Openvino llama support

.ci/docker/ci_commit_pins/optimum-executorch.txt

@@ -1 +1 @@
-bd06b54e627fbfd354a2cffa4c80fb21883209a9
+44d8d54e38c0258357d4e92e1fefe21e845947a3

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-53a2908a10f414a2f85caa06703a26a40e873869
+cf9d09490c7f6685ec68d5db3acf2e0d73c54d00

.ci/scripts/build-qnn-sdk.sh

@@ -38,6 +38,7 @@ set_up_aot() {
       -DEXECUTORCH_BUILD_EXTENSION_EXTENSION_LLM=ON \
       -DEXECUTORCH_BUILD_EXTENSION_EXTENSION_LLM_RUNNER=ON \
       -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
+      -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
       -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
       -DEXECUTORCH_ENABLE_EVENT_TRACER=ON \
       -DPYTHON_EXECUTABLE=python3

.ci/scripts/setup-openvino.sh

@@ -10,19 +10,17 @@ set -ex
 # shellcheck source=/dev/null
 source "$(dirname "${BASH_SOURCE[0]}")/utils.sh"
 
-git clone https://github.com/openvinotoolkit/openvino.git
-cd openvino && git checkout releases/2025/1
-git submodule update --init --recursive
-sudo ./install_build_dependencies.sh
-mkdir build && cd build
-cmake .. -DCMAKE_BUILD_TYPE=Release -DENABLE_PYTHON=ON
-make -j$(nproc)
+# Download and install OpenVINO from release packages
+OPENVINO_VERSION="2025.3"
+OPENVINO_BUILD="2025.3.0.19807.44526285f24"
+OPENVINO_URL="https://storage.openvinotoolkit.org/repositories/openvino/packages/${OPENVINO_VERSION}/linux/openvino_toolkit_ubuntu22_${OPENVINO_BUILD}_x86_64.tgz"
 
-cd ..
-cmake --install build --prefix dist
+curl -Lo /tmp/openvino_toolkit.tgz --retry 3 --fail ${OPENVINO_URL}
+tar -xzf /tmp/openvino_toolkit.tgz
+mv openvino_toolkit_ubuntu22_${OPENVINO_BUILD}_x86_64 openvino
 
-source dist/setupvars.sh
-cd ../backends/openvino
+source openvino/setupvars.sh
+cd backends/openvino
 pip install -r requirements.txt
 cd scripts
 ./openvino_build.sh --enable_python

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

@@ -13,7 +13,7 @@ download_ai_lite_core() {
   API_BASE="https://soc-developer.semiconductor.samsung.com/api/v1/resource/ai-litecore/download"
   API_KEY=$SAMSUNG_AI_LITECORE_KEY
 
-  VERSION="0.5"
+  VERSION="0.7"
   OS_NAME="Ubuntu 22.04"
   OUT_FILE="/tmp/exynos-ai-litecore-v${VERSION}.tar.gz"
   TARGET_PATH="/tmp/exynos_ai_lite_core"
@@ -62,7 +62,7 @@ install_enn_backend() {
   export PYTHONPATH=${PYTHONPATH:-}:${EXECUTORCH_ROOT}/..
 }
 
-AI_LITE_CORE_VERSION=0.5.0
+AI_LITE_CORE_VERSION=0.7.0
 
 download_ai_lite_core ${AI_LITE_CORE_VERSION}
 install_enn_backend

.ci/scripts/test_backend.sh

@@ -1,6 +1,7 @@
 #!/usr/bin/env bash
 # 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.
@@ -58,6 +59,13 @@ fi
 if [[ "$FLOW" == *arm* ]]; then
     # Setup ARM deps.
     .ci/scripts/setup-arm-baremetal-tools.sh
+    source examples/arm/ethos-u-scratch/setup_path.sh
+
+    if [[ "$FLOW" == *ethos_u* ]]; then
+        # Prepare a test runner binary that can run on the Corstone-3x0 FVPs
+        backends/arm/scripts/build_executorch.sh
+        backends/arm/test/setup_testing.sh
+    fi
 fi
 
 if [[ $IS_MACOS -eq 1 ]]; then

.ci/scripts/test_ios_ci.sh

@@ -36,6 +36,7 @@ say() {
 
 say "Cloning the Demo App"
 
