Files
g1_clean_hybrid_25k_8b/train_results.json
ModelHub XC 86555376d8 初始化项目,由ModelHub XC社区提供模型
Model: DCAgent/g1_clean_hybrid_25k_8b
Source: Original Platform
2026-05-01 10:50:12 +08:00

16 lines
542 B
JSON

{
"achieved_tflops_per_gpu": 3.754126430475683,
"achieved_tflops_per_gpu_theoretical": 233.76541483479863,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.022832343354821205,
"mfu_percent": 0.2653092883728398,
"mfu_percent_theoretical": 16.520524016593544,
"total_flos": 4.836413774367818e+18,
"train_loss": 0.19618723029549,
"train_runtime": 26839.4318,
"train_samples_per_second": 6.577,
"train_steps_per_second": 0.069,
"valid_targets_mean": 5367.0,
"valid_targets_min": 1264
}