Update on "[ET-VK] Adding repeat support to add_copy_packed_dim_offset_node function."

This diff adds support for the repeat operation in `add_copy_packed_dim_offset_node` function in the Vulkan backend for Executorch. The function now takes an additional boolean parameter, "repeat", which indicates whether the copy should wrap around the tensor dimension.

`copy_packed_dim_offset` shader now has 2 functions `repeat_copy` and `no_repeat_copy` which are chosen based on specialization constant parameter.
`no_repeat_copy` function has the legacy copy code.
`repeat_copy` function reads input tensor's dim based on output pos and wraps it according to WHCB repetitions.

Push constants `src_offset` and `dst_offset` contains source and destination's WHCB dimensions (and not copy offsets) respectively, when calling repeat function.

Differential Revision: [D71477552](https://our.internmc.facebook.com/intern/diff/D71477552/)

[ghstack-poisoned]

Author: trivedivivek

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-295f2ed4d103017f7e19a7b8263ece606cd629db
+7ae0ce6360b6e4f944906502d20da24c04debee5

.ci/scripts/unittest-linux.sh

@@ -21,8 +21,7 @@ if [[ "$BUILD_TOOL" == "cmake" ]]; then
     source .ci/scripts/setup-vulkan-linux-deps.sh
 
     PYTHON_EXECUTABLE=python \
-    EXECUTORCH_BUILD_PYBIND=ON \
-    CMAKE_ARGS="-DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
+    CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
     .ci/scripts/setup-linux.sh "$@"
 
     # Install llama3_2_vision dependencies.

.ci/scripts/unittest-macos.sh

@@ -21,8 +21,7 @@ trap 'rm -rfv ${TMP_DIR}' EXIT
 
 # Setup MacOS dependencies as there is no Docker support on MacOS atm
 PYTHON_EXECUTABLE=python \
-EXECUTORCH_BUILD_PYBIND=ON \
-CMAKE_ARGS="-DEXECUTORCH_BUILD_COREML=ON -DEXECUTORCH_BUILD_MPS=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
+CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_COREML=ON -DEXECUTORCH_BUILD_MPS=ON -DEXECUTORCH_BUILD_XNNPACK=ON -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON" \
 ${CONDA_RUN} --no-capture-output \
 .ci/scripts/setup-macos.sh "$@"
 

.ci/scripts/wheel/envvar_base.sh

@@ -8,13 +8,10 @@
 # should typically only contain shell variable assignments. Be sure to export
 # any variables so that subprocesses will see them.
 
-# Enable pybindings so that users can execute ExecuTorch programs from python.
-export EXECUTORCH_BUILD_PYBIND=1
-
 # Ensure that CMAKE_ARGS is defined before referencing it. Defaults to empty
 # if not defined.
 export CMAKE_ARGS="${CMAKE_ARGS:-}"
 
 # Link the XNNPACK backend into the pybindings runtime so that users can execute
 # ExecuTorch programs that delegate to it.
-CMAKE_ARGS="${CMAKE_ARGS} -DEXECUTORCH_BUILD_XNNPACK=ON"
+CMAKE_ARGS="${CMAKE_ARGS} -DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_XNNPACK=ON"

.github/workflows/pull.yml

@@ -365,8 +365,7 @@ jobs:
         # build module for executorch.extension.pybindings.portable_lib
         BUILD_TOOL="cmake"
         PYTHON_EXECUTABLE=python \
-        EXECUTORCH_BUILD_XNNPACK=ON \
-        EXECUTORCH_BUILD_PYBIND=ON \
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON -DEXECUTORCH_BUILD_XNNPACK=ON" \
         bash .ci/scripts/setup-linux.sh --build-tool "${BUILD_TOOL}"
 
