Merge branch 'main' into export-D78291986

Author: rodrigos01meta

.ci/scripts/setup-vulkan-linux-deps.sh

@@ -43,7 +43,7 @@ install_vulkan_sdk() {
   export PATH="${PATH}:${_vulkan_sdk_dir}/${VULKAN_SDK_VERSION}/x86_64/bin/"
 }
 
-VULKAN_SDK_VERSION="1.3.296.0"
+VULKAN_SDK_VERSION="1.4.321.1"
 
 install_swiftshader
 install_vulkan_sdk "${VULKAN_SDK_VERSION}"

.github/workflows/apple-perf.yml

@@ -416,7 +416,7 @@ jobs:
       - set-parameters
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}

.github/workflows/apple.yml

@@ -49,7 +49,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -136,7 +136,7 @@ jobs:
     needs: set-version
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -276,7 +276,7 @@ jobs:
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     secrets: inherit
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}

.github/workflows/build-presets.yml

@@ -22,7 +22,7 @@ jobs:
     with:
       job-name: build
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: 3.12
       submodules: recursive
       timeout: 90

.github/workflows/build-wheels-macos.yml

@@ -63,7 +63,7 @@ jobs:
       post-script: ${{ matrix.post-script }}
       package-name: ${{ matrix.package-name }}
       # Meta's macOS runners do not have Xcode, so use GitHub's runners.
-      runner-type: macos-latest-xlarge
+      runner-type: macos-14-xlarge
       setup-miniconda: true
       smoke-test-script: ${{ matrix.smoke-test-script }}
       trigger-event: ${{ github.event_name }}

.github/workflows/pull.yml

@@ -855,13 +855,14 @@ jobs:
         .ci/scripts/setup-linux.sh --build-tool "cmake"
 
         # Install test requirements
-        pip install -r backends/nxp/requirements-tests.txt
+        pip install -r backends/nxp/requirements-tests-pypi.txt
+        pip install -r backends/nxp/requirements-tests-eiq.txt
 
         # Run pytest
         PYTHON_EXECUTABLE=python bash backends/nxp/run_unittests.sh
 
         # Run aot examples:
-        PYTHON_EXECUTABLE=python bash examples/nxp/run_aot_example.sh cifar10        
+        PYTHON_EXECUTABLE=python bash examples/nxp/run_aot_example.sh cifar10
         PYTHON_EXECUTABLE=python bash examples/nxp/run_aot_example.sh mobilenetv2
 
 
@@ -901,6 +902,34 @@ jobs:
         done
 
 
+  test-vulkan-operators-linux:
+    name: test-vulkan-operators-linux
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    with:
+      runner: linux.2xlarge
+      docker-image: ci-image:executorch-ubuntu-22.04-clang12
+      submodules: 'recursive'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 90
+      script: |
+        set -eux
+
+        # The generic Linux job chooses to use base env, not the one setup by the image
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+
+        # Setup swiftshader and Vulkan SDK which are required to build the Vulkan delegate
+        source .ci/scripts/setup-vulkan-linux-deps.sh
+
+        # Setup python
+        PYTHON_EXECUTABLE=python \
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_VULKAN=ON" \
+        .ci/scripts/setup-linux.sh --build-tool "cmake"
+
+        PYTHON_EXECUTABLE=python bash backends/vulkan/test/custom_ops/build_and_run.sh add
 
   nxp-build-test:
     name: nxp-build-test

.github/workflows/trunk.yml

@@ -435,7 +435,7 @@ jobs:
     name: test-coreml-delegate
     uses: pytorch/test-infra/.github/workflows/macos_job.yml@main
     with:
-      runner: macos-latest-xlarge
+      runner: macos-14-xlarge
       python-version: '3.11'
       submodules: 'recursive'
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}

backends/apple/mps/serialization/mps_graph_serialize.py

@@ -1,14 +1,16 @@
 # Copyright (c) Meta Platforms, Inc. and affiliates.
 # All rights reserved.
+# Copyright 2025 Arm Limited and/or its affiliates.
 #
 # This source code is licensed under the BSD-style license found in the
 # LICENSE file in the root directory of this source tree.
 
+import importlib.resources as _resources
 import json
 import os
 import tempfile
 
