{ "achieved_tflops_per_gpu": 0.004432472230677925, "achieved_tflops_per_gpu_theoretical": 1253.5246882445722, "epoch": 7.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.13487885892391205, "mfu_percent": 0.00031324892089596644, "mfu_percent_theoretical": 88.58831719042912, "total_flos": 841177935118336.0, "train_loss": 0.10868970680736577, "train_runtime": 11861.0153, "train_samples_per_second": 5.919, "train_steps_per_second": 0.37, "valid_targets_mean": 2596.9, "valid_targets_min": 1254 }