{ "epoch": 4.0, "eval_loss": 2.323676109313965, "eval_runtime": 85.2953, "eval_samples_per_second": 86.699, "eval_steps_per_second": 0.68, "perplexity": 10.21315003740504, "total_flos": 434462785536000.0, "train_loss": 2.16538261238199, "train_runtime": 15200.3368, "train_samples_per_second": 17.512, "train_steps_per_second": 0.274 }