{ "epoch": 20.0, "eval_loss": 3.125084161758423, "eval_runtime": 16.4094, "eval_samples": 53825, "eval_samples_per_second": 3280.126, "eval_steps_per_second": 205.065, "perplexity": 22.761810686928044, "train_loss": 3.981161423732987, "train_runtime": 4340.1096, "train_samples": 215299, "train_samples_per_second": 992.136, "train_steps_per_second": 62.012 }