Merge branch 'master' into concedo_experimental

# Conflicts:
#	.gitignore
#	CMakeLists.txt
#	Makefile
#	README.md
#	build.zig
#	ggml-opencl.cpp
#	tests/CMakeLists.txt
#	tests/test-double-float.cpp
#	tests/test-sampling.cpp

otherarch/llama_v3.cpp

@@ -3770,8 +3770,10 @@ int llama_v3_apply_lora_from_file_internal(const struct llama_v3_model & model,
                     throw std::runtime_error(format_old(
                         "%s: error: lora failed", __func__));
                 }
+#if defined(GGML_USE_CUBLAS)
                 offload_func = ggml_cuda_assign_buffers;
                 offload_func_force_inplace = ggml_cuda_assign_buffers_force_inplace;
+#endif
             }
 #endif // GGML_USE_CUBLAS