Add PR-blocking CI for QNN model tests, operator/model testsuite, and pass unit tests (#18854)

.github/workflows/pull.yml

@@ -674,6 +674,40 @@ jobs:
       build-tool: buck2
       docker-image: ci-image:executorch-ubuntu-22.04-clang12
 
+  test-qnn-buck-build-linux:
+    name: test-qnn-buck-build-linux
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    strategy:
+      fail-fast: false
+    with:
+      runner: linux.2xlarge
+      docker-image: ci-image:executorch-ubuntu-22.04-qnn-sdk
+      submodules: 'recursive'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 90
+      script: |
+        set -eux
+
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+
+        # Download QNN SDK and get the path
+        QNN_SDK_ROOT=$(python3 backends/qualcomm/scripts/download_qnn_sdk.py --print-sdk-path)
+        echo "QNN_SDK_ROOT=${QNN_SDK_ROOT}"
+
+        # Configure Buck to find the QNN SDK
+        echo "[qualcomm]" >> .buckconfig
+        echo "  qnn_sdk_root = ${QNN_SDK_ROOT}" >> .buckconfig
+
+        # Setup buck2
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool buck2
+
+        # Build QNN backend with Buck
+        buck2 build //backends/qualcomm/...
+
   unittest-arm-backend-with-no-deps:
     name: unittest-arm-backend-with-no-deps
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
@@ -823,6 +857,8 @@ jobs:
       id-token: write
       contents: read
     strategy:
+      matrix:
+        model: [mv2, mv3, dl3]
       fail-fast: false
     with:
       runner: linux.2xlarge
@@ -834,9 +870,55 @@ jobs:
         # The generic Linux job chooses to use base env, not the one setup by the image
         CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
         conda activate "${CONDA_ENV}"
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool cmake
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-qnn-deps.sh
+        PYTHON_EXECUTABLE=python bash .ci/scripts/build-qnn-sdk.sh
+        PYTHON_EXECUTABLE=python bash .ci/scripts/test_model.sh ${{ matrix.model }} "cmake" "qnn"
+
+  test-qnn-testsuite-linux:
+    name: test-qnn-testsuite-linux
+    permissions:
+      id-token: write
+      contents: read
+    uses: ./.github/workflows/_test_backend.yml
+    with:
+      backend: qnn
+      flows: '["qnn"]'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 120
+      run-linux: true
+      runner-linux: linux.2xlarge
+
+  test-qnn-passes-linux:
+    name: test-qnn-passes-linux
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    strategy:
+      fail-fast: false
+    with:
+      runner: linux.2xlarge
+      docker-image: ci-image:executorch-ubuntu-22.04-qnn-sdk
+      submodules: 'recursive'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 30
+      script: |
+        # The generic Linux job chooses to use base env, not the one setup by the image
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+
+        PYTHON_EXECUTABLE=python bash .ci/scripts/setup-qnn-deps.sh
+        # Source (not bash) so QNN_SDK_ROOT stays in the environment
+        PYTHON_EXECUTABLE=python source .ci/scripts/build-qnn-sdk.sh
+
+        # Editable install so the PyQnnManagerAdaptor .so built by build-qnn-sdk.sh
+        # is visible in the source tree (the _passes import chain pulls it in transitively)
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_QNN=ON -DQNN_SDK_ROOT=$QNN_SDK_ROOT" \
+          PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool cmake --editable true
 
-        # placeholder for running test_qnn_delegate.py, can use matrix such that we can trigger different jobs, refers to test-llama-runner-qnn-linux
-        # reminder: make sure each job runs fast
+        # Run QNN pass unit tests
+        pytest -xvs backends/qualcomm/tests/test_passes.py
 
   test-phi-3-mini-runner-linux:
     name: test-phi-3-mini-runner-linux

backends/qualcomm/aot/python/targets.bzl

@@ -3,7 +3,7 @@ load(
     "CXX",
 )
 load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")
-load("@fbsource//xplat/executorch/backends/qualcomm/qnn_version.bzl", "get_qnn_library_version")
+load("@fbsource//xplat/executorch/backends/qualcomm/third-party:third_party_libs.bzl", "qnn_third_party_dep")
 
 PYTHON_MODULE_NAME = "PyQnnManagerAdaptor"
 
@@ -32,17 +32,15 @@ def define_common_targets():
             "//executorch/backends/qualcomm/runtime:logging",
             "//executorch/backends/qualcomm:schema",
             "//executorch/backends/qualcomm/runtime:runtime",
-            "fbsource//third-party/pybind11:pybind11",
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:app_sources".format(get_qnn_library_version()),
+            qnn_third_party_dep("api"),
+            qnn_third_party_dep("app_sources"),
+            qnn_third_party_dep("pybind11"),
         ],
         external_deps = [
             "libtorch_python",
         ],
         use_static_deps = True,
-        visibility = [
-            "//executorch/backends/qualcomm/...",
-        ],
+        visibility = ["PUBLIC"],
     )
 
