-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathEmbedder.py
221 lines (202 loc) · 9.76 KB
/
Embedder.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
import re
import torch
from torch import nn
import os
from ELMoForManyLangs import elmo
import numpy as np
class Embedder(nn.Module):
def __init__(self, seq_len=0, use_cuda=True, run_device=None, target_device=None ,d_model=1024):
super(Embedder, self).__init__()
self.embedder = elmo.Embedder(model_dir="new.model", batch_size=512, use_cuda=use_cuda)
self.seq_len = seq_len
self.device = run_device
self.target_device = target_device
if self.device != None:
self.embedder.model.to(self.device)
self.bos_vec, self.eos_vec, self.pad, self.oov = self.embedder.sents2elmo([["<bos>","<eos>","<pad>","<oov>"]], output_layer=0)[0]
def __call__(self, sents, max_len=0, with_bos_eos=True, layer=-1, pad_matters=False):
seq_lens = np.array([len(x) for x in sents], dtype=np.int64)
sents = [[self.sub_unk(x) for x in sent] for sent in sents]
if max_len != 0:
pass
elif self.seq_len != 0:
max_len = self.seq_len
else:
max_len = seq_lens.max()
emb_list = self.embedder.sents2elmo(sents, output_layer=layer)
if not with_bos_eos:
for i in range(len(emb_list)):
if max_len - seq_lens[i] > 0:
if pad_matters:
emb_list[i] = np.concatenate([emb_list[i], np.tile(self.pad,[max_len - seq_lens[i],1])], axis=0)
else:
emb_list[i] = np.concatenate([emb_list[i], np.zeros((max_len - seq_lens[i], emb_list[i].shape[1]))])
else:
emb_list[i] = emb_list[i][:max_len]
elif with_bos_eos:
for i in range(len(emb_list)):
if max_len - seq_lens[i] > 0:
if pad_matters:
emb_list[i] = np.concatenate([
self.bos_vec[np.newaxis],
emb_list[i],
self.eos_vec[np.newaxis],
np.tile(self.pad, [max_len - seq_lens[i], 1])], axis=0)
else:
emb_list[i] = np.concatenate([
self.bos_vec[np.newaxis],
emb_list[i],
self.eos_vec[np.newaxis],
np.zeros((max_len - seq_lens[i], emb_list[i].shape[1]))], axis=0)
else:
emb_list[i] = np.concatenate([self.bos_vec[np.newaxis], emb_list[i][:max_len],self.eos_vec[np.newaxis]], axis=0)
embedded = np.array(emb_list, dtype=np.float32)
seq_lens = seq_lens+2 if with_bos_eos else seq_lens
return embedded, seq_lens
def forward(self, sents, max_len=0, with_bos_eos=True, layer=-1, pad_matters=False):
return torch.from_numpy(self.__call__(sents, max_len=0, with_bos_eos=True, layer=-1, pad_matters=False)[0]).to(self.target_device)
def sub_unk(self, e):
e = e.replace('，',',')
e = e.replace('：',':')
e = e.replace('；',';')
e = e.replace('？','?')
e = e.replace('！', '!')
return e
class oov_handler():
def __init__(self):
self.ch_gex = re.compile(r'[\u4e00-\u9fff]+')
self.num_gex = re.compile(r'[0-9]+')
self.eng_gex = re.compile(r'[a-zA-Z]+')
self.sym_list = list(np.load(os.path.join(os.path.dirname(__file__),"sym_list.npy")))
def __call__(self, word):
if self.ch_gex.findall(word) != []:
return "<oov>"
if self.eng_gex.findall(word) != []:
return "<eng>"
if self.num_gex.findall(word) != []:
return "<num>"
if word in self.sym_list:
return "<sym>"
else:
return "<unk>"
class invELMo(nn.Module):
def __init__(self,
elmo=None,
batch_size=32,
input_size=1024,
hidden_size=300,
h_size=500,
n_layers=3,
dropout=0.33):
super(invELMo, self).__init__()
self.batch_size = batch_size
self.vocab_lim = 100000
self.n_layers = n_layers
self.hidden_size = hidden_size
# self.load_elmo()
if elmo == None:
print("ELMo model not provided. You can't use this model to train, but you can test.")
self.elmo = elmo
self.total_line = 50563844
self.gru = nn.LSTM(input_size, hidden_size, n_layers,
dropout=(0 if n_layers == 1 else dropout),
bidirectional=True,
batch_first=True)
self.fc1 = nn.Linear(2*hidden_size, self.vocab_lim)
self.criterion = nn.CrossEntropyLoss(ignore_index=0)
self.optimizer = torch.optim.Adam(self.parameters())
self.device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
self.load_corpus_dict(self.vocab_lim)
self.handle_oov = oov_handler()
self.bos_vec, self.eos_vec = np.load(os.path.join(os.path.dirname(__file__),"bos_eos.npy"))
def process(self, sent):
return [x if x in self.word2idx else self.handle_oov(x) for x in sent]
def load_elmo(self):
print("loading ELMo model ...")
