Merge branch 'main' into tosa_tools_update_1

Author: digantdesai

.ci/scripts/build_android_instrumentation.sh

@@ -12,10 +12,10 @@ if [[ -z "${PYTHON_EXECUTABLE:-}" ]]; then
 fi
 which "${PYTHON_EXECUTABLE}"
 
-mkdir -p "${BUILD_AAR_DIR}"/executorch_android/src/androidTest/resources
-cp extension/module/test/resources/add.pte "${BUILD_AAR_DIR}"/executorch_android/src/androidTest/resources
+mkdir -p extension/android/executorch_android/src/androidTest/resources
+cp extension/module/test/resources/add.pte extension/android/executorch_android/src/androidTest/resources
 
-pushd "${BUILD_AAR_DIR}"
+pushd extension/android
 ANDROID_HOME="${ANDROID_SDK:-/opt/android/sdk}" ./gradlew :executorch_android:testDebugUnitTest
 ANDROID_HOME="${ANDROID_SDK:-/opt/android/sdk}" ./gradlew :executorch_android:assembleAndroidTest
 popd

.ci/scripts/test_ios_ci.sh

@@ -7,7 +7,7 @@
 
 set -e
 
-APP_PATH="examples/demo-apps/apple_ios/ExecuTorchDemo/ExecuTorchDemo"
+APP_PATH="executorch-examples/apple/ExecuTorchDemo/ExecuTorchDemo"
 MODEL_NAME="mv3"
 SIMULATOR_NAME="executorch"
 
@@ -34,6 +34,10 @@ say() {
   echo -e "\033[1m\n\t** $1 **\n\033[0m"
 }
 
+say "Cloning the Demo App"
+
+git clone --depth 1 https://github.com/pytorch-labs/executorch-examples.git
+
 say "Installing CoreML Backend Requirements"
 
 ./backends/apple/coreml/scripts/install_requirements.sh

.github/workflows/_android.yml

@@ -37,7 +37,7 @@ jobs:
 
         mkdir -p ${ARTIFACTS_DIR_NAME}/library_test_dir
         bash .ci/scripts/build_android_instrumentation.sh
-        cp ${BUILD_AAR_DIR}/executorch_android/build/outputs/apk/androidTest/debug/executorch_android-debug-androidTest.apk "${ARTIFACTS_DIR_NAME}/library_test_dir"
+        cp extension/android/executorch_android/build/outputs/apk/androidTest/debug/executorch_android-debug-androidTest.apk "${ARTIFACTS_DIR_NAME}/library_test_dir"
 
         mkdir -p ${ARTIFACTS_DIR_NAME}/fp32-xnnpack-custom
         bash examples/models/llama/install_requirements.sh

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

@@ -7,6 +7,10 @@ on:
         description: Version name to be uploaded for AAR release
         required: false
         type: string
+      upload_to_maven:
+        description: Upload the AAR to maven staging repository
+        required: false
+        type: boolean
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
@@ -31,11 +35,14 @@ jobs:
   build-aar:
     name: build-aar
     needs: check-if-aar-exists
-    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    if: ${{ !github.event.pull_request.head.repo.fork }}
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@release/2.7
+    secrets: inherit
     permissions:
       id-token: write
       contents: read
     with:
+      secrets-env: EXECUTORCH_MAVEN_SIGNING_KEYID EXECUTORCH_MAVEN_SIGNING_PASSWORD EXECUTORCH_MAVEN_CENTRAL_PASSWORD EXECUTORCH_MAVEN_CENTRAL_USERNAME EXECUTORCH_MAVEN_SIGNING_GPG_KEY_CONTENTS
       runner: linux.2xlarge
       docker-image: executorch-ubuntu-22.04-clang12-android
       submodules: 'true'
@@ -52,6 +59,16 @@ jobs:
         PYTHON_EXECUTABLE=python bash .ci/scripts/setup-linux.sh --build-tool buck2
         export ARTIFACTS_DIR_NAME=artifacts-to-be-uploaded
 
+        mkdir -p ~/.gradle
+        touch ~/.gradle/gradle.properties
+        echo "signing.keyId=${SECRET_EXECUTORCH_MAVEN_SIGNING_KEYID}" >> ~/.gradle/gradle.properties
+        echo "signing.password=${SECRET_EXECUTORCH_MAVEN_SIGNING_PASSWORD}" >> ~/.gradle/gradle.properties
+        echo "mavenCentralUsername=${SECRET_EXECUTORCH_MAVEN_CENTRAL_USERNAME}" >> ~/.gradle/gradle.properties
+        echo "mavenCentralPassword=${SECRET_EXECUTORCH_MAVEN_CENTRAL_PASSWORD}" >> ~/.gradle/gradle.properties
+        echo "signing.secretKeyRingFile=/tmp/secring.gpg" >> ~/.gradle/gradle.properties
+
+        echo -n "$SECRET_EXECUTORCH_MAVEN_SIGNING_GPG_KEY_CONTENTS" | base64 -d > /tmp/secring.gpg
+
         # Build AAR Package
         mkdir aar-out
         export BUILD_AAR_DIR=aar-out
@@ -61,6 +78,12 @@ jobs:
 
         shasum -a 256 "${ARTIFACTS_DIR_NAME}/executorch.aar"
 
+        # Publish to maven staging
+        UPLOAD_TO_MAVEN="${{ inputs.upload_to_maven }}"
+        if [[ "$UPLOAD_TO_MAVEN" == "true" ]]; then
+          (cd aar-out; ANDROID_HOME="${ANDROID_SDK:-/opt/android/sdk}" ./gradlew :executorch_android:publishToMavenCentral)
+        fi
+
   upload-release-aar:
     name: upload-release-aar
     needs: build-aar

backends/qualcomm/_passes/layout_transform.py

@@ -47,6 +47,7 @@ class LayoutTransform(ExportPass):
     layout_agnostic_ops = {
         exir_ops.edge.aten.abs.default,
         exir_ops.edge.aten.add.Tensor,
+        exir_ops.edge.aten.amax.default,
         exir_ops.edge.aten.bitwise_or.Tensor,
         exir_ops.edge.aten.bmm.default,
         exir_ops.edge.aten.bitwise_and.Tensor,

backends/qualcomm/builders/__init__.py

@@ -9,6 +9,7 @@
     op_abs,
     op_adaptive_avg_pool2d,
     op_add,
+    op_amax,
     op_and,
     op_arange,
     op_argmin,
@@ -95,6 +96,7 @@
     op_abs,
     op_adaptive_avg_pool2d,
     op_add,
+    op_amax,
     op_and,
     op_arange,
     op_argmin,

