| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9947774185525988, | |
| "global_step": 8500, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00015, | |
| "loss": 2.3077, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "eval_loss": 0.9413899183273315, | |
| "eval_phone_accuracy": 0.7518319068307362, | |
| "eval_runtime": 12.1406, | |
| "eval_samples_per_second": 16.474, | |
| "eval_steps_per_second": 4.118, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003, | |
| "loss": 1.1762, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "eval_loss": 0.6205331683158875, | |
| "eval_phone_accuracy": 0.8215267480076658, | |
| "eval_runtime": 10.7397, | |
| "eval_samples_per_second": 18.623, | |
| "eval_steps_per_second": 4.656, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00028011664899257687, | |
| "loss": 0.8235, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "eval_loss": 0.4551730751991272, | |
| "eval_phone_accuracy": 0.8648594743186174, | |
| "eval_runtime": 10.6304, | |
| "eval_samples_per_second": 18.814, | |
| "eval_steps_per_second": 4.703, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0002602332979851537, | |
| "loss": 0.6647, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "eval_loss": 0.3814973533153534, | |
| "eval_phone_accuracy": 0.8815525898176332, | |
| "eval_runtime": 11.7606, | |
| "eval_samples_per_second": 17.006, | |
| "eval_steps_per_second": 4.251, | |
| "step": 2000 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0002403499469777306, | |
| "loss": 0.5809, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "eval_loss": 0.3442990481853485, | |
| "eval_phone_accuracy": 0.8922361837370032, | |
| "eval_runtime": 11.2643, | |
| "eval_samples_per_second": 17.755, | |
| "eval_steps_per_second": 4.439, | |
| "step": 2500 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00022046659597030753, | |
| "loss": 0.5289, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "eval_loss": 0.314848929643631, | |
| "eval_phone_accuracy": 0.9002575509248419, | |
| "eval_runtime": 11.2871, | |
| "eval_samples_per_second": 17.719, | |
| "eval_steps_per_second": 4.43, | |
| "step": 3000 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002005832449628844, | |
| "loss": 0.4896, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "eval_loss": 0.30238407850265503, | |
| "eval_phone_accuracy": 0.9030845408742857, | |
| "eval_runtime": 11.2397, | |
| "eval_samples_per_second": 17.794, | |
| "eval_steps_per_second": 4.449, | |
| "step": 3500 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0001806998939554613, | |
| "loss": 0.4622, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "eval_loss": 0.2689047157764435, | |
| "eval_phone_accuracy": 0.9141496917193476, | |
| "eval_runtime": 12.0191, | |
| "eval_samples_per_second": 16.64, | |
| "eval_steps_per_second": 4.16, | |
| "step": 4000 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00016081654294803816, | |
| "loss": 0.4386, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "eval_loss": 0.2747265696525574, | |
| "eval_phone_accuracy": 0.9112533277834144, | |
| "eval_runtime": 11.09, | |
| "eval_samples_per_second": 18.034, | |
| "eval_steps_per_second": 4.509, | |
| "step": 4500 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00014093319194061505, | |
| "loss": 0.419, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "eval_loss": 0.2555970251560211, | |
| "eval_phone_accuracy": 0.9181213524458666, | |
| "eval_runtime": 11.6344, | |
| "eval_samples_per_second": 17.19, | |
| "eval_steps_per_second": 4.298, | |
| "step": 5000 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012104984093319194, | |
| "loss": 0.4032, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "eval_loss": 0.2447209358215332, | |
| "eval_phone_accuracy": 0.920003121829392, | |
| "eval_runtime": 10.8148, | |
| "eval_samples_per_second": 18.493, | |
| "eval_steps_per_second": 4.623, | |
| "step": 5500 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00010116648992576882, | |
| "loss": 0.3892, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "eval_loss": 0.23665592074394226, | |
| "eval_phone_accuracy": 0.9239054085694217, | |
| "eval_runtime": 10.9341, | |
| "eval_samples_per_second": 18.291, | |
| "eval_steps_per_second": 4.573, | |
| "step": 6000 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 8.12831389183457e-05, | |
| "loss": 0.3786, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "eval_loss": 0.2295096516609192, | |
| "eval_phone_accuracy": 0.9251801555711647, | |
| "eval_runtime": 10.8468, | |
| "eval_samples_per_second": 18.439, | |
| "eval_steps_per_second": 4.61, | |
| "step": 6500 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 6.139978791092258e-05, | |
| "loss": 0.3701, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "eval_loss": 0.22638411819934845, | |
| "eval_phone_accuracy": 0.9254229645238776, | |
| "eval_runtime": 10.8611, | |
| "eval_samples_per_second": 18.414, | |
| "eval_steps_per_second": 4.604, | |
| "step": 7000 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.151643690349946e-05, | |
| "loss": 0.3592, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "eval_loss": 0.2183924913406372, | |
| "eval_phone_accuracy": 0.9286401831473243, | |
| "eval_runtime": 11.5166, | |
| "eval_samples_per_second": 17.366, | |
| "eval_steps_per_second": 4.342, | |
| "step": 7500 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 2.163308589607635e-05, | |
| "loss": 0.3511, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "eval_loss": 0.2142486721277237, | |
| "eval_phone_accuracy": 0.9288309616101702, | |
| "eval_runtime": 11.0635, | |
| "eval_samples_per_second": 18.077, | |
| "eval_steps_per_second": 4.519, | |
| "step": 8000 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.7497348886532342e-06, | |
| "loss": 0.3459, | |
| "step": 8500 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "eval_loss": 0.2112182378768921, | |
| "eval_phone_accuracy": 0.9303051588230703, | |
| "eval_runtime": 10.7118, | |
| "eval_samples_per_second": 18.671, | |
| "eval_steps_per_second": 4.668, | |
| "step": 8500 | |
| } | |
| ], | |
| "max_steps": 8544, | |
| "num_train_epochs": 1, | |
| "total_flos": 1.2317797842209997e+19, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |