Files
100k_warmup0.05__Qwen3-8B/train_results.json

12 lines
397 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 91.47451401577693,
"achieved_tflops_per_gpu_theoretical": 13518477.768859137,
"epoch": 7.0,
"mfu_percent": 6.464629965779288,
"mfu_percent_theoretical": 955369.453629621,
"total_flos": 1.326951254196224e+16,
"train_loss": 0.0,
"train_runtime": 1.1333,
"train_samples_per_second": 510432.869,
"train_steps_per_second": 3990.218
}