初始化项目,由ModelHub XC社区提供模型

Model: laion/sera-3160__Qwen3-8B
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-05-27 08:56:19 +08:00
commit d349d69b7d
23 changed files with 153020 additions and 0 deletions

16
all_results.json Normal file
View File

@@ -0,0 +1,16 @@
{
"achieved_tflops_per_gpu": 11738.976127273472,
"achieved_tflops_per_gpu_theoretical": 163791.63399424887,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.05725972726941109,
"mfu_percent": 829.6096203020122,
"mfu_percent_theoretical": 11575.380494293206,
"total_flos": 1.7948424939556045e+18,
"train_loss": 0.0,
"train_runtime": 4.778,
"train_samples_per_second": 4629.543,
"train_steps_per_second": 48.346,
"valid_targets_mean": 15044.6,
"valid_targets_min": 4391
}