diff --git a/ggml/src/ggml-cann/aclnn_ops.cpp b/ggml/src/ggml-cann/aclnn_ops.cpp
index a98bd49c37e0cf..211de6c8ab4157 100644
--- a/ggml/src/ggml-cann/aclnn_ops.cpp
+++ b/ggml/src/ggml-cann/aclnn_ops.cpp
@@ -109,8 +109,11 @@ void ggml_cann_repeat(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
  * @brief Adds two tensors element-wise and stores the result in a destination
  * tensor.
  *
- * This function performs the operation: dst = acl_src0 + alpha * acl_src1
- * where alpha is a scalar value.
+ * This function performs the operation: 
+ * \f[
+ *    dst = acl\_src0 + alpha \times acl\_src1 
+ * \f]
+ * where alpha is a scalar value and defaults to 1.0f.
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src0 The first source tensor.
@@ -119,7 +122,6 @@ void ggml_cann_repeat(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
  */
 static void aclnn_add(ggml_backend_cann_context& ctx, aclTensor* acl_src0,
                       aclTensor* acl_src1, aclTensor* acl_dst) {
-    // add: dst = acl_src0 + alpha*acl_src1
     aclScalar* alpha = nullptr;
     float alphaValue = 1.0f;
     alpha = aclCreateScalar(&alphaValue, aclDataType::ACL_FLOAT);
@@ -248,8 +250,11 @@ void ggml_cann_concat(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
  * @brief Creates a tensor with values starting from `start`, incremented by
  * `step`, and ending before `stop`.
  *
- * This function performs the operation: [start, stop), out(i+1) = out(i) +
- * step.
+ * This function performs the operation: 
+ * \f[
+ *    \text {out}_{i+1}=\text {out}_i+\text {step} 
+ * \f]
+ * the range is [start, stop).
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_dst The destination tensor where the values will be stored.
@@ -893,7 +898,7 @@ void ggml_cann_dup(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
         ACL_CHECK(aclDestroyTensor(acl_dst));
         return;
     }
-    // TODO: simplefify
+    // TODO: simplify
     if (src->type == GGML_TYPE_F16) {
         if (dst->type == GGML_TYPE_Q8_0) {
             aclrtlaunch_ascendc_quantize_f16_q8_0(
@@ -1396,6 +1401,10 @@ void ggml_cann_im2col(ggml_backend_cann_context& ctx, ggml_tensor* dst) {
  *
  * This function computes the exponential of each element in the source tensor
  * `acl_src` and stores the result back into the same tensor.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_src}_i=e^{acl\_src_i}
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The tensor on which the exponential function will be applied.
@@ -1422,7 +1431,12 @@ static void aclnn_exp(ggml_backend_cann_context& ctx, aclTensor* acl_src) {
  *
  * This function multiplies each element of the source tensor `acl_src` by the
  * scalar `scale` and stores the result in the destination tensor `acl_dst`. If
- * `inplace` is true, the operation is performed in-place on `acl_src`.
+ * `inplace` is true, `acl_dst` will not be used and the operation is performed
+ *  in-place on `acl_src`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}_i=\text {acl_src}_i \times \text {scale}
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The source tensor whose elements will be multiplied.
@@ -1471,6 +1485,10 @@ static void aclnn_muls(ggml_backend_cann_context& ctx, aclTensor* acl_src,
  *
  * This function performs an element-wise multiplication of the tensors
  * `acl_src` and `acl_other` and stores the result in `acl_src`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_src}_i=\text {acl_src}_i \times \text {acl_other}_i
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The source tensor where the multiplication result will be
@@ -1500,6 +1518,10 @@ static void aclnn_inplace_mul(ggml_backend_cann_context& ctx,
  *
  * This function performs element-wise multiplication of the tensors `acl_src`
  * and `acl_other` and stores the result in the destination tensor `acl_dst`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}_i=\text {acl_src}_i \times \text {acl_other}_i
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The first tensor for element-wise multiplication.
@@ -1528,6 +1550,10 @@ static void aclnn_mul(ggml_backend_cann_context& ctx,
  *
  * This function computes the cosine of each element in the source tensor `acl_src`
  * and stores the result in the destination tensor `acl_dst`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}_i=\cos \left(\text {acl_src}_i\right)
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The source tensor on which the cosine function will be applied.
@@ -1554,7 +1580,11 @@ static void aclnn_cos(ggml_backend_cann_context& ctx, aclTensor* acl_src,
  *
  * This function computes the sine of each element in the source tensor `acl_src`
  * and stores the result in the destination tensor `acl_dst`.
- *
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}_i=\sin \left(\text {acl_src}_i\right)
+ * \f]
+
  * @param ctx The context for the CANN backend operations.
  * @param acl_src The source tensor on which the sine function will be applied.
  * @param acl_dst The destination tensor where the sine results will be stored.
@@ -1721,6 +1751,10 @@ static void aclnn_fill_scalar(ggml_backend_cann_context& ctx, float scalar,
  *
  * This function computes the element-wise power of the destination tensor
  * `acl_dst` raised to the power of the exponent tensor `acl_exp`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}_i=acl\_dst_i^{\text {acl_exp}_i}
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_dst The destination tensor, which also serves as the base tensor.
@@ -1764,12 +1798,16 @@ static void aclnn_pow_tensor_tensor(ggml_backend_cann_context& ctx,
  * @param dst          The destination tensor object for additional metadata.
  *
  * The function performs the following steps:
- * 1. Calculates the logarithm floor of the number of heads to determine the base for bias calculation.
- * 2. Initializes arrays with arithmetic sequences and fills them with bias values.
- * 3. Computes the bias tensor based on the calculated biases and arithmetic sequences.
+ * 1. Calculates the logarithm floor of the number of heads to determine the
+      base for bias calculation.
+ * 2. Initializes arrays with arithmetic sequences and fills them with bias
+      values.
+ * 3. Computes the bias tensor based on the calculated biases and arithmetic
+      sequences.
  * 4. Reshapes the bias tensor to match the dimensions of the input tensors.
  * 5. Multiplies the position tensor by the bias tensor.
- * 6. Adds the result of the multiplication to the source tensor to produce the final output.
+ * 6. Adds the result of the multiplication to the source tensor to produce the
+      final output.
  */
 static void aclnn_alibi(ggml_backend_cann_context& ctx, aclTensor* acl_src,
                         aclTensor* acl_position, aclTensor* acl_dst,
@@ -2189,6 +2227,10 @@ static void aclnn_repeat_interleave(ggml_backend_cann_context& ctx,
  * This function computes the matrix multiplication of the input tensor
  * `acl_input` and the weight tensor `acl_weight`, and stores the result in the
  * destination tensor `acl_dst`.
+ * The operation is defined as:
+ * \f[
+ *     \text {acl_dst}=\text {acl_input@acl_weight}
+ * \f]
  *
  * @param ctx The context for the CANN backend operations.
  * @param acl_input The input tensor for the matrix multiplication.