feat: make prebuilt libs

.github/workflows/ci.yml

@@ -35,7 +35,7 @@ jobs:
       - name: Run unit tests
         run: yarn test --maxWorkers=2 --coverage
 
-  build-ios:
+  build-ios-from-source:
     runs-on: macos-latest
     steps:
       - name: Checkout
@@ -69,7 +69,55 @@ jobs:
         run: |
           yarn build:ios
 
-  build-android:
+  build-ios-frameworks:
+    runs-on: macos-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Setup
+        uses: ./.github/actions/setup
+
+      - name: Cache build & pods
+        uses: actions/cache@v3
+        with:
+          path: |
+            example/ios/Pods
+            example/ios/build
+          key: ${{ runner.os }}-pods-${{ hashFiles('example/ios/Podfile.lock') }}
+          restore-keys: |
+            ${{ runner.os }}-pods-
+
+      - name: Upgrade CocoaPods to version 1.15.2
+        run: |
+          gem uninstall cocoapods --ignore-dependencies
+          gem install cocoapods -v 1.15.2
+
+      - name: Install cocoapods
+        run: |
+          yarn example pods
+        env:
+          NO_FLIPPER: 1
+          RNLLAMA_BUILD_FROM_SOURCE: 0
+
+      - name: Build example for iOS
+        run: |
+          yarn build:ios
+
+  build-android-from-source:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v3
+
+      - name: Setup
+        uses: ./.github/actions/setup
+
+      - name: Build example for Android
+        run: |
+          yarn build:android
+
+  build-android-libs:
     runs-on: ubuntu-latest
     steps:
       - name: Checkout
@@ -80,4 +128,5 @@ jobs:
 
       - name: Build example for Android
         run: |
+          sed -i 's/rnllamaBuildFromSource=true/rnllamaBuildFromSource=false/g' example/android/gradle.properties
           yarn build:android

.gitignore

@@ -73,3 +73,11 @@ docs/API/.nojekyll
 
 *.metallib
 .xocde.env.local
+
+build-arm64/
+build-x86_64/
+jniLibs/
+build-ios/
+build-tvos/
+
+ios/rnllama.xcframework/**/*.framework

CONTRIBUTING.md

@@ -17,22 +17,6 @@ yarn bootstrap
 
 While developing, you can run the [example app](/example/) to test your changes. Any changes you make in your library's JavaScript code will be reflected in the example app without a rebuild. If you change any native code, then you'll need to rebuild the example app.
 
-To start the packager:
-
-```sh
-yarn example start
-```
-
-To run the example app on iOS:
-
-```sh
-yarn example pods
-yarn example ios
-```
-
-For test better performance on completion, you can run the app in Release mode:
-- iOS: `yarn example ios --mode Release`
-
 Make sure your code passes TypeScript and ESLint. Run the following to verify:
 
 ```sh

README.md

@@ -18,6 +18,8 @@ npm install llama.rn
 
 Please re-run `npx pod-install` again.
 
+By default, `llama.rn` will use pre-built `rnllama.xcframework` for iOS. If you want to build from source, please set `RNLLAMA_BUILD_FROM_SOURCE` to `1` in your Podfile.
+
 #### Android
 
 Add proguard rule if it's enabled in project (android/app/proguard-rules.pro):
@@ -27,6 +29,8 @@ Add proguard rule if it's enabled in project (android/app/proguard-rules.pro):
 -keep class com.rnllama.** { *; }
 ```
 
+By default, `llama.rn` will use pre-built libraries for Android. If you want to build from source, please set `rnllamaBuildFromSource` to `true` in `android/gradle.properties`.
+
 ## Obtain the model
 
 You can search HuggingFace for available models (Keyword: [`GGUF`](https://huggingface.co/search/full-text?q=GGUF&type=model)).
@@ -118,29 +122,6 @@ Please visit the [Documentation](docs/API) for more details.
 
 You can also visit the [example](example) to see how to use it.
 
-Run the example:
-
-```bash
-yarn && yarn bootstrap
-
-# iOS
-yarn example ios
-# Use device
-yarn example ios --device "<device name>"
-# With release mode
-yarn example ios --mode Release
-
-# Android
-yarn example android
-# With release mode
-yarn example android --mode release
-```
-
-This example used [react-native-document-picker](https://github.com/rnmods/react-native-document-picker) for select model.
-
-- iOS: You can move the model to iOS Simulator, or iCloud for real device.
-- Android: Selected file will be copied or downloaded to cache directory so it may be slow.
-
 ## Grammar Sampling
 
 GBNF (GGML BNF) is a format for defining [formal grammars](https://en.wikipedia.org/wiki/Formal_grammar) to constrain model outputs in `llama.cpp`. For example, you can use it to force the model to generate valid JSON, or speak only in emojis.

android/build.gradle

@@ -54,9 +54,18 @@ android {
       }
     }
   }
-  externalNativeBuild {
-    cmake {
-      path = file('src/main/CMakeLists.txt')
+  def rnllamaBuildFromSource = project.properties["rnllamaBuildFromSource"]
+  if (rnllamaBuildFromSource == "true") {
+    externalNativeBuild {
+      cmake {
+        path = file('src/main/CMakeLists.txt')
+      }
+    }
+    // Exclude jniLibs
+    sourceSets {
+      main {
+        jniLibs.srcDirs = []
+      }
     }
   }
   buildTypes {

android/src/main/CMakeLists.txt

@@ -45,9 +45,9 @@ set(
     ${RNLLAMA_LIB_DIR}/unicode.cpp
     ${RNLLAMA_LIB_DIR}/sgemm.cpp
     ${RNLLAMA_LIB_DIR}/common.cpp
-    ${RNLLAMA_LIB_DIR}/rn-llama.hpp
     ${RNLLAMA_LIB_DIR}/amx/amx.cpp
     ${RNLLAMA_LIB_DIR}/amx/mmq.cpp
+    ${RNLLAMA_LIB_DIR}/rn-llama.cpp
     ${CMAKE_SOURCE_DIR}/jni-utils.h
     ${CMAKE_SOURCE_DIR}/jni.cpp
 )

android/src/main/jni.cpp

@@ -11,7 +11,7 @@
 #include "llama.h"
 #include "llama-impl.h"
 #include "ggml.h"
-#include "rn-llama.hpp"
+#include "rn-llama.h"
 #include "jni-utils.h"
 
 #define UNUSED(x) (void)(x)

cpp/README.md

@@ -1,4 +1,4 @@
 # Note
 
-- Only `rn-llama.hpp` is the specific file for this project, others are sync from [llama.cpp](https://github.com/ggerganov/llama.cpp).
+- Only `rn-llama.h` and `rn-llama.cpp` are the specific files for this folder, others are sync from [llama.cpp](https://github.com/ggerganov/llama.cpp).
 - We can update the native source by using the [bootstrap](../scripts/bootstrap.sh) script.