From d54b9ca656e7eb25fe7cf713b3f6e3483bb1e4b1 Mon Sep 17 00:00:00 2001
From: continue-revolution <3357667574@qq.com>
Date: Sun, 29 Jan 2023 23:42:04 +0800
Subject: [PATCH 01/12] tflite enable on-device training
---
fedscale/edge/tflite/.gitignore | 6 +
fedscale/edge/tflite/README.md | 55 ++
fedscale/edge/tflite/app/build.gradle | 107 ++++
fedscale/edge/tflite/app/proguard-rules.pro | 21 +
.../tflite/app/src/main/AndroidManifest.xml | 64 +++
.../fedscale/android/executor/FLExecutor.java | 489 ++++++++++++++++++
.../android/tflite/TFLiteBackend.java | 288 +++++++++++
.../com/fedscale/android/utils/Backend.java | 42 ++
.../android/utils/ClientConnections.java | 52 ++
.../com/fedscale/android/utils/Common.java | 145 ++++++
.../tflite/app/src/main/proto/job_api.proto | 44 ++
.../drawable-v24/ic_launcher_foreground.xml | 34 ++
.../res/drawable/ic_launcher_background.xml | 170 ++++++
.../app/src/main/res/layout/activity_main.xml | 40 ++
.../res/mipmap-anydpi-v26/ic_launcher.xml | 5 +
.../mipmap-anydpi-v26/ic_launcher_round.xml | 5 +
.../src/main/res/mipmap-hdpi/ic_launcher.png | Bin 0 -> 3056 bytes
.../res/mipmap-hdpi/ic_launcher_round.png | Bin 0 -> 5024 bytes
.../src/main/res/mipmap-mdpi/ic_launcher.png | Bin 0 -> 2096 bytes
.../res/mipmap-mdpi/ic_launcher_round.png | Bin 0 -> 2858 bytes
.../src/main/res/mipmap-xhdpi/ic_launcher.png | Bin 0 -> 4569 bytes
.../res/mipmap-xhdpi/ic_launcher_round.png | Bin 0 -> 7098 bytes
.../main/res/mipmap-xxhdpi/ic_launcher.png | Bin 0 -> 6464 bytes
.../res/mipmap-xxhdpi/ic_launcher_round.png | Bin 0 -> 10676 bytes
.../main/res/mipmap-xxxhdpi/ic_launcher.png | Bin 0 -> 9250 bytes
.../res/mipmap-xxxhdpi/ic_launcher_round.png | Bin 0 -> 15523 bytes
.../tflite/app/src/main/res/values/array.xml | 28 +
.../tflite/app/src/main/res/values/colors.xml | 6 +
.../app/src/main/res/values/strings.xml | 4 +
.../tflite/app/src/main/res/values/styles.xml | 20 +
fedscale/edge/tflite/assets/conf.json | 42 ++
fedscale/edge/tflite/build.gradle | 26 +
fedscale/edge/tflite/gradle.properties | 24 +
.../tflite/gradle/wrapper/gradle-wrapper.jar | Bin 0 -> 59203 bytes
.../gradle/wrapper/gradle-wrapper.properties | 6 +
fedscale/edge/tflite/gradlew | 185 +++++++
fedscale/edge/tflite/gradlew.bat | 89 ++++
fedscale/edge/tflite/local.properties | 8 +
fedscale/edge/tflite/settings.gradle | 2 +
39 files changed, 2007 insertions(+)
create mode 100644 fedscale/edge/tflite/.gitignore
create mode 100644 fedscale/edge/tflite/README.md
create mode 100644 fedscale/edge/tflite/app/build.gradle
create mode 100644 fedscale/edge/tflite/app/proguard-rules.pro
create mode 100644 fedscale/edge/tflite/app/src/main/AndroidManifest.xml
create mode 100644 fedscale/edge/tflite/app/src/main/java/com/fedscale/android/executor/FLExecutor.java
create mode 100644 fedscale/edge/tflite/app/src/main/java/com/fedscale/android/tflite/TFLiteBackend.java
create mode 100644 fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Backend.java
create mode 100644 fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/ClientConnections.java
create mode 100644 fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Common.java
create mode 100644 fedscale/edge/tflite/app/src/main/proto/job_api.proto
create mode 100644 fedscale/edge/tflite/app/src/main/res/drawable-v24/ic_launcher_foreground.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/drawable/ic_launcher_background.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/layout/activity_main.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-anydpi-v26/ic_launcher.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-anydpi-v26/ic_launcher_round.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-hdpi/ic_launcher.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-hdpi/ic_launcher_round.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-mdpi/ic_launcher.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-mdpi/ic_launcher_round.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xhdpi/ic_launcher.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xhdpi/ic_launcher_round.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xxhdpi/ic_launcher.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xxhdpi/ic_launcher_round.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xxxhdpi/ic_launcher.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/mipmap-xxxhdpi/ic_launcher_round.png
create mode 100644 fedscale/edge/tflite/app/src/main/res/values/array.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/values/colors.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/values/strings.xml
create mode 100644 fedscale/edge/tflite/app/src/main/res/values/styles.xml
create mode 100644 fedscale/edge/tflite/assets/conf.json
create mode 100644 fedscale/edge/tflite/build.gradle
create mode 100644 fedscale/edge/tflite/gradle.properties
create mode 100644 fedscale/edge/tflite/gradle/wrapper/gradle-wrapper.jar
create mode 100644 fedscale/edge/tflite/gradle/wrapper/gradle-wrapper.properties
create mode 100644 fedscale/edge/tflite/gradlew
create mode 100644 fedscale/edge/tflite/gradlew.bat
create mode 100644 fedscale/edge/tflite/local.properties
create mode 100644 fedscale/edge/tflite/settings.gradle
diff --git a/fedscale/edge/tflite/.gitignore b/fedscale/edge/tflite/.gitignore
new file mode 100644
index 00000000..c194522a
--- /dev/null
+++ b/fedscale/edge/tflite/.gitignore
@@ -0,0 +1,6 @@
+assets/*train*
+assets/*test*
+assets/*tflite*
+.idea
+.gradle
+app/build
\ No newline at end of file
diff --git a/fedscale/edge/tflite/README.md b/fedscale/edge/tflite/README.md
new file mode 100644
index 00000000..7780f82c
--- /dev/null
+++ b/fedscale/edge/tflite/README.md
@@ -0,0 +1,55 @@
+# TensorFlow Lite Model Personalization Demo
+
+### Overview
+
+This is a camera app that continuously classifies the objects in the frames seen
+by your device's back camera. This example illustrates a way of personalizing a
+TFLite model on-device without sending any data to the server. It can be adapted
+for various tasks and models. These instructions walk you through building and
+running the demo on an Android device.
