{"_step":1,"total_flos":3.35015804141568e+16,"train_steps_per_second":0.088,"_timestamp":1.7312532738257406e+09,"train/global_step":990,"train/loss":Infinity,"_runtime":11255.688741316,"_wandb":{"runtime":11258},"train/epoch":4.982698961937716,"train/grad_norm":NaN,"train_runtime":11256.003,"train_samples_per_second":5.649,"train/learning_rate":2e-05}