{ "epoch": 1.0, "eval_loss": 1.6334210634231567, "eval_runtime": 29.1651, "eval_samples_per_second": 13.784, "eval_steps_per_second": 1.749, "perplexity": 5.121365298786188, "total_flos": 4.12623063416832e+16, "train_loss": 1.9041457484861346, "train_runtime": 710.2567, "train_samples_per_second": 6.393, "train_steps_per_second": 3.197 }