From a7790a92e80b2812a89c30859e3464befb69e6a9 Mon Sep 17 00:00:00 2001
From: Hongxin Liu <lhx0217@gmail.com>
Date: Tue, 26 Mar 2024 15:09:05 +0800
Subject: [PATCH] [devops] fix example test ci (#5504)

---
 .github/workflows/example_check_on_dispatch.yml | 4 +---
 .github/workflows/example_check_on_pr.yml       | 4 +---
 .github/workflows/example_check_on_schedule.yml | 3 +--
 3 files changed, 3 insertions(+), 8 deletions(-)

diff --git a/.github/workflows/example_check_on_dispatch.yml b/.github/workflows/example_check_on_dispatch.yml
index 24e726b4f..d877b06ce 100644
--- a/.github/workflows/example_check_on_dispatch.yml
+++ b/.github/workflows/example_check_on_dispatch.yml
@@ -46,7 +46,7 @@ jobs:
       matrix: ${{fromJson(needs.manual_check_matrix_preparation.outputs.matrix)}}
     container:
       image: hpcaitech/pytorch-cuda:2.1.0-12.1.0
-      options: --gpus all --rm -v /data/scratch/examples-data:/data/
+      options: --gpus all --rm -v /data/scratch/examples-data:/data/ -v /dev/shm
     timeout-minutes: 15
     steps:
       - name: 📚 Checkout
@@ -60,5 +60,3 @@ jobs:
           echo "Testing ${dir} now"
           cd "${PWD}/examples/${dir}"
           bash test_ci.sh
-        env:
-          NCCL_SHM_DISABLE: 1
diff --git a/.github/workflows/example_check_on_pr.yml b/.github/workflows/example_check_on_pr.yml
index 728f059c1..6170628e1 100644
--- a/.github/workflows/example_check_on_pr.yml
+++ b/.github/workflows/example_check_on_pr.yml
@@ -78,7 +78,7 @@ jobs:
       matrix: ${{fromJson(needs.detect-changed-example.outputs.matrix)}}
     container:
       image: hpcaitech/pytorch-cuda:2.1.0-12.1.0
-      options: --gpus all --rm -v /data/scratch/examples-data:/data/
+      options: --gpus all --rm -v /data/scratch/examples-data:/data/ -v /dev/shm
     timeout-minutes: 20
     concurrency:
       group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}-run-example-${{ matrix.directory }}
@@ -95,5 +95,3 @@ jobs:
           example_dir=${{ matrix.directory }}
           cd "${PWD}/examples/${example_dir}"
           bash test_ci.sh
-        env:
-          NCCL_SHM_DISABLE: 1
diff --git a/.github/workflows/example_check_on_schedule.yml b/.github/workflows/example_check_on_schedule.yml
index efb131a86..2588ac824 100644
--- a/.github/workflows/example_check_on_schedule.yml
+++ b/.github/workflows/example_check_on_schedule.yml
@@ -35,6 +35,7 @@ jobs:
       matrix: ${{fromJson(needs.matrix_preparation.outputs.matrix)}}
     container:
       image: hpcaitech/pytorch-cuda:2.1.0-12.1.0
+      options: --gpus all --rm -v /data/scratch/examples-data:/data/ -v /dev/shm
     timeout-minutes: 10
     steps:
       - name: 📚 Checkout
@@ -50,8 +51,6 @@ jobs:
           echo "Testing ${example_dir} now"
           cd "${PWD}/examples/${example_dir}"
           bash test_ci.sh
-        env:
-          NCCL_SHM_DISABLE: 1
 
       - name: Notify Lark
         id: message-preparation