Pytorch：计算批处理张量的标准
我的张量t带有shape（batch_size x dims）和另一种张量v带有shape（vocab_size x dims）。我想用shape（batch_size x vocab_size）产生张量d，以便d [i，j] = norm（t [i] - v [j]）。
对于单个张量（无批次）进行此操作是微不足道的：d = torch.norm（v -t），因为t将被广播。当张量批处理时，该怎么办？
I have tensor t with shape (Batch_Size x Dims) and another tensor v with shape (Vocab_Size x Dims). I'd like to produce a tensor d with shape (Batch_Size x Vocab_Size), such that d[i,j] = norm(t[i] - v[j]).
Doing this for a single tensor (no batches) is trivial: d = torch.norm(v - t), since t would be broadcast. How can I do this when the tensors have batches?
如果你对这篇内容有疑问，欢迎到本站社区发帖提问 参与讨论，获取更多帮助，或者扫码二维码加入 Web 技术交流群。
绑定邮箱获取回复消息
由于您还没有绑定你的真实邮箱，如果其他用户或者作者回复了您的评论，将不能在第一时间通知您！
发布评论
评论（1）
将单位尺寸插入
v和t以使其（1 x vocab_size x dims）和（batch_size x 1 x dims）分别。接下来，以广播的差异获取形状张量（batch_size x vocab_size x dims）。将其传递给torch.norm以及可选的dim = 2参数，以便沿最后一个维度采取规范。这将导致所需的（batch_size x vocab_size）规范的张量。编辑：正如@konstantinoskokos在评论中指出的那样，由于广播规则 Numpy和Pytorch使用的是领先的统一维度
V不需要明确。即您可以使用Insert unitary dimensions into
vandtto make them (1 x Vocab_Size x Dims) and (Batch_Size x 1 x Dims) respectively. Next, take the broadcasted difference to get a tensor of shape (Batch_Size x Vocab_Size x Dims). Pass that totorch.normalong with the optionaldim=2argument so that the norm is taken along the last dimension. This will result in the desired (Batch_Size x Vocab_Size) tensor of norms.Edit: As pointed out by @KonstantinosKokos in the comments, due to the broadcasting rules used by numpy and pytorch, the leading unitary dimension on
vdoes not need to be explicit. I.e. you can use