Files
Qwen3-8B_exp_tas_temp_0.5_t…/train_results.json

16 lines
548 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 0.0010161478779803959,
"achieved_tflops_per_gpu_theoretical": 220.21379658350594,
"epoch": 8.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.004353777505457401,
"mfu_percent": 0.00032568842242961405,
"mfu_percent_theoretical": 70.58134505881601,
"total_flos": 1257015450664960.0,
"train_loss": 0.07835676025835768,
"train_runtime": 38657.4963,
"train_samples_per_second": 2.085,
"train_steps_per_second": 0.065,
"valid_targets_mean": 3003.7,
"valid_targets_min": 671
}