Merge branch 'npu_support' of github.com:hipudding/llama.cpp into npu_support

Author: hipudding

CMakeLists.txt

@@ -1218,6 +1218,7 @@ add_library(ggml OBJECT
             ggml-backend.h
             ggml-quants.c
             ggml-quants.h
+<<<<<<< HEAD
             ${GGML_SOURCES_CUDA}      ${GGML_HEADERS_CUDA}
             ${GGML_SOURCES_OPENCL}    ${GGML_HEADERS_OPENCL}
             ${GGML_SOURCES_METAL}     ${GGML_HEADERS_METAL}
@@ -1228,6 +1229,17 @@ add_library(ggml OBJECT
             ${GGML_SOURCES_VULKAN}    ${GGML_HEADERS_VULKAN}
             ${GGML_SOURCES_ROCM}      ${GGML_HEADERS_ROCM}
             ${GGML_SOURCES_LLAMAFILE} ${GGML_HEADERS_LLAMAFILE}
+=======
+            ${GGML_SOURCES_CUDA}    ${GGML_HEADERS_CUDA}
+            ${GGML_SOURCES_OPENCL}  ${GGML_HEADERS_OPENCL}
+            ${GGML_SOURCES_METAL}   ${GGML_HEADERS_METAL}
+            ${GGML_SOURCES_MPI}     ${GGML_HEADERS_MPI}
+            ${GGML_SOURCES_EXTRA}   ${GGML_HEADERS_EXTRA}
+            ${GGML_SOURCES_SYCL}    ${GGML_HEADERS_SYCL}
+            ${GGML_SOURCES_KOMPUTE} ${GGML_HEADERS_KOMPUTE}
+            ${GGML_SOURCES_VULKAN}  ${GGML_HEADERS_VULKAN}
+            ${GGML_SOURCES_ROCM}    ${GGML_HEADERS_ROCM}
+>>>>>>> a21434a136a034b6c64bd50b17442e36f3d7e3c8
             ${GGML_SOURCES_CANN}    ${GGML_HEADERS_CANN}
             )
 

examples/llava/clip.cpp

@@ -983,6 +983,11 @@ struct clip_ctx * clip_model_load(const char * fname, const int verbosity = 1) {
     LOG_TEE("%s: CLIP using Metal backend\n", __func__);
 #endif
 
+#ifdef GGML_USE_CANN
+    new_clip->backend = ggml_backend_cann_init(0);
+    printf("%s: CLIP using CANN backend\n", __func__);
+#endif
+
 #ifdef GGML_USE_CANN
     new_clip->backend = ggml_backend_cann_init(0);
     printf("%s: CLIP using CANN backend\n", __func__);

ggml-cann.cpp

@@ -450,6 +450,8 @@ static bool ggml_cann_compute_forward(ggml_backend_cann_context& ctx,
             break;
         case GGML_OP_ROPE:
         case GGML_OP_ALIBI:
+            ggml_cann_alibi(ctx, dst);
+            break;
         case GGML_OP_IM2COL:
             ggml_cann_im2col(ctx, dst);
             break;
@@ -685,8 +687,8 @@ GGML_CALL static bool ggml_backend_cann_supports_op(ggml_backend_t backend,
             return true;
         case GGML_OP_SOFT_MAX:
         case GGML_OP_ROPE:
-        case GGML_OP_ALIBI:
             return false;
+        case GGML_OP_ALIBI:
         case GGML_OP_IM2COL:
             return true;
         case GGML_OP_POOL_2D: