{ "achieved_tflops_per_gpu": 3.594365942296465, "achieved_tflops_per_gpu_theoretical": 135.51618251292672, "epoch": 7.0, "loss_nan_ranks": 0, "loss_rank_avg": 0.6334422826766968, "mfu_percent": 0.2540187945085841, "mfu_percent_theoretical": 9.577115371938284, "total_flos": 8.133461226500915e+16, "train_loss": 0.7130181619099208, "train_runtime": 707.1363, "train_samples_per_second": 3.128, "train_steps_per_second": 0.04, "valid_targets_mean": 12522.3, "valid_targets_min": 2429 }