{ "aux_attn_accuracy": 0.9422222224871317, "aux_loss": 0.2298412849009037, "epoch": 1.0, "lm_loss": 0.0855401718375894, "mean_token_accuracy": 0.9700513021151225, "total_flos": 1.066126245903519e+19, "total_loss": 0.10852430010680109, "train_loss": 0.027537114826725155, "train_runtime": 28154.6326, "train_samples_per_second": 1.758, "train_steps_per_second": 0.11 }