| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.11984619738002897, | |
| "eval_steps": 1000, | |
| "global_step": 14400, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 8.322652595835345e-06, | |
| "grad_norm": 14.9375, | |
| "learning_rate": 2e-06, | |
| "loss": 0.6484, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0008322652595835345, | |
| "grad_norm": 0.3671875, | |
| "learning_rate": 0.0002, | |
| "loss": 0.4345, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.001664530519167069, | |
| "grad_norm": 0.1689453125, | |
| "learning_rate": 0.0004, | |
| "loss": 0.2537, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.0024967957787506035, | |
| "grad_norm": 0.55859375, | |
| "learning_rate": 0.0006, | |
| "loss": 0.2344, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.003329061038334138, | |
| "grad_norm": 0.154296875, | |
| "learning_rate": 0.0008, | |
| "loss": 0.222, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.004161326297917672, | |
| "grad_norm": 0.126953125, | |
| "learning_rate": 0.001, | |
| "loss": 0.2142, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.004993591557501207, | |
| "grad_norm": 0.12353515625, | |
| "learning_rate": 0.0012, | |
| "loss": 0.2011, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.005825856817084741, | |
| "grad_norm": 0.10693359375, | |
| "learning_rate": 0.0014, | |
| "loss": 0.1897, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.006658122076668276, | |
| "grad_norm": 0.10595703125, | |
| "learning_rate": 0.0016, | |
| "loss": 0.1775, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.0074903873362518105, | |
| "grad_norm": 0.0859375, | |
| "learning_rate": 0.0018000000000000002, | |
| "loss": 0.1615, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.008322652595835344, | |
| "grad_norm": 0.083984375, | |
| "learning_rate": 0.002, | |
| "loss": 0.1444, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.008322652595835344, | |
| "eval_peoplespeech-clean-transcription_loss": 2.2110702991485596, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.5405, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.708, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.839, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.009154917855418878, | |
| "grad_norm": 0.0771484375, | |
| "learning_rate": 0.001999725185109816, | |
| "loss": 0.1323, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.009987183115002414, | |
| "grad_norm": 0.083984375, | |
| "learning_rate": 0.0019989008914857113, | |
| "loss": 0.1242, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.010819448374585948, | |
| "grad_norm": 0.07861328125, | |
| "learning_rate": 0.00199752757218401, | |
| "loss": 0.1162, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.011651713634169482, | |
| "grad_norm": 0.07275390625, | |
| "learning_rate": 0.001995605982021898, | |
| "loss": 0.1128, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.012483978893753018, | |
| "grad_norm": 0.07666015625, | |
| "learning_rate": 0.0019931371771625545, | |
| "loss": 0.1094, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.013316244153336551, | |
| "grad_norm": 0.06201171875, | |
| "learning_rate": 0.001990122514534651, | |
| "loss": 0.1052, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.014148509412920085, | |
| "grad_norm": 0.058837890625, | |
| "learning_rate": 0.0019865636510865464, | |
| "loss": 0.1022, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.014980774672503621, | |
| "grad_norm": 0.0625, | |
| "learning_rate": 0.001982462542875576, | |
| "loss": 0.1011, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.015813039932087155, | |
| "grad_norm": 0.06787109375, | |
| "learning_rate": 0.001977821443992945, | |
| "loss": 0.0983, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.01664530519167069, | |
