-
Notifications
You must be signed in to change notification settings - Fork 0
/
vocab.py
85 lines (72 loc) · 2.75 KB
/
vocab.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
import torch
class EquivariantVocab(object):
"""
Indexes all n vocabulary words, with a subset of words which will be
equivariant. These equivariant words will be indexed as the first
len(equivariant_words) words in the vocabulary.
"""
def __init__(self, equivariant_words, padding=True):
self.n = 0
self.equivariant_words = equivariant_words
self.word2idx = {}
self.idx2word = {}
self.padding = padding
self.padding_idx = -1
for word in equivariant_words:
self.add_word(word)
def add_word(self, word):
if word in self.word2idx.keys():
return
else:
self.word2idx[word] = self.n
self.idx2word[self.n] = word
self.n += 1
def add_sentence(self, sentence):
for word in sentence.split(" "):
self.add_word(word)
def change_ordering(self):
"""
Ensures equivariant words are in the first p positions.
"""
non_equi_words = [word for word in self.word2idx.keys()
if word not in self.equivariant_words]
for i, word in enumerate(self.equivariant_words):
self.word2idx[word] = i
self.idx2word[i] = word
for j, word in enumerate(non_equi_words):
self.word2idx[word] = j + len(self.equivariant_words)
self.idx2word[j + len(self.equivariant_words)] = word
if self.padding:
self.padding_idx = self.word2idx["<EOS>"]
def __len__(self):
return self.n
def tensor_to_sent(self, ten):
idx_arr = torch.nonzero(ten.squeeze(2))
sent = idx_arr[:,1]
return sent
def idx_to_word(self, idx):
return self.idx2word[idx]
def sent_to_tensor(self, sent):
ten = torch.zeros((len(sent), self.n))
ten[[i for i in range(len(sent))], sent] = 1
return ten.unsqueeze(2)
def word_to_idx(self, word):
return self.word2idx[word]
def batch_tensor_to_sent(self, ten):
idx_arr = torch.nonzero(ten.squeeze(2))
sent = idx_arr[:, 2]
sent = sent.view(ten.shape[0], ten.shape[1])
return sent
def batch_sent_to_tensor(self, sent):
B = sent.shape[0]
N = sent.shape[1]
ten = torch.zeros(B, N, self.n)
ten[[i for i in range(B) for j in range(N)],
[i for j in range(B) for i in range(N)], sent.flatten()] = 1
return ten.unsqueeze(-1)
def words_to_tensor(self, sent):
input_split = sent.split(" ")
idxs = torch.zeros((len(input_split)), dtype=torch.int64)
for i in range(len(input_split)):
idxs[i] = self.word_to_idx(input_split[i])
return self.sent_to_tensor(idxs)