-
Notifications
You must be signed in to change notification settings - Fork 0
/
mst_lstm.py
141 lines (124 loc) · 5.67 KB
/
mst_lstm.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
"""
For Hydrogen;
%load_ext autoreload
%autoreload 2
"""
from typing import List,Tuple,Union,Optional
import numpy as np
import torch
import torch.autograd as autograd
import torch.nn as nn
import torch.optim as optim
from decoder import eisner_decode
from util import set_logger
# set logger
logger = set_logger(__name__)
# Fix seed
torch.manual_seed(1)
class BiLSTM_Parser(nn.Module):
def __init__(self,
vocab_size,
pos_size,
word_embed_dim,
pos_embed_dim,
lstm_hidden_dim,
mlp_hidden_dim,
num_layers):
super(BiLSTM_Parser,self).__init__()
# Hidden dimension must be an even number for now
# This is the TOTAL dimension of the bidirectional hidden layer
assert lstm_hidden_dim % 2 == 0
# Hyperparam
self.vocab_size = vocab_size
self.word_embed_dim = word_embed_dim
self.pos_embed_dim = pos_embed_dim
self.lstm_hidden_dim = lstm_hidden_dim
# Layers
self.word_embeds = nn.Embedding(vocab_size, word_embed_dim)
self.pos_embeds = nn.Embedding(pos_size, pos_embed_dim)
self.lstm = nn.LSTM(input_size = word_embed_dim+pos_embed_dim,
hidden_size = lstm_hidden_dim // 2,
num_layers = num_layers,
bidirectional= True)
self.Linear_head = nn.Linear(lstm_hidden_dim,mlp_hidden_dim // 2)
self.Linear_modif = nn.Linear(lstm_hidden_dim,mlp_hidden_dim // 2)
self.output_layer = nn.Linear(mlp_hidden_dim,1) # output layer
# For test : word_tensor = data[0]; pos_tensor = data[1]
def forward(self,
word_tensor:torch.LongTensor,
pos_tensor :torch.LongTensor,
head_golden:Optional[List[int]] = None) \
-> Tuple[List[int],torch.tensor,torch.tensor]:
sentence_len = len(word_tensor[0])
# Word/POS embedding
word_embeds = self.word_embeds(word_tensor) # word_embeds.shape = (1,sentence_len,word_embed_dim)
pos_embeds = self.pos_embeds(pos_tensor) # pos_embeds.shape = (1,sentence_len,pos_embed_dim)
embeds = torch.cat((word_embeds,pos_embeds),2) # embeds.shape = (1,sentence_len,(word_embed_dim+pos_embed_dim))
embeds = embeds.view(sentence_len,1,-1) # embeds.shape = (sentence_len,1,(word_embed_dim+pos_embed_dim))
# Bidirectional LSTM
lstm_out, _ = self.lstm(embeds) # lstm_out.shape = (sentence_len,1,lstm_hidden_dim)
lstm_out = lstm_out.view(sentence_len, self.lstm_hidden_dim)
# Compute score of h -> m (Hold values in float as well for decoding etc)
## Precompute the necessrary components
head_features = self.Linear_head(lstm_out) # head_features.shape(sentence_len,mlp_hidden_dim//2)
modif_features = self.Linear_modif(lstm_out) # modif_features.shape(sentence_len,mlp_hidden_dim//2)
## Compute score matrix
score_matrix = torch.empty(size=(sentence_len,sentence_len))
for m in range(sentence_len):
for h in range(sentence_len):
feature_func = torch.cat((head_features[h],modif_features[m]))
neuron = torch.tanh(feature_func) # neuron.shape = [mlp_hidden_dim]
score_matrix[m][h] = self.output_layer(neuron)
# Find the best path, given the score_matrix
head_hat,score_hat = eisner_decode(score_matrix,head_golden)
# Score for the golden head
if head_golden is not None:
score_golden = 0
for m,h in enumerate(head_golden):
score_golden += score_matrix[h][m]
else:
score_golden = None
return head_hat,score_hat,score_golden
# util func for debug for test : tensor = embeds
# def get_n_next_function(tensor:torch.Tensor,n:Optional[int]=None):
# """
# Function to check what is the n next function in the backword propagation.
# If n is None, go back as far as possible.
# """
# component = tensor.grad_fn
# logger.debug(component)
# if n is not None:
# for i in range(n):
# if len(component.next_functions) == 0:
# raise ValueError(f"No more function back than {i}")
# component = component.next_functions[0][0]
# logger.debug(component)
# else:
# while len(component.next_functions) != 0:
# component = component.next_functions[0][0]
# logger.debug(component)
# return component
if __name__ == '__main__':
### Script for test
# Load test
from pathlib import Path
from data_processor import ConllDataSet
dev_path = Path("data","en-universal-dev.conll")
dev_data = ConllDataSet(dev_path)
# Init model (self = model)
model = BiLSTM_Parser(vocab_size = dev_data.vocab_size,
pos_size = dev_data.pos_size,
word_embed_dim = 100,
pos_embed_dim = 25,
lstm_hidden_dim = 250,
mlp_hidden_dim = 100,
num_layers = 2)
# Check forward() and loss funtion
data = dev_data[1]
head_hat,score_hat,score_golden = model(data[0],data[1],data[2])
logger.debug("Data flowed through the network")
# Check computational graph
component = loss.grad_fn
while len(component.next_functions) != 0:
component = component.next_functions[0][0]
logger.debug(component)