Files
swesmith-unified-316__Qwen3-8B/train_results.json

16 lines
539 B
JSON
Raw Permalink Normal View History

{
"achieved_tflops_per_gpu": 3.827406304137867,
"achieved_tflops_per_gpu_theoretical": 159.2315144526889,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.17928698658943176,
"mfu_percent": 0.2704880780309447,
"mfu_percent_theoretical": 11.253110562027485,
"total_flos": 3.664642942225613e+16,
"train_loss": 0.2080350241490773,
"train_runtime": 299.2107,
"train_samples_per_second": 2.69,
"train_steps_per_second": 0.047,
"valid_targets_mean": 6241.5,
"valid_targets_min": 1282
}