chore: bump llama.cpp

Signed-off-by: thxCode <thxcode0824@gmail.com>

Author: thxCode

llama-box/patches/llama.cpp/dynamic_link.patch

@@ -1,5 +1,5 @@
 diff --git a/ggml/src/ggml-cpu/ggml-cpu.cpp b/ggml/src/ggml-cpu/ggml-cpu.cpp
-index e16cdc9d4..b8d1c41e5 100644
+index 8dacd3671..daf09e224 100644
 --- a/ggml/src/ggml-cpu/ggml-cpu.cpp
 +++ b/ggml/src/ggml-cpu/ggml-cpu.cpp
 @@ -632,6 +632,18 @@ static void * ggml_backend_cpu_get_proc_address(ggml_backend_reg_t reg, const ch

llama-box/patches/llama.cpp/ggml-hip.patch

@@ -1,5 +1,5 @@
 diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh
-index 8f2725547..47624876e 100644
+index 2e5d48797..a79b27431 100644
 --- a/ggml/src/ggml-cuda/common.cuh
 +++ b/ggml/src/ggml-cuda/common.cuh
 @@ -60,6 +60,10 @@
@@ -22,7 +22,7 @@ index 8f2725547..47624876e 100644
  #define GGML_CUDA_CC_IS_CDNA3(cc) (cc >= GGML_CUDA_CC_CDNA3 && cc < GGML_CUDA_CC_RDNA1)
 
  // Moore Threads
-@@ -512,7 +516,7 @@ static __device__ __forceinline__ uint32_t __hgt2_mask(const half2 a, const half
+@@ -520,7 +524,7 @@ static __device__ __forceinline__ uint32_t __hgt2_mask(const half2 a, const half
 
  static __device__ __forceinline__ int ggml_cuda_dp4a(const int a, const int b, int c) {
  #if defined(GGML_USE_HIP)

llama-box/patches/llama.cpp/ggml-rpc.patch

@@ -1,5 +1,5 @@
 diff --git a/ggml/src/ggml-rpc/ggml-rpc.cpp b/ggml/src/ggml-rpc/ggml-rpc.cpp
-index 29bc421d5..5eaa5479d 100644
+index df6ba5407..9610949b6 100644
 --- a/ggml/src/ggml-rpc/ggml-rpc.cpp
 +++ b/ggml/src/ggml-rpc/ggml-rpc.cpp
 @@ -4,6 +4,8 @@

llama-box/patches/llama.cpp/model_py.patch

@@ -1,8 +1,8 @@
 diff --git a/convert_hf_to_gguf.py b/convert_hf_to_gguf.py
-index 82b005e84..02b01a7d2 100755
+index b8c7d97a7..b2145ce0d 100755
 --- a/convert_hf_to_gguf.py
 +++ b/convert_hf_to_gguf.py
-@@ -4706,6 +4706,123 @@ class XLMRobertaModel(BertModel):
+@@ -4811,6 +4811,123 @@ class XLMRobertaModel(BertModel):
          return super().modify_tensors(data_torch, name, bid)
 
 
@@ -168,7 +168,7 @@ index 911eea504..3b731485e 100644
          MODEL_TENSOR.TOKEN_EMBD,
          MODEL_TENSOR.TOKEN_EMBD_NORM,
 diff --git a/gguf-py/gguf/tensor_mapping.py b/gguf-py/gguf/tensor_mapping.py
-index a0f11563a..159f8c8f0 100644
+index dc7c03b46..6f2d740ec 100644
 --- a/gguf-py/gguf/tensor_mapping.py
 +++ b/gguf-py/gguf/tensor_mapping.py
 @@ -167,6 +167,7 @@ class TensorNameMap:

llama-box/patches/llama.cpp/mrope.patch

@@ -52,10 +52,10 @@ index cb8eff4a7..845c3994a 100644
                  GGML_ASSERT( dst->type == GGML_TYPE_F16 || dst->type == GGML_TYPE_F32);
 
