Files
swesmith-unified-3160__Qwen…/all_results.json
ModelHub XC ac9268b670 初始化项目,由ModelHub XC社区提供模型
Model: laion/swesmith-unified-3160__Qwen3-8B
Source: Original Platform
2026-06-09 15:41:17 +08:00

16 lines
528 B
JSON

{
"achieved_tflops_per_gpu": 15460.43140421341,
"achieved_tflops_per_gpu_theoretical": 339888.1373606715,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09388194978237152,
"mfu_percent": 1092.6099932306297,
"mfu_percent_theoretical": 24020.363064358407,
"total_flos": 4.487730344563835e+17,
"train_loss": 0.0,
"train_runtime": 0.9071,
"train_samples_per_second": 8967.387,
"train_steps_per_second": 100.324,
"valid_targets_mean": 6277.6,
"valid_targets_min": 2461
}