Script and config updates in preparation for 0.5.0 (#341)

This includes:

* Bumping cuda-quantum wheel dependencies from 0.12 to 0.13.
* Updating the GitHub build scripts to accept draft release artifacts
that *only* hold the nv-qldpc-decoder (as opposed to duplicates of the
things that are built on the public repo).
* Adds CUDA 13 to .github/workflows/all_libs_release.yaml
* Update some of our tests to not fail if they are nv-qldpc-decoder
tests that are being run in a CPU-only environment (like some of our
GitHub scripts do)

---------

Signed-off-by: Ben Howe <[email protected]>

Author: bmhowe23

.github/workflows/all_libs.yaml

@@ -106,7 +106,10 @@ jobs:
           LD_LIBRARY_PATH: ${{ env.MPI_PATH }}/lib:${{ env.LD_LIBRARY_PATH }}
         shell: bash
         run: |
-          pip install numpy pytest cupy-cuda${{ steps.config.outputs.cuda_major }}x cuquantum-cu${{ steps.config.outputs.cuda_major }} torch lightning ml_collections mpi4py transformers quimb opt_einsum torch nvidia-cublas cuquantum-python-cu${{ steps.config.outputs.cuda_major }}==25.09
+          # Install the correct torch first.
+          cuda_no_dot=$(echo ${{ matrix.cuda_version }} | sed 's/\.//')
+          pip install torch==2.9.0 --index-url https://download.pytorch.org/whl/cu${cuda_no_dot}
+          pip install numpy pytest cupy-cuda${{ steps.config.outputs.cuda_major }}x cuquantum-cu${{ steps.config.outputs.cuda_major }} lightning ml_collections mpi4py transformers quimb opt_einsum nvidia-cublas cuquantum-python-cu${{ steps.config.outputs.cuda_major }}==25.09
           # The following tests are needed for docs/sphinx/examples/qec/python/tensor_network_decoder.py.
           if [ "$(uname -m)" == "x86_64" ]; then
               # Stim is not currently available on manylinux ARM wheels, so only

.github/workflows/all_libs_release.yaml

@@ -10,10 +10,11 @@ on:
       assets_repo:
         type: string
         description: Retrieve assets from a draft release from this repo (e.g. NVIDIA/cudaqx)
+        default: 'NVIDIA/cudaqx'
         required: false
       assets_tag:
         type: string
-        description: Retrieve assets from a draft release with this tag (e.g. installed_files-1)
+        description: Retrieve assets from a draft release with this tag (e.g. docker-files-123)
         required: false
 
 jobs:
@@ -26,19 +27,29 @@ jobs:
           { arch: arm64, gpu: a100 },
           { arch: amd64, gpu: a100 },
         ]
-        cuda_version: ['12.6']
+        cuda_version: ['12.6', '13.0']
     runs-on: linux-${{ matrix.runner.arch }}-gpu-${{ matrix.runner.gpu }}-latest-1
     container: 
-      image: ${{ format('ghcr.io/nvidia/cudaqx-dev:{0}-{1}', inputs.release-number, matrix.runner.arch) }}
+      image: >-
+        ${{
+          inputs.release-number == '0.0.0'
+          && format('ghcr.io/nvidia/cudaqx-dev:{0}-{1}-cu{2}', 'latest', matrix.runner.arch, matrix.cuda_version)
+          || format('ghcr.io/nvidia/cudaqx-dev:{0}-{1}', inputs.release-number, matrix.runner.arch)
+        }}
       env:
         NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }}
     permissions: write-all
     steps:
+      - name: Install dependencies
+        run: |
+          apt update && apt install -y --no-install-recommends zip unzip patchelf git-lfs
+
       - name: Checkout repository
         uses: actions/checkout@v4
         with:
-          ref: releases/v${{ inputs.release-number }}
+          ref: ${{ inputs.release-number == '0.0.0' && 'main' || format('releases/v{0}', inputs.release-number) }}
           set-safe-directory: true
+          lfs: true # download assets file(s) for TRT tests
 
       - name: Set git safe directory
         run: git config --global --add safe.directory $GITHUB_WORKSPACE
@@ -52,10 +63,6 @@ jobs:
           echo "Setting CUDAQX_QEC_VERSION=${{ inputs.release-number }}" >> $GITHUB_STEP_SUMMARY
           echo "Setting CUDAQX_SOLVERS_VERSION=${{ inputs.release-number }}" >> $GITHUB_STEP_SUMMARY
 