+
+The model files are downloaded via Gradle scripts when you build and run the
+app. You don't need to do any steps to download TFLite models into the project
+explicitly.
+
+This application should be run on a physical Android device.
+
+![App example showing UI controls. Training mode.](screenshot1.jpg?raw=true
+"Training mode")
+
+![App example without UI controls. Inference mode.](screenshot2.jpg?raw=true
+"Inference mode")
+
+## Build the demo using Android Studio
+
+### Prerequisites
+
+* The **[Android Studio](https://developer.android.com/studio/index.html)**
+ IDE (Android Studio 2021.2.1 or newer). This sample has been tested on Android
+ Studio Chipmunk.
+
+* A physical Android device with a minimum OS version of SDK 23 (Android 6.0 -
+ Marshmallow) with developer mode enabled. The process of enabling developer
+ mode may vary by device.
+
+### Building
+
+* Open Android Studio. From the Welcome screen, select Open an existing Android
+ Studio project.
+
+* From the Open File or Project window that appears, navigate to and select the
+ tensorflow-lite/examples/model_personalization/android directory. Click OK.
+
+* If it asks you to do a Gradle Sync, click OK.
+
+* With your Android device connected to your computer and developer mode
+ enabled, click on the green Run arrow in Android Studio.
+
+### Models used
+
+Downloading, extraction, and placing the models into the assets folder is
+managed automatically by the download.gradle file.
+
+### Generate your model
+
+To generate or customize your model you can read [here](../README.md).
diff --git a/fedscale/edge/tflite/app/build.gradle b/fedscale/edge/tflite/app/build.gradle
new file mode 100644
index 00000000..288f77d0
--- /dev/null
+++ b/fedscale/edge/tflite/app/build.gradle
@@ -0,0 +1,107 @@
+buildscript {
+ repositories {
+ maven {
+ url "https://plugins.gradle.org/m2/"
+ }
+ }
+ dependencies {
+ classpath "com.google.protobuf:protobuf-gradle-plugin:0.8.10"
+ }
+}
+
+apply plugin: 'com.android.application'
+apply plugin: 'com.google.protobuf'
+
+android {
+
+ aaptOptions {
+ noCompress "tflite"
+ }
+
+ compileSdkVersion 32
+ compileOptions {
+ sourceCompatibility JavaVersion.VERSION_1_8
+ targetCompatibility JavaVersion.VERSION_1_8
+ }
+
+ defaultConfig {
+ applicationId "com.fedscale.android.executor"
+ minSdkVersion 23
+ targetSdkVersion 32
+ multiDexEnabled true
+ versionCode 1
+ versionName "1.0"
+ testInstrumentationRunner "androidx.test.runner.AndroidJUnitRunner"
+ }
+
+ buildTypes {
+ release {
+ minifyEnabled false
+ proguardFiles getDefaultProguardFile('proguard-android-optimize.txt'), 'proguard-rules.pro'
+ }
+ }
+ sourceSets {
+ main {
+ assets {
+ srcDirs = ["../assets"]
+ }
+ }
+ }
+// dataBinding {
+// enabled = true
+// }
+ packagingOptions {
+ exclude 'META-INF/com.android.tools/proguard/coroutines.pro'
+ }
+}
+
+dependencies {
+
+ implementation fileTree(dir: 'libs', include: ['*.jar'])
+
+ // App compat and UI things
+ implementation 'androidx.appcompat:appcompat:1.5.0'
+ implementation 'com.google.android.material:material:1.6.1'
+ implementation 'androidx.constraintlayout:constraintlayout:2.1.4'
+
+ // You need to build grpc-java to obtain these libraries below.
+ implementation 'javax.annotation:javax.annotation-api:1.3.2'
+ implementation 'io.grpc:grpc-okhttp:1.49.0' // CURRENT_GRPC_VERSION
+ implementation 'io.grpc:grpc-protobuf-lite:1.49.0' // CURRENT_GRPC_VERSION
+ implementation 'io.grpc:grpc-stub:1.49.0' // CURRENT_GRPC_VERSION
+
+ //WindowManager
+ implementation 'androidx.window:window:1.1.0-alpha03'
+
+ // Unit testing
+ testImplementation 'junit:junit:4.13.2'
+
+ // Instrumented testing
+ androidTestImplementation 'androidx.test.ext:junit:1.1.3'
+ androidTestImplementation 'androidx.test.espresso:espresso-core:3.4.0'
+
+ // Tensorflow lite dependencies
+ implementation 'org.tensorflow:tensorflow-lite:2.9.0'
+ implementation 'org.tensorflow:tensorflow-lite-gpu:2.9.0'
+ implementation 'org.tensorflow:tensorflow-lite-support:0.4.2'
+ implementation 'org.tensorflow:tensorflow-lite-select-tf-ops:2.9.0'
+}
+
+protobuf {
+ protoc { artifact = 'com.google.protobuf:protoc:3.21.1' }
+ plugins {
+ grpc { artifact = 'io.grpc:protoc-gen-grpc-java:1.49.0' // CURRENT_GRPC_VERSION
+ }
+ }
+ generateProtoTasks {
+ all().each { task ->
+ task.builtins {
+ java { option 'lite' }
+ }
+ task.plugins {
+ grpc { // Options added to --grpc_out
+ option 'lite' }
+ }
+ }
+ }
+}
diff --git a/fedscale/edge/tflite/app/proguard-rules.pro b/fedscale/edge/tflite/app/proguard-rules.pro
new file mode 100644
index 00000000..f1b42451
--- /dev/null
+++ b/fedscale/edge/tflite/app/proguard-rules.pro
@@ -0,0 +1,21 @@
+# Add project specific ProGuard rules here.
+# You can control the set of applied configuration files using the
+# proguardFiles setting in build.gradle.
+#
+# For more details, see
+# http://developer.android.com/guide/developing/tools/proguard.html
+
+# If your project uses WebView with JS, uncomment the following
+# and specify the fully qualified class name to the JavaScript interface
+# class:
+#-keepclassmembers class fqcn.of.javascript.interface.for.webview {
+# public *;
+#}
+
+# Uncomment this to preserve the line number information for
+# debugging stack traces.
+#-keepattributes SourceFile,LineNumberTable
+
+# If you keep the line number information, uncomment this to
+# hide the original source file name.
