convert : force patch_merger tensors to f16/f32 (#18124)

Author: CISC

convert_hf_to_gguf.py

@@ -1838,7 +1838,7 @@ def _find_param(self, obj: dict[str, Any], keys: Iterable[str], optional: bool =
 
     def tensor_force_quant(self, name, new_name, bid, n_dims):
         del bid, name, n_dims  # unused
-        if ".patch_embd.weight" in new_name:
+        if ".patch_embd.weight" in new_name or ".patch_merger.weight" in new_name:
             return gguf.GGMLQuantizationType.F16 if self.ftype == gguf.LlamaFileType.MOSTLY_F16 else gguf.GGMLQuantizationType.F32
         return False