| "grad_norm": 0.0537109375, | |
| "learning_rate": 0.001972642905324813, | |
| "loss": 0.0975, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.01664530519167069, | |
| "eval_peoplespeech-clean-transcription_loss": 1.6914767026901245, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.5301, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.716, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.839, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.017477570451254223, | |
| "grad_norm": 0.05859375, | |
| "learning_rate": 0.0019669297731502505, | |
| "loss": 0.0947, | |
| "step": 2100 | |
| }, | |
| { | |
| "epoch": 0.018309835710837757, | |
| "grad_norm": 0.062255859375, | |
| "learning_rate": 0.00196068518757684, | |
| "loss": 0.0935, | |
| "step": 2200 | |
| }, | |
| { | |
| "epoch": 0.019142100970421294, | |
| "grad_norm": 0.059326171875, | |
| "learning_rate": 0.001953912580814779, | |
| "loss": 0.0911, | |
| "step": 2300 | |
| }, | |
| { | |
| "epoch": 0.019974366230004828, | |
| "grad_norm": 0.060302734375, | |
| "learning_rate": 0.0019466156752904343, | |
| "loss": 0.0904, | |
| "step": 2400 | |
| }, | |
| { | |
| "epoch": 0.020806631489588362, | |
| "grad_norm": 0.0615234375, | |
| "learning_rate": 0.0019387984816003866, | |
| "loss": 0.0882, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.021638896749171896, | |
| "grad_norm": 0.056884765625, | |
| "learning_rate": 0.0019304652963070869, | |
| "loss": 0.0874, | |
| "step": 2600 | |
| }, | |
| { | |
| "epoch": 0.02247116200875543, | |
| "grad_norm": 0.049072265625, | |
| "learning_rate": 0.0019216206995773372, | |
| "loss": 0.0871, | |
| "step": 2700 | |
| }, | |
| { | |
| "epoch": 0.023303427268338964, | |
| "grad_norm": 0.053955078125, | |
| "learning_rate": 0.0019122695526648968, | |
| "loss": 0.0856, | |
| "step": 2800 | |
| }, | |
| { | |
| "epoch": 0.0241356925279225, | |
| "grad_norm": 0.05517578125, | |
| "learning_rate": 0.0019024169952385887, | |
| "loss": 0.0845, | |
| "step": 2900 | |
| }, | |
| { | |
| "epoch": 0.024967957787506035, | |
| "grad_norm": 0.0546875, | |
| "learning_rate": 0.0018920684425573864, | |
| "loss": 0.0852, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.024967957787506035, | |
| "eval_peoplespeech-clean-transcription_loss": 1.6414048671722412, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.8151, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.521, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.815, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.02580022304708957, | |
| "grad_norm": 0.052978515625, | |
| "learning_rate": 0.0018812295824940284, | |
| "loss": 0.0836, | |
| "step": 3100 | |
| }, | |
| { | |
| "epoch": 0.026632488306673103, | |
| "grad_norm": 0.05126953125, | |
| "learning_rate": 0.0018699063724087904, | |
| "loss": 0.0821, | |
| "step": 3200 | |
| }, | |
| { | |
| "epoch": 0.027464753566256637, | |
| "grad_norm": 0.056884765625, | |
| "learning_rate": 0.0018581050358751443, | |
| "loss": 0.0816, | |
| "step": 3300 | |
| }, | |
| { | |
| "epoch": 0.02829701882584017, | |
| "grad_norm": 0.045654296875, | |
| "learning_rate": 0.0018458320592590974, | |
| "loss": 0.0814, | |
| "step": 3400 | |
| }, | |
| { | |
| "epoch": 0.029129284085423705, | |
| "grad_norm": 0.04638671875, | |
| "learning_rate": 0.0018330941881540914, | |
| "loss": 0.0791, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.029961549345007242, | |
| "grad_norm": 0.043212890625, | |
| "learning_rate": 0.0018198984236734246, | |
| "loss": 0.0804, | |
| "step": 3600 | |
| }, | |
| { | |
| "epoch": 0.030793814604590776, | |
| "grad_norm": 0.054931640625, | |
| "learning_rate": 0.0018062520186022297, | |
| "loss": 0.0802, | |
| "step": 3700 | |
| }, | |
| { | |
