learning_rate": 1.3894178000365508e-05, | |
| "loss": 0.7977, | |
| "step": 104750 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.3283235436379947e-05, | |
| "loss": 0.7995, | |
| "step": 105000 | |
| }, | |
| { | |
| "epoch": 9.08, | |
| "learning_rate": 1.2685724411044718e-05, | |
| "loss": 0.7984, | |
| "step": 105250 | |
| }, | |
| { | |
| "epoch": 9.1, | |
| "learning_rate": 1.2101672915371896e-05, | |
| "loss": 0.7992, | |
| "step": 105500 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "learning_rate": 1.1531108309848192e-05, | |
| "loss": 0.799, | |
| "step": 105750 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 1.0974057323153274e-05, | |
| "loss": 0.7974, | |
| "step": 106000 | |
| }, | |
| { | |
| "epoch": 9.17, | |
| "learning_rate": 1.0430546050907607e-05, | |
| "loss": 0.7993, | |
| "step": 106250 | |
| }, | |
| { | |
| "epoch": 9.19, | |
| "learning_rate": 9.900599954449895e-06, | |
| "loss": 0.7971, | |
| "step": 106500 | |
| }, | |
| { | |
| "epoch": 9.21, | |
| "learning_rate": 9.38424385964437e-06, | |
| "loss": 0.7985, | |
| "step": 106750 | |
| }, | |
| { | |
| "epoch": 9.23, | |
| "learning_rate": 8.881501955717851e-06, | |
| "loss": 0.7973, | |
| "step": 107000 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 8.392397794126493e-06, | |
| "loss": 0.797, | |
| "step": 107250 | |
| }, | |
| { | |
| "epoch": 9.27, | |
| "learning_rate": 7.91695428745256e-06, | |
| "loss": 0.7977, | |
| "step": 107500 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 7.4551937083310195e-06, | |
| "loss": 0.7977, | |
| "step": 107750 | |
| }, | |
| { | |
| "epoch": 9.32, | |
| "learning_rate": 7.007137688406183e-06, | |
| "loss": 0.7967, | |
| "step": 108000 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 6.57280721731831e-06, | |
| "loss": 0.7969, | |
| "step": 108250 | |
| }, | |
| { | |
| "epoch": 9.36, | |
| "learning_rate": 6.152222641720361e-06, | |
| "loss": 0.7979, | |
| "step": 108500 | |
| }, | |
| { | |
| "epoch": 9.38, | |
| "learning_rate": 5.745403664324866e-06, | |
| "loss": 0.7972, | |
| "step": 108750 | |
| }, | |
| { | |
| "epoch": 9.4, | |
| "learning_rate": 5.352369342980855e-06, | |
| "loss": 0.7982, | |
| "step": 109000 | |
| }, | |
| { | |
| "epoch": 9.43, | |
| "learning_rate": 4.973138089781115e-06, | |
| "loss": 0.797, | |
| "step": 109250 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 4.607727670199734e-06, | |
| "loss": 0.7971, | |
| "step": 109500 | |
| }, | |
| { | |
| "epoch": 9.47, | |
| "learning_rate": 4.256155202259682e-06, | |
| "loss": 0.7982, | |
| "step": 109750 | |
| }, | |
| { | |
| "epoch": 9.49, | |
| "learning_rate": 3.918437155731036e-06, | |
| "loss": 0.7968, | |
| "step": 110000 | |
| }, | |
| { | |
| "epoch": 9.51, | |
| "learning_rate": 3.5945893513594116e-06, | |
| "loss": 0.7971, | |
| "step": 110250 | |
| }, | |
| { | |
| "epoch": 9.53, | |
| "learning_rate": 3.2846269601248474e-06, | |
| "loss": 0.7967, | |
| "step": 110500 | |
| }, | |
| { | |
| "epoch": 9.55, | |
| "learning_rate": 2.988564502530977e-06, | |
| "loss": 0.7973, | |
| "step": 110750 | |
| }, | |
| { | |
| "epoch": 9.58, | |
| "learning_rate": 2.706415847925003e-06, | |
| "loss": 0.796, | |
| "step": 111000 | |
| }, | |
| { | |
| "epoch": 9.6, | |
| "learning_rate": 2.4381942138477884e-06, | |
| "loss": 0.7959, | |
| "step": 111250 | |
| }, | |
| { | |
| "epoch": 9.62, | |
| "learning_rate": 2.1839121654147805e-06, | |
| "loss": 0.7966, | |
| "step": 111500 | |
| }, | |
| { | |
| "epoch": 9.64, | |
| "learning_rate": 1.9435816147273187e-06, | |
| "loss": 0.7982, | |
| "step": 111750 | |
| }, | |
| { | |
| "epoch": 9.66, | |
| "learning_rate": 1.7172138203147113e-06, | |
| "loss": 0.7974, | |
| "step": 112000 | |
| }, | |
| { | |
| "epoch": 9.68, | |
| "learning_rate": 1.5048193866066261e-06, | |
| "loss": 0.7962, | |
| "step": 112250 | |
| }, | |
| { | |
| "epoch": 9.71, | |
| "learning_rate": 1.306408263436487e-06, | |
| "loss": 0.796, | |
| "step": 112500 | |
| }, | |
| { | |
| "epoch": 9.73, | |
| "learning_rate": 1.1219897455753134e-06, | |
| "loss": 0.7969, | |
| "step": 112750 | |
| }, | |
| { | |
| "epoch": 9.75, | |
| "learning_rate": 9.515724722962692e-07, | |
| "loss": 0.7966, | |
| "step": 113000 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 7.95164426969952e-07, | |
| "loss": 0.7971, | |
| "step": 113250 | |
| }, | |
| { | |
| "epoch": 9.79, | |
| "learning_rate": 6.5277293669046e-07, | |
| "loss": 0.7964, | |
| "step": 113500 | |
| }, | |
| { | |
| "epoch": 9.81, | |
| "learning_rate": 5.244046719320661e-07, | |
| "loss": 0.7987, | |
| "step": 113750 | |
| }, | |
| { | |
| "epoch": 9.84, | |
| "learning_rate": 4.1006564623676794e-07, | |
| "loss": 0.7974, | |
| "step": 114000 | |
| }, | |
| { | |
| "epoch": 9.86, | |
| "learning_rate": 3.097612159326135e-07, | |
| "loss": 0.7968, | |
| "step": 114250 | |
| }, | |
| { | |
| "epoch": 9.88, | |
| "learning_rate": 2.2349607988273454e-07, | |
| "loss": 0.7967, | |
| "step": 114500 | |
| }, | |
| { | |
| "epoch": 9.9, | |
| "learning_rate": 1.512742792652233e-07, | |
| "loss": 0.7963, | |
| "step": 114750 | |
| }, | |
| { | |
| "epoch": 9.92, | |
| "learning_rate": 9.309919738381688e-08, | |
| "loss": 0.7958, | |
| "step": 115000 | |
| }, | |
| { | |
| "epoch": 9.94, | |
| "learning_rate": 4.897355950945758e-08, | |
| "loss": 0.7967, | |
| "step": 115250 | |
| }, | |
| { | |
| "epoch": 9.96, | |
| "learning_rate": 1.8899432752528255e-08, | |
| "loss": 0.7965, | |
| "step": 115500 | |
| }, | |
| { | |
| "epoch": 9.99, | |
| "learning_rate": 2.8782259661630503e-09, | |
| "loss": 0.7964, | |
| "step": 115750 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_alliteration_score": 0.44889892715979673, | |
| "eval_harmonic_meter_score": 0.22169202291201373, | |
| "eval_harmonic_rhyme_score": 0.5585708483881331, | |
| "eval_meter_score": 0.46651850277597545, | |
| "eval_rhyme_score": 0.8524697393236106, | |
| "eval_runtime": 1181.0934, | |
| "eval_samples_per_second": 2.286, | |
| "eval_steps_per_second": 0.286, | |
| "step": 115910 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "step": 115910, | |
| "total_flos": 1.446063607638917e+18, | |
| "train_loss": 0.9344727087942378, | |
| "train_runtime": 52307.177, | |
| "train_samples_per_second": 283.652, | |
| "train_steps_per_second": 2.216 | |
| } | |
| ], | |
| "max_steps": 115910, | |
| "num_train_epochs": 10, | |
| "total_flos": 1.446063607638917e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |