{ "best_metric": 8.493092929242826, "best_model_checkpoint": "./whisper-small-bn-3ds\\checkpoint-26000", "epoch": 3.1294531839391992, "global_step": 28000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.06, "learning_rate": 5.58659217877095e-07, "loss": 1.8271, "step": 500 }, { "epoch": 0.11, "learning_rate": 1.11731843575419e-06, "loss": 0.6492, "step": 1000 }, { "epoch": 0.17, "learning_rate": 1.675977653631285e-06, "loss": 0.358, "step": 1500 }, { "epoch": 0.22, "learning_rate": 2.23463687150838e-06, "loss": 0.2337, "step": 2000 }, { "epoch": 0.22, "eval_loss": 0.2297607660293579, "eval_runtime": 8187.9402, "eval_samples_per_second": 1.174, "eval_steps_per_second": 0.147, "eval_wer": 31.773352182072205, "step": 2000 }, { "epoch": 0.28, "learning_rate": 2.793296089385475e-06, "loss": 0.2002, "step": 2500 }, { "epoch": 0.34, "learning_rate": 3.35195530726257e-06, "loss": 0.1755, "step": 3000 }, { "epoch": 0.39, "learning_rate": 3.910614525139665e-06, "loss": 0.1562, "step": 3500 }, { "epoch": 0.45, "learning_rate": 4.46927374301676e-06, "loss": 0.1444, "step": 4000 }, { "epoch": 0.45, "eval_loss": 0.1516258269548416, "eval_runtime": 8249.6732, "eval_samples_per_second": 1.165, "eval_steps_per_second": 0.146, "eval_wer": 22.238205070972114, "step": 4000 }, { "epoch": 0.5, "learning_rate": 5.027932960893855e-06, "loss": 0.134, "step": 4500 }, { "epoch": 0.56, "learning_rate": 5.58659217877095e-06, "loss": 0.1231, "step": 5000 }, { "epoch": 0.61, "learning_rate": 6.145251396648045e-06, "loss": 0.1162, "step": 5500 }, { "epoch": 0.67, "learning_rate": 6.70391061452514e-06, "loss": 0.1106, "step": 6000 }, { "epoch": 0.67, "eval_loss": 0.1180974692106247, "eval_runtime": 8273.9229, "eval_samples_per_second": 1.162, "eval_steps_per_second": 0.145, "eval_wer": 17.223035142977604, "step": 6000 }, { "epoch": 0.73, "learning_rate": 7.262569832402235e-06, "loss": 0.1069, "step": 6500 }, { "epoch": 0.78, "learning_rate": 7.82122905027933e-06, "loss": 0.1014, "step": 7000 }, { "epoch": 0.84, "learning_rate": 8.379888268156426e-06, "loss": 0.0964, "step": 7500 }, { "epoch": 0.89, "learning_rate": 8.93854748603352e-06, "loss": 0.0933, "step": 8000 }, { "epoch": 0.89, "eval_loss": 0.10048126429319382, "eval_runtime": 9253.512, "eval_samples_per_second": 1.039, "eval_steps_per_second": 0.13, "eval_wer": 14.800806858295513, "step": 8000 }, { "epoch": 0.95, "learning_rate": 9.497206703910616e-06, "loss": 0.0908, "step": 8500 }, { "epoch": 1.01, "learning_rate": 9.973753280839897e-06, "loss": 0.0859, "step": 9000 }, { "epoch": 1.06, "learning_rate": 9.711286089238846e-06, "loss": 0.0743, "step": 9500 }, { "epoch": 1.12, "learning_rate": 9.448818897637797e-06, "loss": 0.0747, "step": 10000 }, { "epoch": 1.12, "eval_loss": 0.08645793795585632, "eval_runtime": 8690.2436, "eval_samples_per_second": 1.106, "eval_steps_per_second": 0.138, "eval_wer": 12.718351892727528, "step": 10000 }, { "epoch": 1.17, "learning_rate": 9.186351706036746e-06, "loss": 0.0706, "step": 10500 }, { "epoch": 1.23, "learning_rate": 8.923884514435697e-06, "loss": 0.0687, "step": 11000 }, { "epoch": 1.29, "learning_rate": 8.661417322834647e-06, "loss": 0.068, "step": 11500 }, { "epoch": 1.34, "learning_rate": 8.398950131233596e-06, "loss": 0.0642, "step": 12000 }, { "epoch": 1.34, "eval_loss": 0.07882149517536163, "eval_runtime": 9343.6844, "eval_samples_per_second": 1.029, "eval_steps_per_second": 0.129, "eval_wer": 11.644056447012723, "step": 12000 }, { "epoch": 1.4, "learning_rate": 8.136482939632546e-06, "loss": 0.0644, "step": 12500 }, { "epoch": 1.45, "learning_rate": 7.874015748031496e-06, "loss": 0.0631, "step": 13000 }, { "epoch": 1.51, "learning_rate": 7.611548556430447e-06, "loss": 0.0615, "step": 13500 }, { "epoch": 1.56, "learning_rate": 7.349081364829397e-06, "loss": 0.0613, "step": 14000 }, { "epoch": 1.56, "eval_loss": 0.07153687626123428, "eval_runtime": 9348.8718, "eval_samples_per_second": 1.028, "eval_steps_per_second": 0.129, "eval_wer": 10.652430908624952, "step": 14000 }, { "epoch": 1.62, "learning_rate": 7.086614173228347e-06, "loss": 