Feature/add -ms option for modelscope

common/CMakeLists.txt

@@ -77,6 +77,8 @@ add_library(${TARGET}
     fit.h
     hf-cache.cpp
     hf-cache.h
+    ms.cpp
+    ms.h
     http.h
     json-partial.cpp
     json-partial.h

common/arg.cpp

@@ -4,6 +4,7 @@
 #include "chat.h"
 #include "common.h"
 #include "download.h"
+#include "ms.h"
 #include "json-schema-to-grammar.h"
 #include "log.h"
 #include "sampling.h"
@@ -377,6 +378,27 @@ static handle_model_result common_params_handle_model(struct common_params_model
             result.found_mtp = true;
             result.mtp.path  = download_result.mtp_path;
         }
+    } else if (!model.ms_repo.empty()) {
+        // Handle ModelScope repository
+        // Split the repo ID to extract clean repo and quantization tag
+        auto [ms_repo, ms_tag] = common_download_split_repo_tag(model.ms_repo);
+
+        // Use the complete original value for model name
+        model.name = model.ms_repo;
+
+        auto download_result = ms::download_model(ms_repo, model.hf_file, offline, ms_tag, bearer_token);
+
+        if (download_result.model_path.empty()) {
+            throw std::runtime_error("failed to download model from ModelScope");
+        }
+
+        model.path = download_result.model_path;
+
+        // Set mmproj path if available
+        if (!download_result.mmproj_path.empty()) {
+            result.found_mmproj = true;
+            result.mmproj.path = download_result.mmproj_path;
+        }
     } else if (!model.url.empty()) {
         if (model.path.empty()) {
             auto f = string_split<std::string>(model.url, '#').front();
@@ -515,6 +537,14 @@ static bool common_params_parse_ex(int argc, char ** argv, common_params_context
         }
     }
 
+    // Handle MS_TOKEN environment variable for ModelScope authentication
+    if (params.hf_token.empty()) {
+        const char * ms_token = std::getenv("MS_TOKEN");
+        if (ms_token) {
+            params.hf_token = ms_token;
+        }
+    }
+
     // handle command line arguments
     auto check_arg = [&](int i) {
         if (i+1 >= argc) {
@@ -2660,6 +2690,16 @@ common_params_context common_params_parser_init(common_params & params, llama_ex
             params.hf_token = value;
         }
     ).set_env("HF_TOKEN"));
+
+    add_opt(common_arg(
+        {"-ms", "-msr", "--ms-repo"}, "<user>/<model>[:quant]",
+        "ModelScope model repository; quant is optional, case-insensitive, default to Q4_K_M, or falls back to the first file in the repo if Q4_K_M doesn't exist.\n"
+        "example: Qwen/Qwen3-0.6B-GGUF:Q8_0\n"
+        "(default: unused)",
+        [](common_params & params, const std::string & value) {
+            params.model.ms_repo = value;
+        }
+    ).set_env("LLAMA_ARG_MS_REPO"));
     add_opt(common_arg(
         {"--context-file"}, "FNAME",
         "file to load context from (use comma-separated values to specify multiple files)",

common/common.h

@@ -293,6 +293,7 @@ struct common_params_model {
     std::string url         = ""; // model url to download                                  // NOLINT
     std::string hf_repo     = ""; // HF repo                                                // NOLINT
     std::string hf_file     = ""; // HF file                                                // NOLINT
+    std::string ms_repo     = ""; // ModelScope repo                                        // NOLINT
     std::string docker_repo = ""; // Docker repo                                            // NOLINT
     std::string name        = ""; // in format <user>/<model>[:<tag>] (tag is optional)     // NOLINT
 };

common/download.cpp

@@ -499,14 +499,7 @@ int common_download_file_single(const std::string & url,
     return 304; // Not Modified - fake cached response
 }
 
-struct gguf_split_info {
-    std::string prefix; // tag included
-    std::string tag;
-    int index;
-    int count;
-};
-
-static gguf_split_info get_gguf_split_info(const std::string & path) {
+gguf_split_info get_gguf_split_info(const std::string & path) {
     static const std::regex re_split("^(.+)-([0-9]{5})-of-([0-9]{5})$", std::regex::icase);
     static const std::regex re_tag("[-.]([A-Z0-9_]+)$", std::regex::icase);
     std::smatch m;
@@ -619,7 +612,7 @@ static hf_cache::hf_file find_best_mtp(const hf_cache::hf_files & files,
     return find_best_sibling(files, model, "mtp-");
 }
 
-static bool gguf_filename_is_model(const std::string & filepath) {
+bool gguf_filename_is_model(const std::string & filepath) {
     if (!string_ends_with(filepath, ".gguf")) {
         return false;
     }

common/download.h

@@ -108,3 +108,15 @@ int common_download_file_single(const std::string & url,
 // resolve and download model from Docker registry
 // return local path to downloaded model file
 std::string common_docker_resolve_model(const std::string & docker);
+
+// shared by HF (download.cpp) and ModelScope (ms.cpp)
+bool gguf_filename_is_model(const std::string & filepath);
+
+struct gguf_split_info {
+    std::string prefix; // tag included
+    std::string tag;
+    int index;
+    int count;
+};
+
+gguf_split_info get_gguf_split_info(const std::string & path);