Files
sera-3160__Qwen3-8B/train_results.json
ModelHub XC d349d69b7d 初始化项目,由ModelHub XC社区提供模型
Model: laion/sera-3160__Qwen3-8B
Source: Original Platform
2026-05-27 08:56:19 +08:00

12 lines
396 B
JSON

{
"achieved_tflops_per_gpu": 11738.976127273472,
"achieved_tflops_per_gpu_theoretical": 163791.63399424887,
"epoch": 7.0,
"mfu_percent": 829.6096203020122,
"mfu_percent_theoretical": 11575.380494293206,
"total_flos": 1.7948424939556045e+18,
"train_loss": 0.0,
"train_runtime": 4.778,
"train_samples_per_second": 4629.543,
"train_steps_per_second": 48.346
}