+#-renamesourcefileattribute SourceFile
diff --git a/fedscale/edge/tflite/app/src/main/AndroidManifest.xml b/fedscale/edge/tflite/app/src/main/AndroidManifest.xml
new file mode 100644
index 00000000..20fef019
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/AndroidManifest.xml
@@ -0,0 +1,64 @@
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
diff --git a/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/executor/FLExecutor.java b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/executor/FLExecutor.java
new file mode 100644
index 00000000..f07ac375
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/executor/FLExecutor.java
@@ -0,0 +1,489 @@
+package com.fedscale.android.executor;
+
+import android.os.Handler;
+import android.os.HandlerThread;
+import androidx.appcompat.app.AppCompatActivity;
+import android.os.Bundle;
+import android.util.Log;
+import android.view.WindowManager;
+import android.widget.TextView;
+
+import com.fedscale.android.R;
+import com.fedscale.android.utils.Backend;
+import com.google.common.hash.Hashing;
+import com.google.protobuf.ByteString;
+import com.fedscale.android.tflite.TFLiteBackend;
+import com.fedscale.android.utils.ClientConnections;
+import com.fedscale.android.utils.Common;
+
+import org.json.JSONException;
+import org.json.JSONObject;
+import org.tensorflow.lite.support.common.FileUtil;
+
+import io.grpc.executor.CompleteRequest;
+import io.grpc.executor.PingRequest;
+import io.grpc.executor.RegisterRequest;
+import io.grpc.executor.ServerResponse;
+
+import java.io.ByteArrayInputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.ByteBuffer;
+import java.nio.charset.StandardCharsets;
+import java.util.Iterator;
+import java.util.LinkedList;
+import java.util.Queue;
+
+/**
+ * Sample com.fedscale.android executor with MNN backend support.
+ * Training and executing will be handled inside MNN C++.
+ * Server-client communication will be handled in JAVA.
+ */
+public class FLExecutor extends AppCompatActivity {
+ // TODO: 1. How to convert json(bytes) to bytes?
+ // TODO: 2. How to support batch train/test?
+ // TODO: 3. Test client with cloud
+ private JSONObject config;
+
+ private ByteBuffer currentModel;
+
+ private String mExecutorID;
+ private String aggregatorIP;
+ private int aggregatorPort;
+ private ClientConnections communicator;
+
+ private int round = 0;
+ private boolean receivedStopRequest = false;
+ private Queue eventQueue = new LinkedList<>();
+
+ private TextView mUserId;
+ private TextView mExecuteStatus;
+ private TextView mExecuteResult;
+
+ HandlerThread mThread;
+ Handler mHandle;
+
+ /**
+ * Initialization of the User Interface.
+ * Three lines of UI: executor id, status, result (not changed)
+ */
+ private void initUI() {
+ runOnUiThread(()->{
+ String userIdMsg = this.mExecutorID + ": Round " + this.round;
+ this.mUserId.setText(userIdMsg);
+ this.mExecuteStatus.setText(Common.CLIENT_CONNECT);
+ this.mExecuteResult.setText("");
+ });
+ }
+
+ private String initExecutorId(String username) {
+ long currTime = System.currentTimeMillis();
+ ByteBuffer buffer = ByteBuffer.allocate(Long.BYTES);
+ buffer.putLong(currTime);
+ return Hashing
+ .hmacSha256(buffer.array())
+ .hashString(username, StandardCharsets.UTF_8).toString();
+ }
+
+ // No need for setupEnv
+
+ /**
+ * Set up grpc connection.
+ */
+ private void setupCommunication() {
+ this.initControlCommunication();
+ }
+
+ // No need for setupSeed
+
+ /**
+ * Create communication channel between coordinator and executor.
+ * This channel serves control messages.
+ */
+ private void initControlCommunication() {
+ this.communicator.ConnectToServer();
+ }
+
+ // No need for initDataCommunication
+
+ // No need for initModel
+
+ /**
+ * Move all files/directories inside assetsDir into com.fedscale.android cache directory.
+ */
+ private void initData() throws IOException {
+ Log.i(Common.TAG, "Data movement starts ...");
+ Common.copyDir(getBaseContext(), "", getCacheDir());
+ Log.i(Common.TAG, "Data movement completes ...");
+ }
+
+ /**
+ * Initialize variables associated to conf.json.
+ */
+ private void initAsset() throws IOException, JSONException {
+ String configPath = getCacheDir() + "/conf.json";
+ String configStr = Common.readStringFromFile(configPath);
+ this.config = new JSONObject(configStr);
+ this.mExecutorID = this.initExecutorId(this.config.getString("username"));
+ this.aggregatorIP = this.config.getJSONObject("aggregator").getString("ip");
+ this.aggregatorPort = this.config.getJSONObject("aggregator").getInt("port");
+// this.communicator = new ClientConnections(
+// this.aggregatorIP,
+// this.aggregatorPort);
+ }
+
+ /**
+ * Start running the executor by setting up execution and communication environment,
+ * and monitoring the grpc message.
+ */
+ private void runExecutor() throws Exception {
+ this.initData();
+ this.initAsset();
+ this.initUI();
+// this.setupCommunication();
+// this.eventMonitor();
+
+ this.currentModel = FileUtil.loadMappedFile(getBaseContext(), "model.tflite");
+ JSONObject tmp = new JSONObject();
+ tmp.put("client_id", this.mExecutorID);
+ this.FLTrain(tmp.toString());
+ this.FLTest("{}");
+ }
+
+ /**
+ * Add new events to worker queues.
+ *
+ * @param request Add grpc request from server (e.g. MODEL_TEST, MODEL_TRAIN) to event_queue.
+ */
+ private void dispatchWorkerEvents(ServerResponse request) {
+ this.eventQueue.add(request);
+ }
+
+ /**
+ * Deserialize the response from server.
+ *
+ * @param responses Serialized response from server.
+ * @return Deserialized response string from server.
+ */
+ private String deserializeResponse(ByteString responses) throws IOException {
+ String res = responses.toString("UTF-8");
+ Common.largeLog("deserialize", res);
+ return res;
+ }
+
+ /**
+ * Serialize the response to send to server upon assigned job completion
+ *
+ * @param responses Client responses after job completion.
+ * @return The serialized response object to server.
+ */
+ private ByteString serializeResponse(JSONObject responses) {
+ Common.largeLog("serialize", responses.toString());
+ return ByteString.copyFrom(responses.toString().getBytes(StandardCharsets.UTF_8));
+ }
+
+ /**
+ * Receive the broadcast global model for current round.
+ *
+ * @param model The broadcast global model config.