self.elmo = Embedder()
print("ELMo model loaded!")
def load_corpus_dict(self, limit):
self.idx2word = ["<pad>"] + list(np.load(os.path.join(os.path.dirname(__file__),"idx2word_new.npy")))[:limit-1]
self.word2idx = dict([(word, i) for i, word in enumerate(self.idx2word)])
def corpus_generator(self, filename="shuff_corpus.txt"):
f = open(os.path.join(os.path.dirname(__file__),filename))
batch_list = []
for i in f:
batch_list.append(self.process(["<bos>"]+sub_unk(i.strip()).split(' ')+["<eos>"]))
if len(batch_list) == self.batch_size:
yield batch_list
batch_list = []
def padded_corpus_generator(self, filename="shuff_corpus.txt", max_len=25):
f = open(os.path.join(os.path.dirname(__file__),filename))
batch_list = []
for i in f:
org_sent = sub_unk(i.strip()).split(' ')
if len(org_sent) > max_len - 2:
org_sent = org_sent[:max_len - 2]
batch_list.append(self.process(["<bos>"] + org_sent + ["<eos>"] + ["<pad>" for _ in range(max_len - 2 - len(org_sent))]))
if len(batch_list) == self.batch_size:
yield batch_list
batch_list = []
def sent2idx(self, sents, max_len = 0):
if max_len==0:
for i in sents:
if len(i) > max_len:
max_len = len(i)
sents_lens = []
sent_mat = np.zeros((len(sents), max_len), dtype=np.int64)
for i in range(len(sents)):
sents_i_len = len(sents[i])
sents_lens.append(sents_i_len)
for j in range(max_len):
if j < sents_i_len:
sent_mat[i][j] = self.word2idx[sents[i][j]]
return [sent_mat, np.array(sents_lens)]
def forward(self, input_seq, input_lengths, hidden=None):
embedded = torch.from_numpy(input_seq).to(self.device)
packed = torch.nn.utils.rnn.pack_padded_sequence(embedded, input_lengths, batch_first=True)
outputs, hidden = self.gru(packed, hidden) # output: (seq_len, batch, hidden*n_dir)
outputs, _ = torch.nn.utils.rnn.pad_packed_sequence(outputs, batch_first=True)
pred_prob = self.fc1(outputs)#nn.Softmax(dim=-1)(self.fc1(outputs))
embedded.cpu()
return pred_prob
def train_model(self, num_epochs=1, step_num=100000, step_to_save_model=1000, filename="shuff_corpus.txt"):
self.to(self.device)
for epoch in range(num_epochs):
ct = Clock(step_num)
His_loss = History(title="Loss", xlabel="step", ylabel="loss",
item_name=["train_loss"])
His_ppl = History(title="Perplexity", xlabel="step", ylabel="loss",
item_name=["train_ppl"])
for step, batch_x in enumerate(self.padded_corpus_generator(filename=filename)):
batch_y, x_lens = self.sent2idx(batch_x)
elmo_x, x_lens = self.elmo(batch_x)
(elmo_x, x_lens, batch_y), _ind = sort_numpy([elmo_x, x_lens, batch_y], piv=1)
target = torch.from_numpy(batch_y).cuda() if self.device!="cpu" else torch.from_numpy(batch_y)
pred = self.forward(elmo_x, x_lens)
loss = self.criterion(pred.transpose(1, 2), target)
self.optimizer.zero_grad()
loss.backward()
self.optimizer.step()
ppl = math.exp(loss.cpu().item())
info_dict = {"loss": loss, "ppl": ppl}
ct.flush(info=info_dict)
His_loss.append_history(0, (step, loss))
His_ppl.append_history(0, (step, ppl))
target.cpu()
if step == step_num:
break
if (step + 1) % step_to_save_model == 0:
torch.save(self.state_dict(), os.path.join(os.path.dirname(__file__),'model.ckpt'))
His_loss.plot("loss_plot")
His_ppl.plot("pll_plot")
test_corpus(self, "small_cou.txt")
torch.save(self.state_dict(), os.path.join(os.path.dirname(__file__),'model.ckpt'))
His_loss.plot()
His_ppl.plot()
def load_model(self, filename='model.ckpt', device="cuda:0"):
self.load_state_dict(torch.load(os.path.join(os.path.dirname(__file__),filename), map_location=device))
print("model.ckpt load!")
def test(self, input_seq, input_lengths, hidden=None):
pred_prob = self.forward(input_seq, input_lengths)
pred_idx = pred_prob.argmax(2)
return [[self.idx2word[x] for x in r] for r in pred_idx]