Implement llama-pull tool

Complete llama-pull tool with documentation

Signed-off-by: Eric Curtin <[email protected]>

Author: ericcurtin

common/arg.cpp

@@ -1393,6 +1393,11 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
         throw std::invalid_argument("error: --prompt-cache-all not supported in interactive mode yet\n");
     }
 
+    // Both cannot be specified at the same time
+    if (!params.model.hf_repo.empty() && !params.model.docker_repo.empty()) {
+        throw std::invalid_argument("error: cannot specify both -hf and -dr options\n");
+    }
+
     // handle model and download
     {
         auto res = common_params_handle_model(params.model, params.hf_token, DEFAULT_MODEL_PATH, params.offline);

tools/CMakeLists.txt

@@ -18,6 +18,7 @@ else()
     add_subdirectory(gguf-split)
     add_subdirectory(imatrix)
     add_subdirectory(llama-bench)
+    add_subdirectory(pull)
     add_subdirectory(main)
     add_subdirectory(perplexity)
     add_subdirectory(quantize)

tools/pull/CMakeLists.txt

@@ -0,0 +1,8 @@
+set(TARGET llama-pull)
+add_executable(${TARGET} pull.cpp)
+target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
+target_compile_features(${TARGET} PRIVATE cxx_std_17)
+
+if(LLAMA_TOOLS_INSTALL)
+    install(TARGETS ${TARGET} RUNTIME)
+endif()

tools/pull/README.md

@@ -0,0 +1,43 @@
+# llama-pull - Model Download Tool
+
+A command-line tool for downloading AI models from HuggingFace and Docker Hub for use with llama.cpp.
+
+## Usage
+
+```bash
+# Download from HuggingFace
+llama-pull -hf <user>/<model>[:<quant>]
+
+# Download from Docker Hub
+llama-pull -dr [<repo>/]<model>[:<quant>]
+```
+
+## Options
+
+- `-hf, --hf-repo REPO` - Download model from HuggingFace repository
+- `-dr, --docker-repo REPO` - Download model from Docker Hub
+- `--hf-token TOKEN` - HuggingFace token for private repositories
+- `-h, --help` - Show help message
+
+## Examples
+
+```bash
+# Download a HuggingFace model
+llama-pull -hf microsoft/DialoGPT-medium
+
+# Download a Docker model (ai/ repo is default)
+llama-pull -dr gemma3
+
+# Download with specific quantization
+llama-pull -hf bartowski/Llama-3.2-1B-Instruct-GGUF:Q4_K_M
+```
+
+## Model Storage
+
+Downloaded models are stored in the standard llama.cpp cache directory:
+- Linux/macOS: `~/.cache/llama.cpp/`
+- The models can then be used with other llama.cpp tools
+
+## Requirements
+
+- Built with `LLAMA_USE_CURL=ON` (default) for download functionality

tools/pull/pull.cpp

@@ -0,0 +1,69 @@
+#include "arg.h"
+#include "common.h"
+#include "log.h"
+
+#include <cstdio>
+#include <string>
+
+static void print_usage(int, char ** argv) {
+    LOG("Usage: %s [options]\n", argv[0]);
+    LOG("\n");
+    LOG("Download models from HuggingFace or Docker Hub\n");
+    LOG("\n");
+    LOG("Options:\n");
+    LOG("  -h, --help                show this help message and exit\n");
+    LOG("  -hf, -hfr, --hf-repo REPO download model from HuggingFace repository\n");
+    LOG("                            format: <user>/<model>[:<quant>]\n");
+    LOG("                            example: microsoft/DialoGPT-medium\n");
+    LOG("  -dr, --docker-repo REPO   download model from Docker Hub\n");
+    LOG("                            format: [<repo>/]<model>[:<quant>]\n");
+    LOG("                            example: gemma3\n");
+    LOG("  --hf-token TOKEN          HuggingFace token for private repositories\n");
+    LOG("\n");
+    LOG("Examples:\n");
+    LOG("  %s -hf microsoft/DialoGPT-medium\n", argv[0]);
+    LOG("  %s -dr gemma3\n", argv[0]);
+    LOG("  %s -hf microsoft/DialoGPT-medium -o ./my-model.gguf\n", argv[0]);
+    LOG("\n");
+}
+
+int main(int argc, char ** argv) {
+    common_params params;
+
+    // Set up argument parsing context
+    auto ctx = common_params_parser_init(params, LLAMA_EXAMPLE_COMMON, print_usage);
+
+    // Parse command line arguments
+    if (!common_params_parse(argc, argv, params, LLAMA_EXAMPLE_COMMON)) {
+        print_usage(argc, argv);
+        return 1;
+    }
+
+    // Check if help was requested or no download option provided
+    if (params.model.hf_repo.empty() && params.model.docker_repo.empty()) {
+        LOG_ERR("error: must specify either -hf <repo> or -dr <repo>\n");
+        print_usage(argc, argv);
+        return 1;
+    }
+
+    LOG_INF("llama-pull: downloading model...\n");
+    try {
+        // Use the existing model handling logic which downloads the model
+        common_init_result llama_init = common_init_from_params(params);
+
+        if (llama_init.model != nullptr) {
+            LOG_INF("Model downloaded and loaded successfully to: %s\n", params.model.path.c_str());
+
+            // We only want to download, not keep the model loaded
+            // The download happens during common_init_from_params
+        } else {
+            LOG_ERR("Failed to download or load model\n");
+            return 1;
+        }
+    } catch (const std::exception & e) {
+        LOG_ERR("Error: %s\n", e.what());
+        return 1;
+    }
+
+    return 0;
+}