Files
sera-3160__Qwen3-8B/all_results.json

16 lines
529 B
JSON
Raw Permalink Normal View History

{
"achieved_tflops_per_gpu": 11738.976127273472,
"achieved_tflops_per_gpu_theoretical": 163791.63399424887,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05725972726941109,
"mfu_percent": 829.6096203020122,
"mfu_percent_theoretical": 11575.380494293206,
"total_flos": 1.7948424939556045e+18,
"train_loss": 0.0,
"train_runtime": 4.778,
"train_samples_per_second": 4629.543,
"train_steps_per_second": 48.346,
"valid_targets_mean": 15044.6,
"valid_targets_min": 4391
}