backends/qualcomm/builders/op_amax.py

@@ -0,0 +1,84 @@
+# Copyright (c) Qualcomm Innovation Center, Inc.
+# All rights reserved
+#
+# This source code is licensed under the BSD-style license found in the
+# LICENSE file in the root directory of this source tree.
+from typing import cast, Dict, List
+
+import executorch.backends.qualcomm.python.PyQnnWrapperAdaptor as PyQnnWrapper
+
+import numpy as np
+
+import torch
+from executorch.backends.qualcomm.utils.constants import QCOM_AXIS_ORDER, QCOM_DATA
+
+from .node_visitor import NodeVisitor, register_node_visitor
+from .qnn_constants import OpAmax, QNN_OP_PACKAGE_NAME_QTI_AISW
+
+
+@register_node_visitor
+class AMax(NodeVisitor):
+    target = ["aten.amax.default"]
+
+    def __init__(self, *args) -> None:
+        super().__init__(*args)
+
+    def define_node(
+        self,
+        node: torch.fx.Node,
+        nodes_to_wrappers: Dict[torch.fx.Node, PyQnnWrapper.TensorWrapper],
+    ) -> PyQnnWrapper.PyQnnOpWrapper:
+        input_node = node.args[0]
+        input_tensor = self.get_tensor(input_node, node)
+        input_tensor_wrapper = self.define_tensor(
+            input_node,
+            node,
+            input_tensor,
+            PyQnnWrapper.Qnn_TensorType_t.QNN_TENSOR_TYPE_NATIVE,
+            nodes_to_wrappers,
+        )
+
+        # mean dims and keep dims
+        mean_dims = cast(List[int], node.args[1])
+        mean_dims = [
+            mean_dim % len(input_node.meta["val"].shape) for mean_dim in mean_dims
+        ]
+        if QCOM_AXIS_ORDER in node.meta:
+            mean_dims = [
+                node.meta[QCOM_AXIS_ORDER].index(mean_dim) for mean_dim in mean_dims
+            ]
+        mean_dims_shape = [len(mean_dims)]
+
+        output_tensor = self.get_tensor(node, node)
+        output_tensor_wrapper = self.define_tensor(
+            node,
+            node,
+            output_tensor,
+            PyQnnWrapper.Qnn_TensorType_t.QNN_TENSOR_TYPE_NATIVE,
+            nodes_to_wrappers,
+        )
+
+        reduce_max_op = PyQnnWrapper.PyQnnOpWrapper(
+            node.name,
+            QNN_OP_PACKAGE_NAME_QTI_AISW,
+            OpAmax.op_name,
+        )
+        reduce_max_op.AddInputTensors([input_tensor_wrapper])
+        reduce_max_op.AddOutputTensors([output_tensor_wrapper])
+        reduce_max_op.AddTensorParam(
+            OpAmax.param_axes,
+            PyQnnWrapper.Qnn_DataType_t.QNN_DATATYPE_UINT_32,
+            len(mean_dims_shape),
+            mean_dims_shape,
+            np.array(mean_dims, dtype=np.uint32),
+            True,
+        )
+        if len(node.args) > 2:
+            keep_dims = cast(bool, node.args[2])
+            reduce_max_op.AddScalarParam(
+                OpAmax.param_keep_dims,
+                PyQnnWrapper.Qnn_DataType_t.QNN_DATATYPE_BOOL_8,
+                {QCOM_DATA: keep_dims},
+            )
+
+        return reduce_max_op

backends/qualcomm/builders/qnn_constants.py

@@ -14,6 +14,13 @@
 # instead of replicating them here.
 
 
+@dataclass(init=False, frozen=True)
+class OpAmax:
+    op_name: str = "ReduceMax"
+    param_axes: str = "axes"
+    param_keep_dims: str = "keep_dims"
+
+
 @dataclass(init=False, frozen=True)
 class OpBatchnorm:
     op_name: str = "Batchnorm"

backends/qualcomm/quantizer/annotators.py

@@ -182,6 +182,11 @@ def annotate_add(node: Node, quantization_config: QuantizationConfig) -> None:
     annotate_binary(node, quantization_config)
 
 
+@register_annotator([torch.ops.aten.amax.default])
+def annotate_amax(node: Node, quantization_config: QuantizationConfig) -> None:
+    annotate_binary(node, quantization_config)
+
+
 @register_annotator([torch.ops.aten.argmin.default])
 def annotate_argmin(node: Node, quantization_config: QuantizationConfig) -> None:
     if _is_annotated([node]):

backends/qualcomm/tests/models.py

@@ -72,6 +72,16 @@ def forward(self, x):
         return torch.any(x, dim=self.dim, keepdim=self.keepdim)
 
 
+class AMax(torch.nn.Module):
+    def __init__(self, dim=None, keepdim=False):
+        super().__init__()
+        self.dim = dim
+        self.keepdim = keepdim
+
+    def forward(self, x):
+        return torch.amax(x, dim=self.dim, keepdim=self.keepdim)
+
+
 class Arange(torch.nn.Module):
     def __init__(self, start, end, step, dtype):
         super().__init__()