| "epoch": 0.03162607986417431, | |
| "grad_norm": 0.054931640625, | |
| "learning_rate": 0.0017921624734111292, | |
| "loss": 0.0805, | |
| "step": 3800 | |
| }, | |
| { | |
| "epoch": 0.032458345123757844, | |
| "grad_norm": 0.048095703125, | |
| "learning_rate": 0.001777637532133752, | |
| "loss": 0.079, | |
| "step": 3900 | |
| }, | |
| { | |
| "epoch": 0.03329061038334138, | |
| "grad_norm": 0.0458984375, | |
| "learning_rate": 0.0017626851781103819, | |
| "loss": 0.0759, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.03329061038334138, | |
| "eval_peoplespeech-clean-transcription_loss": 1.5952188968658447, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.4647, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.762, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.845, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.03412287564292491, | |
| "grad_norm": 0.06494140625, | |
| "learning_rate": 0.001747313629600077, | |
| "loss": 0.0774, | |
| "step": 4100 | |
| }, | |
| { | |
| "epoch": 0.034955140902508446, | |
| "grad_norm": 0.043701171875, | |
| "learning_rate": 0.001731531335263669, | |
| "loss": 0.0774, | |
| "step": 4200 | |
| }, | |
| { | |
| "epoch": 0.03578740616209198, | |
| "grad_norm": 0.048095703125, | |
| "learning_rate": 0.0017153469695201276, | |
| "loss": 0.0769, | |
| "step": 4300 | |
| }, | |
| { | |
| "epoch": 0.03661967142167551, | |
| "grad_norm": 0.051513671875, | |
| "learning_rate": 0.0016987694277788418, | |
| "loss": 0.0763, | |
| "step": 4400 | |
| }, | |
| { | |
| "epoch": 0.037451936681259054, | |
| "grad_norm": 0.047607421875, | |
| "learning_rate": 0.001681807821550438, | |
| "loss": 0.0774, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.03828420194084259, | |
| "grad_norm": 0.040283203125, | |
| "learning_rate": 0.0016644714734388218, | |
| "loss": 0.0748, | |
| "step": 4600 | |
| }, | |
| { | |
| "epoch": 0.03911646720042612, | |
| "grad_norm": 0.05419921875, | |
| "learning_rate": 0.0016467699120171987, | |
| "loss": 0.0755, | |
| "step": 4700 | |
| }, | |
| { | |
| "epoch": 0.039948732460009656, | |
| "grad_norm": 0.048828125, | |
| "learning_rate": 0.001628712866590885, | |
| "loss": 0.0732, | |
| "step": 4800 | |
| }, | |
| { | |
| "epoch": 0.04078099771959319, | |
| "grad_norm": 0.046875, | |
| "learning_rate": 0.0016103102618497923, | |
| "loss": 0.0757, | |
| "step": 4900 | |
| }, | |
| { | |
| "epoch": 0.041613262979176724, | |
| "grad_norm": 0.042236328125, | |
| "learning_rate": 0.0015915722124135226, | |
| "loss": 0.0738, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.041613262979176724, | |
| "eval_peoplespeech-clean-transcription_loss": 1.5546748638153076, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.345, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.849, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.856, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.04244552823876026, | |
| "grad_norm": 0.045166015625, | |
| "learning_rate": 0.001572509017272072, | |
| "loss": 0.0742, | |
| "step": 5100 | |
| }, | |
| { | |
| "epoch": 0.04327779349834379, | |
| "grad_norm": 0.0556640625, | |
| "learning_rate": 0.0015531311541251993, | |
| "loss": 0.0746, | |
| "step": 5200 | |
| }, | |
| { | |
| "epoch": 0.044110058757927326, | |
| "grad_norm": 0.043212890625, | |
| "learning_rate": 0.0015334492736235703, | |
| "loss": 0.073, | |
| "step": 5300 | |
| }, | |
| { | |
| "epoch": 0.04494232401751086, | |
| "grad_norm": 0.044921875, | |
| "learning_rate": 0.0015134741935148419, | |
| "loss": 0.0727, | |
| "step": 5400 | |
| }, | |
| { | |
| "epoch": 0.045774589277094394, | |
| "grad_norm": 0.05029296875, | |
| "learning_rate": 0.0014932168926979072, | |
| "loss": 0.0726, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.04660685453667793, | |
