Files
sera-316-opt1k__Qwen3-8B/train_results.json

12 lines
393 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 5659.383155844384,
"achieved_tflops_per_gpu_theoretical": 104825.15507831163,
"epoch": 7.0,
"mfu_percent": 399.9564067734547,
"mfu_percent_theoretical": 7408.13816807856,
"total_flos": 1.6389573619325338e+17,
"train_loss": 0.0,
"train_runtime": 0.905,
"train_samples_per_second": 2444.322,
"train_steps_per_second": 30.941
}