fix compile

convert_hf_to_gguf.py

@@ -1151,7 +1151,7 @@ def __init__(self, *args, **kwargs):
         self.global_config = copy.deepcopy(self.hparams)
         self.hparams_vision = self.get_vision_config()
         self.hparams_audio = self.get_audio_config()
-        
+
         if self.hparams_vision is None and self.hparams_audio is None:
             raise ValueError("vision_config / audio_config not found in hparams")
 

tools/mtmd/clip.cpp

@@ -2672,7 +2672,7 @@ struct clip_model_loader {
     }
 };
 
-std::pair<clip_ctx *, clip_ctx *> clip_init(const char * fname, struct clip_context_params ctx_params) {
+struct clip_init_result clip_init(const char * fname, struct clip_context_params ctx_params) {
     g_logger_state.verbosity_thold = ctx_params.verbosity;
     clip_ctx * ctx_vision = nullptr;
     clip_ctx * ctx_audio = nullptr;

tools/mtmd/clip.h

@@ -29,8 +29,12 @@ struct clip_context_params {
     enum ggml_log_level verbosity;
 };
 
-// returns pair of <vision, audio> contexts
-std::pair<clip_ctx *, clip_ctx *> clip_init(const char * fname, struct clip_context_params ctx_params);
+struct clip_init_result {
+    struct clip_ctx * ctx_v; // vision context
+    struct clip_ctx * ctx_a; // audio context
+};
+
+struct clip_init_result clip_init(const char * fname, struct clip_context_params ctx_params);
 
 void clip_free(struct clip_ctx * ctx);
 

tools/mtmd/mtmd.cpp

@@ -150,8 +150,8 @@ struct mtmd_context {
         ctx_clip_params.use_gpu   = ctx_params.use_gpu;
         ctx_clip_params.verbosity = ctx_params.verbosity;
         auto res = clip_init(mmproj_fname, ctx_clip_params);
-        ctx_v = res.first;
-        ctx_a = res.second;
+        ctx_v = res.ctx_v;
+        ctx_a = res.ctx_a;
         if (!ctx_v && !ctx_a) {
             throw std::runtime_error(string_format("Failed to load CLIP model from %s\n", mmproj_fname));
         }