| { | |
| "best_metric": 0.18702100678171693, | |
| "best_model_checkpoint": "./en-xlsr/checkpoint-6000", | |
| "epoch": 30.0, | |
| "eval_steps": 600, | |
| "global_step": 6450, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002983555207517619, | |
| "loss": 4.4931, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 0.0002936570086139389, | |
| "loss": 2.9634, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0002889584964761159, | |
| "loss": 2.8686, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.00028425998433829287, | |
| "loss": 1.9458, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 2.33, | |
| "learning_rate": 0.0002795614722004698, | |
| "loss": 0.8991, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "learning_rate": 0.0002748629600626468, | |
| "loss": 0.6992, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 2.79, | |
| "eval_cer": 0.13698708949899893, | |
| "eval_loss": 0.4980902671813965, | |
| "eval_runtime": 22.423, | |
| "eval_samples_per_second": 77.51, | |
| "eval_steps_per_second": 4.861, | |
| "eval_wer": 0.33759717704140707, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 3.26, | |
| "learning_rate": 0.0002701644479248238, | |
| "loss": 0.5755, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 3.72, | |
| "learning_rate": 0.0002654659357870008, | |
| "loss": 0.502, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 4.19, | |
| "learning_rate": 0.00026076742364917777, | |
| "loss": 0.4539, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 0.0002560689115113547, | |
| "loss": 0.4015, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 5.12, | |
| "learning_rate": 0.0002513703993735317, | |
| "loss": 0.3877, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "learning_rate": 0.0002466718872357087, | |
| "loss": 0.3394, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 5.58, | |
| "eval_cer": 0.10568890525395255, | |
| "eval_loss": 0.393413245677948, | |
| "eval_runtime": 22.3286, | |
| "eval_samples_per_second": 77.838, | |
| "eval_steps_per_second": 4.882, | |
| "eval_wer": 0.24673319733142196, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 0.00024197337509788565, | |
| "loss": 0.336, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 0.00023727486296006264, | |
| "loss": 0.289, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 6.98, | |
| "learning_rate": 0.0002325763508222396, | |
| "loss": 0.2961, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 7.44, | |
| "learning_rate": 0.0002278778386844166, | |
| "loss": 0.2615, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 0.00022317932654659358, | |
| "loss": 0.2656, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "learning_rate": 0.00021848081440877052, | |
| "loss": 0.2376, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 8.37, | |
| "eval_cer": 0.10152348514486917, | |
| "eval_loss": 0.41225335001945496, | |
| "eval_runtime": 22.5992, | |
| "eval_samples_per_second": 76.905, | |
| "eval_steps_per_second": 4.823, | |
| "eval_wer": 0.2355957435077466, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 0.0002137823022709475, | |
| "loss": 0.227, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 9.3, | |
| "learning_rate": 0.00020908379013312447, | |
| "loss": 0.2208, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 9.77, | |
| "learning_rate": 0.00020438527799530146, | |
| "loss": 0.2109, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 10.23, | |
| "learning_rate": 0.00019968676585747845, | |
| "loss": 0.2003, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 10.7, | |
| "learning_rate": 0.00019498825371965542, | |
| "loss": 0.1948, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 11.16, | |
| "learning_rate": 0.0001902897415818324, | |
| "loss": 0.1877, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 11.16, | |
| "eval_cer": 0.09280141762363933, | |
| "eval_loss": 0.4269295632839203, | |
| "eval_runtime": 22.6943, | |
| "eval_samples_per_second": 76.583, | |
| "eval_steps_per_second": 4.803, | |
| "eval_wer": 0.21359651541048685, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 11.63, | |
| "learning_rate": 0.0001855912294440094, | |
| "loss": 0.1777, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 12.09, | |
| "learning_rate": 0.00018089271730618636, | |
| "loss": 0.1749, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 12.56, | |
| "learning_rate": 0.00017619420516836332, | |
| "loss": 0.1607, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 13.02, | |
| "learning_rate": 0.00017149569303054031, | |
| "loss": 0.1595, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 13.49, | |
| "learning_rate": 0.00016679718089271728, | |
| "loss": 0.1513, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 13.95, | |
| "learning_rate": 0.00016209866875489427, | |
| "loss": 0.1494, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 13.95, | |
| "eval_cer": 0.09215704324212368, | |
| "eval_loss": 0.4648122489452362, | |
| "eval_runtime": 22.977, | |
| "eval_samples_per_second": 75.641, | |
| "eval_steps_per_second": 4.744, | |
| "eval_wer": 0.21023322489937696, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 14.42, | |
| "learning_rate": 0.00015740015661707126, | |
| "loss": 0.1414, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 14.88, | |
| "learning_rate": 0.00015270164447924822, | |
| "loss": 0.1363, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 15.35, | |
| "learning_rate": 0.00014800313234142518, | |
| "loss": 0.1319, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 15.81, | |
| "learning_rate": 0.00014330462020360218, | |
| "loss": 0.1244, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 16.28, | |
