| { | |
| "best_metric": 25.731181077193543, | |
| "best_model_checkpoint": "/workspace/whisper/pretrain_base/checkpoint-20000", | |
| "epoch": 9.391727493917275, | |
| "global_step": 32880, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 2.425e-05, | |
| "loss": 1.3775, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 4.9250000000000004e-05, | |
| "loss": 0.7466, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 7.425e-05, | |
| "loss": 0.7189, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 9.925000000000001e-05, | |
| "loss": 0.7172, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00012425, | |
| "loss": 0.73, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00014925, | |
| "loss": 0.7847, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00017424999999999998, | |
| "loss": 0.7678, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00019925000000000002, | |
| "loss": 0.794, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00022425, | |
| "loss": 0.8324, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00024925, | |
| "loss": 0.8312, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00027425, | |
| "loss": 0.8452, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00029925000000000004, | |
| "loss": 0.8678, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00032425, | |
| "loss": 0.8898, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00034925, | |
| "loss": 0.8944, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00037425, | |
| "loss": 0.9126, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00039925000000000003, | |
| "loss": 0.9184, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00042425000000000004, | |
| "loss": 0.9372, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044925, | |
| "loss": 0.9781, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00047425, | |
| "loss": 0.9808, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00049925, | |
| "loss": 1.0286, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0004984617875647668, | |
| "loss": 0.9408, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0004968588082901555, | |
| "loss": 0.917, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0004952396373056995, | |
| "loss": 0.9306, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0004936204663212435, | |
| "loss": 0.9656, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0004920012953367875, | |
| "loss": 0.9078, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0004903821243523316, | |
| "loss": 0.8968, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0004887629533678756, | |
| "loss": 0.8644, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0004871437823834197, | |
| "loss": 0.9126, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0004855246113989638, | |
| "loss": 0.8631, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0004839054404145078, | |
| "loss": 0.8724, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00048228626943005185, | |
| "loss": 0.8416, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00048066709844559586, | |
| "loss": 0.8013, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0004790479274611399, | |
| "loss": 0.7938, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00047742875647668393, | |
| "loss": 0.8173, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.000475809585492228, | |
| "loss": 0.8158, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.000474190414507772, | |
| "loss": 0.7718, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004725712435233161, | |
| "loss": 0.7731, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00047095207253886013, | |
| "loss": 0.7525, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00046933290155440414, | |
| "loss": 0.7354, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0004677137305699482, | |
| "loss": 0.7298, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "eval_loss": 0.7294211983680725, | |
| "eval_runtime": 497.9031, | |
| "eval_samples_per_second": 0.532, | |
| "eval_steps_per_second": 0.034, | |
| "eval_wer": 62.55393958766182, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0004660945595854922, | |
| "loss": 0.7686, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0004644753886010363, | |
| "loss": 0.7231, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0004628562176165803, | |
| "loss": 0.7244, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00046123704663212435, | |
| "loss": 0.7147, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00045963406735751296, | |
| "loss": 0.7183, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00045801489637305697, | |
| "loss": 0.6829, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 0.00045641191709844557, | |
| "loss": 0.6598, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 0.00045479274611398963, | |
| "loss": 0.6828, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 0.00045317357512953364, | |
| "loss": 0.682, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 0.00045155440414507776, | |
| "loss": 0.6589, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 0.00044993523316062177, | |
| "loss": 0.6531, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 0.00044831606217616584, | |
