如何用PyTorch实现递归神经网络？

of stacks, one for each example in

# the batch, that we can pop from independently. The words in

# each example have already been reversed, so that they can

# be read from left to right by popping from the end of each

# list; they have also been prefixed with a null value.

buffers = [list(torch.split(b.squeeze(1), 1, 0))

for b in torch.split(buffers, 1, 1)]

# we also need two null values at the bottom of each stack,

# so we can copy from the nulls in the input; these nulls

# are all needed so that the tracker can run even if the

# buffer or stack is empty

stacks = [[buf[0], buf[0]] for buf in buffers]

if hasattr(self, 'tracker'):

self.tracker.reset_state()

for trans_batch in transitions:

if hasattr(self, 'tracker'):

# I described the Tracker earlier as taking 4

# arguments (context_t, b, s1, s2), but here I

# provide the stack contents as a single argument

# while storing the context inside the Tracker

　　推荐阅读

　　未来的人们，将会如何看待我们现在的机器人？

【义务编辑：wangxueyan TEL：（010）68476606】 >>>详细阅读

地址：http://www.17bianji.com/lsqh/34965.html