clip : enable gpu backend (#4205)

* clip: enable CUDA backend

* add missing kernels

* add enough padding for alignment

* remove ggml_repeat of clip.cpp

* add metal backend

* llava : fixes

- avoid ggml_repeat
- use GGML_USE_ instead of CLIP_USE_ macros
- remove unused vars

---------

Co-authored-by: Georgi Gerganov <[email protected]>

Author: FSSRepo

examples/llava/CMakeLists.txt

@@ -24,7 +24,8 @@ endif()
 
 if (NOT MSVC)
     target_compile_options(llava PRIVATE -Wno-cast-qual) # stb_image.h
-    endif()
+endif()
+
 if(TARGET BUILD_INFO)
     add_dependencies(llava BUILD_INFO)
 endif()