fix(ai): 修复 AI 训练 coreCount 在 gpu 下传参错误问题 (#1176)

PKUHPC · Mar 25, 2024 · 0f5d48f · 0f5d48f
1 parent a4319e9
commit 0f5d48f
Show file tree

Hide file tree

Showing 2 changed files with 8 additions and 1 deletion.
diff --git a/.changeset/ten-flies-notice.md b/.changeset/ten-flies-notice.md
@@ -0,0 +1,5 @@
+---
+"@scow/ai": patch
+---
+
+修复 AI 训练 coreCount 在 gpu 下传参错误问题
diff --git a/apps/ai/src/app/(auth)/jobs/[clusterId]/LaunchAppForm.tsx b/apps/ai/src/app/(auth)/jobs/[clusterId]/LaunchAppForm.tsx
@@ -336,7 +336,9 @@ export const LaunchAppForm = (props: Props) => {
             account: account,
             partition: partition,
             nodeCount: nodeCount,
-            coreCount: coreCount,
+            coreCount: gpuCount ?
+              gpuCount * Math.floor(currentPartitionInfo!.cores / currentPartitionInfo!.gpus) :
+              coreCount,
             gpuCount: gpuCount,
             maxTime: maxTime,
             memory: memorySize,