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

12 lines
395 B
JSON
Raw Permalink Normal View History

{
"achieved_tflops_per_gpu": 7507.569163932233,
"achieved_tflops_per_gpu_theoretical": 298122.04991018743,
"epoch": 7.0,
"mfu_percent": 530.5702589351401,
"mfu_percent_theoretical": 21068.696106727028,
"total_flos": 2.1021193659010253e+17,
"train_loss": 0.0,
"train_runtime": 0.875,
"train_samples_per_second": 7999.638,
"train_steps_per_second": 87.996
}