Files
Qwen3-8B_perturbed-docker-e…/train_results.json

16 lines
545 B
JSON
Raw Permalink Normal View History

{
"achieved_tflops_per_gpu": 0.001239108303511846,
"achieved_tflops_per_gpu_theoretical": 196.49846464374374,
"epoch": 8.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.02538919635117054,
"mfu_percent": 0.0003971500972794378,
"mfu_percent_theoretical": 62.98027712940505,
"total_flos": 1518927301312512.0,
"train_loss": 0.13197601472825876,
"train_runtime": 38306.9648,
"train_samples_per_second": 1.867,
"train_steps_per_second": 0.058,
"valid_targets_mean": 3632.8,
"valid_targets_min": 732
}