| { | |
| "best_metric": 23.061646012074995, | |
| "best_model_checkpoint": "whisper-cy-small-augmented/checkpoint-5000", | |
| "epoch": 30.48780487804878, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 4.7e-06, | |
| "loss": 3.0617, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 9.7e-06, | |
| "loss": 1.6123, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.47e-05, | |
| "loss": 1.1238, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 1.22, | |
| "learning_rate": 1.97e-05, | |
| "loss": 0.7931, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 2.47e-05, | |
| "loss": 0.6475, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 1.83, | |
| "learning_rate": 2.97e-05, | |
| "loss": 0.5611, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 2.13, | |
| "learning_rate": 3.4699999999999996e-05, | |
| "loss": 0.4511, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 2.44, | |
| "learning_rate": 3.97e-05, | |
| "loss": 0.3488, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 2.74, | |
| "learning_rate": 4.47e-05, | |
| "loss": 0.3475, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 3.05, | |
| "learning_rate": 4.97e-05, | |
| "loss": 0.3222, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 3.35, | |
| "learning_rate": 4.947777777777778e-05, | |
| "loss": 0.1932, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 3.66, | |
| "learning_rate": 4.892222222222222e-05, | |
| "loss": 0.2064, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 3.96, | |
| "learning_rate": 4.836666666666667e-05, | |
| "loss": 0.2041, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 4.27, | |
| "learning_rate": 4.781111111111111e-05, | |
| "loss": 0.1152, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 4.57, | |
| "learning_rate": 4.725555555555556e-05, | |
| "loss": 0.1119, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 4.88, | |
| "learning_rate": 4.6700000000000003e-05, | |
| "loss": 0.1057, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 5.18, | |
| "learning_rate": 4.614444444444445e-05, | |
| "loss": 0.0789, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 5.49, | |
| "learning_rate": 4.558888888888889e-05, | |
| "loss": 0.0651, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 5.79, | |
| "learning_rate": 4.5033333333333335e-05, | |
| "loss": 0.0721, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "learning_rate": 4.447777777777778e-05, | |
| "loss": 0.0597, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 6.1, | |
| "eval_loss": 0.46903195977211, | |
| "eval_runtime": 1165.3555, | |
| "eval_samples_per_second": 4.519, | |
| "eval_steps_per_second": 0.142, | |
| "eval_wer": 29.466555449634573, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 6.4, | |
| "learning_rate": 4.392222222222223e-05, | |
| "loss": 0.0433, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 6.71, | |
| "learning_rate": 4.3366666666666666e-05, | |
| "loss": 0.0462, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 7.01, | |
| "learning_rate": 4.281111111111111e-05, | |
| "loss": 0.0449, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 7.32, | |
| "learning_rate": 4.225555555555556e-05, | |
| "loss": 0.0308, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 7.62, | |
| "learning_rate": 4.17e-05, | |
| "loss": 0.0308, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 7.93, | |
| "learning_rate": 4.114444444444445e-05, | |
| "loss": 0.0335, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 8.23, | |
| "learning_rate": 4.058888888888889e-05, | |
| "loss": 0.0267, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 8.54, | |
| "learning_rate": 4.0033333333333335e-05, | |
| "loss": 0.025, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 8.84, | |
| "learning_rate": 3.947777777777778e-05, | |
| "loss": 0.0241, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 9.15, | |
| "learning_rate": 3.892222222222223e-05, | |
| "loss": 0.0216, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 9.45, | |
| "learning_rate": 3.8366666666666666e-05, | |
| "loss": 0.0198, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 9.76, | |
| "learning_rate": 3.781111111111112e-05, | |
| "loss": 0.0193, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 10.06, | |
| "learning_rate": 3.7255555555555554e-05, | |
| "loss": 0.0171, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 10.37, | |
| "learning_rate": 3.6700000000000004e-05, | |
| "loss": 0.013, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 10.67, | |
| "learning_rate": 3.614444444444445e-05, | |
| "loss": 0.0164, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 10.98, | |
| "learning_rate": 3.5588888888888885e-05, | |
| "loss": 0.0172, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 11.28, | |
