{ "epoch": 94.0, "learning_rate": 1.0000000000000002e-07, "total_flos": 9.743166860733663e+19, "train_loss": 0.46343856458701926, "train_runtime": 26810.1674, "train_samples_per_second": 39.181, "train_steps_per_second": 2.451 }