+ */
+ public void FLUpdateModel(ByteString model) throws IOException, JSONException {
+ this.round++;
+ this.setText(this.mExecuteStatus, Common.UPDATE_MODEL);
+ this.setText(this.mUserId, this.mExecutorID + ": Round " + this.round);
+// InputStream is = new ByteArrayInputStream(model.toByteArray());
+// final String fileName = this.config.getJSONObject("model_conf").getString("path");
+// final String modelPath = getCacheDir() + "/" + fileName;
+// Common.inputStream2File(is, modelPath);
+ this.currentModel = model.asReadOnlyByteBuffer();
+ }
+
+ /**
+ * Load train config and data to start training on that client.
+ *
+ * @param config The client training config.
+ * @return The client id and train result.
+ */
+ public JSONObject FLTrain(String config) throws Exception {
+ this.setText(this.mExecuteStatus, Common.CLIENT_TRAIN);
+ JSONObject newTrainingConf = this.overrideConf(
+ this.config.getJSONObject("training_conf"),
+ config);
+ Backend backend = new TFLiteBackend();
+ JSONObject trainResult = backend.MLTrain(
+ getCacheDir().toString(),
+ this.currentModel,
+ this.config.getJSONObject("training_data"),
+ newTrainingConf);
+ Common.largeLog("[TRAIN]", trainResult.toString());
+// CompleteRequest request = CompleteRequest.newBuilder()
+// .setClientId(this.mExecutorID)
+// .setExecutorId(this.mExecutorID)
+// .setEvent(Common.CLIENT_TRAIN)
+// .setStatus(true).build();
+// long startTime = System.currentTimeMillis() / 1000;
+// while (System.currentTimeMillis() / 1000 - startTime < 180) {
+// try {
+// ServerResponse response = this.communicator.stub.cLIENTEXECUTECOMPLETION(request);
+// this.dispatchWorkerEvents(response);
+// break;
+// } catch (Exception e) {
+// Log.w(
+// Common.TAG,
+// String.format("Failed to connect to aggregator %s:%d. Will retry in 5 sec.",
+// this.aggregatorIP, this.aggregatorPort));
+// Thread.sleep(5 * 1000);
+// }
+// }
+ return trainResult;
+ }
+
+ /**
+ * Model Testing. By default, we test the accuracy on all data of clients in the test group
+ *
+ * @param config The client testing config.
+ */
+ public void FLTest(String config) throws Exception {
+ this.setText(this.mExecuteStatus, Common.MODEL_TEST);
+ JSONObject newTestingConf = this.overrideConf(
+ this.config.getJSONObject("testing_conf"),
+ config);
+ Backend backend = new TFLiteBackend();
+ JSONObject testResult = backend.MLTest(
+ getCacheDir().toString(),
+ this.currentModel,
+ this.config.getJSONObject("testing_data"),
+ newTestingConf);
+ JSONObject testRes = new JSONObject();
+ testRes.put("executorId", this.mExecutorID);
+ testRes.put("results", testResult);
+ Common.largeLog("[TEST]", testRes.toString());
+// CompleteRequest request = CompleteRequest.newBuilder()
+// .setClientId(this.mExecutorID)
+// .setExecutorId(this.mExecutorID)
+// .setEvent(Common.MODEL_TEST)
+// .setStatus(true)
+// .setDataResult(this.serializeResponse(testRes.toString())).
+// build();
+// long startTime = System.currentTimeMillis() / 1000;
+// while (System.currentTimeMillis() / 1000 - startTime < 180) {
+// try {
+// ServerResponse response = this.communicator.stub.cLIENTEXECUTECOMPLETION(request);
+// this.dispatchWorkerEvents(response);
+// break;
+// } catch (Exception e) {
+// Log.w(
+// Common.TAG,
+// String.format("Failed to connect to aggregator %s:%d. Will retry in 5 sec.",
+// this.aggregatorIP, this.aggregatorPort));
+// Thread.sleep(5 * 1000);
+// }
+// }
+ }
+
+ /**
+ * Stop the current executor
+ */
+ public void FLStop() throws InterruptedException {
+ this.setText(this.mExecuteStatus, Common.SHUT_DOWN);
+ this.communicator.CloseServerConnection();
+ this.receivedStopRequest = true;
+ }
+
+ /**
+ * Return the statistics of training dataset
+ *
+ * @return Return the statistics of training dataset.
+ */
+ private JSONObject reportExecutorInfoHandler() throws JSONException {
+ JSONObject executorInfo = this.config.getJSONObject("executor_info");
+ return executorInfo;
+ }
+
+ // No need for updateModelHandler
+
+ // No need for loadGlobalModel
+
+ /**
+ * Override the variable arguments for different client.
+ *
+ * @param oldConfJSON The default client runtime config.
+ * @param newConf The server specified client runtime config.
+ * @return The JSONObject of the updated client runtime config.
+ */
+ private JSONObject overrideConf(JSONObject oldConfJSON, String newConf) throws JSONException {
+ JSONObject newConfJSON = new JSONObject(newConf);
+ if (newConfJSON.has("client_id")) {
+ oldConfJSON.put("client_id", newConfJSON.getString("client_id"));
+ }
+ if (newConfJSON.has("task_config")) {
+ JSONObject newTaskConfJSON = newConfJSON.getJSONObject("task_config");
+ Iterator keys = newTaskConfJSON.keys();
+ while(keys.hasNext()) {
+ String key = keys.next();
+ oldConfJSON.put(key, newTaskConfJSON.get(key));
+ }
+ }
+ return oldConfJSON;
+ }
+
+ // No need for clientTrainer
+
+ // No need for trainingHandler
+
+ // No need for testingHandler
+
+ /**
+ * Register the executor information to the aggregator.
+ */
+ private void clientRegister() throws Exception {
+ this.setText(this.mExecuteStatus, "Registering");
+ RegisterRequest request = RegisterRequest.newBuilder()
+ .setExecutorId(this.mExecutorID)
+ .setClientId(this.mExecutorID)
+ .setExecutorInfo(this.serializeResponse(
+ this.reportExecutorInfoHandler()))
+ .build();
+ long startTime = System.currentTimeMillis() / 1000;
+ while (System.currentTimeMillis() / 1000 - startTime < 180) {
+ try {
+ ServerResponse response = this.communicator.stub.cLIENTREGISTER(request);
+ this.dispatchWorkerEvents(response);
+ break;
+ } catch (Exception e) {
+ Log.w(
+ Common.TAG,
+ String.format("Failed to connect to aggregator %s:%d. Will retry in 5 sec.",
+ this.aggregatorIP, this.aggregatorPort));
+ Thread.sleep(5 * 1000);
+ }
+ }
+ }
+
+ /**
+ * Ping the aggregator for new task.
