DCAgent/e1_gpt_long_sandboxes_2x_tacc-Qwen3-8B
1
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

初始化项目，由ModelHub XC社区提供模型

Browse Source 
Model: DCAgent/e1_gpt_long_sandboxes_2x_tacc-Qwen3-8B
Source: Original Platform
This commit is contained in:
ModelHub XC
2026-04-26 21:07:02 +08:00
commit f2421355ca
23 changed files with 174490 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
train_results.json Normal file
View File

@@ -0,0 +1,16 @@
{
"achieved_tflops_per_gpu": 0.0029313409083490745,
"achieved_tflops_per_gpu_theoretical": 470.91873423242043,
"epoch": 7.0,
"loss_nan_ranks": 0,
"loss_rank_avg": 0.4288143217563629,
"mfu_percent": 0.00020716190165011127,
"mfu_percent_theoretical": 33.280475917485546,
"total_flos": 4631781651513344.0,
"train_loss": 0.31498055520031176,
"train_runtime": 65837.0719,
"train_samples_per_second": 3.345,
"train_steps_per_second": 0.139,
"valid_targets_mean": 858.8,
"valid_targets_min": 497
}