| "loss": 0.6384, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 0.00044669689119170985, | |
| "loss": 0.6346, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 1.03, | |
| "learning_rate": 0.0004450777202072539, | |
| "loss": 0.558, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 1.06, | |
| "learning_rate": 0.0004434585492227979, | |
| "loss": 0.5664, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 1.09, | |
| "learning_rate": 0.000441839378238342, | |
| "loss": 0.5653, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 1.13, | |
| "learning_rate": 0.000440220207253886, | |
| "loss": 0.5293, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 1.16, | |
| "learning_rate": 0.00043860103626943006, | |
| "loss": 0.5908, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 1.19, | |
| "learning_rate": 0.00043698186528497407, | |
| "loss": 0.5657, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 0.0004353626943005182, | |
| "loss": 0.5672, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 1.25, | |
| "learning_rate": 0.0004337435233160622, | |
| "loss": 0.5844, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 1.28, | |
| "learning_rate": 0.00043212435233160626, | |
| "loss": 0.549, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 1.31, | |
| "learning_rate": 0.00043050518134715027, | |
| "loss": 0.556, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 1.34, | |
| "learning_rate": 0.00042888601036269433, | |
| "loss": 0.5631, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 1.37, | |
| "learning_rate": 0.00042726683937823834, | |
| "loss": 0.5604, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 1.4, | |
| "learning_rate": 0.0004256476683937824, | |
| "loss": 0.5169, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 1.43, | |
| "learning_rate": 0.0004240284974093264, | |
| "loss": 0.5497, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 1.46, | |
| "learning_rate": 0.0004224093264248705, | |
| "loss": 0.5315, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 1.49, | |
| "learning_rate": 0.0004207901554404145, | |
| "loss": 0.5453, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 0.0004191709844559586, | |
| "loss": 0.5372, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 1.55, | |
| "learning_rate": 0.0004175518134715026, | |
| "loss": 0.5304, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 1.58, | |
| "learning_rate": 0.0004159326424870467, | |
| "loss": 0.5226, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 1.61, | |
| "learning_rate": 0.0004143134715025907, | |
| "loss": 0.5347, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 1.64, | |
| "learning_rate": 0.0004126943005181347, | |
| "loss": 0.5393, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 1.67, | |
| "learning_rate": 0.00041107512953367876, | |
| "loss": 0.505, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 1.7, | |
| "learning_rate": 0.00040945595854922277, | |
| "loss": 0.5213, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 1.73, | |
| "learning_rate": 0.00040783678756476684, | |
| "loss": 0.5188, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 1.76, | |
| "learning_rate": 0.00040621761658031085, | |
| "loss": 0.5056, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 1.79, | |
| "learning_rate": 0.00040459844559585496, | |
| "loss": 0.5106, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "learning_rate": 0.000402979274611399, | |
| "loss": 0.5348, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.82, | |
| "eval_loss": 0.5500301718711853, | |
| "eval_runtime": 455.7092, | |
| "eval_samples_per_second": 0.582, | |
| "eval_steps_per_second": 0.037, | |
| "eval_wer": 46.73166054019498, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 1.86, | |
| "learning_rate": 0.00040136010362694304, | |
| "loss": 0.5169, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 1.89, | |
| "learning_rate": 0.00039974093264248705, | |
| "loss": 0.5017, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 1.92, | |
| "learning_rate": 0.0003981217616580311, | |
| "loss": 0.5438, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 1.95, | |
| "learning_rate": 0.0003965025906735751, | |
| "loss": 0.4846, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 1.98, | |
| "learning_rate": 0.0003948834196891192, | |
| "loss": 0.5027, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 2.01, | |
| "learning_rate": 0.0003932642487046632, | |
| "loss": 0.4913, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 2.04, | |
| "learning_rate": 0.0003916612694300518, | |
| "loss": 0.393, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 2.07, | |
| "learning_rate": 0.00039004209844559586, | |
| "loss": 0.3817, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 2.1, | |
| "learning_rate": 0.00038842292746113987, | |
| "loss": 0.3933, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 0.00038680375647668394, | |
| "loss": 0.4169, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 2.16, | |
| "learning_rate": 0.000385184585492228, | |
| "loss": 0.4097, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 2.19, | |
| "learning_rate": 0.00038356541450777206, | |
| "loss": 0.4028, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 2.22, | |
| "learning_rate": 0.0003819462435233161, | |
