funasr/models/lcbnet/attention.py
@@ -11,6 +11,7 @@ import torch from torch import nn class MultiHeadedAttentionReturnWeight(nn.Module): """Multi-Head Attention layer. @@ -108,5 +109,3 @@ q, k, v = self.forward_qkv(query, key, value) scores = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.d_k) return self.forward_attention(v, scores, mask)