{ "epoch": 1.0, "eval_loss": 2.0057356357574463, "eval_runtime": 250.4952, "eval_samples_per_second": 78.716, "eval_steps_per_second": 9.841, "perplexity": 7.431558806811608, "total_flos": 1.1780246832876093e+18, "train_loss": 2.774943617143574, "train_runtime": 2530.0612, "train_samples_per_second": 31.21, "train_steps_per_second": 3.901 }