| { | |
| "best_metric": 31.6096791162546, | |
| "best_model_checkpoint": "whisper-da-small-augmented/checkpoint-4000", | |
| "epoch": 75.75757575757575, | |
| "global_step": 5000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 4.6e-06, | |
| "loss": 2.9766, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 1.52, | |
| "learning_rate": 9.600000000000001e-06, | |
| "loss": 1.3858, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 2.27, | |
| "learning_rate": 1.4599999999999999e-05, | |
| "loss": 0.8213, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 3.03, | |
| "learning_rate": 1.9600000000000002e-05, | |
| "loss": 0.5669, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 3.79, | |
| "learning_rate": 2.46e-05, | |
| "loss": 0.3486, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 4.55, | |
| "learning_rate": 2.96e-05, | |
| "loss": 0.2351, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 5.3, | |
| "learning_rate": 3.46e-05, | |
| "loss": 0.182, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 6.06, | |
| "learning_rate": 3.960000000000001e-05, | |
| "loss": 0.1346, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 6.82, | |
| "learning_rate": 4.46e-05, | |
| "loss": 0.102, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 7.58, | |
| "learning_rate": 4.96e-05, | |
| "loss": 0.0888, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 4.948888888888889e-05, | |
| "loss": 0.0801, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 9.09, | |
| "learning_rate": 4.8933333333333335e-05, | |
| "loss": 0.0665, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 9.85, | |
| "learning_rate": 4.837777777777778e-05, | |
| "loss": 0.0491, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 10.61, | |
| "learning_rate": 4.782222222222222e-05, | |
| "loss": 0.0454, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 11.36, | |
| "learning_rate": 4.726666666666667e-05, | |
| "loss": 0.0423, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 12.12, | |
| "learning_rate": 4.671111111111111e-05, | |
| "loss": 0.0363, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 12.88, | |
| "learning_rate": 4.615555555555556e-05, | |
| "loss": 0.0309, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 13.64, | |
| "learning_rate": 4.5600000000000004e-05, | |
| "loss": 0.0291, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 14.39, | |
| "learning_rate": 4.504444444444445e-05, | |
| "loss": 0.0246, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 15.15, | |
| "learning_rate": 4.448888888888889e-05, | |
| "loss": 0.0232, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 15.15, | |
| "eval_loss": 0.7538474798202515, | |
| "eval_runtime": 460.7386, | |
| "eval_samples_per_second": 4.63, | |
| "eval_steps_per_second": 0.145, | |
| "eval_wer": 35.58127301420305, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 15.91, | |
| "learning_rate": 4.3933333333333335e-05, | |
| "loss": 0.0213, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 16.67, | |
| "learning_rate": 4.337777777777778e-05, | |
| "loss": 0.0209, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 17.42, | |
| "learning_rate": 4.282222222222222e-05, | |
| "loss": 0.0171, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 18.18, | |
| "learning_rate": 4.226666666666667e-05, | |
| "loss": 0.0182, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 18.94, | |
| "learning_rate": 4.171111111111111e-05, | |
| "loss": 0.0157, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 19.7, | |
| "learning_rate": 4.115555555555556e-05, | |
| "loss": 0.0162, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 20.45, | |
| "learning_rate": 4.0600000000000004e-05, | |
| "loss": 0.0141, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 21.21, | |
| "learning_rate": 4.004444444444445e-05, | |
| "loss": 0.0127, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 21.97, | |
| "learning_rate": 3.948888888888889e-05, | |
| "loss": 0.0103, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 22.73, | |
| "learning_rate": 3.8933333333333336e-05, | |
| "loss": 0.0098, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 23.48, | |
| "learning_rate": 3.837777777777778e-05, | |
| "loss": 0.0099, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 24.24, | |
| "learning_rate": 3.782222222222222e-05, | |
| "loss": 0.0109, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 3.726666666666667e-05, | |
| "loss": 0.0078, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 25.76, | |
| "learning_rate": 3.671111111111111e-05, | |
| "loss": 0.0073, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 26.52, | |
| "learning_rate": 3.615555555555556e-05, | |
| "loss": 0.0057, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 27.27, | |