| "grad_norm": 0.048095703125, | |
| "learning_rate": 0.0014726885051885652, | |
| "loss": 0.0718, | |
| "step": 5600 | |
| }, | |
| { | |
| "epoch": 0.04743911979626146, | |
| "grad_norm": 0.046875, | |
| "learning_rate": 0.0014519003139999338, | |
| "loss": 0.0736, | |
| "step": 5700 | |
| }, | |
| { | |
| "epoch": 0.048271385055845, | |
| "grad_norm": 0.051513671875, | |
| "learning_rate": 0.0014308637449409706, | |
| "loss": 0.0724, | |
| "step": 5800 | |
| }, | |
| { | |
| "epoch": 0.049103650315428536, | |
| "grad_norm": 0.053955078125, | |
| "learning_rate": 0.0014095903603365066, | |
| "loss": 0.0724, | |
| "step": 5900 | |
| }, | |
| { | |
| "epoch": 0.04993591557501207, | |
| "grad_norm": 0.037841796875, | |
| "learning_rate": 0.0013880918526722496, | |
| "loss": 0.071, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.04993591557501207, | |
| "eval_peoplespeech-clean-transcription_loss": 1.5297455787658691, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.3364, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.855, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.857, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.050768180834595604, | |
| "grad_norm": 0.05029296875, | |
| "learning_rate": 0.0013663800381682463, | |
| "loss": 0.0713, | |
| "step": 6100 | |
| }, | |
| { | |
| "epoch": 0.05160044609417914, | |
| "grad_norm": 0.037109375, | |
| "learning_rate": 0.0013444668502843329, | |
| "loss": 0.0701, | |
| "step": 6200 | |
| }, | |
| { | |
| "epoch": 0.05243271135376267, | |
| "grad_norm": 0.03564453125, | |
| "learning_rate": 0.0013223643331611537, | |
| "loss": 0.0708, | |
| "step": 6300 | |
| }, | |
| { | |
| "epoch": 0.053264976613346206, | |
| "grad_norm": 0.04443359375, | |
| "learning_rate": 0.001300084635000341, | |
| "loss": 0.0707, | |
| "step": 6400 | |
| }, | |
| { | |
| "epoch": 0.05409724187292974, | |
| "grad_norm": 0.033935546875, | |
| "learning_rate": 0.0012776400013875004, | |
| "loss": 0.0704, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.054929507132513274, | |
| "grad_norm": 0.046142578125, | |
| "learning_rate": 0.0012550427685616766, | |
| "loss": 0.0707, | |
| "step": 6600 | |
| }, | |
| { | |
| "epoch": 0.05576177239209681, | |
| "grad_norm": 0.04150390625, | |
| "learning_rate": 0.0012323053566349834, | |
| "loss": 0.0696, | |
| "step": 6700 | |
| }, | |
| { | |
| "epoch": 0.05659403765168034, | |
| "grad_norm": 0.042236328125, | |
| "learning_rate": 0.0012094402627661448, | |
| "loss": 0.0705, | |
| "step": 6800 | |
| }, | |
| { | |
| "epoch": 0.057426302911263875, | |
| "grad_norm": 0.035888671875, | |
| "learning_rate": 0.0011864600542916813, | |
| "loss": 0.0686, | |
| "step": 6900 | |
| }, | |
| { | |
| "epoch": 0.05825856817084741, | |
| "grad_norm": 0.0498046875, | |
| "learning_rate": 0.0011633773618185302, | |
| "loss": 0.0687, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.05825856817084741, | |
| "eval_peoplespeech-clean-transcription_loss": 1.518404483795166, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.4703, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.758, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.845, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.05909083343043095, | |
| "grad_norm": 0.034423828125, | |
| "learning_rate": 0.0011402048722818862, | |
| "loss": 0.0693, | |
| "step": 7100 | |
| }, | |
| { | |
| "epoch": 0.059923098690014484, | |
| "grad_norm": 0.047119140625, | |
| "learning_rate": 0.0011169553219720827, | |
| "loss": 0.0697, | |
| "step": 7200 | |
| }, | |
| { | |
| "epoch": 0.06075536394959802, | |
| "grad_norm": 0.0576171875, | |
| "learning_rate": 0.001093641489534351, | |
| "loss": 0.0685, | |
| "step": 7300 | |
| }, | |
| { | |
| "epoch": 0.06158762920918155, | |
