Update Android edge app and documentation (#3651)

Updates Android edge app and documentation.

### Types of changes
<!--- Put an `x` in all the boxes that apply, and remove the not
applicable items -->
- [x] Non-breaking change (fix or new feature that would not break
existing functionality).
- [ ] Breaking change (fix or new feature that would cause existing
functionality to change).
- [ ] New tests added to cover the changes.
- [ ] Quick tests passed locally by running `./runtest.sh`.
- [ ] In-line docstrings updated.
- [ ] Documentation updated.

Author: nvkevlu

nvflare/edge/device/android/README.md

@@ -1,100 +1,192 @@
-# ExecuTorch Libraries
+# NVFlare Android App with ExecuTorch
 
-This directory should contain the ExecuTorch Android libraries (.aar files) that are required for the NVFlare Android app.
+This guide will help you set up and build the NVFlare Android application that uses ExecuTorch for federated learning on mobile devices.
 
-## Required Libraries
+## 🎯 What You'll Build
 
-Based on the CIFAR-10 ExecuTorch example, you need to add the following libraries:
+A complete Android application that:
+- Runs federated learning using ExecuTorch
+- Trains CIFAR-10 models on Android devices
+- Integrates with the NVFlare federated learning framework
+- Supports on-device training and model updates
 
-1. **executorch.aar** - Main ExecuTorch library
-2. **executorch_training.aar** - ExecuTorch training library (if separate)
+## 📋 Prerequisites
 
-## How to Obtain
+Before you begin, ensure you have the following installed:
 
-1. **From ExecuTorch Release**: Download the latest ExecuTorch Android libraries from the official releases
-2. **From CIFAR-10 Example**: Copy the libraries from the working CIFAR-10 example project
-3. **Build from Source**: Build ExecuTorch from source and extract the Android libraries
+### Required Software
+- **Java Development Kit (JDK)**: OpenJDK 17 or later
+- **Android SDK**: API level 29+ with latest build tools
+- **Android NDK**: Version 29.0.13599879 or compatible
+- **Python**: 3.10+ (for ExecuTorch build scripts)
+- **Git**: For repository management
+- **CMake**: For building native components
 
-## Installation Steps
+### Quick Installation
+```bash
+# macOS with Homebrew
+brew install openjdk@17
+brew install cmake
 
-1. Download the required .aar files
-2. Place them in this `libs/` directory
-3. The build.gradle.kts is already configured to include all .aar files from this directory
+# Android SDK/NDK via Android Studio SDK Manager
+# Or download from: https://developer.android.com/studio
+```
 
