{ "epoch": 1.0, "eval_loss": 2.008072853088379, "eval_runtime": 29.0587, "eval_samples_per_second": 40.47, "eval_steps_per_second": 5.059, "perplexity": 7.448948288433685, "total_flos": 9.597350805504e+16, "train_loss": 2.090950048406474, "train_runtime": 846.2233, "train_samples_per_second": 12.453, "train_steps_per_second": 6.226 }