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

12 lines
396 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 11738.976127273472,
"achieved_tflops_per_gpu_theoretical": 163791.63399424887,
"epoch": 7.0,
"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
}