Talking-Heads Attention
paper：Talking-Heads Attention
在CaiT这篇文章中，作用采用了talking-heads attention，这里做一下解释。
在原始multi-head self-attention中，各个head的计算是独立进行的，多个head的输出最后concat到一起，然后再经过一个线性变换得到最终的输出。
本文提出了在softmax操作的前后引入跨注意力头维度的线性变换，从而使每个self-attention函数依赖于所有的key和query。
下面分别是timm中普通Attention和TalkingHeadAttention的实现
# class Attention
def forward(self, x: torch.Tensor) -> torch.Tensor: # (1,197,192)B, N, C = x.shapeqkv = self.qkv(x).reshape(B, N, 3, self.num_heads, self.head_dim).permute(2, 0, 3, 1, 4)# (1,197,576)->(1,197,3,3,64)->(3,1,3,197,64), (3, batch_size, num_heads, seq_len, head_dim), 3表示qkvq, k, v = qkv.unbind(0) # (1,3,197,64)q, k = self.q_norm(q), self.k_norm(k)if self.fused_attn: # Falsex = F.scaled_dot_product_attention(q, k, v,dropout_p=self.attn_drop.p if self.training else 0.,)else:# attn=softmax(qk)q = q * self.scaleattn = q @ k.transpose(-2, -1) # (1,3,197,197)attn = attn.softmax(dim=-1)attn = self.attn_drop(attn)x = attn @ v # (1,3,197,64)x = x.transpose(1, 2).reshape(B, N, C) # (1,197,3,64)->(1,197,192)x = self.proj(x) # (1,197,192)x = self.proj_drop(x)return x# class TalkingHeadAttn
def forward(self, x):B, N, C = x.shape # (1,196,384)qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) # (1,196,1152)->(1,196,3,8,48)->(3,1,8,196,48)q, k, v = qkv[0] * self.scale, qkv[1], qkv[2] # (1,8,196,48)attn = q @ k.transpose(-2, -1) # (1,8,196,196)attn = self.proj_l(attn.permute(0, 2, 3, 1)).permute(0, 3, 1, 2) # (1,196,196,8)->(1,196,196,8)->(1,8,196,196)attn = attn.softmax(dim=-1) # （1,8,196,196）attn = self.proj_w(attn.permute(0, 2, 3, 1)).permute(0, 3, 1, 2) # (1,196,196,8)->(1,196,8,8)->(1,8,196,196)attn = self.attn_drop(attn)x = (attn @ v).transpose(1, 2).reshape(B, N, C) # (1,8,196,48)->(1,196,8,48)->(1,196,384)x = self.proj(x) # (1,196,384)x = self.proj_drop(x)return x从下图的对比看的更加清楚，左边是普通的attention，右边是talking-heads attention。左边的输入shape为(1, 197, 192)，其中197=196+1是添加了class token，192是特征维度。右边的输入shape为(1, 196, 384)，特征维度为384。左边num_heads=3，右边num_heads=8。因为左边的代码来自vision transformer，右边的代码来自CaiT，选择的具体模型variant不同，所以特征维度和head数量也不一样，但不影响。
可以看到，TalkingHeadAttention在计算softmax前后分别引入了一个线性变换self.proj_l和self.proj_w，定义分别为self.proj_l = nn.Linear(num_heads, num_heads)和self.proj_w = nn.Linear(num_heads, num_heads)。在线性变换前先对输入进行维度变换通过.permute(0, 2, 3 ,1)将num_head维度放到最后，因此线性变换是针对num_head维度的，从而实现跨head的交互，最后再permute回去。
