updating code to match to match llamacpp tag b4689

.github/workflows/ci.yml

@@ -1,25 +1,23 @@
-# This work flow runs all Java tests for continuous integration.
-# Since it has to build llama.cpp first, for speed, it only runs / tests on the natively supported GitHub runners.
-
+---
 name: Continuous Integration
-on: [ "pull_request", "workflow_dispatch" ]
+on:
+  - pull_request
+  - workflow_dispatch
 env:
-  MODEL_URL: "https://huggingface.co/TheBloke/CodeLlama-7B-GGUF/resolve/main/codellama-7b.Q2_K.gguf"
-  MODEL_NAME: "codellama-7b.Q2_K.gguf"
+  MODEL_URL: https://huggingface.co/TheBloke/CodeLlama-7B-GGUF/resolve/main/codellama-7b.Q2_K.gguf
+  MODEL_NAME: codellama-7b.Q2_K.gguf
 jobs:
 
-  # don't split build and test jobs to keep the workflow simple
   build-and-test-linux:
     name: ubuntu-latest
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-java@v4
         with:
-          distribution: 'zulu'
-          java-version: '11'
+          distribution: zulu
+          java-version: "11"
       - name: Build libraries
-        # cmake should figure out OS and ARCH automatically when running build.sh (but we need mvn compile for it)
         run: |
           mvn compile
           .github/build.sh -DLLAMA_VERBOSE=ON
@@ -41,20 +39,16 @@ jobs:
       fail-fast: false
       matrix:
         target:
-          - {
-            runner: macos-13,
-            cmake: '-DLLAMA_METAL=OFF -DLLAMA_VERBOSE=ON'
-          }
-          - {
-            runner: macos-14,
-            cmake: '-DLLAMA_METAL_EMBED_LIBRARY=ON -DLLAMA_METAL=OFF -DLLAMA_VERBOSE=ON'
-          }
+          - runner: macos-13
+            cmake: -DLLAMA_METAL=OFF -DLLAMA_VERBOSE=ON
+          - runner: macos-14
+            cmake: -DLLAMA_METAL_EMBED_LIBRARY=ON -DLLAMA_VERBOSE=ON
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-java@v4
         with:
-          distribution: 'zulu'
-          java-version: '11'
+          distribution: zulu
+          java-version: "11"
       - name: Build libraries
         run: |
           mvn compile
@@ -71,8 +65,8 @@ jobs:
           if-no-files-found: warn
 
   build-and-test-windows:
-    name: windows-latest
-    runs-on: windows-latest
+    name: windows-2019
+    runs-on: windows-2019
     steps:
       - uses: actions/checkout@v4
       - uses: actions/setup-java@v4
@@ -90,6 +84,8 @@ jobs:
       - if: failure()
         uses: actions/upload-artifact@v4
         with:
-          name: error-log-windows
-          path: ${{ github.workspace }}\hs_err_pid*.log
+          name: windows-output
+          path: |
+            ${{ github.workspace }}\hs_err_pid*.log
+            ${{ github.workspace }}/src/main/resources/de/kherud/llama/**/*
           if-no-files-found: warn

.github/workflows/release.yaml

@@ -18,6 +18,8 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
+      - name: Install libcurl
+        run: sudo apt-get install -y libcurl4-openssl-dev
       - name: Build libraries
         shell: bash
         run: |
@@ -121,10 +123,12 @@ jobs:
           }
     steps:
       - uses: actions/checkout@v4
+      - name: Install curl
+        run: vcpkg install curl
       - name: Build libraries
         shell: cmd
         run: |
-          .github\build.bat ${{ matrix.target.cmake }} -DOS_NAME=${{ matrix.target.os }} -DOS_ARCH=${{ matrix.target.arch }}
+          .github\build.bat ${{ matrix.target.cmake }} -DOS_NAME=${{ matrix.target.os }} -DOS_ARCH=${{ matrix.target.arch }} -DCURL_LIBRARY=C:/vcpkg/packages/curl_x64-windows/lib/libcurl.lib -DCURL_INCLUDE_DIR=C:/vcpkg/packages/curl_x64-windows/include
       - name: Upload artifacts
         uses: actions/upload-artifact@v4
         with:

.gitignore

@@ -1,6 +1,7 @@
 .idea
 target
 build
+cmake-build-*
 .DS_Store
 .directory
 .vscode

CMakeLists.txt

@@ -6,6 +6,7 @@ include(FetchContent)
 
 set(BUILD_SHARED_LIBS ON)
 set(CMAKE_POSITION_INDEPENDENT_CODE ON)
+set(BUILD_SHARED_LIBS OFF)
 
 option(LLAMA_VERBOSE	"llama: verbose output"		OFF)
 
@@ -20,10 +21,11 @@ FetchContent_MakeAvailable(json)
 
 #################### llama.cpp ####################
 
+set(LLAMA_BUILD_COMMON ON)
 FetchContent_Declare(
 	llama.cpp
 	GIT_REPOSITORY https://github.com/ggerganov/llama.cpp.git
-	GIT_TAG        b3534
+	GIT_TAG        b4831
 )
 FetchContent_MakeAvailable(llama.cpp)
 
@@ -102,9 +104,10 @@ target_compile_definitions(jllama PRIVATE
 )
 
 if(OS_NAME STREQUAL "Windows")
-	set_target_properties(jllama llama ggml PROPERTIES
+    set_target_properties(jllama llama ggml PROPERTIES
 	  RUNTIME_OUTPUT_DIRECTORY_DEBUG ${JLLAMA_DIR}
 	  RUNTIME_OUTPUT_DIRECTORY_RELEASE ${JLLAMA_DIR}
+	  RUNTIME_OUTPUT_DIRECTORY_RELWITHDEBINFO ${JLLAMA_DIR}
 	)
 else()
 	set_target_properties(jllama llama ggml PROPERTIES

README.md

@@ -27,7 +27,7 @@ Access this library via Maven:
 <dependency>
     <groupId>de.kherud</groupId>
     <artifactId>llama</artifactId>
-    <version>3.4.1</version>
+    <version>3.4.2</version>
 </dependency>
 ```
 
@@ -37,7 +37,7 @@ Bu default the default library artifact is built only with CPU inference support
 <dependency>
     <groupId>de.kherud</groupId>
     <artifactId>llama</artifactId>
-    <version>3.4.1</version>
+    <version>3.4.2</version>
     <classifier>cuda12-linux-x86-64</classifier>
 </dependency>
 ```
@@ -78,7 +78,7 @@ cmake --build build --config Release
 ```
 
 > [!TIP]
-> Use `-DGGML_CURL=ON` to download models via Java code using `ModelParameters#setModelUrl(String)`.
+> Use `-DLLAMA_CURL=ON` to download models via Java code using `ModelParameters#setModelUrl(String)`.
 
 All compiled libraries will be put in a resources directory matching your platform, which will appear in the cmake output. For example something like:
 

pom.xml

@@ -4,7 +4,7 @@
 
 	<groupId>de.kherud</groupId>
 	<artifactId>llama</artifactId>
-	<version>3.4.1</version>
+	<version>3.4.2</version>
 	<packaging>jar</packaging>
 
 	<name>${project.groupId}:${project.artifactId}</name>