build qwen-1.8b apk.

README.md

@@ -25,6 +25,7 @@ llm模型导出onnx模型请使用[llm-export](https://github.com/wangzhaode/llm
 
 其他版本：
 - Qwen-1_8B-Chat-int8：[![Download][download-qwen-1.8b-mnn-int8]][release-qwen-1.8b-mnn-int8]
+- Android APK: [![Download][download-qwen-1.8b-apk]][release-qwen-1.8b-apk]
 
 [download-chatglm-6b-onnx]: https://img.shields.io/github/downloads/wangzhaode/llm-export/chatglm-6b-onnx/total
 [download-chatglm2-6b-onnx]: https://img.shields.io/github/downloads/wangzhaode/llm-export/chatglm2-6b-onnx/total
@@ -51,6 +52,8 @@ llm模型导出onnx模型请使用[llm-export](https://github.com/wangzhaode/llm
 [download-llama2-7b-chat-mnn]: https://img.shields.io/github/downloads/wangzhaode/mnn-llm/llama2-7b-chat-mnn/total
 [download-qwen-1.8b-mnn]: https://img.shields.io/github/downloads/wangzhaode/mnn-llm/qwen-1.8b-mnn/total
 [download-qwen-1.8b-mnn-int8]: https://img.shields.io/github/downloads/wangzhaode/mnn-llm/qwen-1.8b-mnn-int8/total
+[download-qwen-1.8b-apk]: https://img.shields.io/github/downloads/wangzhaode/mnn-llm/qwen-1.8b-apk/total
+
 [release-chatglm-6b-mnn]: https://github.com/wangzhaode/mnn-llm/releases/tag/chatglm-6b-mnn
 [release-chatglm2-6b-mnn]: https://github.com/wangzhaode/mnn-llm/releases/tag/chatglm2-6b-mnn
 [release-chatglm3-6b-mnn]: https://github.com/wangzhaode/mnn-llm/releases/tag/chatglm3-6b-mnn
@@ -60,6 +63,8 @@ llm模型导出onnx模型请使用[llm-export](https://github.com/wangzhaode/llm
 [release-llama2-7b-chat-mnn]: https://github.com/wangzhaode/mnn-llm/releases/tag/llama2-7b-chat-mnn
 [release-qwen-1.8b-mnn]: https://github.com/wangzhaode/mnn-llm/releases/tag/qwen-1.8b-mnn
 [release-qwen-1.8b-mnn-int8]: https://github.com/wangzhaode/mnn-llm/releases/tag/qwen-1.8b-mnn-int8
+[release-qwen-1.8b-apk]: https://github.com/wangzhaode/mnn-llm/releases/tag/qwen-1.8b-apk
+
 
 ### 速度
 
@@ -157,12 +162,13 @@ adb shell "cd /data/local/tmp && export LD_LIBRARY_PATH=. && ./cli_demo -m model
 
 
 ## Reference
-- [chatglm-6b](https://huggingface.co/THUDM/chatglm-6b)
-- [chatglm2-6b](https://huggingface.co/THUDM/chatglm2-6b)
-- [codegeex2-6b](https://huggingface.co/THUDM/codegeex2-6b)
-- [Baichuan2-7B-Chat](https://huggingface.co/baichuan-inc/Baichuan2-7B-Chat)
-- [Qwen-7B-Chat](https://huggingface.co/tangger/Qwen-7B-Chat)
+- [chatglm-6b](https://modelscope.cn/models/ZhipuAI/chatglm-6b/summary)
+- [chatglm2-6b](https://modelscope.cn/models/ZhipuAI/chatglm2-6b/summary)
+- [chatglm3-6b](https://modelscope.cn/models/ZhipuAI/chatglm3-6b/summary)
+- [codegeex2-6b](https://modelscope.cn/models/ZhipuAI/codegeex2-6b/summary)
+- [Baichuan2-7B-Chat](https://modelscope.cn/models/baichuan-inc/baichuan-7B/summary)
+- [Qwen-7B-Chat](https://modelscope.cn/models/qwen/Qwen-7B-Chat/summary)
+- [Qwen-1.8B-Chat](https://modelscope.cn/models/qwen/Qwen-1_8B-Chat/summary)
 - [cpp-httplib](https://github.com/yhirose/cpp-httplib)
 - [chatgpt-web](https://github.com/xqdoo00o/chatgpt-web)
-- [cppjieba](https://github.com/yanyiwu/cppjieba)
 - [ChatViewDemo](https://github.com/BrettFX/ChatViewDemo)

android/app/src/main/java/com/mnn/llm/Conversation.java

@@ -31,7 +31,7 @@ public class Conversation extends BaseActivity {
     private Button send;
     private DateFormat mDateFormat;
     private Chat mChat;
-    private boolean mHistory = false;
+    private boolean mHistory = true;
 