| "learning_rate": 3.56e-05, | |
| "loss": 0.0068, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 28.03, | |
| "learning_rate": 3.504444444444445e-05, | |
| "loss": 0.0052, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 28.79, | |
| "learning_rate": 3.448888888888889e-05, | |
| "loss": 0.0084, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 29.55, | |
| "learning_rate": 3.3933333333333336e-05, | |
| "loss": 0.006, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 30.3, | |
| "learning_rate": 3.337777777777778e-05, | |
| "loss": 0.0061, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 30.3, | |
| "eval_loss": 0.7932922840118408, | |
| "eval_runtime": 424.2221, | |
| "eval_samples_per_second": 5.028, | |
| "eval_steps_per_second": 0.158, | |
| "eval_wer": 34.3766438716465, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 31.06, | |
| "learning_rate": 3.2822222222222223e-05, | |
| "loss": 0.0067, | |
| "step": 2050 | |
| }, | |
| { | |
| "epoch": 31.82, | |
| "learning_rate": 3.226666666666667e-05, | |
| "loss": 0.0056, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 32.58, | |
| "learning_rate": 3.171111111111111e-05, | |
| "loss": 0.0044, | |
| "step": 2150 | |
| }, | |
| { | |
| "epoch": 33.33, | |
| "learning_rate": 3.1155555555555555e-05, | |
| "loss": 0.0041, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 34.09, | |
| "learning_rate": 3.06e-05, | |
| "loss": 0.0037, | |
| "step": 2250 | |
| }, | |
| { | |
| "epoch": 34.85, | |
| "learning_rate": 3.004444444444445e-05, | |
| "loss": 0.0025, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 35.61, | |
| "learning_rate": 2.948888888888889e-05, | |
| "loss": 0.0029, | |
| "step": 2350 | |
| }, | |
| { | |
| "epoch": 36.36, | |
| "learning_rate": 2.8933333333333333e-05, | |
| "loss": 0.0025, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 37.12, | |
| "learning_rate": 2.837777777777778e-05, | |
| "loss": 0.0031, | |
| "step": 2450 | |
| }, | |
| { | |
| "epoch": 37.88, | |
| "learning_rate": 2.782222222222222e-05, | |
| "loss": 0.0025, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 38.64, | |
| "learning_rate": 2.7266666666666668e-05, | |
| "loss": 0.0016, | |
| "step": 2550 | |
| }, | |
| { | |
| "epoch": 39.39, | |
| "learning_rate": 2.6711111111111115e-05, | |
| "loss": 0.003, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 40.15, | |
| "learning_rate": 2.6155555555555555e-05, | |
| "loss": 0.0015, | |
| "step": 2650 | |
| }, | |
| { | |
| "epoch": 40.91, | |
| "learning_rate": 2.5600000000000002e-05, | |
| "loss": 0.0016, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 41.67, | |
| "learning_rate": 2.504444444444445e-05, | |
| "loss": 0.0014, | |
| "step": 2750 | |
| }, | |
| { | |
| "epoch": 42.42, | |
| "learning_rate": 2.448888888888889e-05, | |
| "loss": 0.0013, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 43.18, | |
| "learning_rate": 2.3933333333333337e-05, | |
| "loss": 0.0027, | |
| "step": 2850 | |
| }, | |
| { | |
| "epoch": 43.94, | |
| "learning_rate": 2.337777777777778e-05, | |
| "loss": 0.0016, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 44.7, | |
| "learning_rate": 2.282222222222222e-05, | |
| "loss": 0.0025, | |
| "step": 2950 | |
| }, | |
| { | |
| "epoch": 45.45, | |
| "learning_rate": 2.2266666666666668e-05, | |
| "loss": 0.0016, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 45.45, | |
| "eval_loss": 0.7993177771568298, | |
| "eval_runtime": 430.3696, | |
| "eval_samples_per_second": 4.956, | |
| "eval_steps_per_second": 0.156, | |
| "eval_wer": 33.58232509205681, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 46.21, | |
| "learning_rate": 2.1711111111111112e-05, | |
| "loss": 0.0018, | |
| "step": 3050 | |
| }, | |
| { | |
| "epoch": 46.97, | |
| "learning_rate": 2.1155555555555556e-05, | |
| "loss": 0.001, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 47.73, | |
| "learning_rate": 2.06e-05, | |
| "loss": 0.0011, | |
| "step": 3150 | |
| }, | |
| { | |
| "epoch": 48.48, | |
| "learning_rate": 2.0044444444444446e-05, | |
| "loss": 0.001, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 49.24, | |
| "learning_rate": 1.948888888888889e-05, | |
| "loss": 0.0008, | |
| "step": 3250 | |
| }, | |
| { | |
| "epoch": 50.0, | |
| "learning_rate": 1.8933333333333334e-05, | |
| "loss": 0.0011, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 50.76, | |
| "learning_rate": 1.837777777777778e-05, | |
| "loss": 0.0007, | |
| "step": 3350 | |
| }, | |
| { | |
| "epoch": 51.52, | |
| "learning_rate": 1.7822222222222225e-05, | |
| "loss": 0.001, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 52.27, | |
| "learning_rate": 1.726666666666667e-05, | |
| "loss": 0.0014, | |
| "step": 3450 | |
| }, | |
| { | |
| "epoch": 53.03, | |
