Files
nemotron-31600-opt100k__Qwe…/all_results.json
ModelHub XC b271f3bf83 初始化项目,由ModelHub XC社区提供模型
Model: laion/nemotron-31600-opt100k__Qwen3-8B
Source: Original Platform
2026-05-09 17:15:10 +08:00

16 lines
544 B
JSON

{
"achieved_tflops_per_gpu": 15.292781648433825,
"achieved_tflops_per_gpu_theoretical": 423.488070352896,
"epoch": 5.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.38277196884155273,
"mfu_percent": 1.0807619539529205,
"mfu_percent_theoretical": 29.928485537307136,
"total_flos": 6.453442034819662e+18,
"train_loss": 0.1001221828027205,
"train_runtime": 13187.2715,
"train_samples_per_second": 11.981,
"train_steps_per_second": 0.125,
"valid_targets_mean": 11949.7,
"valid_targets_min": 2234
}