Automate aarch64-cuda-wheels tests

.github/CHANGELOG.md

@@ -29,6 +29,9 @@
 
 ### Improvements
 
+* Add CI wheels checks for `aarch64` wheels of Lightning-GPU and Lightning-Tensor.
+  [(#1031)](https://github.com/PennyLaneAI/pennylane-lightning/pull/1031/)
+
 * Replace the `dummy_tensor_update` method with the `cutensornetStateCaptureMPS`API to ensure that further gates apply is allowed after the `cutensornetStateCompute` call.
   [(#1028)](https://github.com/PennyLaneAI/pennylane-lightning/pull/1028/)
 

.github/workflows/wheel_linux_aarch64_cuda.yml

@@ -34,15 +34,17 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [pl-4-core-large-runner]
+        os: [arm-gpu]
         arch: [aarch64]
         pl_backend: ["lightning_gpu", "lightning_tensor"]
         cuda_version: ["12"]
         cibw_build: ${{ fromJson(needs.set_wheel_build_matrix.outputs.python_version) }}
         container_img: ["quay.io/pypa/manylinux_2_28_aarch64"]
     timeout-minutes: 45
     name: ${{ matrix.os }}::${{ matrix.arch }} - ${{ matrix.pl_backend }} (Python ${{ fromJson('{ "cp310-*":"3.10","cp311-*":"3.11", "cp312-*":"3.12" }')[matrix.cibw_build] }})
-    runs-on: ${{ matrix.os }}
+    runs-on: 
+      - self-hosted
+      - ${{ matrix.os }}
 
     steps:
       - name: Checkout PennyLane-Lightning
@@ -59,9 +61,6 @@ jobs:
       - name: Configure pyproject.toml file
         run: PL_BACKEND="${{ matrix.pl_backend }}" python scripts/configure_pyproject_toml.py
 
-      - uses: docker/setup-qemu-action@v3
-        name: Set up QEMU
-
       - name: Build wheels
         env:
           CIBW_ARCHS_LINUX: ${{matrix.arch}}
@@ -94,6 +93,35 @@ jobs:
 
         run: python3 -m cibuildwheel --output-dir wheelhouse
 
+      - name: Determine Python version
+        id: pyvs
+        shell: bash
+        run: |
+          echo "version=$(echo  ${{ matrix.cibw_build }} | tr -cd '[:digit:].' | sed 's/./&./1')" >> $GITHUB_OUTPUT
+
+      - uses: actions/setup-python@v5
+        name: Install Python
+        with:
+          python-version: ${{ steps.pyvs.outputs.version }}
+
+      - name: Test wheels
+        run: |
+          python -m ensurepip --upgrade
+          python -m pip install setuptools
+          python -m pip install -r requirements-tests.txt
+          PL_BACKEND="lightning_qubit" python scripts/configure_pyproject_toml.py
+          SKIP_COMPILATION=True python -m pip install . -vv
+          python -m pip install ./wheelhouse/*.whl
+          DEVICENAME=`echo ${{ matrix.pl_backend }} | sed "s/_/./g"`
+          if (${{ matrix.pl_backend == 'lightning_tensor' }}) then
+            PL_DEVICE=${DEVICENAME} python -m pytest tests/
+          else
+            pl-device-test --device=${DEVICENAME} --skip-ops -x --tb=short --no-flaky-report
+            # MCM tests are slow and skipped. get_c_interface() API is not supported with current test setup and skipped.
+            PL_DEVICE=${DEVICENAME} python -m pytest tests/ -k "not test_supported_linux_platform_gpu and not test_native_mcm"
+          fi
+
+
       - name: Validate wheels
         run: |
          python3 -m pip install twine

CMakeLists.txt

@@ -153,7 +153,7 @@ if(ENABLE_PYTHON)
     if("${PL_BACKEND}" STREQUAL "lightning_gpu" OR "${PL_BACKEND}" STREQUAL "lightning_tensor")
         # Allow pip installation of cuQuantum & CUDA 12 libs to be accessible without setting LD_LIBRARY_PATH for lightning_gpu
         # BUILD_RPATH only works for the last call
-        set_target_properties("${PL_BACKEND}_ops" PROPERTIES BUILD_RPATH "$ORIGIN/../cuquantum/lib:$ORIGIN/../nvidia/cuda_runtime/lib:$ORIGIN/../nvidia/cublas/lib:$ORIGIN/../nvidia/cusparse/lib:${SCIPY_OPENBLAS32_RUNTIME_LIB_PATH}:$ORIGIN")
+        set_target_properties("${PL_BACKEND}_ops" PROPERTIES BUILD_RPATH "$ORIGIN/../cuquantum/lib:$ORIGIN/../nvidia/cuda_runtime/lib:$ORIGIN/../nvidia/nvjitlink/lib:$ORIGIN/../nvidia/cublas/lib:$ORIGIN/../nvidia/cusparse/lib:${SCIPY_OPENBLAS32_RUNTIME_LIB_PATH}:$ORIGIN")
     else()
         set_target_properties("${PL_BACKEND}_ops" PROPERTIES BUILD_RPATH "${SCIPY_OPENBLAS32_RUNTIME_LIB_PATH}")
     endif()

pennylane_lightning/core/_version.py

@@ -16,4 +16,4 @@
    Version number (major.minor.patch[-label])
 """
 
-__version__ = "0.40.0-rc0"
+__version__ = "0.40.0-rc1"

tests/lightning_tensor/test_measurements_class.py

@@ -140,7 +140,7 @@ def test_probs_many_wires(self, method, n_qubits, n_targets, tol):
             pytest.skip("Number of targets cannot exceed the number of wires.")
 
         dev = qml.device(device_name, wires=n_qubits, **method)
-        dq = qml.device("lightning.qubit", wires=n_qubits)
+        dq = qml.device("default.qubit", wires=n_qubits)
 
         init_state = np.random.rand(2**n_qubits) + 1.0j * np.random.rand(2**n_qubits)
         init_state /= np.linalg.norm(init_state)
@@ -168,7 +168,7 @@ def test_state_many_wires(self, method, n_qubits, n_targets, tol):
             pytest.skip("Number of targets cannot exceed the number of wires.")
 
         dev = qml.device(device_name, wires=n_qubits, **method)
-        dq = qml.device("lightning.qubit", wires=n_qubits)
+        dq = qml.device("default.qubit", wires=n_qubits)
 
         init_state = np.random.rand(2**n_qubits) + 1.0j * np.random.rand(2**n_qubits)
         init_state /= np.linalg.norm(init_state)