| "loss": 0.409, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 2.25, | |
| "learning_rate": 0.00038032707253886014, | |
| "loss": 0.4108, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 2.28, | |
| "learning_rate": 0.00037870790155440415, | |
| "loss": 0.4117, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 2.31, | |
| "learning_rate": 0.0003770887305699482, | |
| "loss": 0.3998, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 2.34, | |
| "learning_rate": 0.0003754695595854922, | |
| "loss": 0.4065, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 2.37, | |
| "learning_rate": 0.0003738503886010363, | |
| "loss": 0.4243, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 2.4, | |
| "learning_rate": 0.0003722312176165803, | |
| "loss": 0.4074, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 2.43, | |
| "learning_rate": 0.00037061204663212436, | |
| "loss": 0.4109, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 2.46, | |
| "learning_rate": 0.0003689928756476684, | |
| "loss": 0.4029, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 2.49, | |
| "learning_rate": 0.0003673737046632125, | |
| "loss": 0.388, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 2.52, | |
| "learning_rate": 0.0003657545336787565, | |
| "loss": 0.4309, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 2.55, | |
| "learning_rate": 0.0003641353626943005, | |
| "loss": 0.4171, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 2.59, | |
| "learning_rate": 0.00036251619170984457, | |
| "loss": 0.4115, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 2.62, | |
| "learning_rate": 0.0003608970207253886, | |
| "loss": 0.4077, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 2.65, | |
| "learning_rate": 0.00035927784974093264, | |
| "loss": 0.3923, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 2.68, | |
| "learning_rate": 0.00035765867875647665, | |
| "loss": 0.4002, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 2.71, | |
| "learning_rate": 0.0003560395077720207, | |
| "loss": 0.4078, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 0.0003544203367875648, | |
| "loss": 0.4, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 2.77, | |
| "learning_rate": 0.00035280116580310884, | |
| "loss": 0.401, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 2.8, | |
| "learning_rate": 0.00035118199481865285, | |
| "loss": 0.3729, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 2.83, | |
| "learning_rate": 0.0003495628238341969, | |
| "loss": 0.402, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 2.86, | |
| "learning_rate": 0.00034794365284974093, | |
| "loss": 0.3933, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 2.89, | |
| "learning_rate": 0.000346324481865285, | |
| "loss": 0.397, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 2.92, | |
| "learning_rate": 0.000344705310880829, | |
| "loss": 0.3834, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 2.95, | |
| "learning_rate": 0.00034308613989637307, | |
| "loss": 0.3744, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 2.98, | |
| "learning_rate": 0.0003414669689119171, | |
| "loss": 0.3905, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 3.01, | |
| "learning_rate": 0.00033984779792746114, | |
| "loss": 0.3626, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "learning_rate": 0.0003382286269430052, | |
| "loss": 0.2828, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 3.04, | |
| "eval_loss": 0.5047640800476074, | |
| "eval_runtime": 460.4394, | |
| "eval_samples_per_second": 0.576, | |
| "eval_steps_per_second": 0.037, | |
| "eval_wer": 34.3455330030366, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 3.07, | |
| "learning_rate": 0.00033660945595854927, | |
| "loss": 0.2933, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 3.1, | |
| "learning_rate": 0.0003349902849740933, | |
| "loss": 0.2943, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 3.13, | |
| "learning_rate": 0.00033337111398963734, | |
| "loss": 0.2911, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 3.16, | |
| "learning_rate": 0.00033175194300518135, | |
| "loss": 0.2928, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 3.19, | |
| "learning_rate": 0.0003301327720207254, | |
| "loss": 0.2983, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 3.22, | |
| "learning_rate": 0.0003285136010362694, | |
| "loss": 0.2919, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 3.25, | |
| "learning_rate": 0.0003268944300518135, | |
| "loss": 0.3078, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 3.28, | |
| "learning_rate": 0.0003252752590673575, | |
| "loss": 0.3201, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 3.32, | |
| "learning_rate": 0.00032365608808290156, | |
| "loss": 0.3012, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 0.0003220369170984456, | |
| "loss": 0.312, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 3.38, | |
| "learning_rate": 0.00032041774611398963, | |
| "loss": 0.3063, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 3.41, | |
| "learning_rate": 0.0003187985751295337, | |
| "loss": 0.3139, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 3.44, | |
| "learning_rate": 0.0003171794041450777, | |