+git config --global http.postBuffer 524288000
 git clone --depth 1 https://github.com/meta-pytorch/executorch-examples.git
 
 say "Installing CoreML Backend Requirements"

.ci/scripts/test_llama_torchao_lowbit.sh

@@ -31,6 +31,7 @@ cmake -DPYTHON_EXECUTABLE=python \
     -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
     -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
     -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
+    -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
     -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
     -DEXECUTORCH_BUILD_XNNPACK=OFF \
     -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \

.ci/scripts/test_llava.sh

@@ -38,6 +38,7 @@ EXECUTORCH_COMMON_CMAKE_ARGS="                      \
         -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON      \
         -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
         -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
+        -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
         -DEXECUTORCH_BUILD_EXTENSION_LLM=ON \
         -DEXECUTORCH_BUILD_EXTENSION_LLM_RUNNER=ON \
         -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON      \

.ci/scripts/test_model.sh

@@ -48,22 +48,33 @@ prepare_artifacts_upload() {
   fi
 }
 
+
 build_cmake_executor_runner() {
   local backend_string_select="${1:-}"
   echo "Building executor_runner"
   rm -rf ${CMAKE_OUTPUT_DIR}
   mkdir ${CMAKE_OUTPUT_DIR}
+  # Common options:
+  COMMON="-DPYTHON_EXECUTABLE=$PYTHON_EXECUTABLE"
   if [[ "$backend_string_select" == "XNNPACK" ]]; then
     echo "Backend $backend_string_select selected"
-    (cd ${CMAKE_OUTPUT_DIR} \
-      && cmake -DCMAKE_BUILD_TYPE=Release \
+    cmake -DCMAKE_BUILD_TYPE=Release \
         -DEXECUTORCH_BUILD_XNNPACK=ON \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" ..)
+        ${COMMON} \
+        -B${CMAKE_OUTPUT_DIR} .
+    cmake --build ${CMAKE_OUTPUT_DIR} -j4
+  elif [[ "$backend_string_select" == "CUDA" ]]; then
+    echo "Backend $backend_string_select selected"
+    cmake -DCMAKE_BUILD_TYPE=Release \
+        -DEXECUTORCH_BUILD_CUDA=ON \
+        -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
+        ${COMMON} \
+        -B${CMAKE_OUTPUT_DIR} .
     cmake --build ${CMAKE_OUTPUT_DIR} -j4
   else
     cmake -DCMAKE_BUILD_TYPE=Debug \
         -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \
+        ${COMMON} \
         -B${CMAKE_OUTPUT_DIR} .
     cmake --build ${CMAKE_OUTPUT_DIR} -j4 --config Debug
   fi
@@ -320,6 +331,13 @@ test_model_with_mediatek() {
   EXPORTED_MODEL=$(find "./${EXPORT_SCRIPT}" -type f -name "*.pte" -print -quit)
 }
 
+test_model_with_cuda() {
+  # Export a basic .pte and .ptd, then run the model.
+  "${PYTHON_EXECUTABLE}" -m examples.cuda.scripts.export --model_name="${MODEL_NAME}" --output_dir "./"
+  build_cmake_executor_runner "CUDA"
+  ./${CMAKE_OUTPUT_DIR}/executor_runner --model_path "./${MODEL_NAME}.pte" --data_path "./aoti_cuda_blob.ptd"
+}
+
 
 if [[ "${BACKEND}" == "portable" ]]; then
   echo "Testing ${MODEL_NAME} with portable kernels..."
@@ -372,6 +390,12 @@ elif [[ "${BACKEND}" == "mediatek" ]]; then
   if [[ $? -eq 0 ]]; then
     prepare_artifacts_upload
   fi
+elif [[ "${BACKEND}" == "cuda" ]]; then
+  echo "Testing ${MODEL_NAME} with cuda..."
+  test_model_with_cuda
+  if [[ $? -eq 0 ]]; then
+    prepare_artifacts_upload
+  fi
 else
   set +e
   if [[ "${BACKEND}" == *"quantization"* ]]; then