-      - name: Install dependencies
-        run: |
-          apt update && apt install -y --no-install-recommends zip unzip patchelf
-
       - name: Fetch assets and set QEC_EXTERNAL_DECODERS
         env:
           GH_TOKEN: ${{ github.token }}
@@ -71,8 +78,8 @@ jobs:
             gh release view -R ${{ inputs.assets_repo }} ${{ inputs.assets_tag }} >> $GITHUB_STEP_SUMMARY
             # Extract the decoder that needs to be embedded in the release
             mkdir -p tmp
-            unzip -d tmp installed_files-${{ matrix.runner.arch }}.zip
-            echo "QEC_EXTERNAL_DECODERS=$(pwd)/tmp/lib/decoder-plugins/libcudaq-qec-nv-qldpc-decoder.so" >> $GITHUB_ENV
+            tar -C tmp -xzvf nv-qldpc-decoder-${{ matrix.runner.arch }}_ubuntu24.04_cuda${{ matrix.cuda_version }}_release.tar.gz
+            echo "QEC_EXTERNAL_DECODERS=$(pwd)/tmp/libcudaq-qec-nv-qldpc-decoder.so" >> $GITHUB_ENV
           fi
         shell: bash
 
@@ -97,15 +104,15 @@ jobs:
 
       - name: Save build artifacts
         run: |
-          cmake --build ${{ steps.build.outputs.build-dir }} --target zip_installed_files
+          cmake --build ${{ steps.build.outputs.build-dir }} --target zip_installed_files --parallel
           cd ${{ steps.build.outputs.build-dir }}
-          mv installed_files.zip installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version}}.zip
+          mv installed_files.zip installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version }}.zip
       
       - name: Upload build artifacts
         uses: actions/upload-artifact@v4
         with:
-          name: installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version}}
-          path: ${{ steps.build.outputs.build-dir }}/installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version}}.zip
+          name: installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version }}
+          path: ${{ steps.build.outputs.build-dir }}/installed_files-${{ matrix.runner.arch }}-cu${{ matrix.cuda_version }}.zip
 
       # ========================================================================
       # Run tests
@@ -121,7 +128,18 @@ jobs:
       - name: Install python requirements
         env:
           LD_LIBRARY_PATH: ${{ env.MPI_PATH }}/lib:${{ env.LD_LIBRARY_PATH }}
-        run: pip install numpy pytest cupy-cuda${{ steps.config.outputs.cuda_major }}x cuquantum-cu${{ steps.config.outputs.cuda_major }} torch lightning ml_collections mpi4py transformers quimb opt_einsum torch
+        shell: bash
+        run: |
+          # Install the correct torch first.
+          cuda_no_dot=$(echo ${{ matrix.cuda_version }} | sed 's/\.//')
+          pip install torch==2.9.0 --index-url https://download.pytorch.org/whl/cu${cuda_no_dot}
+          pip install numpy pytest cupy-cuda${{ steps.config.outputs.cuda_major }}x cuquantum-cu${{ steps.config.outputs.cuda_major }} lightning ml_collections mpi4py transformers quimb opt_einsum nvidia-cublas cuquantum-python-cu${{ steps.config.outputs.cuda_major }}==25.09
+          # The following tests are needed for docs/sphinx/examples/qec/python/tensor_network_decoder.py.
+          if [ "$(uname -m)" == "x86_64" ]; then
+              # Stim is not currently available on manylinux ARM wheels, so only
+              # install for x86_64.
+              pip install stim beliefmatching
+          fi
 
       - name: Run Python tests
         env:

.github/workflows/build_wheels.yaml

@@ -26,10 +26,11 @@ on:
       assets_repo:
         type: string
         description: Retrieve assets from a draft release from this repo (e.g. NVIDIA/cudaqx)
+        default: 'NVIDIA/cudaqx'
         required: false
       assets_tag:
         type: string
-        description: Retrieve assets from a draft release with this tag (e.g. installed_files-1)
+        description: Retrieve assets from a draft release with this tag (e.g. wheels-123)
         required: false
       artifacts_from_run:
         type: string
@@ -147,16 +148,13 @@ jobs:
           .github/workflows/scripts/build_cudaq.sh --python-version ${{ matrix.python }}
 
