mtmd: fix "v.patch_embd" quant and unsupported im2col ops on Metal for deepseek-ocr...

* mtmd: fix "v.patch_embd" quant and unsupported im2col ops on Metal for deepseek-ocr

* Update src/llama-quant.cpp

Co-authored-by: Sigbjørn Skjæret <redacted>
---------

Co-authored-by: Sigbjørn Skjæret <redacted>

diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
index 82d1004c65cc03f613dc7f0c5c532ef3a90026ae..bcf98cfae76f48a2035d9e9af7acc8cb65ac273c 100755 (executable)
--- a/convert_hf_to_gguf.py
+++ b/convert_hf_to_gguf.py
@@ -7150,6 +7150,8 @@ class DeepseekOCRVisionModel(MmprojModel):
             return gguf.GGMLQuantizationType.F32
         if ".rel_pos_h" in name or '.rel_pos_w' in name:
             return gguf.GGMLQuantizationType.F32
+        if ".neck." in name or ".net_" in name:
+            return gguf.GGMLQuantizationType.F32
         return super().tensor_force_quant(name, new_name, bid, n_dims)
 
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:

diff --git a/src/llama-quant.cpp b/src/llama-quant.cpp
index a1f87b3b356e65bd6eacfa3d444194ef3e9112ff..3c8b32be08487215f00979679a01b9412eba6fe8 100644 (file)
--- a/src/llama-quant.cpp
+++ b/src/llama-quant.cpp
@@ -345,9 +345,12 @@ static bool tensor_allows_quantization(const llama_model_quantize_params * param
 
     // do not quantize specific multimodal tensors
     quantize &= name.find(".position_embd") == std::string::npos;
-    quantize &= name.find("sam.patch_embd") == std::string::npos;
     quantize &= name.find("sam.pos_embd")   == std::string::npos;
+    quantize &= name.find("sam.neck.")      == std::string::npos;
+    quantize &= name.find("sam.net_")       == std::string::npos;
     quantize &= name.find(".rel_pos")       == std::string::npos;
+    quantize &= name.find(".patch_embd")    == std::string::npos;
+    quantize &= name.find(".patch_merger")  == std::string::npos;
 
     return quantize;
 }