{ "epoch": 3.0, "eval_loss": 2.7752132415771484, "eval_runtime": 177.8773, "eval_samples_per_second": 10.592, "eval_steps_per_second": 2.648, "perplexity": 16.042047462451425, "total_flos": 83202240675840.0, "train_loss": 1.594443890733539, "train_runtime": 15602.0514, "train_samples_per_second": 3.26, "train_steps_per_second": 0.102 }