{ "epoch": 1.0, "eval_loss": 2.0564818382263184, "eval_runtime": 17.4644, "eval_samples_per_second": 22.217, "eval_steps_per_second": 2.806, "perplexity": 7.818414918958686, "total_flos": 4.092853911552e+16, "train_loss": 2.2367699406228705, "train_runtime": 595.6661, "train_samples_per_second": 7.544, "train_steps_per_second": 3.772 }