{ "epoch": 4.637143519591931, "eval_loss": 0.15925397718542236, "eval_runtime": 61.5353, "eval_samples_per_second": 675.807, "eval_steps_per_second": 0.666, "perplexity": 1.1726357315864648, "total_flos": 2.3231400526217216e+17, "train_loss": 0.13326009378433226, "train_runtime": 41368.3669, "train_samples_per_second": 495.064, "train_steps_per_second": 0.242 }