Files
g1_gptlong_top8_8b/train_results.json
ModelHub XC 3ac888f875 初始化项目,由ModelHub XC社区提供模型
Model: DCAgent/g1_gptlong_top8_8b
Source: Original Platform
2026-06-23 12:12:40 +08:00

16 lines
542 B
JSON

{
"achieved_tflops_per_gpu": 13.233347482143916,
"achieved_tflops_per_gpu_theoretical": 718.2357673185115,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.01995828002691269,
"mfu_percent": 0.9352189033317254,
"mfu_percent_theoretical": 50.758711471272896,
"total_flos": 6.2847112177181e+18,
"train_loss": 0.030045701052888524,
"train_runtime": 9894.0562,
"train_samples_per_second": 20.297,
"train_steps_per_second": 0.212,
"valid_targets_mean": 2201.4,
"valid_targets_min": 325
}