+ */
+ private void clientPing() throws InterruptedException {
+ this.setText(this.mExecuteStatus, "Pinging");
+ PingRequest request = PingRequest.newBuilder()
+ .setClientId(this.mExecutorID)
+ .setExecutorId(this.mExecutorID).build();
+ long startTime = System.currentTimeMillis() / 1000;
+ while (System.currentTimeMillis() / 1000 - startTime < 180) {
+ try {
+ ServerResponse response = this.communicator.stub.cLIENTPING(request);
+ Log.i(Common.TAG, "Get EVENT " + response.getEvent());
+ this.dispatchWorkerEvents(response);
+ break;
+ } catch (Exception e) {
+ Log.w(
+ Common.TAG,
+ String.format("Failed to connect to aggregator %s:%d. Will retry in 5 sec.",
+ this.aggregatorIP, this.aggregatorPort));
+ Thread.sleep(5 * 1000);
+ }
+ }
+ }
+
+ /**
+ * Activate event handler once receiving new message
+ */
+ private void eventMonitor() throws Exception {
+ Log.i(Common.TAG, "Start monitoring events ...");
+ this.clientRegister();
+ while (!this.receivedStopRequest) {
+ if (this.eventQueue.size() > 0) {
+ ServerResponse request = this.eventQueue.remove();
+ String currentEvent = request.getEvent();
+ Log.i(Common.TAG, "Handling EVENT " + currentEvent);
+ if (currentEvent.equals(Common.CLIENT_TRAIN)) {
+ String trainConfigStr = this.deserializeResponse(request.getMeta());
+ JSONObject trainResult = this.FLTrain(trainConfigStr);
+ CompleteRequest cRequest = CompleteRequest.newBuilder()
+ .setClientId(this.mExecutorID)
+ .setExecutorId(this.mExecutorID)
+ .setEvent(Common.UPLOAD_MODEL)
+ .setStatus(true)
+ .setDataResult(this.serializeResponse(trainResult))
+ .build();
+ long startTime = System.currentTimeMillis() / 1000;
+ while (System.currentTimeMillis() / 1000 - startTime < 180) {
+ try {
+ ServerResponse response =
+ this.communicator.stub.cLIENTEXECUTECOMPLETION(cRequest);
+ this.dispatchWorkerEvents(response);
+ break;
+ } catch (Exception e) {
+ Log.w(
+ Common.TAG,
+ String.format(
+ "Failed to connect to aggregator %s:%d. Will retry in 5 sec.",
+ this.aggregatorIP, this.aggregatorPort));
+ Thread.sleep(5 * 1000);
+ }
+ }
+ } else if (currentEvent.equals(Common.MODEL_TEST)) {
+ this.FLTest(this.deserializeResponse(request.getMeta()));
+ } else if (currentEvent.equals(Common.UPDATE_MODEL)) {
+ this.FLUpdateModel(request.getData());
+ } else if (currentEvent.equals(Common.SHUT_DOWN)) {
+ this.FLStop();
+ }
+ } else {
+ Thread.sleep(1000);
+ this.clientPing();
+ }
+ }
+ }
+
+ /**
+ * Set text displayed on User Interface.
+ *
+ * @param text The associated TextView object.
+ * @param value The value of the updated text.
+ */
+ private void setText(final TextView text,final String value){
+ runOnUiThread(() -> text.setText(value));
+ }
+
+ @Override
+ protected void onCreate(Bundle savedInstanceState) {
+ super.onCreate(savedInstanceState);
+ getWindow().setFlags(WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON,
+ WindowManager.LayoutParams.FLAG_KEEP_SCREEN_ON);
+ getWindow().addFlags(WindowManager.LayoutParams.FLAG_FULLSCREEN);
+ setContentView(R.layout.activity_main);
+
+ this.mUserId = findViewById(R.id.userId);
+ this.mExecuteStatus = findViewById(R.id.executeStatus);
+ this.mExecuteResult = findViewById(R.id.executeResult);
+
+ mThread = new HandlerThread("MNNTrain");
+ mThread.start();
+ mHandle = new Handler(mThread.getLooper());
+
+ mHandle.post(() -> {
+ try {
+ runExecutor();
+ } catch (Exception e) {
+ e.printStackTrace();
+ }
+ });
+ }
+
+ @Override
+ protected void onPause() {
+ super.onPause();
+ }
+
+ @Override
+ protected void onResume() {
+ super.onResume();
+ }
+
+ @Override
+ protected void onDestroy() {
+ mThread.interrupt();
+ super.onDestroy();
+ }
+}
diff --git a/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/tflite/TFLiteBackend.java b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/tflite/TFLiteBackend.java
new file mode 100644
index 00000000..bff9fb36
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/tflite/TFLiteBackend.java
@@ -0,0 +1,288 @@
+package com.fedscale.android.tflite;
+
+import static java.lang.Math.min;
+
+import android.graphics.Bitmap;
+import android.graphics.BitmapFactory;
+import android.util.Log;
+import android.util.Pair;
+
+import com.fedscale.android.utils.Backend;
+import com.fedscale.android.utils.Common;
+
+import org.json.JSONException;
+import org.json.JSONObject;
+import org.tensorflow.lite.DataType;
+import org.tensorflow.lite.Interpreter;
+import org.tensorflow.lite.support.common.FileUtil;
+import org.tensorflow.lite.support.common.ops.NormalizeOp;
+import org.tensorflow.lite.support.image.ImageProcessor;
+import org.tensorflow.lite.support.image.TensorImage;
+import org.tensorflow.lite.support.image.ops.ResizeOp;
+import org.tensorflow.lite.support.image.ops.ResizeWithCropOrPadOp;
+
+import java.io.File;
+import java.io.FileInputStream;
+import java.io.IOException;
+import java.nio.ByteBuffer;
+import java.nio.FloatBuffer;
+import java.nio.IntBuffer;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Collections;
+import java.util.HashMap;
+import java.util.List;
+import java.util.Map;
+
+public class TFLiteBackend implements Backend {
+// private final Handler handler = new Handler(Looper.getMainLooper());
+
+ @Override
+ public JSONObject MLTrain(
+ String directory,
+ ByteBuffer model,
+ JSONObject trainingDataConf,
+ JSONObject trainingConf) throws JSONException, IOException {
+ final String trainImagesFolder = directory + "/" + trainingDataConf.getString("data" ) + "/";
+ final String trainImagesTxt = directory + "/" + trainingDataConf.getString("label");
+
+ final String clientId = trainingConf.getString("client_id");
+ final int numClasses = trainingConf.getInt("num_classes");
+ final int trainEpochs = trainingConf.getInt("epoch");
+ final int trainBatchSize = trainingConf.getInt("batch_size");
+ final int trainNumWorkers = trainingConf.getInt("num_workers");
+ final float learningRate = (float) trainingConf.getDouble("learning_rate");
+ final float lossDecay = (float) trainingConf.getDouble("loss_decay");
+ final float momentum = (float) trainingConf.getDouble("momentum");
+ final float weightDecay = (float) trainingConf.getDouble("weight_decay");
+ final int channel = trainingConf.getInt("channel");
+ final int height = trainingConf.getInt("height");
+ final int width = trainingConf.getInt("width");
+
