{ "epoch": 88.0, "learning_rate": 1.0000000000000002e-07, "total_flos": 1.500719176717825e+20, "train_loss": 0.3047179739897961, "train_runtime": 30841.8451, "train_samples_per_second": 56.038, "train_steps_per_second": 0.88 }