Dev gzf exp (#1665)

zhifu gao

2024-04-26 1cdb3cc28d4d89a576cc06e5cd8eb80da1f3a3aa

 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)