 diff --git a/ggml/src/ggml-opencl/ggml-opencl.cpp b/ggml/src/ggml-opencl/ggml-opencl.cpp
-index bb8b310b9..f777e7d40 100644
+index 8ba1e00df..d6267161c 100644
 --- a/ggml/src/ggml-opencl/ggml-opencl.cpp
 +++ b/ggml/src/ggml-opencl/ggml-opencl.cpp
-@@ -6684,10 +6684,6 @@ static void ggml_cl_rope(ggml_backend_t backend, const ggml_tensor * src0, const
+@@ -6783,10 +6783,6 @@ static void ggml_cl_rope(ggml_backend_t backend, const ggml_tensor * src0, const
      const bool is_mrope = mode & GGML_ROPE_TYPE_MROPE;
      const bool is_vision = mode == GGML_ROPE_TYPE_VISION;
 

llama-box/patches/llama.cpp/tool_calling.patch

@@ -1,5 +1,5 @@
 diff --git a/common/chat.cpp b/common/chat.cpp
-index 60805ab3b..ecdd0bbb7 100644
+index 316bd2417..bb97ac3a6 100644
 --- a/common/chat.cpp
 +++ b/common/chat.cpp
 @@ -502,6 +502,13 @@ void common_chat_templates_free(struct common_chat_templates * tmpls) {
@@ -16,7 +16,7 @@ index 60805ab3b..ecdd0bbb7 100644
  bool common_chat_templates_was_explicit(const struct common_chat_templates * tmpls) {
      return tmpls->has_explicit_template;
  }
-@@ -1857,6 +1864,83 @@ static common_chat_params common_chat_templates_apply_jinja(
+@@ -1982,6 +1989,83 @@ static common_chat_params common_chat_templates_apply_jinja(
      return common_chat_params_init_generic(tmpl, params);
  }
 
@@ -100,7 +100,7 @@ index 60805ab3b..ecdd0bbb7 100644
  // Legacy template route (adhoc C++ implementation of known templates), forward to llama_chat_apply_template.
  static common_chat_params common_chat_templates_apply_legacy(
      const struct common_chat_templates * tmpls,
-@@ -1915,6 +1999,17 @@ static common_chat_params common_chat_templates_apply_legacy(
+@@ -2041,6 +2125,17 @@ static common_chat_params common_chat_templates_apply_legacy(
      return params;
  }
 
@@ -118,7 +118,7 @@ index 60805ab3b..ecdd0bbb7 100644
  common_chat_params common_chat_templates_apply(
      const struct common_chat_templates * tmpls,
      const struct common_chat_templates_inputs & inputs)
-@@ -1980,7 +2075,7 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
+@@ -2109,7 +2204,7 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
      try {
          common_chat_parse(builder);
      } catch (const common_chat_msg_partial_exception & ex) {
@@ -127,7 +127,7 @@ index 60805ab3b..ecdd0bbb7 100644
          if (!is_partial) {
              builder.clear_tools();
              builder.move_to(0);
-@@ -1988,8 +2083,8 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
+@@ -2117,8 +2212,8 @@ common_chat_msg common_chat_parse(const std::string & input, bool is_partial, co
          }
      }
      auto msg = builder.result();
@@ -140,10 +140,10 @@ index 60805ab3b..ecdd0bbb7 100644
      return msg;
  }
 diff --git a/common/chat.h b/common/chat.h
-index b014f9f0a..f709ef596 100644
+index eb628d8bc..a80089c1b 100644
 --- a/common/chat.h
 +++ b/common/chat.h
-@@ -167,10 +167,15 @@ common_chat_templates_ptr common_chat_templates_init(
+@@ -168,10 +168,15 @@ common_chat_templates_ptr common_chat_templates_init(
                                             const std::string & bos_token_override = "",
                                             const std::string & eos_token_override = "");