 
@@ -61,8 +59,8 @@ def define_common_targets():
             "//executorch/backends/qualcomm/runtime:logging",
             "//executorch/backends/qualcomm:schema",
             "//executorch/backends/qualcomm/runtime:runtime",
-            "fbsource//third-party/pybind11:pybind11",
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:app_sources".format(get_qnn_library_version()),
+            qnn_third_party_dep("api"),
+            qnn_third_party_dep("app_sources"),
+            qnn_third_party_dep("pybind11"),
         ],
     )

backends/qualcomm/aot/wrappers/targets.bzl

@@ -3,7 +3,7 @@ load(
     "ANDROID",
 )
 load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")
-load("@fbsource//xplat/executorch/backends/qualcomm/qnn_version.bzl", "get_qnn_library_version")
+load("@fbsource//xplat/executorch/backends/qualcomm/third-party:third_party_libs.bzl", "qnn_third_party_dep")
 
 def define_common_targets():
     """Defines targets that should be shared between fbcode and xplat.
@@ -23,8 +23,8 @@ def define_common_targets():
         platforms = [ANDROID],
         visibility = ["PUBLIC"],
         deps = [
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:app_sources".format(get_qnn_library_version()),
+            qnn_third_party_dep("api"),
+            qnn_third_party_dep("app_sources"),
             "//executorch/runtime/backend:interface",
             "//executorch/runtime/core:core",
         ],

backends/qualcomm/recipes/BUCK

@@ -23,15 +23,15 @@ fbcode_target(_kind = runtime.python_library,
     ],
     visibility = ["PUBLIC"],
     deps = [
-        "//caffe2:torch",
         "//executorch/export:lib",
-        "//executorch/runtime:runtime",  # @manual
         "//executorch/backends/qualcomm/partition:partition",
         "//executorch/backends/qualcomm/serialization:serialization",
         "//executorch/backends/qualcomm/utils:utils",
         "//executorch/backends/qualcomm/_passes:passes",
         ":qnn_recipe_types",
-    ],
+    ] + ([] if runtime.is_oss else [
+        "//executorch/runtime:runtime",  # @manual
+    ]),
 )
 
 fbcode_target(_kind = runtime.python_library,

backends/qualcomm/runtime/targets.bzl

@@ -3,7 +3,7 @@ load(
     "ANDROID",
 )
 load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")
-load("@fbsource//xplat/executorch/backends/qualcomm/qnn_version.bzl", "get_qnn_library_version")
+load("@fbsource//xplat/executorch/backends/qualcomm/third-party:third_party_libs.bzl", "qnn_third_party_dep")
 
 def define_common_targets():
     """Defines targets that should be shared between fbcode and xplat.
@@ -24,12 +24,12 @@ def define_common_targets():
         platforms = [ANDROID],
         visibility = ["PUBLIC"],
         deps = [
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:app_sources".format(get_qnn_library_version()),
+            qnn_third_party_dep("api"),
+            qnn_third_party_dep("app_sources"),
             "//executorch/runtime/backend:interface",
         ],
         exported_deps = [
-            "fbsource//third-party/toolchains:log",
+            qnn_third_party_dep("log"),
             "//executorch/backends/qualcomm:schema",
             "//executorch/runtime/core:core",
         ],
@@ -71,12 +71,12 @@ def define_common_targets():
             platforms = [ANDROID],
             visibility = ["PUBLIC"],
             resources = ({
-                "qnn_lib": "fbsource//third-party/qualcomm/qnn/qnn-{0}:qnn_offline_compile_libs".format(get_qnn_library_version()),
+                "qnn_lib": qnn_third_party_dep("qnn_offline_compile_libs"),
                 } if include_aot_qnn_lib else {
             }),
             deps = [
-                "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
-                "fbsource//third-party/qualcomm/qnn/qnn-{0}:app_sources".format(get_qnn_library_version()),
+                qnn_third_party_dep("api"),
+                qnn_third_party_dep("app_sources"),
                 ":logging",
                 "//executorch/backends/qualcomm:schema",
                 "//executorch/backends/qualcomm/aot/wrappers:wrappers",

backends/qualcomm/serialization/targets.bzl

@@ -7,10 +7,17 @@ def define_common_targets():
     TARGETS and BUCK files that call this function.
     """
 
+    # Used by the INPUT_SCHEMA genrule in targets.bzl for OSS builds,
+    # where cross-package relative paths don't work in genrule srcs.
+    export_file(
+        name = "qc_compiler_spec.fbs",
+        visibility = ["PUBLIC"],
+    )
+
     export_file(
         name = "qc_compiler_spec_schema",
         src = "qc_compiler_spec.fbs",
-        visibility = ["//executorch/backends/qualcomm/serialization/..."],
+        visibility = ["PUBLIC"],
     )
 
     runtime.python_library(

backends/qualcomm/targets.bzl

@@ -3,12 +3,15 @@ load(
     "ANDROID",
 )
 load("@fbsource//xplat/executorch/build:runtime_wrapper.bzl", "runtime")
-load("@fbsource//xplat/executorch/backends/qualcomm/qnn_version.bzl", "get_qnn_library_version")
+load("@fbsource//xplat/executorch/backends/qualcomm/third-party:third_party_libs.bzl", "qnn_third_party_dep")
 
