{ "achieved_tflops_per_gpu": 8.960699285844338, "achieved_tflops_per_gpu_theoretical": 338.7904562823168, "epoch": 5.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.15064625442028046, "mfu_percent": 0.6332649671974797, "mfu_percent_theoretical": 23.94278842984571, "total_flos": 1.157917389896024e+19, "train_loss": 0.1381187802789235, "train_runtime": 40381.802, "train_samples_per_second": 9.6, "train_steps_per_second": 0.1, "valid_targets_mean": 4476.7, "valid_targets_min": 258 }