[ci] Adding rocm-less docker image for tests (ROCm#1218)

Ensuring that the tests work without ROCm and get ROCm from `TheRock`,
this PR ensures this!

Changes:
- Adding container for pytorch tests and CI tests
- Doing some organization for workflows, to ensure that sanity checks is
indeed checking from TheRock

PyTorch run:
https://github.com/ROCm/TheRock/actions/runs/16951040104/job/48043659985

Author: geomin12

.github/workflows/test_packages.yml

@@ -24,21 +24,14 @@ on:
       - ADHOCBUILD
 
 jobs:
-  test_sanity_check:
-    name: 'Test Sanity Check'
-    if: ${{ inputs.test_runs_on != '' }} # if there is a test machine available
-    uses: './.github/workflows/test_sanity_check.yml'
-    with:
-      artifact_run_id: ${{ inputs.artifact_run_id }}
-      amdgpu_families: ${{ inputs.amdgpu_families }}
-      test_runs_on: ${{ inputs.test_runs_on }}
-
   configure_test_matrix:
     name: "Configure test matrix"
-    needs: test_sanity_check
+    # if there is a test machine available
+    if: ${{ inputs.test_runs_on != '' }}
     runs-on: ${{ inputs.test_runs_on }}
     outputs:
       components: ${{ steps.configure.outputs.components }}
+      platform: ${{ steps.configure.outputs.platform }}
     steps:
       - name: "Checking out repository"
         uses: actions/checkout@08c6903cd8c0fde910a37f88322edcfb5dd907a8 # v5.0.0
@@ -52,10 +45,28 @@ jobs:
         id: configure
         run: python ./build_tools/github_actions/fetch_test_configurations.py
 
+  test_sanity_check:
+    name: 'Test Sanity Check'
+    needs: configure_test_matrix
+    uses: './.github/workflows/test_sanity_check.yml'
+    with:
+      artifact_run_id: ${{ inputs.artifact_run_id }}
+      amdgpu_families: ${{ inputs.amdgpu_families }}
+      test_runs_on: ${{ inputs.test_runs_on }}
+      platform: ${{ needs.configure_test_matrix.outputs.platform }}
+
   test_components:
     name: 'Test ${{ matrix.components.job_name }}'
     runs-on: ${{ inputs.test_runs_on }}
-    needs: configure_test_matrix
+    container:
+      image: ${{ needs.configure_test_matrix.outputs.platform == 'linux' && 'ghcr.io/rocm/no_rocm_image_ubuntu24_04@sha256:405945a40deaff9db90b9839c0f41d4cba4a383c1a7459b28627047bf6302a26' || null }}
+      options: --ipc host
+        --group-add video
+        --device /dev/kfd
+        --device /dev/dri
+        --group-add 992
+        --env-file /etc/podinfo/gha-gpu-isolation-settings
+    needs: [test_sanity_check, configure_test_matrix]
     # skip tests if no test matrix to run
     if: ${{ needs.configure_test_matrix.outputs.components != '[]' }}
     strategy:

.github/workflows/test_pytorch_wheels.yml

@@ -57,6 +57,14 @@ jobs:
   test_wheels:
     name: Test PyTorch Wheels | ${{ inputs.amdgpu_family }}
     runs-on: ${{ inputs.test_runs_on }}
+    container:
+      image: ${{ contains(inputs.test_runs_on, 'linux') && 'ghcr.io/rocm/no_rocm_image_ubuntu24_04@sha256:405945a40deaff9db90b9839c0f41d4cba4a383c1a7459b28627047bf6302a26' || null }}
+      options: --ipc host
+        --group-add video
+        --device /dev/kfd
+        --device /dev/dri
+        --group-add 992
+        --env-file /etc/podinfo/gha-gpu-isolation-settings
     defaults:
       run:
         shell: bash

.github/workflows/test_sanity_check.yml

@@ -11,6 +11,8 @@ on:
         default: ""
       test_runs_on:
         type: string
+      platform:
+        type: string
   workflow_call:
     inputs:
       artifact_run_id:
@@ -21,6 +23,8 @@ on:
         default: ""
       test_runs_on:
         type: string
+      platform:
+        type: string
   push:
     branches:
       - ADHOCBUILD
@@ -32,8 +36,14 @@ jobs:
   test_sanity_check:
     name: "Sanity ROCM Test"
     runs-on: ${{ inputs.test_runs_on }}
-    strategy:
-      fail-fast: false
+    container:
+      image: ${{ inputs.platform == 'linux' && 'ghcr.io/rocm/no_rocm_image_ubuntu24_04@sha256:405945a40deaff9db90b9839c0f41d4cba4a383c1a7459b28627047bf6302a26' || null }}
+      options: --ipc host
+        --group-add video
+        --device /dev/kfd
+        --device /dev/dri
+        --group-add 992
+        --env-file /etc/podinfo/gha-gpu-isolation-settings
     defaults:
       run:
         shell: bash

build_tools/github_actions/fetch_test_configurations.py

@@ -130,7 +130,7 @@ def run():
             logging.info(f"Including job {job_name}")
             output_matrix.append(test_matrix[key])
 
-    gha_set_output({"components": json.dumps(output_matrix)})
+    gha_set_output({"components": json.dumps(output_matrix), "platform": platform})
 
 
 if __name__ == "__main__":