 # Construct the input and output file names. All input and output files rely on scalar_type file.
 SCHEMA_NAME = "qc_compiler_spec"
 
-INPUT_SCHEMA = "serialization/" + SCHEMA_NAME + ".fbs"
+# In OSS, genrule srcs can't use cross-package relative paths, so use
+# an export_file target. In fbcode/xplat, relative path works and avoids
+# the srcs patching gap in _patch_executorch_references.
+INPUT_SCHEMA = "//backends/qualcomm/serialization:qc_compiler_spec.fbs" if runtime.is_oss else "serialization/" + SCHEMA_NAME + ".fbs"
 
 OUTPUT_SCHEMA_HEADER = SCHEMA_NAME + "_generated.h"
 
@@ -60,15 +63,7 @@ def define_common_targets():
     runtime.cxx_library(
         name = "schema",
         srcs = [],
-        visibility = [
-            # Lock this down as tightly as possible to ensure that flatbuffers
-            # are an implementation detail. Ideally this list would only include
-            # //executorch/runtime/executor/...
-            "//executorch/codegen/tools/...",
-            "//executorch/runtime/executor/...",
-            "//executorch/backends/qualcomm/...",
-            "//executorch/backends/qualcomm/runtime/...",
-        ],
+        visibility = ["PUBLIC"],
         exported_headers = {
             OUTPUT_SCHEMA_HEADER: ":{}[{}]".format(SCHEMA_GEN_RULE_NAME, OUTPUT_SCHEMA_HEADER),
         },
@@ -84,7 +79,7 @@ def define_common_targets():
         define_static_target = True,
         visibility = ["PUBLIC"],
         deps = [
-            "fbsource//third-party/qualcomm/qnn/qnn-{0}:api".format(get_qnn_library_version()),
+            qnn_third_party_dep("api"),
             "//executorch/runtime/backend:interface",
             "//executorch/runtime/core:core",
             "//executorch/backends/qualcomm/runtime:runtime_android_build",

backends/qualcomm/tests/BUCK

@@ -13,6 +13,8 @@ fbcode_target(_kind = runtime.python_library,
     ]
 )
 
+# Test targets have transitive deps (caffe2, torchvision, torchaudio, etc.)
+# that are not available in OSS Buck builds.
 fbcode_target(_kind = runtime.python_library,
     name = "test_qnn_delegate",
     srcs = [
@@ -28,18 +30,22 @@ fbcode_target(_kind = runtime.python_library,
         "//caffe2/functorch:functorch_src",
         "//executorch/exir/backend:partitioner",
         "//executorch/exir/dialects:lib",
-        "//executorch/extension/pybindings:portable_lib",  # @manual
         "//executorch/extension/pytree:pylib",
         "//executorch/backends/qualcomm/partition:partition",
         "//executorch/backends/qualcomm/quantizer:quantizer",
         "//executorch/backends/qualcomm/serialization:serialization",
         "//executorch/backends/qualcomm/utils:utils",
         "//executorch/devtools:lib",
         "//executorch/examples/qualcomm:utils",
-        "//executorch/examples/models:models",
         "//executorch/backends/qualcomm/debugger:utils",
         "//executorch/backends/qualcomm/debugger:qnn_intermediate_debugger",
-    ],
+    ] + ([] if runtime.is_oss else [
+        # These deps have pre-existing issues in OSS Buck:
+        # portable_lib: shim//build_defs/cpp_python_extension.bzl missing
+        # models: examples/models/BUCK uses raw python_library (no dep patching)
+        "//executorch/extension/pybindings:portable_lib",  # @manual
+        "//executorch/examples/models:models",
+    ]),
 )
 
 fbcode_target(_kind = runtime.python_library,
@@ -67,9 +73,11 @@ fbcode_target(_kind = runtime.python_test,
         "//executorch/backends/qualcomm/partition:partition",
         "//executorch/backends/qualcomm/serialization:serialization",
         "//executorch/backends/qualcomm/utils:utils",
+        "//executorch/backends/qualcomm/builders:builders",
+    ] + ([] if runtime.is_oss else [
+        # These deps fail in OSS: keep_gpu_sections kwarg breaks TARGETS evaluation
         "//executorch/examples/models/llama:transformer_modules",
         "//executorch/examples/qualcomm/oss_scripts/llama:masking_utils",
         "//executorch/examples/qualcomm/oss_scripts/llama:static_llama",
-        "//executorch/backends/qualcomm/builders:builders",
-    ],
+    ]),
 )