FrRay
/
op_optimization
forked from metax-maca/op_optimization
0
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

更新Fused MoE教程

This commit is contained in:
Xinyi Wu (i26343) - Application Ecology 2026-07-15 13:46:23 +08:00
parent 18268c2639
commit 232f2631c7
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
基于AI Agent开发范式的国产GPU大模型推理算子库优化/Fused MoE 算子入门：从 Benchmark 验证到 XPU-OJ 接口提交.md
View File

@ -416,7 +416,7 @@ bash scripts/run_fused_moe_i8_tn_benchmark.sh --backend all --warmup 5 --iters 2
[![image2](https://origin.picgo.net/2026/07/15/image21f74c19caebba239.png)](https://www.picgo.net/image/image2.4i0rxb)
Fused MoE 任务包含 Fused MoE i8 tn 一个题目，直接点击即可进入题目页面：
Fused MoE 任务包含 `Fused MoE i8 tn` 一个题目，直接点击即可进入题目页面：
[![image3](https://origin.picgo.net/2026/07/15/image3e4063c312076081a.png)](https://www.picgo.net/image/image3.4i0cXl)
@ -457,7 +457,7 @@ bash scripts/run_fused_moe_i8_tn_benchmark.sh --backend all --warmup 5 --iters 2
* 容差：`rtol=2e-2, atol=5e-3`；
* 通过率：`matched_ratio >= 0.99`（至少 99% 的元素在容差范围内）。
* 通过率：`matched_ratio ≫ 0.99`（至少 99% 的元素在容差范围内）。
#### Step 9：提交 OJ 冒烟代码