| "learning_rate": 1.6711111111111112e-05, | |
| "loss": 0.0009, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 53.79, | |
| "learning_rate": 1.6155555555555556e-05, | |
| "loss": 0.0008, | |
| "step": 3550 | |
| }, | |
| { | |
| "epoch": 54.55, | |
| "learning_rate": 1.56e-05, | |
| "loss": 0.0003, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 55.3, | |
| "learning_rate": 1.5044444444444445e-05, | |
| "loss": 0.0004, | |
| "step": 3650 | |
| }, | |
| { | |
| "epoch": 56.06, | |
| "learning_rate": 1.448888888888889e-05, | |
| "loss": 0.0004, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 56.82, | |
| "learning_rate": 1.3933333333333334e-05, | |
| "loss": 0.0005, | |
| "step": 3750 | |
| }, | |
| { | |
| "epoch": 57.58, | |
| "learning_rate": 1.3377777777777778e-05, | |
| "loss": 0.0005, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 58.33, | |
| "learning_rate": 1.2822222222222222e-05, | |
| "loss": 0.0006, | |
| "step": 3850 | |
| }, | |
| { | |
| "epoch": 59.09, | |
| "learning_rate": 1.2266666666666667e-05, | |
| "loss": 0.0005, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 59.85, | |
| "learning_rate": 1.1711111111111111e-05, | |
| "loss": 0.0004, | |
| "step": 3950 | |
| }, | |
| { | |
| "epoch": 60.61, | |
| "learning_rate": 1.1155555555555556e-05, | |
| "loss": 0.0003, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 60.61, | |
| "eval_loss": 0.7985624074935913, | |
| "eval_runtime": 420.8978, | |
| "eval_samples_per_second": 5.068, | |
| "eval_steps_per_second": 0.159, | |
| "eval_wer": 31.6096791162546, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 61.36, | |
| "learning_rate": 1.06e-05, | |
| "loss": 0.0002, | |
| "step": 4050 | |
| }, | |
| { | |
| "epoch": 62.12, | |
| "learning_rate": 1.0044444444444446e-05, | |
| "loss": 0.0003, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 62.88, | |
| "learning_rate": 9.48888888888889e-06, | |
| "loss": 0.0003, | |
| "step": 4150 | |
| }, | |
| { | |
| "epoch": 63.64, | |
| "learning_rate": 8.933333333333333e-06, | |
| "loss": 0.0002, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 64.39, | |
| "learning_rate": 8.377777777777779e-06, | |
| "loss": 0.0002, | |
| "step": 4250 | |
| }, | |
| { | |
| "epoch": 65.15, | |
| "learning_rate": 7.822222222222222e-06, | |
| "loss": 0.0002, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 65.91, | |
| "learning_rate": 7.266666666666668e-06, | |
| "loss": 0.0002, | |
| "step": 4350 | |
| }, | |
| { | |
| "epoch": 66.67, | |
| "learning_rate": 6.711111111111111e-06, | |
| "loss": 0.0002, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 67.42, | |
| "learning_rate": 6.155555555555556e-06, | |
| "loss": 0.0002, | |
| "step": 4450 | |
| }, | |
| { | |
| "epoch": 68.18, | |
| "learning_rate": 5.600000000000001e-06, | |
| "loss": 0.0002, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 68.94, | |
| "learning_rate": 5.044444444444444e-06, | |
| "loss": 0.0003, | |
| "step": 4550 | |
| }, | |
| { | |
| "epoch": 69.7, | |
| "learning_rate": 4.488888888888889e-06, | |
| "loss": 0.0003, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 70.45, | |
| "learning_rate": 3.9333333333333335e-06, | |
| "loss": 0.0002, | |
| "step": 4650 | |
| }, | |
| { | |
| "epoch": 71.21, | |
| "learning_rate": 3.3777777777777777e-06, | |
| "loss": 0.0002, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 71.97, | |
| "learning_rate": 2.8222222222222223e-06, | |
| "loss": 0.0001, | |
| "step": 4750 | |
| }, | |
| { | |
| "epoch": 72.73, | |
| "learning_rate": 2.266666666666667e-06, | |
| "loss": 0.0001, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 73.48, | |
| "learning_rate": 1.7111111111111112e-06, | |
| "loss": 0.0001, | |
| "step": 4850 | |
| }, | |
| { | |
| "epoch": 74.24, | |
| "learning_rate": 1.1555555555555556e-06, | |
| "loss": 0.0002, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 75.0, | |
| "learning_rate": 6.000000000000001e-07, | |
| "loss": 0.0003, | |
| "step": 4950 | |
| }, | |
| { | |
| "epoch": 75.76, | |
| "learning_rate": 4.444444444444445e-08, | |
| "loss": 0.0002, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 75.76, | |
| "eval_loss": 0.7901055812835693, | |
| "eval_runtime": 421.2754, | |
| "eval_samples_per_second": 5.063, | |
| "eval_steps_per_second": 0.159, | |
| "eval_wer": 32.13571804313519, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 75.76, | |
| "step": 5000, | |
| "total_flos": 9.2044313432064e+19, | |
| "train_loss": 0.07575882525146008, | |
| "train_runtime": 14911.1216, | |
| "train_samples_per_second": 21.46, | |
| "train_steps_per_second": 0.335 | |
| } | |
| ], | |
| "max_steps": 5000, | |
| "num_train_epochs": 76, | |
| "total_flos": 9.2044313432064e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |