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

12 lines
396 B
JSON
Raw Permalink Normal View History

{
"achieved_tflops_per_gpu": 2164.831536269501,
"achieved_tflops_per_gpu_theoretical": 110811.58244082017,
"epoch": 7.0,
"mfu_percent": 152.99162800491175,
"mfu_percent_theoretical": 7831.2072396339345,
"total_flos": 5.930599290241024e+16,
"train_loss": 0.0,
"train_runtime": 0.8561,
"train_samples_per_second": 2583.896,
"train_steps_per_second": 32.708
}