0.0597, "step": 14500 }, { "epoch": 1.68, "learning_rate": 6.824146981627298e-06, "loss": 0.0586, "step": 15000 }, { "epoch": 1.73, "learning_rate": 6.561679790026248e-06, "loss": 0.0569, "step": 15500 }, { "epoch": 1.79, "learning_rate": 6.299212598425197e-06, "loss": 0.0565, "step": 16000 }, { "epoch": 1.79, "eval_loss": 0.0665777176618576, "eval_runtime": 8488.0108, "eval_samples_per_second": 1.132, "eval_steps_per_second": 0.142, "eval_wer": 9.88070732372709, "step": 16000 }, { "epoch": 1.84, "learning_rate": 6.036745406824147e-06, "loss": 0.0559, "step": 16500 }, { "epoch": 1.9, "learning_rate": 5.774278215223098e-06, "loss": 0.056, "step": 17000 }, { "epoch": 1.96, "learning_rate": 5.511811023622048e-06, "loss": 0.0537, "step": 17500 }, { "epoch": 2.01, "learning_rate": 5.2493438320209976e-06, "loss": 0.0494, "step": 18000 }, { "epoch": 2.01, "eval_loss": 0.06404710561037064, "eval_runtime": 9359.4591, "eval_samples_per_second": 1.027, "eval_steps_per_second": 0.128, "eval_wer": 9.52646677083075, "step": 18000 }, { "epoch": 2.07, "learning_rate": 4.986876640419948e-06, "loss": 0.0369, "step": 18500 }, { "epoch": 2.12, "learning_rate": 4.724409448818898e-06, "loss": 0.0368, "step": 19000 }, { "epoch": 2.18, "learning_rate": 4.461942257217848e-06, "loss": 0.0367, "step": 19500 }, { "epoch": 2.24, "learning_rate": 4.199475065616798e-06, "loss": 0.037, "step": 20000 }, { "epoch": 2.24, "eval_loss": 0.06194682419300079, "eval_runtime": 9390.5519, "eval_samples_per_second": 1.023, "eval_steps_per_second": 0.128, "eval_wer": 9.031687375478452, "step": 20000 }, { "epoch": 2.29, "learning_rate": 3.937007874015748e-06, "loss": 0.0363, "step": 20500 }, { "epoch": 2.35, "learning_rate": 3.6745406824146986e-06, "loss": 0.0358, "step": 21000 }, { "epoch": 2.4, "learning_rate": 3.412073490813649e-06, "loss": 0.0364, "step": 21500 }, { "epoch": 2.46, "learning_rate": 3.1496062992125985e-06, "loss": 0.0348, "step": 22000 }, { "epoch": 2.46, "eval_loss": 0.05995701625943184, "eval_runtime": 8568.802, "eval_samples_per_second": 1.122, "eval_steps_per_second": 0.14, "eval_wer": 9.032927424088358, "step": 22000 }, { "epoch": 2.51, "learning_rate": 2.887139107611549e-06, "loss": 0.0348, "step": 22500 }, { "epoch": 2.57, "learning_rate": 2.6246719160104988e-06, "loss": 0.0334, "step": 23000 }, { "epoch": 2.63, "learning_rate": 2.362204724409449e-06, "loss": 0.0333, "step": 23500 }, { "epoch": 2.68, "learning_rate": 2.099737532808399e-06, "loss": 0.0329, "step": 24000 }, { "epoch": 2.68, "eval_loss": 0.058194514364004135, "eval_runtime": 7732.5254, "eval_samples_per_second": 1.243, "eval_steps_per_second": 0.155, "eval_wer": 8.840719889553004, "step": 24000 }, { "epoch": 2.74, "learning_rate": 1.8372703412073493e-06, "loss": 0.0332, "step": 24500 }, { "epoch": 2.79, "learning_rate": 1.5748031496062992e-06, "loss": 0.0326, "step": 25000 }, { "epoch": 2.85, "learning_rate": 1.3123359580052494e-06, "loss": 0.0321, "step": 25500 }, { "epoch": 2.91, "learning_rate": 1.0498687664041996e-06, "loss": 0.0321, "step": 26000 }, { "epoch": 2.91, "eval_loss": 0.05638869106769562, "eval_runtime": 7807.0833, "eval_samples_per_second": 1.231, "eval_steps_per_second": 0.154, "eval_wer": 8.493092929242826, "step": 26000 }, { "epoch": 2.96, "learning_rate": 7.874015748031496e-07, "loss": 0.0318, "step": 26500 }, { "epoch": 3.02, "learning_rate": 5.249343832020998e-07, "loss": 0.0285, "step": 27000 }, { "epoch": 3.07, "learning_rate": 2.624671916010499e-07, "loss": 0.0228, "step": 27500 }, { "epoch": 3.13, "learning_rate": 0.0, "loss": 0.0227, "step": 28000 }, { "epoch": 3.13, "eval_loss": 0.05735579878091812, "eval_runtime": 7923.7534, "eval_samples_per_second": 1.213, "eval_steps_per_second": 0.152, "eval_wer": 8.50384001719534, "step": 28000 }, { "epoch": 3.13, "step": 28000, "total_flos": 2.5857251868672e+20, "train_loss": 0.11752363123212542, "train_runtime": 521775.8612, "train_samples_per_second": 1.717, "train_steps_per_second": 0.054 } ], "max_steps": 28000, "num_train_epochs": 4, "total_flos": 2.5857251868672e+20, "trial_name": null, "trial_params": null }