| "grad_norm": 0.052490234375, | |
| "learning_rate": 0.001070276188945293, | |
| "loss": 0.0666, | |
| "step": 7400 | |
| }, | |
| { | |
| "epoch": 0.062419894468765086, | |
| "grad_norm": 0.043212890625, | |
| "learning_rate": 0.00104687226246994, | |
| "loss": 0.0692, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.06325215972834862, | |
| "grad_norm": 0.04052734375, | |
| "learning_rate": 0.0010234425736032607, | |
| "loss": 0.069, | |
| "step": 7600 | |
| }, | |
| { | |
| "epoch": 0.06408442498793215, | |
| "grad_norm": 0.06298828125, | |
| "learning_rate": 0.001, | |
| "loss": 0.0682, | |
| "step": 7700 | |
| }, | |
| { | |
| "epoch": 0.06491669024751569, | |
| "grad_norm": 0.04736328125, | |
| "learning_rate": 0.0009765574263967396, | |
| "loss": 0.0676, | |
| "step": 7800 | |
| }, | |
| { | |
| "epoch": 0.06574895550709922, | |
| "grad_norm": 0.064453125, | |
| "learning_rate": 0.0009531277375300599, | |
| "loss": 0.0685, | |
| "step": 7900 | |
| }, | |
| { | |
| "epoch": 0.06658122076668276, | |
| "grad_norm": 0.037109375, | |
| "learning_rate": 0.0009297238110547074, | |
| "loss": 0.0677, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.06658122076668276, | |
| "eval_peoplespeech-clean-transcription_loss": 1.5178765058517456, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.3361, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.855, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.857, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.06741348602626629, | |
| "grad_norm": 0.041259765625, | |
| "learning_rate": 0.0009063585104656494, | |
| "loss": 0.068, | |
| "step": 8100 | |
| }, | |
| { | |
| "epoch": 0.06824575128584982, | |
| "grad_norm": 0.0390625, | |
| "learning_rate": 0.0008830446780279176, | |
| "loss": 0.0681, | |
| "step": 8200 | |
| }, | |
| { | |
| "epoch": 0.06907801654543336, | |
| "grad_norm": 0.043212890625, | |
| "learning_rate": 0.0008597951277181142, | |
| "loss": 0.0677, | |
| "step": 8300 | |
| }, | |
| { | |
| "epoch": 0.06991028180501689, | |
| "grad_norm": 0.054443359375, | |
| "learning_rate": 0.0008366226381814697, | |
| "loss": 0.0679, | |
| "step": 8400 | |
| }, | |
| { | |
| "epoch": 0.07074254706460043, | |
| "grad_norm": 0.036865234375, | |
| "learning_rate": 0.000813539945708319, | |
| "loss": 0.0685, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.07157481232418396, | |
| "grad_norm": 0.036865234375, | |
| "learning_rate": 0.0007905597372338558, | |
| "loss": 0.0678, | |
| "step": 8600 | |
| }, | |
| { | |
| "epoch": 0.07240707758376749, | |
| "grad_norm": 0.03466796875, | |
| "learning_rate": 0.0007676946433650169, | |
| "loss": 0.0671, | |
| "step": 8700 | |
| }, | |
| { | |
| "epoch": 0.07323934284335103, | |
| "grad_norm": 0.03857421875, | |
| "learning_rate": 0.0007449572314383236, | |
| "loss": 0.0659, | |
| "step": 8800 | |
| }, | |
| { | |
| "epoch": 0.07407160810293456, | |
| "grad_norm": 0.04248046875, | |
| "learning_rate": 0.0007223599986124993, | |
| "loss": 0.0657, | |
| "step": 8900 | |
| }, | |
| { | |
| "epoch": 0.07490387336251811, | |
| "grad_norm": 0.0517578125, | |
| "learning_rate": 0.0006999153649996595, | |
| "loss": 0.0662, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.07490387336251811, | |
| "eval_peoplespeech-clean-transcription_loss": 1.49990975856781, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.6057, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.663, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.833, | |
| "step": 9000 | |
| }, | |
| { | |
| "epoch": 0.07573613862210164, | |
| "grad_norm": 0.05419921875, | |
| "learning_rate": 0.0006776356668388464, | |
| "loss": 0.0659, | |
| "step": 9100 | |
| }, | |
| { | |
| "epoch": 0.07656840388168518, | |
| "grad_norm": 0.041015625, | |
| "learning_rate": 0.0006555331497156671, | |