| "learning_rate": 0.00013860610806577917, | |
| "loss": 0.1194, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 16.74, | |
| "learning_rate": 0.00013390759592795613, | |
| "loss": 0.1186, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 16.74, | |
| "eval_cer": 0.09190389616367109, | |
| "eval_loss": 0.48351120948791504, | |
| "eval_runtime": 22.4373, | |
| "eval_samples_per_second": 77.46, | |
| "eval_steps_per_second": 4.858, | |
| "eval_wer": 0.20576721618790317, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 17.21, | |
| "learning_rate": 0.00012920908379013312, | |
| "loss": 0.118, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 17.67, | |
| "learning_rate": 0.00012451057165231008, | |
| "loss": 0.1098, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 18.14, | |
| "learning_rate": 0.00011981205951448707, | |
| "loss": 0.1048, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 18.6, | |
| "learning_rate": 0.00011511354737666405, | |
| "loss": 0.1051, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 19.07, | |
| "learning_rate": 0.00011041503523884103, | |
| "loss": 0.1014, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 19.53, | |
| "learning_rate": 0.000105716523101018, | |
| "loss": 0.0966, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 19.53, | |
| "eval_cer": 0.08753135571767197, | |
| "eval_loss": 0.49859338998794556, | |
| "eval_runtime": 22.5733, | |
| "eval_samples_per_second": 76.994, | |
| "eval_steps_per_second": 4.829, | |
| "eval_wer": 0.19782764514528312, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "learning_rate": 0.00010101801096319498, | |
| "loss": 0.097, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 20.47, | |
| "learning_rate": 9.631949882537196e-05, | |
| "loss": 0.09, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 20.93, | |
| "learning_rate": 9.162098668754895e-05, | |
| "loss": 0.0904, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 21.4, | |
| "learning_rate": 8.692247454972591e-05, | |
| "loss": 0.0822, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 21.86, | |
| "learning_rate": 8.222396241190289e-05, | |
| "loss": 0.0869, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 22.33, | |
| "learning_rate": 7.752545027407988e-05, | |
| "loss": 0.083, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 22.33, | |
| "eval_cer": 0.08620808689848802, | |
| "eval_loss": 0.5179353952407837, | |
| "eval_runtime": 22.6454, | |
| "eval_samples_per_second": 76.748, | |
| "eval_steps_per_second": 4.813, | |
| "eval_wer": 0.192700005513591, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 22.79, | |
| "learning_rate": 7.282693813625684e-05, | |
| "loss": 0.0817, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 23.26, | |
| "learning_rate": 6.812842599843383e-05, | |
| "loss": 0.0774, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 23.72, | |
| "learning_rate": 6.34299138606108e-05, | |
| "loss": 0.0743, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 24.19, | |
| "learning_rate": 5.873140172278778e-05, | |
| "loss": 0.073, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 24.65, | |
| "learning_rate": 5.4032889584964754e-05, | |
| "loss": 0.0699, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 25.12, | |
| "learning_rate": 4.933437744714174e-05, | |
| "loss": 0.071, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 25.12, | |
| "eval_cer": 0.08569028605619865, | |
| "eval_loss": 0.5539018511772156, | |
| "eval_runtime": 22.6635, | |
| "eval_samples_per_second": 76.687, | |
| "eval_steps_per_second": 4.81, | |
| "eval_wer": 0.19082538457297238, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 25.58, | |
| "learning_rate": 4.463586530931871e-05, | |
| "loss": 0.0706, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 26.05, | |
| "learning_rate": 3.993735317149569e-05, | |
| "loss": 0.0696, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 26.51, | |
| "learning_rate": 3.523884103367267e-05, | |
| "loss": 0.0702, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 26.98, | |
| "learning_rate": 3.0540328895849644e-05, | |
| "loss": 0.0656, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 27.44, | |
| "learning_rate": 2.584181675802662e-05, | |
| "loss": 0.0662, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 27.91, | |
| "learning_rate": 2.1143304620203598e-05, | |
| "loss": 0.0648, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 27.91, | |
| "eval_cer": 0.08437852392239892, | |
| "eval_loss": 0.5582728981971741, | |
| "eval_runtime": 22.4539, | |
| "eval_samples_per_second": 77.403, | |
| "eval_steps_per_second": 4.854, | |
| "eval_wer": 0.18702100678171693, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 28.37, | |
| "learning_rate": 1.6444792482380578e-05, | |
| "loss": 0.0622, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 28.84, | |
| "learning_rate": 1.1746280344557556e-05, | |
| "loss": 0.0602, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 29.3, | |
| "learning_rate": 7.047768206734534e-06, | |
| "loss": 0.06, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 29.77, | |
| "learning_rate": 2.349256068911511e-06, | |
| "loss": 0.0615, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 30.0, | |
| "step": 6450, | |
| "total_flos": 4.472815672614715e+19, | |
| "train_loss": 0.3642665664539781, | |
| "train_runtime": 8377.9265, | |
| "train_samples_per_second": 49.251, | |
| "train_steps_per_second": 0.77 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 6450, | |
| "num_train_epochs": 30, | |
| "save_steps": 600, | |
| "total_flos": 4.472815672614715e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |