Add YAML configuration support to llama-cli with --config flag

.github/workflows/config.yml

@@ -0,0 +1,41 @@
+name: YAML Config Tests
+
+on:
+  push:
+    branches: [ "master", "devin/*" ]
+  pull_request:
+    branches: [ "master" ]
+
+jobs:
+  test-yaml-config:
+    runs-on: ubuntu-latest
+
+    steps:
+    - uses: actions/checkout@v4
+      with:
+        submodules: recursive
+
+    - name: Dependencies
+      id: depends
+      run: |
+        sudo apt-get update
+        sudo apt-get install -y build-essential cmake git-lfs
+        git lfs install
+
+    - name: Download tiny model (stories15M)
+      run: |
+        mkdir -p models
+        # Download only the specific model file we need (19MB) to avoid disk space issues
+        wget -q "https://huggingface.co/ggml-org/models/resolve/main/tinyllamas/stories15M-q4_0.gguf" -O models/stories15M-q4_0.gguf
+        ls -lh models/stories15M-q4_0.gguf
+
+    - name: Build
+      id: cmake_build
+      run: |
+        cmake -B build -DLLAMA_BUILD_TESTS=ON -DLLAMA_BUILD_TOOLS=ON -DLLAMA_FATAL_WARNINGS=ON -DLLAMA_CURL=OFF
+        cmake --build build --config Release -j $(nproc)
+
+    - name: Test YAML config functionality
+      run: |
+        cd build
+        ctest -R "test-config-yaml|test-config-yaml-cli-.*|test-config-yaml-parity" --output-on-failure --timeout 300

README.md

@@ -54,6 +54,38 @@ llama-cli -hf ggml-org/gemma-3-1b-it-GGUF
 llama-server -hf ggml-org/gemma-3-1b-it-GGUF
 ```
 
+### YAML Configuration
+
+You can use YAML configuration files to set parameters instead of command-line flags:
+
+```bash
+llama-cli --config configs/minimal.yaml
+```
+
+Example `minimal.yaml`:
+```yaml
+model:
+  path: models/tinyllama-1.1b-chat-v1.0.Q4_K_M.gguf
+n_ctx: 256
+sampling:
+  seed: 42
+  temp: 0.0
+prompt: "Hello from YAML"
+n_predict: 16
+simple_io: true
+```
+
+You can override YAML values with command-line flags:
+```bash
+llama-cli --config configs/minimal.yaml -n 32 --temp 0.8
+```
+
+**Precedence rules:** Command-line flags > YAML config > defaults
+
+**Path resolution:** Relative paths in YAML files are resolved relative to the YAML file's directory.
+
+**Error handling:** Unknown YAML keys will cause an error with a list of valid keys.
+
 ## Description
 
 The main goal of `llama.cpp` is to enable LLM inference with minimal setup and state-of-the-art performance on a wide

common/CMakeLists.txt

@@ -133,6 +133,32 @@ if (LLAMA_LLGUIDANCE)
     set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} llguidance ${LLGUIDANCE_PLATFORM_LIBS})
 endif ()
 
+if (LLAMA_BUILD_TOOLS)
+    # yaml-cpp for YAML config (CLI-only)
+    find_package(yaml-cpp QUIET)
+    if (NOT yaml-cpp_FOUND)
+        include(FetchContent)
+        FetchContent_Declare(yaml-cpp
+            GIT_REPOSITORY https://github.com/jbeder/yaml-cpp.git
+            GIT_TAG        0.8.0)
+        set(YAML_CPP_BUILD_TESTS OFF CACHE BOOL "" FORCE)
+        set(YAML_CPP_BUILD_TOOLS OFF CACHE BOOL "" FORCE)
+        set(YAML_CPP_BUILD_CONTRIB OFF CACHE BOOL "" FORCE)
+        FetchContent_MakeAvailable(yaml-cpp)
+        # Suppress all warnings for yaml-cpp to avoid -Werror failures
+        if(TARGET yaml-cpp)
+            target_compile_options(yaml-cpp PRIVATE -w)
+        endif()
+    endif()
+
+    target_sources(${TARGET} PRIVATE
+        ${CMAKE_CURRENT_SOURCE_DIR}/config.cpp
+        ${CMAKE_CURRENT_SOURCE_DIR}/config.h
+    )
+    target_link_libraries(${TARGET} PRIVATE yaml-cpp)
+    target_compile_definitions(${TARGET} PUBLIC LLAMA_ENABLE_CONFIG_YAML)
+endif()
+
 target_include_directories(${TARGET} PUBLIC . ../vendor)
 target_compile_features   (${TARGET} PUBLIC cxx_std_17)
 target_link_libraries     (${TARGET} PRIVATE ${LLAMA_COMMON_EXTRA_LIBS} PUBLIC llama Threads::Threads)

common/arg.cpp

@@ -2,6 +2,9 @@
 
 #include "chat.h"
 #include "common.h"
+#ifdef LLAMA_ENABLE_CONFIG_YAML
+#include "config.h"
+#endif
 #include "gguf.h" // for reading GGUF splits
 #include "json-schema-to-grammar.h"
 #include "log.h"
@@ -1223,6 +1226,26 @@ bool common_params_parse(int argc, char ** argv, common_params & params, llama_e
     const common_params params_org = ctx_arg.params; // the example can modify the default params
 
     try {
+#ifdef LLAMA_ENABLE_CONFIG_YAML
+        for (int i = 1; i < argc; ++i) {
+            if (std::string(argv[i]) == "--config") {
+                if (i + 1 >= argc) {
+                    throw std::invalid_argument("error: --config requires a file path");
+                }
+                std::string cfg_path = argv[++i];
+                if (!common_load_yaml_config(cfg_path, ctx_arg.params)) {
+                    throw std::invalid_argument("error: failed to load YAML config: " + cfg_path);
+                }
+                break;
+            }
+        }
+#else
+        for (int i = 1; i < argc; ++i) {
+            if (std::string(argv[i]) == "--config") {
+                throw std::invalid_argument("error: this build does not include YAML config support (LLAMA_BUILD_TOOLS=OFF)");
+            }
+        }
+#endif
         if (!common_params_parse_ex(argc, argv, ctx_arg)) {
             ctx_arg.params = params_org;
             return false;
@@ -1317,6 +1340,16 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.completion = true;
         }
     ));
+
+#ifdef LLAMA_ENABLE_CONFIG_YAML
+    add_opt(common_arg(
+        {"--config"},
+        "<path/to/config.yaml>",
+        "Load parameters from a YAML config file; flags passed on the command line override values from the YAML file.",
+        [](common_params &, const std::string &) {
+        }
+    ));
+#endif
     add_opt(common_arg(
         {"--verbose-prompt"},
         string_format("print a verbose prompt before generation (default: %s)", params.verbose_prompt ? "true" : "false"),