{ "all_params": 6051429888, "epoch": 0.9945945945945946, "memory_footprint": 8753740800, "total_flos": 1.3431114641260646e+17, "train_loss": 0.4029887131374815, "train_runtime": 1125.7865, "train_samples_per_second": 0.822, "train_steps_per_second": 0.102, "trainable_params": 54018048, "trainable_params_percent": 0.8926493242054728 }