diff --git a/3rdparty/llama.cpp b/3rdparty/llama.cpp
index 390c30775..1d75155fc 160000
--- a/3rdparty/llama.cpp
+++ b/3rdparty/llama.cpp
@@ -1 +1 @@
-Subproject commit 390c307752ab78fd8189f359d6954c9ba1be74af
+Subproject commit 1d75155fcc8e882c9585f733704153784cddae04
diff --git a/utils/convert-hf-to-gguf-bitnet.py b/utils/convert-hf-to-gguf-bitnet.py
index b11e831b9..dd3a9c7f3 100644
--- a/utils/convert-hf-to-gguf-bitnet.py
+++ b/utils/convert-hf-to-gguf-bitnet.py
@@ -1045,6 +1045,8 @@ def set_gguf_parameters(self):
 
         self.gguf_writer.add_vocab_size(self.hparams["vocab_size"])
 
+        self.gguf_writer.add_hidden_act("relu2")
+
         # rope dimension count (required for correct positional encoding)
         if "head_dim" in self.hparams:
             rope_dim = self.hparams["head_dim"]
diff --git a/utils/convert-ms-to-gguf-bitnet.py b/utils/convert-ms-to-gguf-bitnet.py
index edf702788..755729bb0 100644
--- a/utils/convert-ms-to-gguf-bitnet.py
+++ b/utils/convert-ms-to-gguf-bitnet.py
@@ -1178,6 +1178,8 @@ def add_meta_arch(self, params: Params) -> None:
         else:
             raise ValueError('f_norm_eps is None')
 
+        self.gguf.add_hidden_act("relu2")
+
         if params.f_rope_freq_base is not None:
             self.gguf.add_rope_freq_base(params.f_rope_freq_base)