Files
a1-qasper/train_results.json

16 lines
544 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 0.0029185296391530946,
"achieved_tflops_per_gpu_theoretical": 514.9614935491215,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.19753853976726532,
"mfu_percent": 0.00020625651160092543,
"mfu_percent_theoretical": 36.39303841336548,
"total_flos": 1412455386578944.0,
"train_loss": 0.2913600100100261,
"train_runtime": 30247.581,
"train_samples_per_second": 2.432,
"train_steps_per_second": 0.152,
"valid_targets_mean": 3447.5,
"valid_targets_min": 627
}