+ List labels = FileUtil.loadLabels(new FileInputStream(trainImagesTxt));
+ final int dataCount = labels.size();
+ final int trainIterations = (int)Math.ceil((double)dataCount / (double)trainBatchSize);
+ Collections.shuffle(labels);
+
+ Pair, List> pair = dataLoader(
+ labels,
+ trainImagesFolder,
+ trainBatchSize,
+ height,
+ width,
+ channel,
+ numClasses);
+ List trainImageBatches = pair.first;
+ List trainLabelBatches = pair.second;
+
+ float epochTrainLoss = 0;
+ float currentLoss = 0;
+
+ Interpreter.Options options = new Interpreter.Options();
+ options.setNumThreads(trainNumWorkers);
+ Interpreter interpreter = new Interpreter(model, options);
+
+ for (int epoch = 0; epoch < trainEpochs; ++epoch) {
+ for (int batchIdx = 0; batchIdx < trainIterations; ++batchIdx) {
+ Map inputs = new HashMap<>();
+ inputs.put("data", trainImageBatches.get(batchIdx));
+ inputs.put("label", trainLabelBatches.get(batchIdx));
+
+ Map outputs = new HashMap<>();
+ FloatBuffer loss = FloatBuffer.allocate(1);
+ outputs.put("loss", loss);
+
+ interpreter.runSignature(inputs, outputs, "train");
+ currentLoss = loss.get(0);
+
+ if (epochTrainLoss == 0) {
+ epochTrainLoss = currentLoss;
+ } else {
+ epochTrainLoss = (1 - lossDecay) * epochTrainLoss +
+ lossDecay * currentLoss;
+ }
+
+ Log.i(
+ "Train",
+ String.format(
+ "[train][epoch][%d][batch][%d][loss][%f]",
+ epoch, batchIdx, currentLoss
+ )
+ );
+ }
+ }
+ Map inputs = new HashMap<>();
+ File outputFile = new File(directory + "/model.ckpt");
+ inputs.put("checkpoint_path", outputFile.getAbsolutePath());
+ Map outputs = new HashMap<>();
+ interpreter.runSignature(inputs, outputs, "save");
+
+ JSONObject results = new JSONObject();
+ results.put("client_id", clientId);
+ results.put("moving_loss", epochTrainLoss);
+ results.put("trained_size", trainEpochs * dataCount);
+ results.put("success", true);
+ byte[] newModel = Common.readBytes(new FileInputStream(outputFile.getAbsolutePath()));
+ results.put("update_weight", newModel);
+ results.put("utility", currentLoss * currentLoss * dataCount);
+ results.put("wall_duration", 0);
+ return results;
+ }
+
+ @Override
+ public JSONObject MLTest(
+ String directory,
+ ByteBuffer model,
+ JSONObject testingDataConf,
+ JSONObject testingConf) throws JSONException, IOException {
+ final String testImagesFolder = directory + "/" + testingDataConf.getString("data" ) + "/";
+ final String testImagesTxt = directory + "/" + testingDataConf.getString("label");
+
+ final int numClasses = testingConf.getInt("num_classes");
+ final int testBatchSize = testingConf.getInt("test_bsz");
+ final int testNumWorkers = testingConf.getInt("num_workers");
+ final int channel = testingConf.getInt("channel");
+ final int height = testingConf.getInt("height");
+ final int width = testingConf.getInt("width");
+
+ List labels = FileUtil.loadLabels(new FileInputStream(testImagesTxt));
+ final int dataCount = labels.size();
+ final int testIterations = (int)Math.ceil((double)dataCount / (double)testBatchSize);
+
+ Pair, List> pair = dataLoader(
+ labels,
+ testImagesFolder,
+ testBatchSize,
+ height,
+ width,
+ channel,
+ numClasses);
+ List testImageBatches = pair.first;
+ List testLabelBatches = pair.second;
+
+ Interpreter.Options options = new Interpreter.Options();
+ options.setNumThreads(testNumWorkers);
+ Interpreter interpreter = new Interpreter(model, options);
+
+ float testLoss = 0;
+ int correctTop1 = 0;
+ int correctTop5 = 0;
+ int sampleCount = 0;
+
+ for (int batchIdx = 0; batchIdx < testIterations; ++batchIdx) {
+ Map inputs = new HashMap<>();
+ inputs.put("data", testImageBatches.get(batchIdx));
+ inputs.put("label", testLabelBatches.get(batchIdx));
+
+ Map outputs = new HashMap<>();
+ FloatBuffer loss = FloatBuffer.allocate(1);
+ outputs.put("loss", loss);
+ IntBuffer top1 = IntBuffer.allocate(1);
+ outputs.put("top1", top1);
+ IntBuffer top5 = IntBuffer.allocate(1);
+ outputs.put("top5", top5);
+
+ interpreter.runSignature(inputs, outputs, "test");
+
+ testLoss += loss.get(0);
+ correctTop1 += top1.get(0);
+ correctTop5 += top5.get(0);
+ sampleCount += testImageBatches.get(batchIdx).length;
+
+ Log.i(
+ "Test",
+ String.format(
+ "[test][batch][%d][loss][%f][top 1][%d/%d=%f%%][top 5][%d/%d=%f%%]",
+ batchIdx, testLoss,
+ correctTop1, sampleCount, (float)correctTop1 / (float)sampleCount * 100,
+ correctTop5, sampleCount, (float)correctTop5 / (float)sampleCount * 100
+ )
+ );
+ }
+ float accuracyFinalTop1 = (float)correctTop1 / dataCount;
+ float accuracyFinalTop5 = (float)correctTop5 / dataCount;
+
+ JSONObject results = new JSONObject();
+ results.put("top_1", accuracyFinalTop1);
+ results.put("top_5", accuracyFinalTop5);
+ results.put("test_loss", testLoss);
+ results.put("test_len", dataCount);
+ return results;
+ }
+
+ private Pair, List> dataLoader(
+ List labels,
+ String imagesFolder,
+ int batchSize,
+ int height,
+ int width,
+ int channel,
+ int numClasses
+ ) {
+ List> imageBatches = new ArrayList<>();
+ List> labelBatches = new ArrayList<>();
+ for (String label: labels) {
+ final String imageFileName = imagesFolder + label.split(" ")[0];
+ final float[] imageBuffer = processImage(
+ BitmapFactory.decodeFile(imageFileName), height, width);
+ final float[] imageLabel = encodeLabel(
+ Integer.parseInt(label.split(" ")[1]), numClasses);
+ int currentLastIdx = imageBatches.size() - 1;
+ if (imageBatches.size() == 0 || imageBatches.get(currentLastIdx).size() == batchSize) {
+ List newImageBatch = new ArrayList<>();
+ imageBatches.add(newImageBatch);
+ List newLabelBatch = new ArrayList<>();
+ labelBatches.add(newLabelBatch);
+ currentLastIdx++;
+ }
+ imageBatches.get(currentLastIdx).add(imageBuffer);
+ labelBatches.get(currentLastIdx).add(imageLabel);
+ }
+ return new Pair<>(
+ this.generateBatchBuffers(imageBatches, channel * width * height),
+ this.generateBatchBuffers(labelBatches, numClasses));
+ }
+
+ private List generateBatchBuffers(
+ List> batches,
+ int allocateSizePerData) {
+ List batchBuffers = new ArrayList<>();
+ for (List batch: batches) {
+ float[][] newBatchBuffer = new float[batch.size()][allocateSizePerData];
+ int iter = 0;
+ for (float[] data: batch) {
+ newBatchBuffer[iter++] = data;
+ }
+ batchBuffers.add(newBatchBuffer);
+ }
+ return batchBuffers;
+ }
+
+ // Preprocess the image and convert it into a TensorImage for classification.