| "loss": 0.0656, | |
| "step": 9200 | |
| }, | |
| { | |
| "epoch": 0.07740066914126871, | |
| "grad_norm": 0.08203125, | |
| "learning_rate": 0.0006336199618317538, | |
| "loss": 0.0663, | |
| "step": 9300 | |
| }, | |
| { | |
| "epoch": 0.07823293440085224, | |
| "grad_norm": 0.03271484375, | |
| "learning_rate": 0.0006119081473277501, | |
| "loss": 0.0663, | |
| "step": 9400 | |
| }, | |
| { | |
| "epoch": 0.07906519966043578, | |
| "grad_norm": 0.043212890625, | |
| "learning_rate": 0.0005904096396634935, | |
| "loss": 0.0642, | |
| "step": 9500 | |
| }, | |
| { | |
| "epoch": 0.07989746492001931, | |
| "grad_norm": 0.03271484375, | |
| "learning_rate": 0.0005691362550590297, | |
| "loss": 0.0644, | |
| "step": 9600 | |
| }, | |
| { | |
| "epoch": 0.08072973017960285, | |
| "grad_norm": 0.047607421875, | |
| "learning_rate": 0.0005480996860000663, | |
| "loss": 0.0648, | |
| "step": 9700 | |
| }, | |
| { | |
| "epoch": 0.08156199543918638, | |
| "grad_norm": 0.03955078125, | |
| "learning_rate": 0.0005273114948114346, | |
| "loss": 0.0655, | |
| "step": 9800 | |
| }, | |
| { | |
| "epoch": 0.08239426069876991, | |
| "grad_norm": 0.048095703125, | |
| "learning_rate": 0.0005067831073020928, | |
| "loss": 0.0651, | |
| "step": 9900 | |
| }, | |
| { | |
| "epoch": 0.08322652595835345, | |
| "grad_norm": 0.0299072265625, | |
| "learning_rate": 0.00048652580648515787, | |
| "loss": 0.066, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.08322652595835345, | |
| "eval_peoplespeech-clean-transcription_loss": 1.489192247390747, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.4911, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.743, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.843, | |
| "step": 10000 | |
| }, | |
| { | |
| "epoch": 0.08405879121793698, | |
| "grad_norm": 0.055908203125, | |
| "learning_rate": 0.0004665507263764299, | |
| "loss": 0.0653, | |
| "step": 10100 | |
| }, | |
| { | |
| "epoch": 0.08489105647752052, | |
| "grad_norm": 0.039794921875, | |
| "learning_rate": 0.0004468688458748006, | |
| "loss": 0.065, | |
| "step": 10200 | |
| }, | |
| { | |
| "epoch": 0.08572332173710405, | |
| "grad_norm": 0.05322265625, | |
| "learning_rate": 0.0004274909827279283, | |
| "loss": 0.0654, | |
| "step": 10300 | |
| }, | |
| { | |
| "epoch": 0.08655558699668758, | |
| "grad_norm": 0.0419921875, | |
| "learning_rate": 0.0004084277875864776, | |
| "loss": 0.0648, | |
| "step": 10400 | |
| }, | |
| { | |
| "epoch": 0.08738785225627112, | |
| "grad_norm": 0.050537109375, | |
| "learning_rate": 0.00038968973815020803, | |
| "loss": 0.0643, | |
| "step": 10500 | |
| }, | |
| { | |
| "epoch": 0.08822011751585465, | |
| "grad_norm": 0.0458984375, | |
| "learning_rate": 0.00037128713340911534, | |
| "loss": 0.0669, | |
| "step": 10600 | |
| }, | |
| { | |
| "epoch": 0.08905238277543819, | |
| "grad_norm": 0.05029296875, | |
| "learning_rate": 0.00035323008798280133, | |
| "loss": 0.0657, | |
| "step": 10700 | |
| }, | |
| { | |
| "epoch": 0.08988464803502172, | |
| "grad_norm": 0.039794921875, | |
| "learning_rate": 0.00033552852656117837, | |
| "loss": 0.0646, | |
| "step": 10800 | |
| }, | |
| { | |
| "epoch": 0.09071691329460525, | |
| "grad_norm": 0.04248046875, | |
| "learning_rate": 0.00031819217844956217, | |
| "loss": 0.0636, | |
| "step": 10900 | |
| }, | |
| { | |
| "epoch": 0.09154917855418879, | |
| "grad_norm": 0.05029296875, | |
| "learning_rate": 0.00030123057222115836, | |
| "loss": 0.0658, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.09154917855418879, | |
| "eval_peoplespeech-clean-transcription_loss": 1.4850726127624512, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.5339, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.713, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.839, | |
