fix grammar stack empty error for qwen3.5 (#1490)

* fix grammar stack empty error for qwen3.5

* Add to --help

---------

Co-authored-by: firecoperana <firecoperana>

common/chat.cpp

@@ -3171,10 +3171,9 @@ static common_chat_params common_chat_templates_apply_jinja(
         workaround::func_args_not_string(params.messages);
         // Models with <think> support (Step-3.5-Flash, Nemotron 3 Nano) use the
         // Nemotron v3 PEG parser for streaming and schema-aware parameter parsing.
-        // Qwen3-Coder has no <think> in its template.
-        //if (src.find("<think>") != std::string::npos) {
-        //    return common_chat_params_init_qwen3_coder(tmpl, params);
-        //}
+        if (inputs.use_peg) {
+            return common_chat_params_init_qwen3_coder(tmpl, params);
+        }
         return common_chat_params_init_qwen3_coder_xml(tmpl, params);
     }
 

common/chat.h

@@ -157,6 +157,7 @@ struct common_chat_templates_inputs {
     std::map<std::string, std::string> chat_template_kwargs;
     bool add_bos = false;
     bool add_eos = false;
+    bool use_peg = false;
 };
 
 struct common_chat_params {

common/common.cpp

@@ -1959,6 +1959,10 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa
         params.use_jinja = true;
         return true;
     }
+    if (arg == "--peg") {
+        params.use_peg = true;
+        return true;
+    }
     if (arg == "--chat-template-kwargs") {
         CHECK_ARG
         std::string value = argv[i];
@@ -2246,6 +2250,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
     options.push_back({ "*",           "-h,    --help, --usage",        "print usage and exit" });
     options.push_back({ "*",           "       --version",              "show version and build info" });
     options.push_back({ "*",           "-v,    --verbose",              "print verbose information" });
+    options.push_back({ "*",           "       --minilog",              "print important information" });
     options.push_back({ "*",           "       --verbosity N",          "set specific verbosity level (default: %d)", params.verbosity });
     options.push_back({ "*",           "       --verbose-prompt",       "print a verbose prompt before generation (default: %s)", params.verbose_prompt ? "true" : "false" });
     options.push_back({ "*",           "-dr,   --dry-run",       "skip loading tensors in the files"});
@@ -2374,6 +2379,9 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param
                                                                         "if suffix/prefix are specified, template will be disabled\n"
                                                                         "only commonly used templates are accepted:\n"
                                                                         "https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template" });
+    options.push_back({ "main",        "       --peg",
+                                                                    "use peg parser for qwen3.5 models.\n"
+                                                                    "https://github.com/ikawrakow/ik_llama.cpp/pull/1490" });
     options.push_back({ "main",        "       --chat-template JINJA_TEMPLATE",
                                                                         "use jinja template for chat (default: disabled)\n" });
     options.push_back({ "main",        "       --chat-template-file file_with_JINJA_TEMPLATE",

common/common.h

@@ -392,6 +392,7 @@ struct gpt_params {
     std::string public_path   = "";
     std::string chat_template = "";
     bool use_jinja = false;                                                                                 // NOLINT
+    bool use_peg = false;
     std::string system_prompt = "";
     bool enable_chat_template = true;
     common_reasoning_format reasoning_format = COMMON_REASONING_FORMAT_DEEPSEEK;

examples/server/server-common.cpp

@@ -784,6 +784,7 @@ json oaicompat_chat_params_parse(
     inputs.parallel_tool_calls = json_value(body, "parallel_tool_calls", false);
     inputs.add_generation_prompt = json_value(body, "add_generation_prompt", true);
     inputs.reasoning_format = opt.reasoning_format;
+    inputs.use_peg = opt.use_peg;
     if (body.contains("reasoning_format")) {
         inputs.reasoning_format = common_reasoning_format_from_name(body.at("reasoning_format").get<std::string>());
     }

examples/server/server-common.h

@@ -245,6 +245,7 @@ json oaicompat_chat_params_parse(const json& body);
 
 struct server_chat_params {
     bool use_jinja;
+    bool use_peg;
     bool prefill_assistant;
     common_reasoning_format reasoning_format;
     std::map<std::string, std::string> chat_template_kwargs;

examples/server/server-context.cpp

@@ -314,6 +314,7 @@ void server_context::init() {
 
         chat_params = {
             /* use_jinja             */ params_base.use_jinja,
+            /* use_peg               */ params_base.use_peg,
             /* prefill_assistant     */ params_base.prefill_assistant,
             /* reasoning_format      */ params_base.reasoning_format,
             /* chat_template_kwargs  */ params_base.default_template_kwargs,