[CANN]Modify the code according to the review comments

Signed-off-by: noemotiovon <[email protected]>

Author: noemotiovon

.devops/llama-cli-cann.Dockerfile

@@ -1,4 +1,4 @@
-ARG ASCEND_VERSION=8.0.rc2.alpha003-910b-openeuler22.03-py3.8
+ARG ASCEND_VERSION=8.1.RC1.alpha001-910b-openeuler22.03-py3.10
 
 FROM ascendai/cann:$ASCEND_VERSION AS build
 

.github/workflows/build.yml

@@ -1771,7 +1771,7 @@ jobs:
     strategy:
       matrix:
         cann:
-          - '8.0.rc3.beta1-910b-openeuler22.03-py3.10'
+          - '8.1.RC1.alpha001-910b-openeuler22.03-py3.10'
         device:
           - 'ascend910b3'
         build:

ggml/src/ggml-cann/aclnn_ops.cpp

@@ -88,6 +88,20 @@ void bcast_shape(ggml_tensor * src0, ggml_tensor * src1, ggml_tensor * dst, aclT
     }
 }
 
+void ggml_cann_unary_op(
+    std::function<void(ggml_backend_cann_context&, aclTensor*, aclTensor*)> unary_op,
+    ggml_backend_cann_context& ctx, ggml_tensor* dst) {
+    ggml_tensor* src = dst->src[0];
+
+    aclTensor* acl_src = ggml_cann_create_tensor(src);
+    aclTensor* acl_dst = ggml_cann_create_tensor(dst);
+
+    unary_op(ctx, acl_src, acl_dst);
+
+    ACL_CHECK(aclDestroyTensor(acl_src));
+    ACL_CHECK(aclDestroyTensor(acl_dst));
+}
+
 /**
  * @brief Repeats elements of a tensor along each dimension according to the
  * specified repeat array.

ggml/src/ggml-cann/aclnn_ops.h

@@ -756,19 +756,9 @@ template <void unary_op(ggml_backend_cann_context&, aclTensor*, aclTensor*)>
  * @param dst The destination tensor where the result will be stored.
  *            The source tensor is retrieved from `dst->src[0]`.
  */
-static void ggml_cann_unary_op_func(
+void ggml_cann_unary_op(
     std::function<void(ggml_backend_cann_context&, aclTensor*, aclTensor*)> unary_op,
-    ggml_backend_cann_context& ctx, ggml_tensor* dst) {
-    ggml_tensor* src = dst->src[0];
-
-    aclTensor* acl_src = ggml_cann_create_tensor(src);
-    aclTensor* acl_dst = ggml_cann_create_tensor(dst);
-
-    unary_op(ctx, acl_src, acl_dst);
-
-    ACL_CHECK(aclDestroyTensor(acl_src));
-    ACL_CHECK(aclDestroyTensor(acl_dst));
-}
+    ggml_backend_cann_context& ctx, ggml_tensor* dst);
 
 /**
  * @brief Helper macro to invoke a unary ACL operation using ggml_cann_unary_op.
@@ -794,7 +784,7 @@ static void ggml_cann_unary_op_func(
             aclTensor* acl_dst) {                                \
             GGML_CANN_CALL_ACLNN_OP(OP_NAME, acl_src, acl_dst);  \
         };                                                       \
-        ggml_cann_unary_op_func(lambda, ctx, dst);               \
+        ggml_cann_unary_op(lambda, ctx, dst);                    \
     }                                                            \
     while (0)
 #endif  // CANN_ACLNN_OPS

ggml/src/ggml-cann/ggml-cann.cpp

@@ -1335,7 +1335,7 @@ static bool ggml_cann_compute_forward(ggml_backend_cann_context& ctx,
                         aclTensor* acl_dst) {
                         GGML_CANN_CALL_ACLNN_OP(GeluV2, acl_src, 0, acl_dst);
                     };
-                    ggml_cann_unary_op_func(lambda, ctx, dst);
+                    ggml_cann_unary_op(lambda, ctx, dst);
                 } break;
                 case GGML_UNARY_OP_TANH:
                     GGML_CANN_CALL_UNARY_OP(Tanh);