     @Override
     protected void onCreate(Bundle savedInstanceState) {
@@ -138,12 +138,16 @@ public boolean onCreateOptionsMenu(Menu menu) {
 
     @Override
     public boolean onOptionsItemSelected(MenuItem item) {
+        /*
         if (mHistory) {
             Toast.makeText(getBaseContext(), "关闭上下文", Toast.LENGTH_SHORT).show();
         } else {
             Toast.makeText(getBaseContext(), "打开上下文", Toast.LENGTH_SHORT).show();
         }
         mHistory = !mHistory;
+        */
+        Toast.makeText(getBaseContext(), "清空记忆", Toast.LENGTH_SHORT).show();
+        mChat.Reset();
         return true;
     }
 }
@@ -168,7 +172,7 @@ public void run() {
         System.out.println("[MNN_DEBUG] start response\n");
         while (!last_response.contains("<eop>")) {
             try {
-                Thread.sleep(200);
+                Thread.sleep(50);
             } catch (Exception e) {}
             String response = new String(mChat.Response());
             if (response.equals(last_response)) {
@@ -177,6 +181,7 @@ public void run() {
                 last_response = response;
             }
             Message msg = new Message();
+            System.out.println("[MNN_DEBUG] " + response);
             msg.obj = response.replaceFirst("<eop>", "");
             mHandler.sendMessage(msg);
         }

android/app/src/main/java/com/mnn/llm/DownloadModel.java

@@ -2,11 +2,8 @@
 
 import android.app.AlertDialog;
 import android.content.DialogInterface;
-import android.content.Intent;
 import android.graphics.Color;
 import android.os.Bundle;
-import android.os.Handler;
-import android.os.Message;
 import android.view.View;
 import android.widget.Button;
 
@@ -29,8 +26,7 @@ protected void onCreate(Bundle savedInstanceState) {
         mDownloadAll = (Button)findViewById(R.id.download_all);
         // init Data
         String[] modelArray = this.getResources().getStringArray(R.array.model_list);
-        int[] modelSize = this.getResources().getIntArray(R.array.model_size);
-        mAdapter = new DownloadRecyclerView(this, modelArray, modelSize);
+        mAdapter = new DownloadRecyclerView(this, modelArray);
         mRecyclerView.setAdapter(mAdapter);
     }
     public void downloadAll(View view) {

android/app/src/main/java/com/mnn/llm/MainActivity.java

@@ -13,9 +13,9 @@
 import android.widget.RelativeLayout;
 import android.widget.TextView;
 
-import org.w3c.dom.Text;
-
 import java.io.File;
+import java.io.IOException;
+
 
 public class MainActivity extends AppCompatActivity {
     private Chat mChat;
@@ -28,8 +28,9 @@ public class MainActivity extends AppCompatActivity {
     private TextView mProcessName;
     private TextView mProcessPercent;
     // resource files
-    private String mModelDir = "/data/local/tmp/model";
-    private boolean mModelNeedDownload = true;
+    private String mModelName = "qwen-1.8b-int4";
+    private String mModelDir = "/data/local/tmp/chat/" + mModelName; // default dir
+    private boolean mModelReady = true;
     @Override
     protected void onCreate(Bundle savedInstanceState) {
         super.onCreate(savedInstanceState);
@@ -41,7 +42,8 @@ protected void onCreate(Bundle savedInstanceState) {
         mProcessBar = (ProgressBar)findViewById(R.id.process_bar);
         mProcessName = (TextView)findViewById(R.id.process_name);
         mProcessPercent = (TextView)findViewById(R.id.process_percent);
-        mModelDir = this.getCacheDir().toString() + "/model";
+        // using assert file
+        mModelDir = this.getCacheDir() + "/" + mModelName;
         mProcessHandler = new Handler() {
             @Override
             public void handleMessage(Message msg) {
@@ -58,55 +60,56 @@ public void handleMessage(Message msg) {
                 }
             }
         };
-        /*
-        File model = new File(mModelDir, "glm_block_0.mnn");
-        if (model.exists()) {
-            model.delete();
-        }
-         */
-        onCheckModels();
     }
 