| "loss": 0.2951, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 3.47, | |
| "learning_rate": 0.00031556023316062177, | |
| "loss": 0.2936, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 3.5, | |
| "learning_rate": 0.0003139410621761658, | |
| "loss": 0.3005, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 3.53, | |
| "learning_rate": 0.00031232189119170985, | |
| "loss": 0.3074, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 3.56, | |
| "learning_rate": 0.00031070272020725386, | |
| "loss": 0.3124, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 3.59, | |
| "learning_rate": 0.0003090835492227979, | |
| "loss": 0.3213, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 3.62, | |
| "learning_rate": 0.0003074805699481865, | |
| "loss": 0.305, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 3.65, | |
| "learning_rate": 0.00030586139896373053, | |
| "loss": 0.3053, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 3.68, | |
| "learning_rate": 0.00030424222797927465, | |
| "loss": 0.2989, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 3.71, | |
| "learning_rate": 0.00030262305699481866, | |
| "loss": 0.3046, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 3.74, | |
| "learning_rate": 0.0003010038860103627, | |
| "loss": 0.2977, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 3.77, | |
| "learning_rate": 0.00029938471502590673, | |
| "loss": 0.3002, | |
| "step": 14400 | |
| }, | |
| { | |
| "epoch": 3.8, | |
| "learning_rate": 0.0002977655440414508, | |
| "loss": 0.318, | |
| "step": 14500 | |
| }, | |
| { | |
| "epoch": 3.83, | |
| "learning_rate": 0.0002961463730569948, | |
| "loss": 0.3133, | |
| "step": 14600 | |
| }, | |
| { | |
| "epoch": 3.86, | |
| "learning_rate": 0.00029452720207253887, | |
| "loss": 0.3196, | |
| "step": 14700 | |
| }, | |
| { | |
| "epoch": 3.89, | |
| "learning_rate": 0.0002929080310880829, | |
| "loss": 0.3131, | |
| "step": 14800 | |
| }, | |
| { | |
| "epoch": 3.92, | |
| "learning_rate": 0.00029128886010362695, | |
| "loss": 0.3102, | |
| "step": 14900 | |
| }, | |
| { | |
| "epoch": 3.95, | |
| "learning_rate": 0.00028966968911917095, | |
| "loss": 0.3007, | |
| "step": 15000 | |
| }, | |
| { | |
| "epoch": 3.98, | |
| "learning_rate": 0.0002880505181347151, | |
| "loss": 0.2975, | |
| "step": 15100 | |
| }, | |
| { | |
| "epoch": 4.01, | |
| "learning_rate": 0.0002864313471502591, | |
| "loss": 0.2461, | |
| "step": 15200 | |
| }, | |
| { | |
| "epoch": 4.05, | |
| "learning_rate": 0.00028481217616580315, | |
| "loss": 0.1983, | |
| "step": 15300 | |
| }, | |
| { | |
| "epoch": 4.08, | |
| "learning_rate": 0.00028319300518134716, | |
| "loss": 0.2076, | |
| "step": 15400 | |
| }, | |
| { | |
| "epoch": 4.11, | |
| "learning_rate": 0.0002815738341968912, | |
| "loss": 0.2054, | |
| "step": 15500 | |
| }, | |
| { | |
| "epoch": 4.14, | |
| "learning_rate": 0.00027995466321243523, | |
| "loss": 0.214, | |
| "step": 15600 | |
| }, | |
| { | |
| "epoch": 4.17, | |
| "learning_rate": 0.0002783354922279793, | |
| "loss": 0.2259, | |
| "step": 15700 | |
| }, | |
| { | |
| "epoch": 4.2, | |
| "learning_rate": 0.0002767163212435233, | |
| "loss": 0.2121, | |
| "step": 15800 | |
| }, | |
| { | |
| "epoch": 4.23, | |
| "learning_rate": 0.00027509715025906737, | |
| "loss": 0.2172, | |
| "step": 15900 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "learning_rate": 0.0002734779792746114, | |
| "loss": 0.2182, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 4.26, | |
| "eval_loss": 0.47591984272003174, | |
| "eval_runtime": 430.927, | |
| "eval_samples_per_second": 0.615, | |
| "eval_steps_per_second": 0.039, | |
| "eval_wer": 27.457247882371743, | |
| "step": 16000 | |
| }, | |
| { | |
| "epoch": 4.29, | |
| "learning_rate": 0.0002718588082901555, | |
| "loss": 0.213, | |
| "step": 16100 | |
| }, | |
| { | |
| "epoch": 4.32, | |
| "learning_rate": 0.0002702396373056995, | |
| "loss": 0.2187, | |
| "step": 16200 | |
| }, | |
| { | |
| "epoch": 4.35, | |
| "learning_rate": 0.0002686366580310881, | |
| "loss": 0.2357, | |
| "step": 16300 | |
| }, | |
| { | |
| "epoch": 4.38, | |
| "learning_rate": 0.0002670174870466322, | |
| "loss": 0.2168, | |
| "step": 16400 | |
| }, | |
| { | |
| "epoch": 4.41, | |
| "learning_rate": 0.0002653983160621762, | |
| "loss": 0.2099, | |
| "step": 16500 | |
| }, | |
| { | |
| "epoch": 4.44, | |
| "learning_rate": 0.0002637791450777202, | |
| "loss": 0.2252, | |
| "step": 16600 | |
| }, | |
| { | |
| "epoch": 4.47, | |
| "learning_rate": 0.00026215997409326426, | |
| "loss": 0.2218, | |
| "step": 16700 | |
| }, | |
| { | |
| "epoch": 4.5, | |
| "learning_rate": 0.00026054080310880827, | |
| "loss": 0.2186, | |
| "step": 16800 | |
| }, | |
| { | |
| "epoch": 4.53, | |
| "learning_rate": 0.00025892163212435233, | |
| "loss": 0.2165, | |
| "step": 16900 | |
| }, | |
| { | |
| "epoch": 4.56, | |
| "learning_rate": 0.00025730246113989634, | |
| "loss": 0.2275, | |
| "step": 17000 | |
| }, | |
| { | |
| "epoch": 4.59, | |
| "learning_rate": 0.0002556832901554404, | |
| "loss": 0.2414, | |
| "step": 17100 | |
| }, | |
| { | |
| "epoch": 4.62, | |
| "learning_rate": 0.00025406411917098447, | |
| "loss": 0.2358, | |
| "step": 17200 | |
| }, | |
| { | |
| "epoch": 4.65, | |