+ private float[] processImage(
+ Bitmap image,
+ int targetHeight,
+ int targetWidth) {
+ int height = image.getHeight();
+ int width = image.getWidth();
+ int cropSize = min(height, width);
+ ImageProcessor.Builder builder = new ImageProcessor.Builder();
+ ImageProcessor imageProcessor = builder
+ .add(new ResizeWithCropOrPadOp(cropSize, cropSize))
+ .add(new ResizeOp(targetHeight, targetWidth, ResizeOp.ResizeMethod.BILINEAR))
+ .add(new NormalizeOp(0f, 255f))
+ .build();
+ TensorImage tensorImage = new TensorImage(DataType.FLOAT32);
+ tensorImage.load(image);
+ return imageProcessor.process(tensorImage).getTensorBuffer().getFloatArray();
+ }
+
+
+ // encode the classes name to float array
+ private float[] encodeLabel(int id, int numClasses) {
+ float[] classArray = new float[numClasses];
+ Arrays.fill(classArray, 0f);
+ classArray[id] = 1f;
+ return classArray;
+ }
+}
\ No newline at end of file
diff --git a/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Backend.java b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Backend.java
new file mode 100644
index 00000000..b9a2a206
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Backend.java
@@ -0,0 +1,42 @@
+package com.fedscale.android.utils;
+
+import org.json.JSONException;
+import org.json.JSONObject;
+
+import java.io.IOException;
+import java.nio.ByteBuffer;
+
+/**
+ * Training and testing interface from Java to JNI C++ backend.
+ */
+public interface Backend {
+ /**
+ * Training interface from Java to JNI C++ backend.
+ *
+ * @param directory Directory of model and data.
+ * @param model Model.
+ * @param trainingDataConf Training set and label path config.
+ * @param trainingConf Stringified JSON config for training.
+ * @return Stringified JSON training result containing updated model in JSON.
+ */
+ JSONObject MLTrain(
+ String directory,
+ ByteBuffer model,
+ JSONObject trainingDataConf,
+ JSONObject trainingConf) throws JSONException, IOException;
+
+ /**
+ * Testing interface from Java to JNI C++ backend.
+ *
+ * @param directory Directory of model and data.
+ * @param model Model.
+ * @param testingDataConf Testing set and label path config.
+ * @param testingConf Stringified JSON config for testing.
+ * @return Stringified JSON testing result.
+ */
+ JSONObject MLTest(
+ String directory,
+ ByteBuffer model,
+ JSONObject testingDataConf,
+ JSONObject testingConf) throws JSONException, IOException;
+}
diff --git a/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/ClientConnections.java b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/ClientConnections.java
new file mode 100644
index 00000000..dd97d36f
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/ClientConnections.java
@@ -0,0 +1,52 @@
+package com.fedscale.android.utils;
+
+import android.util.Log;
+
+import java.util.concurrent.TimeUnit;
+
+import io.grpc.ManagedChannel;
+import io.grpc.ManagedChannelBuilder;
+import io.grpc.executor.JobServiceGrpc;
+
+/**
+ * GRPC connection from client to server.
+ */
+public class ClientConnections {
+ public int port;
+ public String address;
+ public ManagedChannel channel;
+ public JobServiceGrpc.JobServiceBlockingStub stub;
+ public final int MAX_MESSAGE_LENGTH = 1*1024*1024*1024;
+
+ /**
+ * Initialization of IP and port.
+ *
+ * @param address server IP.
+ * @param port server port.
+ */
+ public ClientConnections(String address, int port) {
+ this.address = address;
+ this.port = port;
+ }
+
+ /**
+ * Initialize connection to server.
+ */
+ public void ConnectToServer() {
+ Log.i("GRPC", "%%%%%%%%%% Opening grpc connection to " +
+ this.address + " %%%%%%%%%%");
+ this.channel = ManagedChannelBuilder
+ .forAddress(this.address, this.port)
+ .maxInboundMessageSize(this.MAX_MESSAGE_LENGTH)
+ .usePlaintext().build();
+ this.stub = JobServiceGrpc.newBlockingStub(this.channel);
+ }
+
+ /**
+ * Stop connection from server.
+ */
+ public void CloseServerConnection() throws InterruptedException {
+ Log.i("GRPC", "%%%%%%%%%% Closing grpc connection to the aggregator %%%%%%%%%%");
+ this.channel.shutdown().awaitTermination(5, TimeUnit.SECONDS);
+ }
+}
diff --git a/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Common.java b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Common.java
new file mode 100644
index 00000000..4b5e0c1d
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/java/com/fedscale/android/utils/Common.java
@@ -0,0 +1,145 @@
+package com.fedscale.android.utils;
+
+import android.content.Context;
+import android.graphics.BitmapFactory;
+import android.util.Log;
+
+import java.io.ByteArrayInputStream;
+import java.io.ByteArrayOutputStream;
+import java.io.File;
+import java.io.FileOutputStream;
+import java.io.IOException;
+import java.io.InputStream;
+import java.nio.FloatBuffer;
+import java.nio.charset.Charset;
+
+import com.google.common.io.Files;
+
+/**
+ * Const string representing status and tools for file operations.
