{ "epoch": 1.0, "eval_loss": 1.7617119550704956, "eval_runtime": 10.1058, "eval_samples_per_second": 13.754, "eval_steps_per_second": 1.781, "perplexity": 5.822396548451661, "total_flos": 1.17580762841088e+16, "train_loss": 2.0752741063396565, "train_runtime": 203.1277, "train_samples_per_second": 6.37, "train_steps_per_second": 3.185 }