Merge pull request #1084 from TingquanGao/dev/fix_matmul

fix: fix @ -> matmul

docs/en/models/Twins.md

@@ -3,9 +3,9 @@
 ## Overview
 The Twins network includes Twins-PCPVT and Twins-SVT, which focuses on the meticulous design of the spatial attention mechanism, resulting in a simple but more effective solution. Since the architecture only involves matrix multiplication, and the current deep learning framework has a high degree of optimization for matrix multiplication, the architecture is very efficient and easy to implement. Moreover, this architecture can achieve excellent performance in a variety of downstream vision tasks such as image classification, target detection, and semantic segmentation. [Paper](https://arxiv.org/abs/2104.13840).
 
-## Accuracy, FLOPS and Parameters
+## Accuracy, FLOPs and Parameters
 
-| Models        | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Params<br>(M) |
+| Models        | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPs<br>(G) | Params<br>(M) |
 |:--:|:--:|:--:|:--:|:--:|:--:|:--:|
 | pcpvt_small   | 0.8082 | 0.9552 | 0.812 | - | 3.7 | 24.1   |
 | pcpvt_base    | 0.8242 | 0.9619 | 0.827 | - | 6.4 | 43.8   |

docs/zh_CN/models/Twins.md

@@ -3,9 +3,9 @@
 ## 概述
 Twins网络包括Twins-PCPVT和Twins-SVT，其重点对空间注意力机制进行了精心设计，得到了简单却更为有效的方案。由于该体系结构仅涉及矩阵乘法，而目前的深度学习框架中对矩阵乘法有较高的优化程度，因此该体系结构十分高效且易于实现。并且，该体系结构在图像分类、目标检测和语义分割等多种下游视觉任务中都能够取得优异的性能。[论文地址](https://arxiv.org/abs/2104.13840)。
 
-## 精度、FLOPS和参数量
+## 精度、FLOPs和参数量
 
-| Models        | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPS<br>(G) | Params<br>(M) |
+| Models        | Top1 | Top5 | Reference<br>top1 | Reference<br>top5 | FLOPs<br>(G) | Params<br>(M) |
 |:--:|:--:|:--:|:--:|:--:|:--:|:--:|
 | pcpvt_small   | 0.8082 | 0.9552 | 0.812 | - | 3.7 | 24.1   |
 | pcpvt_base    | 0.8242 | 0.9619 | 0.827 | - | 6.4 | 43.8   |

ppcls/arch/backbone/model_zoo/gvt.py

@@ -82,11 +82,11 @@ class GroupAttention(nn.Layer):
             B, total_groups, self.ws**2, 3, self.num_heads, C // self.num_heads
         ]).transpose([3, 0, 1, 4, 2, 5])
         q, k, v = qkv[0], qkv[1], qkv[2]
-        attn = (q @ k.transpose([0, 1, 2, 4, 3])) * self.scale
+        attn = paddle.matmul(q, k.transpose([0, 1, 2, 4, 3])) * self.scale
 
         attn = nn.Softmax(axis=-1)(attn)
         attn = self.attn_drop(attn)
-        attn = (attn @ v).transpose([0, 1, 3, 2, 4]).reshape(
+        attn = paddle.matmul(attn, v).transpose([0, 1, 3, 2, 4]).reshape(
             [B, h_group, w_group, self.ws, self.ws, C])
 
         x = attn.transpose([0, 1, 3, 2, 4, 5]).reshape([B, N, C])
@@ -147,11 +147,11 @@ class Attention(nn.Layer):
                     [2, 0, 3, 1, 4])
         k, v = kv[0], kv[1]
 
-        attn = (q @ k.transpose([0, 1, 3, 2])) * self.scale
+        attn = paddle.matmul(q, k.transpose([0, 1, 3, 2])) * self.scale
         attn = nn.Softmax(axis=-1)(attn)
         attn = self.attn_drop(attn)
 
-        x = (attn @ v).transpose([0, 2, 1, 3]).reshape([B, N, C])
+        x = paddle.matmul(attn, v).transpose([0, 2, 1, 3]).reshape([B, N, C])
         x = self.proj(x)
         x = self.proj_drop(x)
         return x