-import pkg_resources
+import executorch.backends.apple.mps.serialization as serialization_package
 from executorch.backends.apple.mps.serialization.mps_graph_schema import MPSGraph
 from executorch.exir._serialize._dataclass import _DataclassEncoder
 from executorch.exir._serialize._flatbuffer import _flatc_compile
@@ -19,7 +21,9 @@ def convert_to_flatbuffer(mps_graph: MPSGraph) -> bytes:
     with tempfile.TemporaryDirectory() as d:
         schema_path = os.path.join(d, "schema.fbs")
         with open(schema_path, "wb") as schema_file:
-            schema_file.write(pkg_resources.resource_string(__name__, "schema.fbs"))
+            schema_file.write(
+                _resources.read_binary(serialization_package, "schema.fbs")
+            )
         json_path = os.path.join(d, "schema.json")
         with open(json_path, "wb") as json_file:
             json_file.write(mps_graph_json.encode("ascii"))

backends/arm/CMakeLists.txt

@@ -73,9 +73,10 @@ if(EXECUTORCH_BUILD_VGF)
   # vgf backend
   list(TRANSFORM _vgf_backend_sources PREPEND "${EXECUTORCH_ROOT}/")
   add_library(vgf_backend ${_vgf_backend_sources})
+  install(TARGETS vgf_backend EXPORT ExecuTorchTargets)
   target_include_directories(
-    vgf_backend PUBLIC ${_common_include_directories} ${VULKAN_HEADERS_PATH}
-                       ${VOLK_HEADERS_PATH}
+    vgf_backend PRIVATE ${_common_include_directories} ${VULKAN_HEADERS_PATH}
+                        ${VOLK_HEADERS_PATH}
   )
   target_compile_options(
     vgf_backend PRIVATE -DUSE_VULKAN_WRAPPER -DUSE_VULKAN_VOLK

backends/arm/README.md

@@ -88,6 +88,19 @@ You can test to run some models with the full fvp test flow
 backends/arm/test/test_arm_baremetal.sh test_full_ethosu_fvp
 ```
 
+To run the unit test suite with VKML use the following. Note Vulkan SDK need to be installed.
+Have a look at install_vulkan_sdk() in .ci/scripts/setup-vulkan-linux-deps.sh on how to install Vulkan SDK.
+
+```
+backends/arm/test/test_arm_baremetal.sh test_pytest_vkml
+```
+
+You can test to run some models with the full VKML flow
+
+```
+backends/arm/test/test_arm_baremetal.sh test_full_vkml
+```
+
 ## Unit tests
 
 This is the structure of the test directory
@@ -102,6 +115,7 @@ test                            #  Root test folder
 ├── tosautil                    #  Utility functions for TOSA artifacts
 ├ common.py                     #  Common functions and definitions used by many tests
 ├ setup_testing.sh              #  Script to prepare testing for using the Corstone 3x0 FVP
+├ setup_testing_vkml.sh         #  Script to prepare testing for using the VKML
 ├ test_arm_baremetal.sh         #  Help script to trigger testing
 ```
 
@@ -123,7 +137,7 @@ first you need to build and prepare some used target libs
 
 ```
 examples/arm/run.sh --model_name=add --build_only
-backends/arm/test/setup_testing.sh
+backends/arm/test/setup_testing.sh and/or backends/arm/test/setup_testing_vkml.sh
 ```
 
 The you can run the tests with
@@ -195,6 +209,38 @@ List of model specific and optional passes:
 - InsertCastForOpsWithInt64InputPass
     - Functionality:
         - For LLMs such as LLama, some opeartors like aten.embedding have int64 input. In order to lower these operators to TOSA, this pass will insert a casting node that converts the input from int64 to int32.
-        - Example usage: backends/arm/test/models/test_llama.py
     - Supported Ops:
         - aten.embedding.default, aten.slice_copy.Tensor
+    - Example usage:
+        - backends/arm/test/models/test_llama.py
+
+- ConvertInt64ConstOpsToInt32Pass
+    - Functionalities:
+      - Rewrites constant-producing ops that output int64 to instead output int32, when values are within int32 bounds.
+    - Supported Ops:
+      - `torch.full`, `torch.arange`, `torch.eye`, `torch.linspace`, `torch.tensor`
+    - Example usage:
+        - backends/arm/test/models/stable_diffusion/test_CLIPTextModelWithProjection.py
+        - backends/arm/test/models/stable_diffusion/test_T5EncoderModel.py
+
+- ConvertInt64OutputOpsToInt32Pass
+    - Overview:
+      - Rewrites or removes operations that produce int64 outputs, converting them to int32 where possible.
+      - Overflow checks are applied selectively; for ops without such checks, users need to ensure values fit within the int32 range.
+    - Functionalities:
+        1. Handling casting to int64:
+            - (1) int32 -> int64:
+                - Removes the cast and redirect uses of int64 to int32
+            - (2) other types -> int64:
+                - Rewrites the cast to other types -> int32
+            - Supported Ops:
+              - torch.ops.aten.to.\[dtype|dtype_layout\]
+              - exir_ops.edge.dim_order_ops._to_dim_order_copy.default
+        2. Post-process argmax outputs:
+            - Inserts an int64->int32 cast after the argmax operations that produce int64 outputs:
+            - Supported Ops:
+              - torch.ops.aten.argmax.default
+              - exir_ops.edge.aten.argmax.default
+    - Example usage:
+      - (Functionality 1) backends/arm/test/models/stable_diffusion/test_T5EncoderModel.py
+      - (Functionality 2) backends/arm/test/models/stable_diffusion/test_CLIPTextModelWithProjection.py