| "step": 11000 | |
| }, | |
| { | |
| "epoch": 0.09238144381377232, | |
| "grad_norm": 0.039794921875, | |
| "learning_rate": 0.0002846530304798727, | |
| "loss": 0.0641, | |
| "step": 11100 | |
| }, | |
| { | |
| "epoch": 0.09321370907335585, | |
| "grad_norm": 0.048095703125, | |
| "learning_rate": 0.00026846866473633125, | |
| "loss": 0.065, | |
| "step": 11200 | |
| }, | |
| { | |
| "epoch": 0.09404597433293939, | |
| "grad_norm": 0.0458984375, | |
| "learning_rate": 0.00025268637039992293, | |
| "loss": 0.065, | |
| "step": 11300 | |
| }, | |
| { | |
| "epoch": 0.09487823959252292, | |
| "grad_norm": 0.045166015625, | |
| "learning_rate": 0.00023731482188961818, | |
| "loss": 0.0654, | |
| "step": 11400 | |
| }, | |
| { | |
| "epoch": 0.09571050485210646, | |
| "grad_norm": 0.04150390625, | |
| "learning_rate": 0.00022236246786624792, | |
| "loss": 0.0661, | |
| "step": 11500 | |
| }, | |
| { | |
| "epoch": 0.09654277011169, | |
| "grad_norm": 0.036376953125, | |
| "learning_rate": 0.00020783752658887068, | |
| "loss": 0.066, | |
| "step": 11600 | |
| }, | |
| { | |
| "epoch": 0.09737503537127354, | |
| "grad_norm": 0.044189453125, | |
| "learning_rate": 0.0001937479813977703, | |
| "loss": 0.0649, | |
| "step": 11700 | |
| }, | |
| { | |
| "epoch": 0.09820730063085707, | |
| "grad_norm": 0.037353515625, | |
| "learning_rate": 0.00018010157632657541, | |
| "loss": 0.0636, | |
| "step": 11800 | |
| }, | |
| { | |
| "epoch": 0.0990395658904406, | |
| "grad_norm": 0.0341796875, | |
| "learning_rate": 0.00016690581184590858, | |
| "loss": 0.0649, | |
| "step": 11900 | |
| }, | |
| { | |
| "epoch": 0.09987183115002414, | |
| "grad_norm": 0.052001953125, | |
| "learning_rate": 0.00015416794074090258, | |
| "loss": 0.0663, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.09987183115002414, | |
| "eval_peoplespeech-clean-transcription_loss": 1.485172986984253, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.8864, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.474, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.809, | |
| "step": 12000 | |
| }, | |
| { | |
| "epoch": 0.10070409640960767, | |
| "grad_norm": 0.0400390625, | |
| "learning_rate": 0.00014189496412485593, | |
| "loss": 0.0634, | |
| "step": 12100 | |
| }, | |
| { | |
| "epoch": 0.10153636166919121, | |
| "grad_norm": 0.049072265625, | |
| "learning_rate": 0.00013009362759120978, | |
| "loss": 0.0641, | |
| "step": 12200 | |
| }, | |
| { | |
| "epoch": 0.10236862692877474, | |
| "grad_norm": 0.054931640625, | |
| "learning_rate": 0.00011877041750597173, | |
| "loss": 0.0635, | |
| "step": 12300 | |
| }, | |
| { | |
| "epoch": 0.10320089218835828, | |
| "grad_norm": 0.041015625, | |
| "learning_rate": 0.00010793155744261352, | |
| "loss": 0.0656, | |
| "step": 12400 | |
| }, | |
| { | |
| "epoch": 0.10403315744794181, | |
| "grad_norm": 0.042724609375, | |
| "learning_rate": 9.758300476141169e-05, | |
| "loss": 0.0658, | |
| "step": 12500 | |
| }, | |
| { | |
| "epoch": 0.10486542270752534, | |
| "grad_norm": 0.046875, | |
| "learning_rate": 8.773044733510338e-05, | |
| "loss": 0.0655, | |
| "step": 12600 | |
| }, | |
| { | |
| "epoch": 0.10569768796710888, | |
| "grad_norm": 0.052001953125, | |
| "learning_rate": 7.837930042266262e-05, | |
| "loss": 0.0652, | |
| "step": 12700 | |
| }, | |
| { | |
| "epoch": 0.10652995322669241, | |
| "grad_norm": 0.044677734375, | |
| "learning_rate": 6.953470369291348e-05, | |
| "loss": 0.0652, | |
| "step": 12800 | |
| }, | |
| { | |
| "epoch": 0.10736221848627595, | |
| "grad_norm": 0.04443359375, | |
| "learning_rate": 6.120151839961363e-05, | |
| "loss": 0.0639, | |
| "step": 12900 | |
| }, | |
| { | |
| "epoch": 0.10819448374585948, | |
| "grad_norm": 0.03955078125, | |
