| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 25.0, | |
| "global_step": 1500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 1.0, | |
| "eval_loss": 2.974224090576172, | |
| "eval_runtime": 5.5654, | |
| "eval_samples_per_second": 113.379, | |
| "eval_steps_per_second": 14.195, | |
| "eval_wer": 1.0, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 2.0, | |
| "eval_loss": 2.769304037094116, | |
| "eval_runtime": 5.1383, | |
| "eval_samples_per_second": 122.802, | |
| "eval_steps_per_second": 15.375, | |
| "eval_wer": 0.99901655466317, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 3.0, | |
| "eval_loss": 1.608498215675354, | |
| "eval_runtime": 5.1567, | |
| "eval_samples_per_second": 122.366, | |
| "eval_steps_per_second": 15.32, | |
| "eval_wer": 0.929191935748238, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 4.0, | |
| "eval_loss": 1.4975178241729736, | |
| "eval_runtime": 5.1972, | |
| "eval_samples_per_second": 121.412, | |
| "eval_steps_per_second": 15.201, | |
| "eval_wer": 0.9142763481396492, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 5.0, | |
| "eval_loss": 1.4740536212921143, | |
| "eval_runtime": 5.1873, | |
| "eval_samples_per_second": 121.642, | |
| "eval_steps_per_second": 15.229, | |
| "eval_wer": 0.8701852155384363, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 6.0, | |
| "eval_loss": 1.5547226667404175, | |
| "eval_runtime": 5.2242, | |
| "eval_samples_per_second": 120.784, | |
| "eval_steps_per_second": 15.122, | |
| "eval_wer": 0.88264218980495, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 7.0, | |
| "eval_loss": 1.385581374168396, | |
| "eval_runtime": 5.194, | |
| "eval_samples_per_second": 121.486, | |
| "eval_steps_per_second": 15.21, | |
| "eval_wer": 0.8431404687756106, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 8.0, | |
| "eval_loss": 1.6811991930007935, | |
| "eval_runtime": 5.2358, | |
| "eval_samples_per_second": 120.516, | |
| "eval_steps_per_second": 15.088, | |
| "eval_wer": 0.8677266021963612, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 8.33, | |
| "learning_rate": 0.00021716666666666667, | |
| "loss": 1.4916, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 9.0, | |
| "eval_loss": 1.651587724685669, | |
| "eval_runtime": 5.1966, | |
| "eval_samples_per_second": 121.426, | |
| "eval_steps_per_second": 15.202, | |
| "eval_wer": 0.8567447959350926, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 10.0, | |
| "eval_loss": 1.698971152305603, | |
| "eval_runtime": 5.2046, | |
| "eval_samples_per_second": 121.24, | |
| "eval_steps_per_second": 15.179, | |
| "eval_wer": 0.8611702999508277, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 11.0, | |
| "eval_loss": 1.6570122241973877, | |
| "eval_runtime": 5.2467, | |
| "eval_samples_per_second": 120.267, | |
| "eval_steps_per_second": 15.057, | |
| "eval_wer": 0.8464186198983773, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 12.0, | |
| "eval_loss": 1.7964118719100952, | |
| "eval_runtime": 5.2019, | |
| "eval_samples_per_second": 121.302, | |
| "eval_steps_per_second": 15.187, | |
| "eval_wer": 0.8636289132929028, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 13.0, | |
| "eval_loss": 1.72978937625885, | |
| "eval_runtime": 5.2508, | |
| "eval_samples_per_second": 120.172, | |
| "eval_steps_per_second": 15.045, | |
| "eval_wer": 0.8367480740862153, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 14.0, | |
| "eval_loss": 1.6711727380752563, | |
| "eval_runtime": 5.2066, | |
| "eval_samples_per_second": 121.192, | |
| "eval_steps_per_second": 15.173, | |
| "eval_wer": 0.8236354695951483, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 15.0, | |
| "eval_loss": 1.6306202411651611, | |
| "eval_runtime": 5.2186, | |
| "eval_samples_per_second": 120.913, | |
| "eval_steps_per_second": 15.138, | |
| "eval_wer": 0.8242910998197017, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 16.0, | |
| "eval_loss": 1.6020203828811646, | |
| "eval_runtime": 5.2523, | |
| "eval_samples_per_second": 120.138, | |
| "eval_steps_per_second": 15.041, | |
| "eval_wer": 0.812817570890018, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 16.67, | |
| "learning_rate": 0.00013383333333333332, | |
| "loss": 0.1786, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 17.0, | |
| "eval_loss": 1.7215255498886108, | |
| "eval_runtime": 5.2057, | |
| "eval_samples_per_second": 121.213, | |
| "eval_steps_per_second": 15.176, | |
| "eval_wer": 0.8308474020652352, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 18.0, | |
| "eval_loss": 1.7741730213165283, | |
| "eval_runtime": 5.2226, | |
| "eval_samples_per_second": 120.821, | |
| "eval_steps_per_second": 15.127, | |
| "eval_wer": 0.8380593345353221, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 19.0, | |
| "eval_loss": 1.7588073015213013, | |
| "eval_runtime": 5.2348, | |
| "eval_samples_per_second": 120.539, | |
| "eval_steps_per_second": 15.091, | |
| "eval_wer": 0.833633830519587, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 20.0, | |
| "eval_loss": 1.7859772443771362, | |
| "eval_runtime": 5.211, | |
| "eval_samples_per_second": 121.089, | |
| "eval_steps_per_second": 15.16, | |
| "eval_wer": 0.8344533683002786, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 21.0, | |
| "eval_loss": 1.8294274806976318, | |
| "eval_runtime": 5.2508, | |
| "eval_samples_per_second": 120.173, | |
| "eval_steps_per_second": 15.045, | |
| "eval_wer": 0.8367480740862153, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 22.0, | |
| "eval_loss": 1.7776710987091064, | |
| "eval_runtime": 5.2165, | |
| "eval_samples_per_second": 120.962, | |
| "eval_steps_per_second": 15.144, | |
| "eval_wer": 0.8226520242583183, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 23.0, | |
| "eval_loss": 1.8306204080581665, | |
| "eval_runtime": 5.2033, | |
| "eval_samples_per_second": 121.27, | |
| "eval_steps_per_second": 15.183, | |
| "eval_wer": 0.8290444189477135, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 24.0, | |
| "eval_loss": 1.858679175376892, | |
| "eval_runtime": 5.2509, | |
| "eval_samples_per_second": 120.17, | |
| "eval_steps_per_second": 15.045, | |
| "eval_wer": 0.8247828224881167, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 25.0, | |
| "learning_rate": 5.0499999999999994e-05, | |
| "loss": 0.1062, | |
| "step": 1500 | |
| } | |
| ], | |
| "max_steps": 1800, | |
| "num_train_epochs": 30, | |
| "total_flos": 5.842748350654745e+18, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |