[ROCm] Update periodic.yml to use 2GPU runners (#146839)

amdfaa · pytorchmergebot · commit b6273d7f4ba4 · 2025-02-10T23:41:11.000Z
Temporary fix for rocm workflow. The 4-GPU runners are all taken offline due to (network timeout issue), and so we aren't able to run any periodic jobs. Pull Request resolved: #146839 Approved by: https://github.com/jeffdaily
diff --git a/.github/workflows/periodic.yml b/.github/workflows/periodic.yml
@@ -149,9 +149,9 @@ jobs:
       docker-image-name: pytorch-linux-focal-rocm-n-py3
       test-matrix: |
         { include: [
-          { config: "distributed", shard: 1, num_shards: 3, runner: "linux.rocm.gpu.4", owners: ["module:rocm", "oncall:distributed"] },
-          { config: "distributed", shard: 2, num_shards: 3, runner: "linux.rocm.gpu.4", owners: ["module:rocm", "oncall:distributed"] },
-          { config: "distributed", shard: 3, num_shards: 3, runner: "linux.rocm.gpu.4", owners: ["module:rocm", "oncall:distributed"] },
+          { config: "distributed", shard: 1, num_shards: 3, runner: "linux.rocm.gpu.2", owners: ["module:rocm", "oncall:distributed"] },
+          { config: "distributed", shard: 2, num_shards: 3, runner: "linux.rocm.gpu.2", owners: ["module:rocm", "oncall:distributed"] },
+          { config: "distributed", shard: 3, num_shards: 3, runner: "linux.rocm.gpu.2", owners: ["module:rocm", "oncall:distributed"] },
         ]}
     secrets: inherit
 
