{ "achieved_tflops_per_gpu": 0.0037947999271627413, "achieved_tflops_per_gpu_theoretical": 508.1856844234752, "epoch": 7.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.13111908733844757, "mfu_percent": 0.0002681837404355294, "mfu_percent_theoretical": 35.914182644768566, "total_flos": 2030698249125888.0, "train_loss": 0.1592172774357645, "train_runtime": 33445.4103, "train_samples_per_second": 2.404, "train_steps_per_second": 0.15, "valid_targets_mean": 5482.0, "valid_targets_min": 621 }