| "learning_rate": 0.00025244494818652853, | |
| "loss": 0.226, | |
| "step": 17300 | |
| }, | |
| { | |
| "epoch": 4.68, | |
| "learning_rate": 0.00025082577720207254, | |
| "loss": 0.2181, | |
| "step": 17400 | |
| }, | |
| { | |
| "epoch": 4.71, | |
| "learning_rate": 0.0002492066062176166, | |
| "loss": 0.226, | |
| "step": 17500 | |
| }, | |
| { | |
| "epoch": 4.74, | |
| "learning_rate": 0.0002475874352331606, | |
| "loss": 0.2275, | |
| "step": 17600 | |
| }, | |
| { | |
| "epoch": 4.77, | |
| "learning_rate": 0.0002459682642487047, | |
| "loss": 0.2238, | |
| "step": 17700 | |
| }, | |
| { | |
| "epoch": 4.81, | |
| "learning_rate": 0.0002443490932642487, | |
| "loss": 0.2139, | |
| "step": 17800 | |
| }, | |
| { | |
| "epoch": 4.84, | |
| "learning_rate": 0.00024272992227979275, | |
| "loss": 0.2268, | |
| "step": 17900 | |
| }, | |
| { | |
| "epoch": 4.87, | |
| "learning_rate": 0.0002411107512953368, | |
| "loss": 0.2272, | |
| "step": 18000 | |
| }, | |
| { | |
| "epoch": 4.9, | |
| "learning_rate": 0.00023949158031088083, | |
| "loss": 0.2203, | |
| "step": 18100 | |
| }, | |
| { | |
| "epoch": 4.93, | |
| "learning_rate": 0.00023787240932642486, | |
| "loss": 0.2316, | |
| "step": 18200 | |
| }, | |
| { | |
| "epoch": 4.96, | |
| "learning_rate": 0.0002362532383419689, | |
| "loss": 0.2238, | |
| "step": 18300 | |
| }, | |
| { | |
| "epoch": 4.99, | |
| "learning_rate": 0.00023463406735751296, | |
| "loss": 0.2134, | |
| "step": 18400 | |
| }, | |
| { | |
| "epoch": 5.02, | |
| "learning_rate": 0.000233014896373057, | |
| "loss": 0.1554, | |
| "step": 18500 | |
| }, | |
| { | |
| "epoch": 5.05, | |
| "learning_rate": 0.00023139572538860104, | |
| "loss": 0.135, | |
| "step": 18600 | |
| }, | |
| { | |
| "epoch": 5.08, | |
| "learning_rate": 0.00022979274611398964, | |
| "loss": 0.1328, | |
| "step": 18700 | |
| }, | |
| { | |
| "epoch": 5.11, | |
| "learning_rate": 0.00022817357512953368, | |
| "loss": 0.145, | |
| "step": 18800 | |
| }, | |
| { | |
| "epoch": 5.14, | |
| "learning_rate": 0.00022655440414507771, | |
| "loss": 0.1353, | |
| "step": 18900 | |
| }, | |
| { | |
| "epoch": 5.17, | |
| "learning_rate": 0.00022493523316062175, | |
| "loss": 0.146, | |
| "step": 19000 | |
| }, | |
| { | |
| "epoch": 5.2, | |
| "learning_rate": 0.00022331606217616581, | |
| "loss": 0.1508, | |
| "step": 19100 | |
| }, | |
| { | |
| "epoch": 5.23, | |
| "learning_rate": 0.00022169689119170985, | |
| "loss": 0.1587, | |
| "step": 19200 | |
| }, | |
| { | |
| "epoch": 5.26, | |
| "learning_rate": 0.0002200777202072539, | |
| "loss": 0.1504, | |
| "step": 19300 | |
| }, | |
| { | |
| "epoch": 5.29, | |
| "learning_rate": 0.00021845854922279793, | |
| "loss": 0.15, | |
| "step": 19400 | |
| }, | |
| { | |
| "epoch": 5.32, | |
| "learning_rate": 0.00021683937823834196, | |
| "loss": 0.1537, | |
| "step": 19500 | |
| }, | |
| { | |
| "epoch": 5.35, | |
| "learning_rate": 0.00021522020725388603, | |
| "loss": 0.1467, | |
| "step": 19600 | |
| }, | |
| { | |
| "epoch": 5.38, | |
| "learning_rate": 0.00021360103626943006, | |
| "loss": 0.155, | |
| "step": 19700 | |
| }, | |
| { | |
| "epoch": 5.41, | |
| "learning_rate": 0.0002119818652849741, | |
| "loss": 0.148, | |
| "step": 19800 | |
| }, | |
| { | |
| "epoch": 5.44, | |
| "learning_rate": 0.00021036269430051814, | |
| "loss": 0.1419, | |
| "step": 19900 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "learning_rate": 0.00020874352331606217, | |
| "loss": 0.154, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 5.47, | |
| "eval_loss": 0.5057598948478699, | |
| "eval_runtime": 487.4254, | |
| "eval_samples_per_second": 0.544, | |
| "eval_steps_per_second": 0.035, | |
| "eval_wer": 25.731181077193543, | |
| "step": 20000 | |
| }, | |
| { | |
| "epoch": 5.5, | |
| "learning_rate": 0.00020712435233160624, | |
| "loss": 0.1565, | |
| "step": 20100 | |
| }, | |
| { | |
| "epoch": 5.54, | |
| "learning_rate": 0.00020550518134715027, | |
| "loss": 0.1496, | |
| "step": 20200 | |
| }, | |
| { | |
| "epoch": 5.57, | |
| "learning_rate": 0.0002038860103626943, | |
| "loss": 0.1639, | |
| "step": 20300 | |
| }, | |
| { | |
| "epoch": 5.6, | |
| "learning_rate": 0.00020226683937823835, | |
| "loss": 0.1572, | |
| "step": 20400 | |
| }, | |
| { | |
| "epoch": 5.63, | |
| "learning_rate": 0.0002006476683937824, | |
| "loss": 0.1423, | |
| "step": 20500 | |
| }, | |
| { | |
| "epoch": 5.66, | |
| "learning_rate": 0.00019902849740932645, | |
| "loss": 0.1494, | |
| "step": 20600 | |
| }, | |
| { | |
| "epoch": 5.69, | |
| "learning_rate": 0.00019740932642487048, | |
| "loss": 0.1487, | |
| "step": 20700 | |
| }, | |
| { | |
| "epoch": 5.72, | |
| "learning_rate": 0.00019579015544041452, | |
| "loss": 0.1569, | |
| "step": 20800 | |
| }, | |
| { | |
| "epoch": 5.75, | |
| "learning_rate": 0.00019417098445595853, | |
| "loss": 0.1472, | |
| "step": 20900 | |
| }, | |
| { | |
| "epoch": 5.78, | |
| "learning_rate": 0.0001925518134715026, | |
| "loss": 0.1432, | |
| "step": 21000 | |
| }, | |
| { | |
| "epoch": 5.81, | |
| "learning_rate": 0.00019093264248704663, | |
| "loss": 0.1551, | |
| "step": 21100 | |
| }, | |
| { | |
| "epoch": 5.84, | |
| "learning_rate": 0.00018931347150259067, | |
| "loss": 0.1505, | |
| "step": 21200 | |
