-
Notifications
You must be signed in to change notification settings - Fork 0
/
load_data.py
151 lines (130 loc) · 5.62 KB
/
load_data.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
import numpy as np
from yaml_parser import Input, Output
def compute_dimensionality(src):
if isinstance(src, Input):
if src.type == "vocabulary":
j = 0
with open(src.vocab_file_path, encoding="utf-8") as f:
for line in f:
j += 1
j += len(src.additional_labels)
return j
elif src.type == "embeddings":
with open(src.embeds_file_path, encoding="utf-8") as f:
line = next(f)
line = line.strip().split(' ')
return len(line) - 1
elif isinstance(src, Output):
j = 0
with open(src.vocab_file_path, encoding="utf-8") as f:
for line in f:
j += 1
return j
else:
raise Exception("Can only compute dimensionality of inputs and outputs")
def create_lookup_matrix(_input):
if _input.type == "vocabulary":
num_lines = 0
with open(_input.vocab_file_path, encoding="utf-8") as f:
for line in f:
num_lines += 1
return np.identity(num_lines + len(_input.additional_labels)) # One-hot encoding
elif _input.type == "embeddings":
embeds = list()
with open(_input.embeds_file_path, encoding="utf-8") as f:
for line in f:
line = line.split(' ')
vals = list(map(float, line[1:]))
embeds.append(vals)
for label in _input.additional_labels:
vals = np.random.normal(scale=0.5, size=(len(embeds[0]),))
embeds.append(vals)
return np.array(embeds) # Embedding matrix
else:
raise Exception("Unknown input type '{}'".format(_input.type))
def load_mappings(model):
input_labels_to_ix = list()
input_ix_to_labels = list()
output_labels_to_ix = list()
output_ix_to_labels = list()
for i in range(len(model.inputs)):
input_labels_to_ix.append(dict())
input_ix_to_labels.append(dict())
j = 0
if model.inputs[i].type == "vocabulary":
with open(model.inputs[i].vocab_file_path, encoding="utf-8") as f:
for line in f:
line = line.strip()
input_labels_to_ix[i][line] = j
input_ix_to_labels[i][j] = line
j += 1
elif model.inputs[i].type == "embeddings":
with open(model.inputs[i].embeds_file_path, encoding="utf-8") as f:
for line in f:
line = line.split(' ')
label = line[0]
input_labels_to_ix[i][label] = j
input_ix_to_labels[i][j] = label
j += 1
else:
assert False # Handle other input types here
for label in model.inputs[i].additional_labels:
input_labels_to_ix[i][label] = j
input_ix_to_labels[i][j] = label
j += 1
for i in range(len(model.outputs)):
output_labels_to_ix.append(dict())
output_ix_to_labels.append(dict())
j = 0
with open(model.outputs[i].vocab_file_path, encoding="utf-8") as f:
for line in f:
line = line.strip()
output_labels_to_ix[i][line] = j
output_ix_to_labels[i][j] = line
j += 1
return input_labels_to_ix, input_ix_to_labels, output_labels_to_ix, output_ix_to_labels
class SequenceBatchContainer():
def __init__(self, input_batches, output_batches):
self.input_batches = input_batches
self.output_batches = output_batches
for inp_batch in self.input_batches:
assert len(inp_batch[0]) == len(self.input_batches[0][0])
for outp_batch in self.output_batches:
assert len(outp_batch[0]) == len(self.input_batches[0][0])
def load_sequences(data_source, mappings):
'''Load batches of sequences from the specified source.
Labels are converted to numbers according to given mappings.'''
input_labels_to_ix, _, output_labels_to_ix, _ = mappings
sequences = list()
input_files = list()
output_files = list()
for inp in data_source.inputs:
input_files.append(open(inp, encoding="utf-8"))
for outp in data_source.outputs:
output_files.append(open(outp, encoding="utf-8"))
for global_line in input_files[0]: # Iterate over all files simultaneously
input_lines = list()
output_lines = list()
global_line = global_line.strip().split('\t')
global_line = labels_as_numbers(global_line, input_labels_to_ix[0])
input_lines.append([global_line]) # TODO: Batching/padding (batch size hardcoded to 1 for now)
for i in range(1, len(input_files)):
line = next(input_files[i]) # TODO when batching: Pad several sequences to the same length and check via assertion
line = line.strip().split('\t')
line = labels_as_numbers(line, input_labels_to_ix[i])
input_lines.append([line])
for i in range(len(output_files)):
line = next(output_files[i])
line = line.strip().split('\t')
line = labels_as_numbers(line, output_labels_to_ix[i])
output_lines.append([line])
sequences.append(SequenceBatchContainer(input_lines, output_lines))
for i_f in input_files:
i_f.close()
for o_f in output_files:
o_f.close()
return sequences
def labels_as_numbers(seq, lbl_to_ix):
return [lbl_to_ix[lbl] for lbl in seq]
def numbers_as_labels(seq, ix_to_lbl):
return [ix_to_lbl[ix] for ix in seq]