正文共5225个字，预计阅读时间12分钟。

data = [ ("me gusta comer en la cafeteria".split(), "SPANISH"),         ("Give it to me".split(), "ENGLISH"),         ("No creo que sea una buena idea".split(), "SPANISH"),         ("No it is not a good idea to get lost at sea".split(), "ENGLISH") ] test_data = [("Yo creo que si".split(), "SPANISH"),              ("it is lost on me".split(), "ENGLISH")]

word_to_ix = {}for sent, _ in data + test_data:    f

or word in sent:

if word not in word_to_ix: word_to_ix[word] = len(word_to_ix) print(word_to_ix)

{'me': 0, 'gusta': 1, 'comer': 2, 'en': 3, 'la': 4, 'cafeteria': 5, 'Give': 6, 'it': 7, 'to': 8, 'No': 9, 'creo': 10, 'que': 11, 'sea': 12, 'una': 13, 'buena': 14, 'idea': 15, 'is': 16, 'not': 17, 'a': 18, 'good': 19, 'get': 20, 'lost': 21, 'at': 22, 'Yo': 23, 'si': 24, 'on': 25}

VOCAB_SIZE = len(word_to_ix) NUM_LABELS = 2#只有两类 ENGLISH  SPANISH

class BoWClassifier(nn.Module):

#nn.Module 这是继承torch的神经网络模板

def __init__(self, num_labels, vocab_size):

super(BoWClassifier, self).__init__()

self.linear = nn.Linear(vocab_size, num_labels)

def forward(self, bow_vec):

return F.log_softmax(self.linear(bow_vec))

def make_bow_vector(sentence, word_to_ix)

def make_bow_vector(sentence, word_to_ix):    vec = torch.zeros(len(word_to_ix))    for word in sentence:        vec[word_to_ix[word]] += 1    return vec.view(1, -1)

def make_target(label, label_to_ix):    return torch.LongTensor([label_to_ix[label]])

model = BoWClassifier(NUM_LABELS, VOCAB_SIZE)

for param in model.parameters():

print("param:", param)

param: Parameter containing:Columns 0 to 9 0.0786  0.1596  0.1259  0.0054  0.0558 -0.0911 -0.1804 -0.1526 -0.0287 -0.1086-0.0651 -0.1096 -0.1807 -0.1907 -0.0727 -0.0179  0.1530 -0.0910  0.1943 -0.1148Columns 10 to 19 0.0452 -0.0786  0.1776  0.0425  0.1194 -0.1330 -0.1877 -0.0412 -0.0269 -0.1572-0.0361  0.1909  0.1558  0.1309  0.1461 -0.0822  0.1078 -0.1354 -0.1877  0.0184Columns 20 to 25 0.1818 -0.1401  0.1118  0.1002  0.1438  0.0790 0.1812 -0.1414 -0.1876  0.1569  0.0804 -0.1897

[torch.FloatTensor of size 2x26]param: Parameter containing: 0.1859 0.1245[torch.FloatTensor of size 2]

sample = data[0] bow_vector = make_bow_vector(sample[0], word_to_ix) log_probs = model(autograd.Variable(bow_vector))

print("log_probs", log_probs)

log_probs Variable containing:-0.6160 -0.7768[torch.FloatTensor of size 1x2]

label_to_ix = { "SPANISH": 0, "ENGLISH": 1 }

for instance, label in test_data:

bow_vec = autograd.Variable(make_bow_vector(instance, word_to_ix))

log_probs = model(bow_vec)

print log_probsprint next(model.parameters())[:,word_to_ix["creo"]]

Variable containing:

-0.5431 -0.8698

[torch.FloatTensor of size 1x2]

Variable containing:

-0.7405 -0.6480

[torch.FloatTensor of size 1x2]

Variable containing:

-0.0467 0.1065

[torch.FloatTensor of size 2]

loss_function = nn.NLLLoss() optimizer = optim.SGD(model.parameters(), lr=0.1)for epoch in range(100):    for instance, label in data:        model.zero_grad()        bow_vec = autograd.Variable(make_bow_vector(instance, word_to_ix))        target = autograd.Variable(make_target(label, label_to_ix))        log_probs = model(bow_vec)        loss = loss_function(log_probs, target)        loss.backward()        optimizer.step()

for instance, label in test_data:

bow_vec = autograd.Variable(make_bow_vector(instance, word_to_ix))

log_probs = model(bow_vec)

print log_probs

Variable containing:-0.0842 -2.5161[torch.FloatTensor of size 1x2]Variable containing:-2.4886 -0.0867[torch.FloatTensor of size 1x2]