temp update

Author: samuraieng

convert_hf_to_gguf.py

@@ -4217,12 +4217,13 @@ def set_gguf_parameters(self):
         super().set_gguf_parameters()
         assert self.hparams_vision is not None
         hparams = self.hparams_vision
-        if "sarashina2_vision" in self.global_config['model_type']:
-            self.global_config['model_type'] = "qwen2_vl"
         model_type = self.global_config['model_type']
         if model_type == 'sarashina2_vision':
-            model_type = "qwen2_vl"
-        if model_type == 'qwen2_vl':
+            model_type = 'qwen2_vl'
+            self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.QWEN2VL)
+            spatial_merge_size = self.hparams.get("spatial_merge_size", 2)
+            self.gguf_writer.add_uint32("clip.vision.spatial_merge_size", spatial_merge_size)
+        elif model_type == 'qwen2_vl':
             self.gguf_writer.add_clip_projector_type(gguf.VisionProjectorType.QWEN2VL)
         elif model_type == 'qwen2_5_vl' or model_type == 'qwen2_5_omni':
             if model_type == 'qwen2_5_omni':

tools/mtmd/mtmd.cpp

@@ -1250,9 +1250,9 @@ mtmd_decoder_pos mtmd_image_tokens_get_decoder_pos(const mtmd_image_tokens * ima
     mtmd_decoder_pos pos;
     // M-RoPE logic
     // TODO: support other types of position encoding if needed
-    pos.t = pos_0;
-    pos.x = pos_0 + (i % image_tokens->nx);
-    pos.y = pos_0 + (i / image_tokens->nx);
+    pos.t = 0;
+    pos.x = i % image_tokens->nx;
+    pos.y = i / image_tokens->nx;
     pos.z = 0; // unused for now
     return pos;
 }