| "learning_rate": 5.338432470956589e-05, | |
| "loss": 0.0648, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.10819448374585948, | |
| "eval_peoplespeech-clean-transcription_loss": 1.4848434925079346, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.5215, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.722, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.84, | |
| "step": 13000 | |
| }, | |
| { | |
| "epoch": 0.10902674900544301, | |
| "grad_norm": 0.04541015625, | |
| "learning_rate": 4.6087419185220966e-05, | |
| "loss": 0.0632, | |
| "step": 13100 | |
| }, | |
| { | |
| "epoch": 0.10985901426502655, | |
| "grad_norm": 0.050048828125, | |
| "learning_rate": 3.931481242315993e-05, | |
| "loss": 0.0624, | |
| "step": 13200 | |
| }, | |
| { | |
| "epoch": 0.11069127952461008, | |
| "grad_norm": 0.0439453125, | |
| "learning_rate": 3.307022684974936e-05, | |
| "loss": 0.0649, | |
| "step": 13300 | |
| }, | |
| { | |
| "epoch": 0.11152354478419362, | |
| "grad_norm": 0.047119140625, | |
| "learning_rate": 2.7357094675186987e-05, | |
| "loss": 0.0644, | |
| "step": 13400 | |
| }, | |
| { | |
| "epoch": 0.11235581004377715, | |
| "grad_norm": 0.042236328125, | |
| "learning_rate": 2.2178556007054874e-05, | |
| "loss": 0.0644, | |
| "step": 13500 | |
| }, | |
| { | |
| "epoch": 0.11318807530336068, | |
| "grad_norm": 0.03564453125, | |
| "learning_rate": 1.7537457124423894e-05, | |
| "loss": 0.0664, | |
| "step": 13600 | |
| }, | |
| { | |
| "epoch": 0.11402034056294422, | |
| "grad_norm": 0.05859375, | |
| "learning_rate": 1.3436348913453578e-05, | |
| "loss": 0.0653, | |
| "step": 13700 | |
| }, | |
| { | |
| "epoch": 0.11485260582252775, | |
| "grad_norm": 0.046875, | |
| "learning_rate": 9.877485465349056e-06, | |
| "loss": 0.0646, | |
| "step": 13800 | |
| }, | |
| { | |
| "epoch": 0.11568487108211128, | |
| "grad_norm": 0.030517578125, | |
| "learning_rate": 6.862822837445881e-06, | |
| "loss": 0.0663, | |
| "step": 13900 | |
| }, | |
| { | |
| "epoch": 0.11651713634169482, | |
| "grad_norm": 0.03759765625, | |
| "learning_rate": 4.394017978101905e-06, | |
| "loss": 0.0648, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.11651713634169482, | |
| "eval_peoplespeech-clean-transcription_loss": 1.4849343299865723, | |
| "eval_peoplespeech-clean-transcription_model_preparation_time": 0.0065, | |
| "eval_peoplespeech-clean-transcription_runtime": 9.3501, | |
| "eval_peoplespeech-clean-transcription_samples_per_second": 6.845, | |
| "eval_peoplespeech-clean-transcription_steps_per_second": 0.856, | |
| "step": 14000 | |
| }, | |
| { | |
| "epoch": 0.11734940160127835, | |
| "grad_norm": 0.045166015625, | |
| "learning_rate": 2.472427815989886e-06, | |
| "loss": 0.0636, | |
| "step": 14100 | |
| }, | |
| { | |
| "epoch": 0.1181816668608619, | |
| "grad_norm": 0.0361328125, | |
| "learning_rate": 1.099108514288627e-06, | |
| "loss": 0.0646, | |
| "step": 14200 | |
| }, | |
| { | |
| "epoch": 0.11901393212044543, | |
| "grad_norm": 0.04052734375, | |
| "learning_rate": 2.748148901841052e-07, | |
| "loss": 0.0645, | |
| "step": 14300 | |
| }, | |
| { | |
| "epoch": 0.11984619738002897, | |
| "grad_norm": 0.056396484375, | |
| "learning_rate": 0.0, | |
| "loss": 0.0661, | |
| "step": 14400 | |
| } | |
| ], | |
| "logging_steps": 100, | |
| "max_steps": 14400, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 3600, | |
| "stateful_callbacks": { | |
| "TrainerControl": { | |
| "args": { | |
| "should_epoch_stop": false, | |
| "should_evaluate": false, | |
| "should_log": false, | |
| "should_save": true, | |
| "should_training_stop": true | |
| }, | |
| "attributes": {} | |
| } | |
| }, | |
| "total_flos": 9.254469835026432e+17, | |
| "train_batch_size": 24, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |