From a7ab4708320102268fca8f0aa8960a85d56db14b Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Sigbj=C3=B8rn=20Skj=C3=A6ret?= Date: Wed, 17 Dec 2025 05:36:01 +0100 Subject: [PATCH] force patch_merger tensors to f16/f32 --- convert_hf_to_gguf.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py index bd16ba312f..806b3d7b47 100755 --- a/convert_hf_to_gguf.py +++ b/convert_hf_to_gguf.py @@ -1838,7 +1838,7 @@ class MmprojModel(ModelBase): 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