{ "achieved_tflops_per_gpu": 0.0016902134410301918, "achieved_tflops_per_gpu_theoretical": 931.6737547763712, "epoch": 6.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.33341556787490845, "mfu_percent": 0.00011944971314701003, "mfu_percent_theoretical": 65.8426681820757, "total_flos": 235219454066688.0, "train_loss": 0.41497956364475497, "train_runtime": 8697.8458, "train_samples_per_second": 4.4, "train_steps_per_second": 0.275, "valid_targets_mean": 1707.1, "valid_targets_min": 654 }