.ci/scripts/test_openvino.sh

@@ -10,7 +10,7 @@ set -ex
 # shellcheck source=/dev/null
 source "$(dirname "${BASH_SOURCE[0]}")/utils.sh"
 
-source openvino/dist/setupvars.sh
+source openvino/setupvars.sh
 cd backends/openvino/tests
 python test_runner.py --test_type ops
 python test_runner.py --test_type models

.ci/scripts/test_torchao_huggingface_checkpoints.sh

@@ -129,6 +129,7 @@ if [[ "$TEST_WITH_RUNNER" -eq 1 ]]; then
         -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
         -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
         -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
+        -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
         -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
         -DEXECUTORCH_BUILD_XNNPACK=ON \
         -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \

.ci/scripts/test_yolo12.sh

@@ -119,6 +119,8 @@ cmake_install_executorch_libraries() {
           -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
           -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
           -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
+          -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
+          -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
           -DEXECUTORCH_BUILD_EXTENSION_RUNNER_UTIL=ON \
           -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
           -B"${build_dir}"
@@ -131,6 +133,8 @@ cmake_install_executorch_libraries() {
                        -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
                        -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
                        -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
+                       -DEXECUTORCH_BUILD_EXTENSION_FLAT_TENSOR=ON \
+                       -DEXECUTORCH_BUILD_EXTENSION_NAMED_DATA_MAP=ON \
                        -DEXECUTORCH_BUILD_EXTENSION_RUNNER_UTIL=ON \
                        -DEXECUTORCH_ENABLE_LOGGING=ON \
                        -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \

.ci/scripts/utils.sh

@@ -125,14 +125,15 @@ build_executorch_runner_cmake() {
   clean_executorch_install_folders
   mkdir "${CMAKE_OUTPUT_DIR}"
 
-  pushd "${CMAKE_OUTPUT_DIR}" || return
   if [[ $1 == "Debug" ]]; then
       CXXFLAGS="-fsanitize=address,undefined"
   else
       CXXFLAGS=""
   fi
-  CXXFLAGS="$CXXFLAGS" retry cmake -DPYTHON_EXECUTABLE="${PYTHON_EXECUTABLE}" -DCMAKE_BUILD_TYPE="${1:-Release}" ..
-  popd || return
+  CXXFLAGS="$CXXFLAGS" retry cmake \
+    -DPYTHON_EXECUTABLE="${PYTHON_EXECUTABLE}" \
+    -DCMAKE_BUILD_TYPE="${1:-Release}" \
+    -B${CMAKE_OUTPUT_DIR} .
 
   if [ "$(uname)" == "Darwin" ]; then
     CMAKE_JOBS=$(( $(sysctl -n hw.ncpu) - 1 ))

.github/workflows/android-release-artifacts.yml

@@ -90,6 +90,10 @@ jobs:
         fi
 
         FLAVOR="${{ inputs.flavor }}"
+        if [ ! -z "$FLAVOR" ]; then
+          GRADLE_ARGS+=" -Dflavor=${FLAVOR}"
+        fi
+
         if [[ "$FLAVOR" == "vulkan" || -z "$FLAVOR" ]]; then
           curl -O https://sdk.lunarg.com/sdk/download/1.4.321.1/linux/vulkansdk-linux-x86_64-1.4.321.1.tar.xz
           tar xf vulkansdk-linux-x86_64-1.4.321.1.tar.xz -C /tmp