{ "all_params": 45957632, "epoch": 3.0, "memory_footprint": 136644608, "total_flos": 366452498497536.0, "train_loss": 2.2660372549837287, "train_runtime": 171.2707, "train_samples_per_second": 15.397, "train_steps_per_second": 1.927, "trainable_params": 1179648, "trainable_params_percent": 2.5668163233475565 }