         # see if we can import the module successfully
@@ -504,7 +503,7 @@ jobs:
 
         # Setup MacOS dependencies as there is no Docker support on MacOS atm
         PYTHON_EXECUTABLE=python \
-        EXECUTORCH_BUILD_PYBIND=ON \
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON" \
         EXECUTORCH_BUILD_ARM_BAREMETAL=ON \
         .ci/scripts/setup-linux.sh --build-tool "${BUILD_TOOL}"
 

.github/workflows/trunk.yml

@@ -261,7 +261,7 @@ jobs:
 
         # build module for executorch.extension.pybindings.portable_lib
         BUILD_TOOL=${{ matrix.build-tool }}
-        EXECUTORCH_BUILD_PYBIND=ON PYTHON_EXECUTABLE=python ${CONDA_RUN} bash .ci/scripts/setup-macos.sh --build-tool "${BUILD_TOOL}"
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_PYBIND=ON" PYTHON_EXECUTABLE=python ${CONDA_RUN} bash .ci/scripts/setup-macos.sh --build-tool "${BUILD_TOOL}"
 
         # see if we can import the module successfully
         ${CONDA_RUN} python -c "from executorch.extension.pybindings import portable_lib; print('success!')"
@@ -536,9 +536,8 @@ jobs:
         git clone https://github.com/huggingface/optimum-executorch
         cd optimum-executorch
         # There is no release yet, for CI stability, always test from the same commit on main
-        git checkout 6a7e83f3eee2976fa809335bfb78a45b1ea1cb25
-        pip install .
-        pip install accelerate sentencepiece
+        git checkout 1907349524b5f2d61f9c04d2e985da826d4308ba
+        pip install .[tests]
         pip list
         echo "::endgroup::"
 

Package.swift

@@ -82,5 +82,24 @@ let package = Package(
           (value["libraries"] as? [String] ?? []).map { .linkedLibrary($0) }
       ),
     ]
-  }
+  } + [
+    .testTarget(
+      name: "tests",
+      dependencies: [
+        .target(name: "executorch_debug"),
+        .target(name: "kernels_portable"),
+      ],
+      path: "extension/apple/ExecuTorch/__tests__",
+      resources: [
+        .copy("resources/add.pte")
+      ],
+      linkerSettings: [
+        .linkedLibrary("c++"),
+        .unsafeFlags([
+          "-Xlinker", "-force_load",
+          "-Xlinker", "cmake-out/kernels_portable.xcframework/macos-arm64/libkernels_portable_macos.a",
+        ])
+      ]
+    )
+  ]
 )

backends/apple/coreml/runtime/delegate/ETCoreMLModelManager.mm

@@ -208,7 +208,7 @@ void set_outputs(std::vector<executorchcoreml::MultiArray>& outputs,
                                     const inmemoryfs::InMemoryFileSystem *inmemory_fs,
                                     NSError * __autoreleasing *error) {
     NSError *local_error = nil;
-    if (![fm createDirectoryAtURL:dst_url withIntermediateDirectories:NO attributes:@{} error:error]) {
+    if (![fm createDirectoryAtURL:dst_url withIntermediateDirectories:YES attributes:@{} error:error]) {
         ETCoreMLLogUnderlyingErrorAndSetNSError(error,
                                                 ETCoreMLErrorModelSaveFailed,
                                                 local_error,

backends/xnnpack/test/models/deeplab_v3.py

@@ -23,6 +23,9 @@ def forward(self, *args):
 
 
 class TestDeepLabV3(unittest.TestCase):
+    def setUp(self):
+        torch._dynamo.reset()
+
     dl3 = DL3Wrapper()
     dl3 = dl3.eval()
     model_inputs = (torch.randn(1, 3, 224, 224),)

backends/xnnpack/test/models/edsr.py

@@ -14,6 +14,9 @@
 
 
 class TestEDSR(unittest.TestCase):
+    def setUp(self):
+        torch._dynamo.reset()
+
     edsr = edsr_r16f64(2, False).eval()  # noqa
     model_inputs = (torch.randn(1, 3, 224, 224),)