CI: Skip triton setup in Aiter standard/multigpu tests and add retries when setting up triton

.github/scripts/build_aiter_triton.sh

@@ -0,0 +1,37 @@
+#!/bin/bash
+
+set -ex
+
+echo
+echo "==== ROCm Packages Installed ===="
+dpkg -l | grep rocm || echo "No ROCm packages found."
+
+echo
+echo "==== Install dependencies and aiter ===="
+pip install --upgrade pandas zmq einops numpy==1.26.2
+pip uninstall -y aiter || true
+pip install --upgrade "pybind11>=3.0.1"
+pip install --upgrade "ninja>=1.11.1"
+python3 setup.py develop
+
+# Read BUILD_TRITON env var, default to 1. If 1, install Triton; if 0, skip installation.
+BUILD_TRITON=${BUILD_TRITON:-1}
+
+if [[ "$BUILD_TRITON" == "1" ]]; then
+    echo
+    echo "==== Install triton ===="
+    pip uninstall -y triton || true
+    git clone --depth=1 https://github.com/triton-lang/triton || true
+    cd triton
+    pip install -r python/requirements.txt
+    pip install filecheck
+    MAX_JOBS=64 pip --retries=5 install .
+    cd ..
+else
+    echo
+    echo "[SKIP] Triton installation skipped because BUILD_TRITON=$BUILD_TRITON"
+fi
+
+echo
+echo "==== Show installed packages ===="
+pip list

.github/workflows/aiter-test.yaml

@@ -104,14 +104,14 @@ jobs:
           --name aiter_test \
           ${{ env.DOCKER_IMAGE }}
 
-      - name: Setup-Triton
+      - name: Setup Aiter
         run: |
           set -ex
-          echo "Setting up Triton..."
+          echo "Setting up Aiter..."
           docker exec \
           -w /workspace \
           aiter_test \
-          ./.github/scripts/build_triton.sh
+          bash -c "BUILD_TRITON=0 ./.github/scripts/build_aiter_triton.sh"
 
       - name: Tests
         run: |
@@ -190,7 +190,7 @@ jobs:
           docker exec \
           -w /workspace \
           aiter_test \
-          ./.github/scripts/build_triton.sh
+          bash -c "BUILD_TRITON=0 ./.github/scripts/build_aiter_triton.sh"
 
       - name: Tests
         run: |

.github/workflows/operators-tuning.yaml

@@ -54,14 +54,14 @@ jobs:
           --name operators_tuning_test \
           rocm/pytorch:latest
 
-      - name: Setup-Triton
+      - name: Setup Aiter and Triton
         run: |
           set -ex
-          echo "Setting up Triton..."
+          echo "Setting up Aiter and Triton..."
           docker exec \
           -w /workspace \
           operators_tuning_test \
-          ./.github/scripts/build_triton.sh
+          bash -c "BUILD_TRITON=0 ./.github/scripts/build_aiter_triton.sh"
 
       - name: Show Computing Units
         run: |

.github/workflows/triton-test.yaml

@@ -65,14 +65,14 @@ jobs:
           triton_test \
           bash -c "pip install speedtest-cli && speedtest-cli --simple" || true
 
-      - name: Setup-Triton
+      - name: Setup Aiter and Triton
         run: |
           set -ex
-          echo "Setuping Triton..."
+          echo "Setting up Aiter and Triton..."
           docker exec \
           -w /workspace \
           triton_test \
-          ./.github/scripts/build_triton.sh
+          ./.github/scripts/build_aiter_triton.sh
 
       - name: Install Pytest
         run: |