| }, | |
| { | |
| "epoch": 5.87, | |
| "learning_rate": 0.0001876943005181347, | |
| "loss": 0.1487, | |
| "step": 21300 | |
| }, | |
| { | |
| "epoch": 5.9, | |
| "learning_rate": 0.00018607512953367874, | |
| "loss": 0.1602, | |
| "step": 21400 | |
| }, | |
| { | |
| "epoch": 5.93, | |
| "learning_rate": 0.0001844559585492228, | |
| "loss": 0.147, | |
| "step": 21500 | |
| }, | |
| { | |
| "epoch": 5.96, | |
| "learning_rate": 0.00018283678756476684, | |
| "loss": 0.1437, | |
| "step": 21600 | |
| }, | |
| { | |
| "epoch": 5.99, | |
| "learning_rate": 0.00018121761658031088, | |
| "loss": 0.1611, | |
| "step": 21700 | |
| }, | |
| { | |
| "epoch": 6.02, | |
| "learning_rate": 0.00017959844559585492, | |
| "loss": 0.0933, | |
| "step": 21800 | |
| }, | |
| { | |
| "epoch": 6.05, | |
| "learning_rate": 0.00017797927461139895, | |
| "loss": 0.0719, | |
| "step": 21900 | |
| }, | |
| { | |
| "epoch": 6.08, | |
| "learning_rate": 0.00017636010362694302, | |
| "loss": 0.0872, | |
| "step": 22000 | |
| }, | |
| { | |
| "epoch": 6.11, | |
| "learning_rate": 0.0001747571243523316, | |
| "loss": 0.0816, | |
| "step": 22100 | |
| }, | |
| { | |
| "epoch": 6.14, | |
| "learning_rate": 0.00017313795336787566, | |
| "loss": 0.0808, | |
| "step": 22200 | |
| }, | |
| { | |
| "epoch": 6.17, | |
| "learning_rate": 0.0001715187823834197, | |
| "loss": 0.0817, | |
| "step": 22300 | |
| }, | |
| { | |
| "epoch": 6.2, | |
| "learning_rate": 0.00016989961139896373, | |
| "loss": 0.0832, | |
| "step": 22400 | |
| }, | |
| { | |
| "epoch": 6.23, | |
| "learning_rate": 0.00016828044041450777, | |
| "loss": 0.0941, | |
| "step": 22500 | |
| }, | |
| { | |
| "epoch": 6.27, | |
| "learning_rate": 0.0001666612694300518, | |
| "loss": 0.0931, | |
| "step": 22600 | |
| }, | |
| { | |
| "epoch": 6.3, | |
| "learning_rate": 0.00016504209844559587, | |
| "loss": 0.0843, | |
| "step": 22700 | |
| }, | |
| { | |
| "epoch": 6.33, | |
| "learning_rate": 0.00016343911917098445, | |
| "loss": 0.0846, | |
| "step": 22800 | |
| }, | |
| { | |
| "epoch": 6.36, | |
| "learning_rate": 0.0001618199481865285, | |
| "loss": 0.0942, | |
| "step": 22900 | |
| }, | |
| { | |
| "epoch": 6.39, | |
| "learning_rate": 0.00016020077720207255, | |
| "loss": 0.0901, | |
| "step": 23000 | |
| }, | |
| { | |
| "epoch": 6.42, | |
| "learning_rate": 0.00015858160621761658, | |
| "loss": 0.0882, | |
| "step": 23100 | |
| }, | |
| { | |
| "epoch": 6.45, | |
| "learning_rate": 0.00015696243523316062, | |
| "loss": 0.0854, | |
| "step": 23200 | |
| }, | |
| { | |
| "epoch": 6.48, | |
| "learning_rate": 0.00015534326424870468, | |
| "loss": 0.0908, | |
| "step": 23300 | |
| }, | |
| { | |
| "epoch": 6.51, | |
| "learning_rate": 0.00015372409326424872, | |
| "loss": 0.0853, | |
| "step": 23400 | |
| }, | |
| { | |
| "epoch": 6.54, | |
| "learning_rate": 0.00015210492227979276, | |
| "loss": 0.088, | |
| "step": 23500 | |
| }, | |
| { | |
| "epoch": 6.57, | |
| "learning_rate": 0.0001504857512953368, | |
| "loss": 0.0928, | |
| "step": 23600 | |
| }, | |
| { | |
| "epoch": 6.6, | |
| "learning_rate": 0.00014886658031088083, | |
| "loss": 0.0987, | |
| "step": 23700 | |
| }, | |
| { | |
| "epoch": 6.63, | |
| "learning_rate": 0.0001472474093264249, | |
| "loss": 0.0913, | |
| "step": 23800 | |
| }, | |
| { | |
| "epoch": 6.66, | |
| "learning_rate": 0.00014562823834196893, | |
| "loss": 0.0932, | |
| "step": 23900 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "learning_rate": 0.00014400906735751297, | |
| "loss": 0.0941, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 6.69, | |
| "eval_loss": 0.5185205340385437, | |
| "eval_runtime": 511.8024, | |
| "eval_samples_per_second": 0.518, | |
| "eval_steps_per_second": 0.033, | |
| "eval_wer": 29.47099248841298, | |
| "step": 24000 | |
| }, | |
| { | |
| "epoch": 6.72, | |
| "learning_rate": 0.000142389896373057, | |
| "loss": 0.0883, | |
| "step": 24100 | |
| }, | |
| { | |
| "epoch": 6.75, | |
| "learning_rate": 0.00014077072538860102, | |
| "loss": 0.0963, | |
| "step": 24200 | |
| }, | |
| { | |
| "epoch": 6.78, | |
| "learning_rate": 0.00013915155440414508, | |
| "loss": 0.0915, | |
| "step": 24300 | |
| }, | |
| { | |
| "epoch": 6.81, | |
| "learning_rate": 0.00013753238341968912, | |
| "loss": 0.098, | |
| "step": 24400 | |
| }, | |
| { | |
| "epoch": 6.84, | |
| "learning_rate": 0.00013591321243523315, | |
| "loss": 0.0911, | |
| "step": 24500 | |
| }, | |
| { | |
| "epoch": 6.87, | |
| "learning_rate": 0.0001342940414507772, | |
| "loss": 0.0931, | |
| "step": 24600 | |
| }, | |
| { | |
| "epoch": 6.9, | |
| "learning_rate": 0.00013267487046632123, | |
| "loss": 0.0873, | |
| "step": 24700 | |
| }, | |
| { | |
| "epoch": 6.93, | |
| "learning_rate": 0.0001310556994818653, | |
| "loss": 0.0932, | |
| "step": 24800 | |
| }, | |
| { | |
| "epoch": 6.96, | |
| "learning_rate": 0.00012943652849740933, | |
| "loss": 0.0887, | |
| "step": 24900 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "learning_rate": 0.00012781735751295336, | |
| "loss": 0.0974, | |
| "step": 25000 | |
| }, | |
| { | |
| "epoch": 7.03, | |
| "learning_rate": 0.0001261981865284974, | |
| "loss": 0.0446, | |
| "step": 25100 | |
| }, | |
| { | |
| "epoch": 7.06, | |
| "learning_rate": 0.00012457901554404146, | |
