| {"eval/runtime":4130.3962,"train_samples_per_second":1.426,"total_flos":5.435317790834688e+20,"_step":420,"eval/steps_per_second":0.413,"train/epoch":5.148,"train_runtime":112211.4993,"train_loss":0.08380846776664257,"eval/wer":7.07244677342519,"_wandb":{"runtime":112523},"_runtime":112210.651756658,"train/grad_norm":1.3872036933898926,"eval/samples_per_second":3.3,"eval/loss":0.12359699606895447,"train/learning_rate":1.381578947368421e-09,"_timestamp":1.728249865647773e+09,"train_steps_per_second":0.089,"train/loss":0.0504,"train/global_step":10000} |