+ */
+public class Common {
+ public static String TAG = "Demo";
+ public static String UPDATE_MODEL = "update_model";
+ public static String MODEL_TEST = "model_test";
+ public static String SHUT_DOWN = "terminate_executor";
+ public static String START_ROUND = "start_round";
+ public static String CLIENT_CONNECT = "client_connect";
+ public static String CLIENT_TRAIN = "client_train";
+ public static String DUMMY_EVENT = "dummy_event";
+ public static String UPLOAD_MODEL = "upload_model";
+
+ /**
+ * Read file into string.
+ *
+ * @param filename The name of the file.
+ * @return The string content of the file.
+ * @throws IOException
+ */
+ public static String readStringFromFile(String filename) throws IOException {
+ return Files.toString(new File(filename), Charset.forName("UTF-8"));
+ }
+
+ /**
+ * Copy file from asset to com.fedscale.android.
+ *
+ * @param context Android context.
+ * @param assetsFile Asset filename.
+ * @param outFile Destination filename.
+ * @throws IOException
+ */
+ public static void copyAssetResource2File(Context context, String assetsFile, String outFile) throws IOException {
+ InputStream is = context.getAssets().open(assetsFile);
+ inputStream2File(is, outFile);
+ }
+
+ /**
+ * Write string content to file.
+ *
+ * @param str String content.
+ * @param outFile Destination filename.
+ * @throws IOException
+ */
+ public static void writeString2File(String str, String outFile) throws IOException {
+ InputStream is = new ByteArrayInputStream(str.getBytes("UTF-8"));
+ inputStream2File(is, outFile);
+ }
+
+ /**
+ * Write InputStream content into file.
+ *
+ * @param is InputStream content.
+ * @param outFile Destination filename.
+ * @throws IOException
+ */
+ public static void inputStream2File(InputStream is, String outFile) throws IOException {
+ File outF = new File(outFile);
+ FileOutputStream fos = new FileOutputStream(outF);
+
+ int byteCount;
+ byte[] buffer = new byte[1024];
+ while ((byteCount = is.read(buffer)) != -1) {
+ fos.write(buffer, 0, byteCount);
+ }
+ fos.flush();
+ is.close();
+ fos.close();
+ outF.setReadable(true);
+ }
+
+ public static byte[] readBytes(InputStream inputStream) throws IOException {
+ byte[] buffer = new byte[1024];
+ ByteArrayOutputStream os = new ByteArrayOutputStream();
+ int byteCount;
+ while ((byteCount = inputStream.read(buffer)) != -1) {
+ os.write(buffer, 0, byteCount);
+ }
+ inputStream.close();
+ return os.toByteArray();
+ }
+
+ /**
+ * Copy asset directory into com.fedscale.android.
+ *
+ * @param context Android context.
+ * @param assetsDir Asset subdirectory.
+ * @param outDir Destination directory.
+ * @throws IOException
+ */
+ public static void copyDir(Context context, String assetsDir, File outDir) throws IOException {
+ if (!outDir.exists()) {
+ outDir.mkdir();
+ }
+ String[] assets = context.getAssets().list(assetsDir);
+ for (int i = 0; i < assets.length; ++i) {
+ String inPath = assetsDir + "/" + assets[i];
+ if (assetsDir.equals("")) {
+ inPath = assets[i];
+ }
+ String outPath = outDir + "/" + assets[i];
+ String[] subAssets = context.getAssets().list(inPath);
+ if (subAssets.length == 0) {
+ // copy file
+ copyAssetResource2File(context, inPath, outPath);
+ } else {
+ // copy directory
+ copyDir(context, inPath, new File(outPath));
+ }
+ }
+ }
+
+ /**
+ * Log large content.
+ *
+ * @param tag Tag for this log.
+ * @param content Content of the log.
+ */
+ public static void largeLog(String tag, String content) {
+ if (content.length() > 4000) {
+ Log.d(tag, content.substring(0, 4000));
+ largeLog(tag, content.substring(4000));
+ } else {
+ Log.d(tag, content);
+ }
+ }
+}
diff --git a/fedscale/edge/tflite/app/src/main/proto/job_api.proto b/fedscale/edge/tflite/app/src/main/proto/job_api.proto
new file mode 100644
index 00000000..b5f2287b
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/proto/job_api.proto
@@ -0,0 +1,44 @@
+// GRPC definition for the communication between the aggregator and the executors.
+// After modifying this file, run
+// $ python -m grpc_tools.protoc -I../communication/ --python_out=.. --grpc_python_out=.. job_api.proto
+syntax = "proto3";
+
+option java_multiple_files = true;
+option java_package = "io.grpc.executor";
+option java_outer_classname = "Executor";
+option objc_class_prefix = "EXE";
+
+package fedscale;
+
+service JobService {
+ rpc CLIENT_REGISTER (RegisterRequest) returns (ServerResponse) {}
+ rpc CLIENT_PING (PingRequest) returns (ServerResponse) {}
+ rpc CLIENT_EXECUTE_COMPLETION (CompleteRequest) returns (ServerResponse) {}
+}
+
+message ServerResponse {
+ string event = 1;
+ bytes meta = 2;
+ bytes data = 3;
+}
+
+message RegisterRequest {
+ string client_id = 1;
+ string executor_id = 2;
+ bytes executor_info = 3;
+}
+
+message PingRequest {
+ string client_id = 1;
+ string executor_id = 2;
+}
+
+message CompleteRequest {
+ string client_id = 1;
+ string executor_id = 2;
+ string event = 3;
+ bool status = 4;
+ string msg = 5;
+ string meta_result = 6;
+ bytes data_result = 7;
+}
\ No newline at end of file
diff --git a/fedscale/edge/tflite/app/src/main/res/drawable-v24/ic_launcher_foreground.xml b/fedscale/edge/tflite/app/src/main/res/drawable-v24/ic_launcher_foreground.xml
new file mode 100644
index 00000000..c7bd21db
--- /dev/null
+++ b/fedscale/edge/tflite/app/src/main/res/drawable-v24/ic_launcher_foreground.xml
@@ -0,0 +1,34 @@
+
+
+
+
+