| "loss": 0.0377, | |
| "step": 25200 | |
| }, | |
| { | |
| "epoch": 7.09, | |
| "learning_rate": 0.0001229598445595855, | |
| "loss": 0.038, | |
| "step": 25300 | |
| }, | |
| { | |
| "epoch": 7.12, | |
| "learning_rate": 0.00012134067357512954, | |
| "loss": 0.0402, | |
| "step": 25400 | |
| }, | |
| { | |
| "epoch": 7.15, | |
| "learning_rate": 0.00011972150259067357, | |
| "loss": 0.0425, | |
| "step": 25500 | |
| }, | |
| { | |
| "epoch": 7.18, | |
| "learning_rate": 0.00011810233160621763, | |
| "loss": 0.0398, | |
| "step": 25600 | |
| }, | |
| { | |
| "epoch": 7.21, | |
| "learning_rate": 0.00011648316062176166, | |
| "loss": 0.0441, | |
| "step": 25700 | |
| }, | |
| { | |
| "epoch": 7.24, | |
| "learning_rate": 0.0001148639896373057, | |
| "loss": 0.0439, | |
| "step": 25800 | |
| }, | |
| { | |
| "epoch": 7.27, | |
| "learning_rate": 0.00011324481865284975, | |
| "loss": 0.0412, | |
| "step": 25900 | |
| }, | |
| { | |
| "epoch": 7.3, | |
| "learning_rate": 0.00011162564766839379, | |
| "loss": 0.0435, | |
| "step": 26000 | |
| }, | |
| { | |
| "epoch": 7.33, | |
| "learning_rate": 0.00011000647668393784, | |
| "loss": 0.0462, | |
| "step": 26100 | |
| }, | |
| { | |
| "epoch": 7.36, | |
| "learning_rate": 0.00010838730569948187, | |
| "loss": 0.0414, | |
| "step": 26200 | |
| }, | |
| { | |
| "epoch": 7.39, | |
| "learning_rate": 0.0001067681347150259, | |
| "loss": 0.0432, | |
| "step": 26300 | |
| }, | |
| { | |
| "epoch": 7.42, | |
| "learning_rate": 0.00010514896373056995, | |
| "loss": 0.0428, | |
| "step": 26400 | |
| }, | |
| { | |
| "epoch": 7.45, | |
| "learning_rate": 0.00010352979274611398, | |
| "loss": 0.0456, | |
| "step": 26500 | |
| }, | |
| { | |
| "epoch": 7.48, | |
| "learning_rate": 0.00010191062176165803, | |
| "loss": 0.046, | |
| "step": 26600 | |
| }, | |
| { | |
| "epoch": 7.51, | |
| "learning_rate": 0.00010029145077720207, | |
| "loss": 0.0408, | |
| "step": 26700 | |
| }, | |
| { | |
| "epoch": 7.54, | |
| "learning_rate": 9.867227979274612e-05, | |
| "loss": 0.0437, | |
| "step": 26800 | |
| }, | |
| { | |
| "epoch": 7.57, | |
| "learning_rate": 9.705310880829016e-05, | |
| "loss": 0.0468, | |
| "step": 26900 | |
| }, | |
| { | |
| "epoch": 7.6, | |
| "learning_rate": 9.54339378238342e-05, | |
| "loss": 0.0402, | |
| "step": 27000 | |
| }, | |
| { | |
| "epoch": 7.63, | |
| "learning_rate": 9.381476683937824e-05, | |
| "loss": 0.0416, | |
| "step": 27100 | |
| }, | |
| { | |
| "epoch": 7.66, | |
| "learning_rate": 9.219559585492228e-05, | |
| "loss": 0.0431, | |
| "step": 27200 | |
| }, | |
| { | |
| "epoch": 7.69, | |
| "learning_rate": 9.057642487046633e-05, | |
| "loss": 0.0426, | |
| "step": 27300 | |
| }, | |
| { | |
| "epoch": 7.73, | |
| "learning_rate": 8.895725388601037e-05, | |
| "loss": 0.0459, | |
| "step": 27400 | |
| }, | |
| { | |
| "epoch": 7.76, | |
| "learning_rate": 8.73380829015544e-05, | |
| "loss": 0.0454, | |
| "step": 27500 | |
| }, | |
| { | |
| "epoch": 7.79, | |
| "learning_rate": 8.571891191709846e-05, | |
| "loss": 0.0436, | |
| "step": 27600 | |
| }, | |
| { | |
| "epoch": 7.82, | |
| "learning_rate": 8.409974093264248e-05, | |
| "loss": 0.043, | |
| "step": 27700 | |
| }, | |
| { | |
| "epoch": 7.85, | |
| "learning_rate": 8.248056994818653e-05, | |
| "loss": 0.0464, | |
| "step": 27800 | |
| }, | |
| { | |
| "epoch": 7.88, | |
| "learning_rate": 8.086139896373057e-05, | |
| "loss": 0.0426, | |
| "step": 27900 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "learning_rate": 7.924222797927462e-05, | |
| "loss": 0.0379, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 7.91, | |
| "eval_loss": 0.6318183541297913, | |
| "eval_runtime": 486.5478, | |
| "eval_samples_per_second": 0.545, | |
| "eval_steps_per_second": 0.035, | |
| "eval_wer": 26.897874380693622, | |
| "step": 28000 | |
| }, | |
| { | |
| "epoch": 7.94, | |
| "learning_rate": 7.762305699481865e-05, | |
| "loss": 0.038, | |
| "step": 28100 | |
| }, | |
| { | |
| "epoch": 7.97, | |
| "learning_rate": 7.600388601036269e-05, | |
| "loss": 0.0413, | |
| "step": 28200 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "learning_rate": 7.438471502590674e-05, | |
| "loss": 0.043, | |
| "step": 28300 | |
| }, | |
| { | |
| "epoch": 8.03, | |
| "learning_rate": 7.276554404145078e-05, | |
| "loss": 0.0169, | |
| "step": 28400 | |
| }, | |
| { | |
| "epoch": 8.06, | |
| "learning_rate": 7.114637305699483e-05, | |
| "loss": 0.0178, | |
| "step": 28500 | |
| }, | |
| { | |
| "epoch": 8.09, | |
| "learning_rate": 6.952720207253886e-05, | |
| "loss": 0.0138, | |
| "step": 28600 | |
| }, | |
| { | |
| "epoch": 8.12, | |
| "learning_rate": 6.79080310880829e-05, | |
| "loss": 0.0134, | |
| "step": 28700 | |
| }, | |
| { | |
| "epoch": 8.15, | |
| "learning_rate": 6.628886010362695e-05, | |
| "loss": 0.0131, | |
| "step": 28800 | |
| }, | |
| { | |
| "epoch": 8.18, | |
| "learning_rate": 6.468588082901554e-05, | |
| "loss": 0.0165, | |
| "step": 28900 | |
| }, | |
| { | |
| "epoch": 8.21, | |
| "learning_rate": 6.306670984455959e-05, | |
| "loss": 0.0159, | |
| "step": 29000 | |
| }, | |
| { | |
| "epoch": 8.24, | |
| "learning_rate": 6.144753886010363e-05, | |
| "loss": 0.0135, | |
| "step": 29100 | |
| }, | |
| { | |
| "epoch": 8.27, | |
