{ "epoch": 4.0, "eval_loss": 2.4195876121520996, "eval_runtime": 85.2391, "eval_samples_per_second": 85.63, "eval_steps_per_second": 0.68, "perplexity": 11.241222615286187, "total_flos": 429332983971840.0, "train_loss": 2.284057284167586, "train_runtime": 15107.0718, "train_samples_per_second": 17.393, "train_steps_per_second": 0.272 }