modify chinese comment

2025-12-24 03:59:23 +01:00 · 2022-06-30 14:51:09 +08:00
parent 99772b1acc
commit d361c1730b
1 changed files with 0 additions and 2 deletions
--- a/modelscope/models/nlp/space/modules/multihead_attention.py
+++ b/modelscope/models/nlp/space/modules/multihead_attention.py
@@ -51,8 +51,6 @@ class MultiheadAttention(nn.Module):
        if mask is not None:
            '''
            mask: [batch size, num_heads, seq_len, seq_len]
-            mask后两维(seq_len, seq_len)矩阵来看，其中有的行可能都是true(1)，对应句子中<pad>位看的行
-            导致softmax后该行的每个位置的attn prob都为1/n而非0，所以此处需重置为0

            >>> F.softmax([-1e10, -100, -100])
            >>> [0.00, 0.50, 0.50]