{ "epoch": 1.0, "eval_loss": 1.9345422983169556, "eval_runtime": 24.7701, "eval_samples_per_second": 40.412, "eval_steps_per_second": 5.087, "perplexity": 6.920875633911888, "total_flos": 9.788605661184e+16, "train_loss": 2.1113919824927456, "train_runtime": 862.752, "train_samples_per_second": 12.458, "train_steps_per_second": 6.229 }