| "learning_rate": 5.9828367875647666e-05, | |
| "loss": 0.0164, | |
| "step": 29200 | |
| }, | |
| { | |
| "epoch": 8.3, | |
| "learning_rate": 5.820919689119171e-05, | |
| "loss": 0.0138, | |
| "step": 29300 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 5.659002590673575e-05, | |
| "loss": 0.0147, | |
| "step": 29400 | |
| }, | |
| { | |
| "epoch": 8.36, | |
| "learning_rate": 5.498704663212435e-05, | |
| "loss": 0.015, | |
| "step": 29500 | |
| }, | |
| { | |
| "epoch": 8.39, | |
| "learning_rate": 5.3367875647668394e-05, | |
| "loss": 0.0144, | |
| "step": 29600 | |
| }, | |
| { | |
| "epoch": 8.42, | |
| "learning_rate": 5.174870466321244e-05, | |
| "loss": 0.0171, | |
| "step": 29700 | |
| }, | |
| { | |
| "epoch": 8.45, | |
| "learning_rate": 5.012953367875648e-05, | |
| "loss": 0.014, | |
| "step": 29800 | |
| }, | |
| { | |
| "epoch": 8.49, | |
| "learning_rate": 4.8510362694300525e-05, | |
| "loss": 0.0128, | |
| "step": 29900 | |
| }, | |
| { | |
| "epoch": 8.52, | |
| "learning_rate": 4.6891191709844555e-05, | |
| "loss": 0.017, | |
| "step": 30000 | |
| }, | |
| { | |
| "epoch": 8.55, | |
| "learning_rate": 4.52720207253886e-05, | |
| "loss": 0.014, | |
| "step": 30100 | |
| }, | |
| { | |
| "epoch": 8.58, | |
| "learning_rate": 4.365284974093264e-05, | |
| "loss": 0.0147, | |
| "step": 30200 | |
| }, | |
| { | |
| "epoch": 8.61, | |
| "learning_rate": 4.2033678756476686e-05, | |
| "loss": 0.0143, | |
| "step": 30300 | |
| }, | |
| { | |
| "epoch": 8.64, | |
| "learning_rate": 4.041450777202073e-05, | |
| "loss": 0.016, | |
| "step": 30400 | |
| }, | |
| { | |
| "epoch": 8.67, | |
| "learning_rate": 3.879533678756477e-05, | |
| "loss": 0.0146, | |
| "step": 30500 | |
| }, | |
| { | |
| "epoch": 8.7, | |
| "learning_rate": 3.717616580310881e-05, | |
| "loss": 0.0152, | |
| "step": 30600 | |
| }, | |
| { | |
| "epoch": 8.73, | |
| "learning_rate": 3.5556994818652846e-05, | |
| "loss": 0.013, | |
| "step": 30700 | |
| }, | |
| { | |
| "epoch": 8.76, | |
| "learning_rate": 3.393782383419689e-05, | |
| "loss": 0.013, | |
| "step": 30800 | |
| }, | |
| { | |
| "epoch": 8.79, | |
| "learning_rate": 3.2318652849740933e-05, | |
| "loss": 0.0106, | |
| "step": 30900 | |
| }, | |
| { | |
| "epoch": 8.82, | |
| "learning_rate": 3.069948186528497e-05, | |
| "loss": 0.015, | |
| "step": 31000 | |
| }, | |
| { | |
| "epoch": 8.85, | |
| "learning_rate": 2.9080310880829017e-05, | |
| "loss": 0.0116, | |
| "step": 31100 | |
| }, | |
| { | |
| "epoch": 8.88, | |
| "learning_rate": 2.7461139896373057e-05, | |
| "loss": 0.0136, | |
| "step": 31200 | |
| }, | |
| { | |
| "epoch": 8.91, | |
| "learning_rate": 2.5841968911917097e-05, | |
| "loss": 0.0121, | |
| "step": 31300 | |
| }, | |
| { | |
| "epoch": 8.94, | |
| "learning_rate": 2.422279792746114e-05, | |
| "loss": 0.0131, | |
| "step": 31400 | |
| }, | |
| { | |
| "epoch": 8.97, | |
| "learning_rate": 2.2603626943005185e-05, | |
| "loss": 0.0155, | |
| "step": 31500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "learning_rate": 2.098445595854922e-05, | |
| "loss": 0.0111, | |
| "step": 31600 | |
| }, | |
| { | |
| "epoch": 9.03, | |
| "learning_rate": 1.9365284974093265e-05, | |
| "loss": 0.004, | |
| "step": 31700 | |
| }, | |
| { | |
| "epoch": 9.06, | |
| "learning_rate": 1.774611398963731e-05, | |
| "loss": 0.0034, | |
| "step": 31800 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 1.6126943005181345e-05, | |
| "loss": 0.0041, | |
| "step": 31900 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "learning_rate": 1.4507772020725389e-05, | |
| "loss": 0.0039, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 9.12, | |
| "eval_loss": 0.7823485732078552, | |
| "eval_runtime": 510.0267, | |
| "eval_samples_per_second": 0.52, | |
| "eval_steps_per_second": 0.033, | |
| "eval_wer": 26.306536678919613, | |
| "step": 32000 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 1.288860103626943e-05, | |
| "loss": 0.003, | |
| "step": 32100 | |
| }, | |
| { | |
| "epoch": 9.18, | |
| "learning_rate": 1.1269430051813473e-05, | |
| "loss": 0.0036, | |
| "step": 32200 | |
| }, | |
| { | |
| "epoch": 9.22, | |
| "learning_rate": 9.650259067357513e-06, | |
| "loss": 0.0043, | |
| "step": 32300 | |
| }, | |
| { | |
| "epoch": 9.25, | |
| "learning_rate": 8.031088082901555e-06, | |
| "loss": 0.0044, | |
| "step": 32400 | |
| }, | |
| { | |
| "epoch": 9.28, | |
| "learning_rate": 6.4119170984455965e-06, | |
| "loss": 0.003, | |
| "step": 32500 | |
| }, | |
| { | |
| "epoch": 9.31, | |
| "learning_rate": 4.7927461139896375e-06, | |
| "loss": 0.0032, | |
| "step": 32600 | |
| }, | |
| { | |
| "epoch": 9.34, | |
| "learning_rate": 3.1735751295336785e-06, | |
| "loss": 0.0032, | |
| "step": 32700 | |
| }, | |
| { | |
| "epoch": 9.37, | |
| "learning_rate": 1.5544041450777201e-06, | |
| "loss": 0.003, | |
| "step": 32800 | |
| }, | |
| { | |
| "epoch": 9.39, | |
| "step": 32880, | |
| "total_flos": 3.411337442033664e+19, | |
| "train_loss": 0.25426562409440095, | |
| "train_runtime": 60905.8756, | |
| "train_samples_per_second": 8.635, | |
| "train_steps_per_second": 0.54 | |
| } | |
| ], | |
| "max_steps": 32880, | |
| "num_train_epochs": 10, | |
| "total_flos": 3.411337442033664e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |