初始化项目,由ModelHub XC社区提供模型

Model: laion/swesmith-1000__Qwen3-8B
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-05-05 05:35:25 +08:00
commit 0e6a69ed25
23 changed files with 152651 additions and 0 deletions

16
train_results.json Normal file
View File

@@ -0,0 +1,16 @@
{
"achieved_tflops_per_gpu": 2.9646103807892437,
"achieved_tflops_per_gpu_theoretical": 118.57665969881089,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.09160405397415161,
"mfu_percent": 0.20951310111584762,
"mfu_percent_theoretical": 8.379975950446,
"total_flos": 2.1021193659010253e+17,
"train_loss": 0.20924820180063125,
"train_runtime": 2215.847,
"train_samples_per_second": 3.159,
"train_steps_per_second": 0.035,
"valid_targets_mean": 6769.5,
"valid_targets_min": 1162
}