-## Note
+## 🚀 Quick Start
 
-The build.gradle.kts file uses:
-```kotlin
-implementation(fileTree(mapOf("dir" to "libs", "include" to listOf("*.jar", "*.aar"))))
+Follow these steps to get your NVFlare Android app running:
+
+### Step 1: Set Up Environment Variables
+
+```bash
+# Set your Android development environment
+export JAVA_HOME=/opt/homebrew/Cellar/openjdk@17/17.0.15/libexec/openjdk.jdk/Contents/Home
+export ANDROID_HOME=/Users/$(whoami)/Library/Android/sdk
+export ANDROID_NDK=/Users/$(whoami)/Library/Android/sdk/ndk/29.0.13599879
+export ANDROID_SDK=/Users/$(whoami)/Library/Android/sdk
+export PATH=$PATH:$ANDROID_HOME/tools:$ANDROID_HOME/platform-tools
 ```
 
-This will automatically include all .jar and .aar files in this directory.
+> **Note**: Adjust the paths above to match your system configuration.
 
-## Current Status
+### Step 2: Build ExecuTorch Libraries
 
-⚠️ **Libraries not yet added** - You need to add the ExecuTorch libraries before the app will compile and run successfully.
+```bash
+# Create Python environment
+python3.12 -m venv androidexecutorchenv
+source androidexecutorchenv/bin/activate
 
-# Clone ExecuTorch repository
+# Clone and build ExecuTorch
 git clone https://github.com/pytorch/executorch.git --recurse-submodules
 cd executorch
 
-# Setup Python environment (optional but recommended)
-uv venv --seed --prompt et --python 3.10
-source .venv/bin/activate
+# Update and install dependencies (clean if you have attempted installing previously)
+git pull
+./install_executorch.sh --clean
+git submodule sync --recursive
+git submodule update --init --recursive
 
-# Install build tools and ExecuTorch pip wheel
-./install_executorch.sh
+# Install with training extensions
+CMAKE_ARGS="-DEXECUTORCH_BUILD_EXTENSION_TRAINING=ON,-DEXECUTORCH_BUILD_PYBIND=ON" ./install_executorch.sh
 
-# Build ExecuTorch for Android
-./scripts/build_android_library.sh
+# Build Android library
+EXECUTORCH_BUILD_EXTENSION_LLM=OFF ./scripts/build_android_library.sh
+```
 
-# Create libs directory in NVFlare Android app
-mkdir -p nvflare/edge/device/android/app/libs
+### Step 3: Copy ExecuTorch Libraries
 
-# Copy the ExecuTorch AAR - make sure the app directory is what you are actually using
-cp ./extension/android/executorch_android/build/outputs/aar/executorch_android-debug.aar nvflare/edge/device/android/app/libs/executorch.aar
+```bash
+# Create libs directory
+mkdir -p /path/to/NVFlare/nvflare/edge/device/android/app/libs
 
-Copy nvflare/edge/ios/NVFlareMobile/NVFlareMobile/Assets.xcassets/cifar10/data_batch_1.dataset/data_batch_1.bin to nvflare/edge/device/android/app/src/main/assets/data_batch_1.bin
+# Copy the ExecuTorch AAR file
+cp ./extension/android/executorch_android/build/outputs/aar/executorch_android-debug.aar \
+   /path/to/NVFlare/nvflare/edge/device/android/app/libs/executorch.aar
+```
 
-## ⚠️ Important: SDK Copy Requirement
+### Step 4: Set Up NVFlare SDK
 
-**Critical Setup Step**: The NVFlare Android SDK contents must be copied to the app's source directory for the app to build and function correctly.
+**Critical**: The NVFlare Android SDK must be copied to the app's source directory.
 
-### SDK Source Location
-The SDK source code is located at: `nvflare/edge/device/android/sdk/`
+```bash
+# Copy SDK to app source directory
+cp -r nvflare/edge/device/android/sdk \
+      nvflare/edge/device/android/app/src/main/java/com/nvidia/nvflare/
+```
+
+### Step 5: Add Training Data
+
+```bash
+# Copy CIFAR-10 dataset
+cp nvflare/edge/ios/NVFlareMobile/NVFlareMobile/Assets.xcassets/cifar10/data_batch_1.dataset/data_batch_1.bin \
+   nvflare/edge/device/android/app/src/main/assets/data_batch_1.bin
+```
 
-This directory contains:
-- `core/` - Core SDK functionality
-- `training/` - Training-related components  
-- `utils/` - Utility functions
-- `models/` - Model definitions
-- `config/` - Configuration files
-- `ETTrainerExecutor.kt` - Main training executor
+### Step 6: Build and Run
 
-### Required Action: Copy SDK to App
-**You must copy the contents of the SDK directory to the app's source directory:**
+Open the project in Android Studio and build the app, or use the command line:
 
