Files
a1-nemo_prism_math/train_results.json

16 lines
545 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 0.0038796584681248967,
"achieved_tflops_per_gpu_theoretical": 1656.685331220529,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.16949467360973358,
"mfu_percent": 0.0002741808104681906,
"mfu_percent_theoretical": 117.08023542194552,
"total_flos": 488803397074944.0,
"train_loss": 0.26370995024582006,
"train_runtime": 7874.4592,
"train_samples_per_second": 7.812,
"train_steps_per_second": 0.489,
"valid_targets_mean": 2481.9,
"valid_targets_min": 1022
}