| "learning_rate": 3.5033333333333336e-05, | |
| "loss": 0.0114, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 11.59, | |
| "learning_rate": 3.447777777777778e-05, | |
| "loss": 0.0116, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 11.89, | |
| "learning_rate": 3.392222222222222e-05, | |
| "loss": 0.0126, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 12.2, | |
| "learning_rate": 3.336666666666667e-05, | |
| "loss": 0.0107, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 12.2, | |
| "eval_loss": 0.47065049409866333, | |
| "eval_runtime": 1115.4091, | |
| "eval_samples_per_second": 4.721, | |
| "eval_steps_per_second": 0.148, | |
| "eval_wer": 26.267079758500163, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 12.5, | |
| "learning_rate": 3.281111111111112e-05, | |
| "loss": 0.0091, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 12.8, | |
| "learning_rate": 3.2255555555555554e-05, | |
| "loss": 0.0102, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 13.11, | |
| "learning_rate": 3.1700000000000005e-05, | |
| "loss": 0.0092, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 13.41, | |
| "learning_rate": 3.114444444444445e-05, | |
| "loss": 0.0067, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 13.72, | |
| "learning_rate": 3.058888888888889e-05, | |
| "loss": 0.0086, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 14.02, | |
| "learning_rate": 3.0033333333333336e-05, | |
| "loss": 0.0082, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 14.33, | |
| "learning_rate": 2.9477777777777783e-05, | |
| "loss": 0.0064, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 14.63, | |
| "learning_rate": 2.8922222222222224e-05, | |
| "loss": 0.0065, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 14.94, | |
| "learning_rate": 2.836666666666667e-05, | |
| "loss": 0.0051, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 15.24, | |
| "learning_rate": 2.781111111111111e-05, | |
| "loss": 0.0055, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 15.55, | |
| "learning_rate": 2.7255555555555555e-05, | |
| "loss": 0.0051, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 15.85, | |
| "learning_rate": 2.6700000000000002e-05, | |
| "loss": 0.0039, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 16.16, | |
| "learning_rate": 2.6144444444444442e-05, | |
| "loss": 0.0038, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 16.46, | |
| "learning_rate": 2.558888888888889e-05, | |
| "loss": 0.0028, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 16.77, | |
| "learning_rate": 2.5033333333333336e-05, | |
| "loss": 0.0025, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 17.07, | |
| "learning_rate": 2.447777777777778e-05, | |
| "loss": 0.0024, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 17.38, | |
| "learning_rate": 2.3922222222222224e-05, | |
| "loss": 0.0021, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 17.68, | |
| "learning_rate": 2.3366666666666668e-05, | |
| "loss": 0.0022, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 17.99, | |
| "learning_rate": 2.281111111111111e-05, | |
| "loss": 0.0025, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 18.29, | |
| "learning_rate": 2.225555555555556e-05, | |
| "loss": 0.0026, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 18.29, | |
| "eval_loss": 0.46428239345550537, | |
| "eval_runtime": 1090.3134, | |
| "eval_samples_per_second": 4.83, | |
| "eval_steps_per_second": 0.151, | |
| "eval_wer": 24.67627899586908, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 18.6, | |
| "learning_rate": 2.1700000000000002e-05, | |
| "loss": 0.0018, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 18.9, | |
| "learning_rate": 2.1144444444444446e-05, | |
| "loss": 0.0015, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 19.21, | |
| "learning_rate": 2.058888888888889e-05, | |
| "loss": 0.0019, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 19.51, | |
| "learning_rate": 2.0033333333333334e-05, | |
| "loss": 0.0011, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 19.82, | |
| "learning_rate": 1.9477777777777777e-05, | |
| "loss": 0.0012, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 20.12, | |
| "learning_rate": 1.8922222222222224e-05, | |
| "loss": 0.001, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 20.43, | |
| "learning_rate": 1.8366666666666668e-05, | |
| "loss": 0.0014, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 20.73, | |
| "learning_rate": 1.7811111111111112e-05, | |
| "loss": 0.0015, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 21.04, | |
| "learning_rate": 1.7255555555555556e-05, | |
| "loss": 0.0015, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 21.34, | |
| "learning_rate": 1.6700000000000003e-05, | |
