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

12 lines
393 B
JSON
Raw Normal View History

{
"achieved_tflops_per_gpu": 241.4141263070502,
"achieved_tflops_per_gpu_theoretical": 4685.472010384441,
"epoch": 7.0,
"mfu_percent": 17.06106899696468,
"mfu_percent_theoretical": 331.1287639847662,
"total_flos": 1.5644485162421453e+17,
"train_loss": 0.0,
"train_runtime": 20.2511,
"train_samples_per_second": 109.229,
"train_steps_per_second": 1.383
}