{ "achieved_tflops_per_gpu": 0.0021852834915228347, "achieved_tflops_per_gpu_theoretical": 470.9187342324204, "epoch": 7.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.2484266757965088, "mfu_percent": 0.00015443699586733815, "mfu_percent_theoretical": 33.28047591748554, "total_flos": 645639129989120.0, "train_loss": 0.38789241854065165, "train_runtime": 18465.5427, "train_samples_per_second": 2.225, "train_steps_per_second": 0.139, "valid_targets_mean": 4023.9, "valid_targets_min": 865 }