| "loss": 0.0017, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 21.65, | |
| "learning_rate": 1.6144444444444446e-05, | |
| "loss": 0.0016, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 21.95, | |
| "learning_rate": 1.558888888888889e-05, | |
| "loss": 0.0014, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 22.26, | |
| "learning_rate": 1.5033333333333336e-05, | |
| "loss": 0.001, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 22.56, | |
| "learning_rate": 1.447777777777778e-05, | |
| "loss": 0.0009, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 22.87, | |
| "learning_rate": 1.3922222222222223e-05, | |
| "loss": 0.001, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 23.17, | |
| "learning_rate": 1.3366666666666667e-05, | |
| "loss": 0.001, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 23.48, | |
| "learning_rate": 1.2811111111111112e-05, | |
| "loss": 0.0009, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 23.78, | |
| "learning_rate": 1.2255555555555556e-05, | |
| "loss": 0.0009, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 24.09, | |
| "learning_rate": 1.1700000000000001e-05, | |
| "loss": 0.001, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 24.39, | |
| "learning_rate": 1.1144444444444445e-05, | |
| "loss": 0.0007, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 24.39, | |
| "eval_loss": 0.46286216378211975, | |
| "eval_runtime": 1094.5709, | |
| "eval_samples_per_second": 4.811, | |
| "eval_steps_per_second": 0.151, | |
| "eval_wer": 23.802430886558625, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 24.7, | |
| "learning_rate": 1.058888888888889e-05, | |
| "loss": 0.0009, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 1.0033333333333333e-05, | |
| "loss": 0.0007, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 25.3, | |
| "learning_rate": 9.477777777777778e-06, | |
| "loss": 0.0005, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 25.61, | |
| "learning_rate": 8.922222222222222e-06, | |
| "loss": 0.0006, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 25.91, | |
| "learning_rate": 8.366666666666667e-06, | |
| "loss": 0.0004, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 26.22, | |
| "learning_rate": 7.811111111111113e-06, | |
| "loss": 0.0004, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 26.52, | |
| "learning_rate": 7.255555555555556e-06, | |
| "loss": 0.0005, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 26.83, | |
| "learning_rate": 6.700000000000001e-06, | |
| "loss": 0.0004, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 27.13, | |
| "learning_rate": 6.144444444444445e-06, | |
| "loss": 0.0005, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 27.44, | |
| "learning_rate": 5.588888888888889e-06, | |
| "loss": 0.0005, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 27.74, | |
| "learning_rate": 5.033333333333334e-06, | |
| "loss": 0.0006, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 28.05, | |
| "learning_rate": 4.477777777777778e-06, | |
| "loss": 0.0004, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 28.35, | |
| "learning_rate": 3.922222222222222e-06, | |
| "loss": 0.0003, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 28.66, | |
| "learning_rate": 3.3666666666666665e-06, | |
| "loss": 0.0005, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 28.96, | |
| "learning_rate": 2.811111111111111e-06, | |
| "loss": 0.0003, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 29.27, | |
| "learning_rate": 2.2555555555555557e-06, | |
| "loss": 0.0003, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 29.57, | |
| "learning_rate": 1.7000000000000002e-06, | |
| "loss": 0.0003, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 29.88, | |
| "learning_rate": 1.1444444444444446e-06, | |
| "loss": 0.0004, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 30.18, | |
| "learning_rate": 5.888888888888889e-07, | |
| "loss": 0.0004, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 30.49, | |
| "learning_rate": 3.3333333333333334e-08, | |
| "loss": 0.0004, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 30.49, | |
| "eval_loss": 0.46102383732795715, | |
| "eval_runtime": 1096.0988, | |
| "eval_samples_per_second": 4.804, | |
| "eval_steps_per_second": 0.151, | |
| "eval_wer": 23.061646012074995, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 30.49, | |
| "step": 5000, | |
| "total_flos": 9.23300129783808e+19, | |
| "train_loss": 0.11097193325944245, | |
| "train_runtime": 18201.3214, | |
| "train_samples_per_second": 17.581, | |
| "train_steps_per_second": 0.275 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 31, | |
| "total_flos": 9.23300129783808e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |