Files
sera-1000-opt1k__Qwen3-8B/train_results.json

12 lines
397 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 18537.362817194047,
"achieved_tflops_per_gpu_theoretical": 270754.5568517019,
"epoch": 7.0,
"mfu_percent": 1310.0609764801447,
"mfu_percent_theoretical": 19134.597657364095,
"total_flos": 5.7154397037972685e+17,
"train_loss": 0.0,
"train_runtime": 0.9635,
"train_samples_per_second": 7265.245,
"train_steps_per_second": 79.918
}