     @Override
     protected void onResume() {
         super.onResume();
-        onCheckModels();
     }
 
     public void onCheckModels() {
-        mModelNeedDownload = checkModelsNeedDownload();
-        if (mModelNeedDownload) {
+        mModelReady = checkModelsReady();
+        // try copy from assert file
+        if (!mModelReady) {
+            try {
+                mModelDir = Common.copyAssetResource2File(this, mModelName);
+            } catch (IOException e) {
+                throw new RuntimeException(e);
+            } catch (InterruptedException e) {
+                throw new RuntimeException(e);
+            }
+            mModelReady = checkModelsReady();
+        }
+        // download models
+        if (!mModelReady) {
             mModelInfo.setVisibility(View.VISIBLE);
-            mModelInfo.setText("使用前请先下载模型！");
+            mModelInfo.setText("请下载模型文件");
             mLoadButton.setText("下载模型");
         } else {
             mModelInfo.setVisibility(View.VISIBLE);
-            mModelInfo.setText("模型下载完毕，请加载模型！");
+            mModelInfo.setText(mModelName + "模型文件就绪，模型加载中");
             mLoadButton.setText("加载模型");
         }
     }
-    public boolean checkModelsNeedDownload() {
+    public boolean checkModelsReady() {
         System.out.println("### Check Models!");
         File dir = new File(mModelDir);
         if (!dir.exists()) {
-            return true;
+            return false;
         }
         String[] modelArray = this.getResources().getStringArray(R.array.model_list);
-        int[] modelSize = this.getResources().getIntArray(R.array.model_size);
         for (int i = 0; i < modelArray.length; i++) {
             File model = new File(mModelDir, modelArray[i]);
             if (!model.exists()) {
-                return true;
-            }
-            if (model.length() != modelSize[i]) {
-                return true;
+                return false;
             }
         }
-        return false;
+        return true;
     }
 
     public void loadModel(View view) {
-        if (mModelNeedDownload) {
+        onCheckModels();
+        if (!mModelReady) {
             startActivity(new Intent(this, DownloadModel.class));
             return;
         }
@@ -115,7 +118,6 @@ public void loadModel(View view) {
         mLoadButton.setText("模型加载中 ...");
         mProcessView.setVisibility(View.VISIBLE);
         mChat = new Chat();
-        System.out.println("[MNN_DEBUG] is chat Ready: " + mChat.Ready());
         Handler handler = new Handler() {
             @Override
             public void handleMessage(Message msg) {

android/app/src/main/java/com/mnn/llm/recyclerdownload/DownloadRecyclerView.java

@@ -25,7 +25,7 @@ public class DownloadRecyclerView extends RecyclerView.Adapter<RecyclerView.View
     private Handler mHandler;
     private Button mButton;
 
-    public DownloadRecyclerView(Context context, String[] models, int[] modelSize) {
+    public DownloadRecyclerView(Context context, String[] models) {
         this.mContext = context;
         this.mItems = new ArrayList<DownloadData>();
         final String modelDir = context.getCacheDir().toString() + "/model";
@@ -48,7 +48,7 @@ public void handleMessage(Message msg) {
             }
         };
         for (int i = 0; i < models.length; i++) {
-            this.mItems.add(new DownloadData(mHandler, modelDir, models[i], i, modelSize[i]));
+            this.mItems.add(new DownloadData(mHandler, modelDir, models[i], i, 25751300));
         }
     }
 

android/app/src/main/java/com/mnn/llm/recylcerchat/ConversationRecyclerView.java

@@ -101,7 +101,6 @@ private void configureViewHolder3(HolderMe vh1, int position) {
     }
 
     private void configureViewHolder2(HolderYou vh1, int position) {
-            vh1.getTime().setText(items.get(position).getTime());
             vh1.getChatText().setText(items.get(position).getText());
     }
     private void configureViewHolder1(HolderDate vh1, int position) {

android/app/src/main/java/com/mnn/llm/recylcerchat/HolderYou.java

@@ -9,22 +9,13 @@
 
 public class HolderYou extends RecyclerView.ViewHolder {
 
-    private TextView time, chatText;
+    private TextView chatText;
 
     public HolderYou(View v) {
         super(v);
-        time = (TextView) v.findViewById(R.id.tv_time);
         chatText = (TextView) v.findViewById(R.id.tv_chat_text);
     }
 