 ```bash
-# Copy the SDK directory to the proper package structure
-cp -r nvflare/edge/device/android/sdk nvflare/edge/device/android/app/src/main/java/com/nvidia/nvflare
+cd nvflare/edge/device/android
+./gradlew assembleDebug
+```
+
+## 📁 Project Structure
+
+```
+nvflare/edge/device/android/
+├── app/
+│   ├── libs/                    # ExecuTorch libraries (.aar files)
+│   ├── src/main/
+│   │   ├── assets/              # Training data (CIFAR-10)
+│   │   └── java/com/nvidia/nvflare/
+│   │       ├── app/             # Main application code
+│   │       └── sdk/             # NVFlare Android SDK (copied from sdk/)
+├── sdk/                         # NVFlare Android SDK source
+│   ├── core/                    # Core SDK functionality
+│   ├── training/                # Training components
+│   ├── utils/                   # Utility functions
+│   └── models/                  # Model definitions
+└── README.md                    # This file
 ```
 
-### Why This Is Required
-The app code in `nvflare/edge/device/android/app/src/main/java/` does not contain the SDK code. The SDK is maintained separately in the `sdk/` directory, but its contents must be copied to the app's source directory during the build process.
+## 🔧 Configuration
+
+### Build Configuration
+
+The `build.gradle.kts` file is pre-configured to automatically include all `.aar` files from the `libs/` directory:
+
+```kotlin
+implementation(fileTree(mapOf("dir" to "libs", "include" to listOf("*.jar", "*.aar"))))
+```
+
+### Environment Setup
+
+Make sure your `local.properties` file contains the correct Android SDK path:
+
+```properties
+sdk.dir=/Users/yourusername/Library/Android/sdk
+```
+
+## 🐛 Troubleshooting
+
+### Common Issues
+
+**Build fails with missing classes**
+- Ensure the SDK has been copied to the app's source directory (Step 4)
+- Verify all required `.aar` files are in the `libs/` directory
+
+**ExecuTorch build fails**
+- Check that all environment variables are set correctly
+- Ensure you have sufficient disk space (build requires ~10GB)
+- Try building with `EXECUTORCH_BUILD_EXTENSION_LLM=OFF` if LLM extensions cause issues
+
+**App crashes on startup**
+- Verify the CIFAR-10 dataset is in the correct location
+- Check that all required permissions are granted in `AndroidManifest.xml`
+
+### Getting Help
+
+1. Check the [NVFlare documentation](https://nvflare.readthedocs.io/)
+2. Review the [ExecuTorch Android examples](https://github.com/pytorch/executorch/tree/main/examples)
+3. Open an issue in the NVFlare repository for Android-specific problems
+
+## 📚 Additional Resources
+
+- [NVFlare Documentation](https://nvflare.readthedocs.io/)
+- [ExecuTorch Documentation](https://pytorch.org/executorch/)
+- [Android Development Guide](https://developer.android.com/guide)
+- [Federated Learning Concepts](https://nvflare.readthedocs.io/en/latest/fl_introduction.html)
+
+## 🎉 Next Steps
 
-### What Happens If SDK Is Not Copied
-If the SDK contents are not copied to the app's source directory:
-- The app will fail to compile with missing class errors
-- Runtime errors will occur when trying to access SDK functionality
-- Training operations will fail with `TrainingError.DATASET_CREATION_FAILED` or similar errors
+Once your app is running:
 
-### Verification
-To ensure the SDK is properly set up:
-1. Verify the `sdk/` directory exists at `nvflare/edge/device/android/sdk/`
-2. Copy the SDK directory to `nvflare/edge/device/android/app/src/main/java/com/nvidia/nvflare`
-3. Check that all required SDK files are now present at `nvflare/edge/device/android/app/src/main/java/com/nvidia/nvflare/sdk/`
-4. Ensure the app can compile successfully
-5. Test that the app can import and use SDK classes
+1. **Test Training**: Run a simple training session to verify everything works
+2. **Connect to Server**: Set up connection to an NVFlare server for federated learning
+3. **Customize Models**: Modify the model architecture for your specific use case
+4. **Deploy**: Package and deploy your app to Android devices
 
-**Note**: The SDK contents must be copied to the app's source directory before building. This is a manual step that must be performed when setting up the development environment.
+---
 
+**Happy Federated Learning! 🚀**

nvflare/edge/device/android/app/build.gradle.kts

@@ -69,7 +69,7 @@ dependencies {
     // ExecuTorch dependencies
     implementation(fileTree(mapOf("dir" to "libs", "include" to listOf("*.jar", "*.aar"))))
     // Add sources for better IDE support
-    implementation(files("libs/executorch-sources.jar"))
+    //implementation(files("libs/executorch-sources.jar"))
     // Add SoLoader dependency for PyTorch native library loading
     implementation("com.facebook.soloader:nativeloader:0.10.5")
     // Add Facebook JNI dependency required by executorch