Files
sera-316__Qwen3-8B/train_results.json
ModelHub XC 1987e8e177 初始化项目,由ModelHub XC社区提供模型
Model: laion/sera-316__Qwen3-8B
Source: Original Platform
2026-05-27 08:56:19 +08:00

16 lines
542 B
JSON

{
"achieved_tflops_per_gpu": 6.997790647079837,
"achieved_tflops_per_gpu_theoretical": 128.74037338728039,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.27273428440093994,
"mfu_percent": 0.4945435086275503,
"mfu_percent_theoretical": 9.09825960334137,
"total_flos": 1.6389573619325338e+17,
"train_loss": 0.31328126788139343,
"train_runtime": 731.9084,
"train_samples_per_second": 3.022,
"train_steps_per_second": 0.038,
"valid_targets_mean": 13154.9,
"valid_targets_min": 5475
}