pytorch attend操作 代码

# seq: [batch,sequence_len,hidden_dim]
# cond: [batch,hidden_dim]
# lens: [batch] 

# context: [batch,hidden_dim]
# scores: [batch,sequence_len] 
def attend(seq, cond, lens):
    """
    attend over the sequences `seq` using the condition `cond`.
    """
    scores = cond.unsqueeze(1).expand_as(seq).mul(seq).sum(2)
    max_len = max(lens)
    for i, l in enumerate(lens):
        if l < max_len:
            scores.data[i, l:] = -np.inf
    scores = F.softmax(scores, dim=1)
    context = scores.unsqueeze(2).expand_as(seq).mul(seq).sum(1)
    return context,

这个attend的输入是一个sequence和一个item的，
输出的​​​context​​​是[batch_size, hidden_dim]
输出的​​​score​​​是[batch_size, sequence_len] 不同于