{"train/epoch":2.98961937716263,"_step":1,"train/loss":Infinity,"train/grad_norm":NaN,"train_runtime":6182.0596,"train/learning_rate":2e-05,"_wandb":{"runtime":6182},"_runtime":6181.362831337,"train_steps_per_second":0.096,"train_samples_per_second":6.171,"train/global_step":594,"_timestamp":1.7311751374258363e+09,"total_flos":2.010094824849408e+16}