在理解CRF的時候費了一些功夫，將一些難以理解的地方稍微做了下標注，隔三差五看看加強記憶, 代碼是pytorch文檔上的example

import torch
import torch.autograd as autograd
import torch.nn as nn
import torch.optim as optimdef to_scalar(var): #var是Variable,維度是１# returns a python floatreturn var.view(-1).data.tolist()[0]def argmax(vec):# return the argmax as a python int_, idx = torch.max(vec, 1)return to_scalar(idx)def prepare_sequence(seq, to_ix):idxs = [to_ix[w] for w in seq]tensor = torch.LongTensor(idxs)return autograd.Variable(tensor)# Compute log sum exp in a numerically stable way for the forward algorithm
def log_sum_exp(vec): #vec是1*5, type是Variablemax_score = vec[0, argmax(vec)]#max_score維度是１，　max_score.view(1,-1)維度是１＊１，max_score.view(1, -1).expand(1, vec.size()[1])的維度是１＊５max_score_broadcast = max_score.view(1, -1).expand(1, vec.size()[1]) # vec.size()維度是1*5return max_score + torch.log(torch.sum(torch.exp(vec - max_score_broadcast)))#為什么指數之后再求和，而后才log呢class BiLSTM_CRF(nn.Module):def 

總結

以上是生活随笔為你收集整理的pytorch版的bilstm+crf实现sequence label的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。