       - name: Build CUDA-QX wheels
-        env:
-          SUFFIX: ${{ matrix.platform == 'amd64' && 'x86_64' || 'aarch64' }}
         shell: bash
         run: |
           if [[ -n "${{ inputs.assets_repo }}" ]] && [[ -n "${{ inputs.assets_tag }}" ]]; then
             # Extract the decoder that needs to be embedded in the wheel
             mkdir -p tmp
-            PYVER=$(echo ${{ matrix.python }} | tr -d '.')
-            unzip -d tmp cudaq_qec-*-cp${PYVER}-cp${PYVER}-manylinux*${SUFFIX}*.whl
-            export QEC_EXTERNAL_DECODERS=$(pwd)/tmp/cudaq_qec/lib/decoder-plugins/libcudaq-qec-nv-qldpc-decoder.so
+            tar -C tmp -xzvf nv-qldpc-decoder-${{ matrix.platform }}_ubuntu24.04_cuda${{ matrix.cuda_version }}_py${{ matrix.python }}_release.tar.gz
+            export QEC_EXTERNAL_DECODERS=$(pwd)/tmp/libcudaq-qec-nv-qldpc-decoder.so
           fi
           # This is needed to allow the "git rev-parse" commands in the build
           # scripts to work.

libs/qec/pyproject.toml.cu12

@@ -11,7 +11,7 @@ maintainers = [{name = "NVIDIA Corporation & Affiliates"}]
 requires-python = ">=3.11"
 readme = "README.md"
 dependencies = [
-  'cuda-quantum-cu12 >= 0.12',
+  'cuda-quantum-cu12 >= 0.13',
 ]
 classifiers = [
     'Intended Audience :: Science/Research',

libs/qec/pyproject.toml.cu13

@@ -11,7 +11,7 @@ maintainers = [{name = "NVIDIA Corporation & Affiliates"}]
 requires-python = ">=3.11"
 readme = "README.md"
 dependencies = [
-  'cuda-quantum-cu13 >= 0.12',
+  'cuda-quantum-cu13 >= 0.13',
 ]
 classifiers = [
     'Intended Audience :: Science/Research',

libs/qec/python/tests/test_decoding_config.py

@@ -14,6 +14,21 @@
 
 # nv_qldpc_decoder_config tests
 
+
+def is_nv_qldpc_decoder_available():
+    """
+    Helper function to check if the NV-QLDPC decoder is available.
+    """
+    try:
+        H_list = [[1, 0, 0, 1, 0, 1, 1], [0, 1, 0, 1, 1, 0, 1],
+                  [0, 0, 1, 0, 1, 1, 1]]
+        H_np = np.array(H_list, dtype=np.uint8)
+        nv_dec_gpu_and_cpu = qec.get_decoder("nv-qldpc-decoder", H_np)
+        return True
+    except Exception as e:
+        return False
+
+
 FIELDS = {
     "use_sparsity": (bool, True, False),
     "error_rate": (float, 1e-3, 5e-2),
@@ -280,6 +295,9 @@ def test_configure_decoders_from_str_smoke():
     decoder_config.H_sparse = [1, 2, 3, -1, 6, 7, 8, -1, -1]
     decoder_config.set_decoder_custom_args(nv)
     yaml_str = decoder_config.to_yaml_str()
+    # Do not instantiate the decoder if it is not available.
+    if not is_nv_qldpc_decoder_available():
+        return
     status = qec.configure_decoders_from_str(yaml_str)
     assert isinstance(status, int)
     qec.finalize_decoders()

libs/solvers/pyproject.toml.cu12

@@ -11,7 +11,7 @@ maintainers = [{name = "NVIDIA Corporation & Affiliates"}]
 requires-python = ">=3.11"
 readme = "README.md"
 dependencies = [
-  'cuda-quantum-cu12 >= 0.12',
+  'cuda-quantum-cu12 >= 0.13',
   'fastapi',
   'networkx',
   'pyscf',

libs/solvers/pyproject.toml.cu13

@@ -11,7 +11,7 @@ maintainers = [{name = "NVIDIA Corporation & Affiliates"}]
 requires-python = ">=3.11"
 readme = "README.md"
 dependencies = [
-  'cuda-quantum-cu13 >= 0.12',
+  'cuda-quantum-cu13 >= 0.13',
   'fastapi',
   'networkx',
   'pyscf',