-    public TextView getTime() {
-        return time;
-    }
-
-    public void setTime(TextView time) {
-        this.time = time;
-    }
-
     public TextView getChatText() {
         return chatText;
     }

android/app/src/main/jni/CMakeLists.txt

@@ -7,6 +7,7 @@ cmake_minimum_required(VERSION 3.10)
 include_directories(${CMAKE_CURRENT_LIST_DIR}/../../../../../include/)
 link_directories(${CMAKE_CURRENT_LIST_DIR}/libs/arm64-v8a)
 
+add_definitions(-DUSING_DISK_EMBED)
 FILE(GLOB SRCS ../../../../../src/*.cpp)
 add_library(llm_mnn SHARED llm_mnn_jni.cpp ${SRCS})
 

android/app/src/main/jni/llm_mnn_jni.cpp

@@ -10,7 +10,7 @@
 
 #include "llm.hpp"
 
-static Llm* llm;
+static std::unique_ptr<Llm> llm(nullptr);
 static std::stringstream response_buffer;
 
 extern "C" {
@@ -25,32 +25,33 @@ JNIEXPORT void JNI_OnUnload(JavaVM* vm, void* reserved) {
 }
 
 JNIEXPORT jboolean JNICALL Java_com_mnn_llm_Chat_Init(JNIEnv* env, jobject thiz, jstring modelDir) {
-    if (llm->load_progress() < 100) {
-        const char* model_dir = env->GetStringUTFChars(modelDir, 0);
-        llm = Llm::createLLM(model_dir);
+    const char* model_dir = env->GetStringUTFChars(modelDir, 0);
+    if (!llm.get()) {
+        llm.reset(Llm::createLLM(model_dir));
         llm->load(model_dir);
     }
     return JNI_TRUE;
 }
 
 JNIEXPORT jboolean JNICALL Java_com_mnn_llm_Chat_Ready(JNIEnv* env, jobject thiz) {
-    if (llm->load_progress() >= 100) {
+    if (llm.get() && llm->load_progress() >= 100) {
         return JNI_TRUE;
     }
     return JNI_FALSE;
 }
 
 JNIEXPORT jfloat JNICALL Java_com_mnn_llm_Chat_Progress(JNIEnv* env, jobject thiz) {
+    if (!llm.get()) return jfloat(0);
     return jfloat(llm->load_progress());
 }
 
 JNIEXPORT jstring JNICALL Java_com_mnn_llm_Chat_Submit(JNIEnv* env, jobject thiz, jstring inputStr) {
-    if (llm->load_progress() < 100) {
+    if (!llm.get()) {
         return env->NewStringUTF("Failed, Chat is not ready!");
     }
     const char* input_str = env->GetStringUTFChars(inputStr, 0);
     auto chat = [&](std::string str) {
-        llm->response(str, &response_buffer);
+        llm->response(str, &response_buffer, "<eop>");
     };
     std::thread chat_thread(chat, input_str);
     chat_thread.detach();

android/app/src/main/res/layout/activity_main.xml

@@ -28,7 +28,7 @@
             android:layout_height="wrap_content"
             android:layout_gravity="center"
             android:textSize="8pt"
-            android:text="ChatGLM-MNN" />
+            android:text="mnn-llm" />
     </LinearLayout>
 
     <LinearLayout

android/app/src/main/res/layout/layout_holder_you.xml

@@ -9,8 +9,7 @@
         android:layout_width="wrap_content"
         android:layout_height="wrap_content"
         android:layout_alignParentLeft="true"
-        android:orientation="horizontal"
-        android:gravity="center_vertical">
+        android:orientation="horizontal">
         <ImageView
             android:layout_width="40dp"
             android:layout_height="40dp"
@@ -19,7 +18,6 @@
         <TextView
             android:id="@+id/tv_chat_text"
             android:layout_toLeftOf="@id/tv_time"
-            android:maxWidth="@dimen/maxwidth"
             android:layout_width="wrap_content"
             android:layout_height="wrap_content"
             android:padding="@dimen/space10"
@@ -28,17 +26,5 @@
             android:textSize="@dimen/h3"
             android:textAppearance="@style/Light"
             android:background="@drawable/bg_chat_you"/>
-        <TextView
-            android:id="@+id/tv_time"
-            android:layout_marginLeft="@dimen/space10"
-            android:layout_width="wrap_content"
-            android:layout_height="wrap_content"
-            android:drawablePadding="@dimen/space5"
-            android:drawableLeft="@drawable/ic_clock"
-            android:text="5:04pm"
-            android:textColor="@color/colorTextInfo"
-            android:textSize="@dimen/h4"
-            android:textAppearance="@style/ExtraLight"
-            android:layout_centerVertical="true"/>
     </LinearLayout>
 </RelativeLayout>