-
Notifications
You must be signed in to change notification settings - Fork 0
/
intent_snips_class_byparts.py
237 lines (196 loc) · 10.4 KB
/
intent_snips_class_byparts.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
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
import tensorflow as tf
from keras.backend.tensorflow_backend import set_session
config = tf.ConfigProto()
# config.gpu_options.per_process_gpu_memory_fraction = 0.95
config.gpu_options.allow_growth = True
config.gpu_options.visible_device_list = '0'
set_session(tf.Session(config=config))
import pandas as pd
import numpy as np
import fasttext
from metrics import fmeasure
from intent_models import cnn_word_model,cnn_word_model_ner
from intent_recognizer_class import IntentRecognizer
import sys, os
sys.path.append('/home/dilyara/Documents/GitHub/general_scripts')
from random_search_class import param_gen
from save_load_model import init_from_scratch, init_from_saved, save
from save_predictions import save_predictions
SEED = 42
np.random.seed(SEED)
tf.set_random_seed(SEED)
FIND_BEST_PARAMS = False
AVERAGE_FOR_PARAMS = True
NUM_OF_CALCS = 16
VERSION = '_findbest_byparts_paraphrases_2_nobpe'
path = '/home/dilyara/data/data_files/snips'
train_data = []
train_data.append(pd.read_csv("/home/dilyara/data/data_files/snips/snips_ner_gold/snips_ner_gold_0/snips_train_0"))
test_data = []
test_data.append(pd.read_csv("/home/dilyara/data/data_files/snips/snips_ner_gold/snips_ner_gold_0/snips_test_0"))
fasttext_model_file = '/home/dilyara/data/data_files/embeddings/reddit_fasttext_model.bin'
fasttext_model = fasttext.load_model(fasttext_model_file)
#-------------PARAMETERS----------------
text_size = 25
embedding_size = 100
n_splits = 1
kernel_sizes=[1,2,3]
train_sizes = [10, 25, 50, 100, 200, 500, 1000] # per intent
# train_sizes = [10, 25, 50] # per intent
intents = ['AddToPlaylist', 'BookRestaurant', 'GetWeather',
'PlayMusic', 'RateBook', 'SearchCreativeWork',
'SearchScreeningEvent']
#---------------------------------------
train_requests = [train_data[i].loc[:,'request'].values for i in range(n_splits)]
train_classes = [train_data[i].loc[:,intents].values for i in range(n_splits)]
test_requests = [test_data[i].loc[:2761, 'request'].values for i in range(n_splits)]
test_classes = [test_data[i].loc[:2761, intents].values for i in range(n_splits)]
f1_mean_per_size = []
f1_std_per_size = []
for n_size, train_size in enumerate(train_sizes):
print("\n\n______NUMBER OF TRAIN SAMPLES PER INTENT = %d___________" % train_size)
train_index_parts = []
for model_ind in range(n_splits):
train_part = []
for i, intent in enumerate(intents):
samples_intent = np.nonzero(train_classes[model_ind][:,i])[0]
train_part.extend(list(np.random.choice(samples_intent, size=train_size)))
train_index_parts.append(train_part)
train_requests_part = [train_requests[model_ind][train_index_parts[model_ind]] for model_ind in range(n_splits)]
train_classes_part = [train_classes[model_ind][train_index_parts[model_ind]] for model_ind in range(n_splits)]
if FIND_BEST_PARAMS:
print("___TO FIND APPROPRIATE PARAMETERS____")
FindBestRecognizer = IntentRecognizer(intents, fasttext_embedding_model=fasttext_model, n_splits=n_splits)
best_mean_f1 = 0.
best_network_params = dict()
best_learning_params = dict()
params_f1 = []
for p in range(20):
FindBestRecognizer.gener_network_parameters(coef_reg_cnn={'range': [0.0001,0.01], 'scale': 'log'},
coef_reg_den={'range': [0.0001,0.01], 'scale': 'log'},
filters_cnn={'range': [50,200], 'discrete': True},
dense_size={'range': [50,200], 'discrete': True},
dropout_rate={'range': [0.4,0.6]})
FindBestRecognizer.gener_learning_parameters(batch_size={'range': [16,64], 'discrete': True},
lear_rate={'range': [0.01,0.1], 'scale': 'log'},
lear_rate_decay={'range': [0.01,0.1], 'scale': 'log'},
epochs={'range': [50,100], 'discrete': True, 'scale': 'log'})
FindBestRecognizer.init_model(cnn_word_model, text_size, embedding_size, kernel_sizes, add_network_params=None)
FindBestRecognizer.fit_model(train_requests_part, train_classes_part, verbose=True, to_use_kfold=False)
train_predictions = FindBestRecognizer.predict(train_requests_part)
FindBestRecognizer.report(np.vstack([train_classes_part[i] for i in range(n_splits)]),
np.vstack([train_predictions[i] for i in range(n_splits)]),
mode='TRAIN')
test_predictions = FindBestRecognizer.predict(test_requests)
f1_test = FindBestRecognizer.report(np.vstack([test_classes[i] for i in range(n_splits)]),
np.vstack([test_predictions[i] for i in range(n_splits)]),
mode='TEST')[0]
mean_f1 = np.mean(f1_test)
params_dict = FindBestRecognizer.all_params_to_dict()
params_dict['mean_f1'] = mean_f1
params_f1.append(params_dict)
params_f1_dataframe = pd.DataFrame(params_f1)
params_f1_dataframe.to_csv("/home/dilyara/data/outputs/intent_snips/depend_" +
VERSION + '_' + str(train_size) + '.txt')
if mean_f1 > best_mean_f1:
FindBestRecognizer.save_models(fname='/home/dilyara/data/models/intent_models/snips_models_softmax/best_model_' +
VERSION + '_' + str(train_size))
print('___BETTER PARAMETERS FOUND!___\n')
print('___THESE PARAMETERS ARE:___', params_dict)
best_mean_f1 = mean_f1
if AVERAGE_FOR_PARAMS:
params = [
# 10
[{'coef_reg_cnn': 0.0002240188358941768,
'coef_reg_den': 0.00013254278511375586,
'filters_cnn': 220,
'dense_size': 80,
'dropout_rate': 0.439508706178354},
{'batch_size': 17,
'lear_rate': 0.014911813954885302,
'lear_rate_decay': 0.011552169958875022,
'epochs': 22}],
# 25
[{'coef_reg_cnn': 0.00012373572818256555,
'coef_reg_den': 0.00017171259810186691,
'filters_cnn': 202,
'dense_size': 67,
'dropout_rate': 0.5603207356574003},
{'batch_size': 26,
'lear_rate': 0.054040612295756969,
'lear_rate_decay': 0.084926115338805563,
'epochs': 24}],
# 50
[{'coef_reg_cnn': 0.00015919311850687678,
'coef_reg_den': 0.00016115679404622989,
'filters_cnn': 290,
'dense_size': 54,
'dropout_rate': 0.5852312361349971},
{'batch_size': 23,
'lear_rate': 0.048151980276947157,
'lear_rate_decay': 0.029064116214377402,
'epochs': 33}],
# 100
[{'coef_reg_cnn': 0.00033168959552320646,
'coef_reg_den': 0.00044867444269376276,
'filters_cnn': 234,
'dense_size': 95,
'dropout_rate': 0.4171426478913063},
{'batch_size': 32,
'lear_rate': 0.034295802954288496,
'lear_rate_decay': 0.067480368299883756,
'epochs': 50}],
# 200
[{'coef_reg_cnn': 0.00020510867913527356,
'coef_reg_den': 0.00030370411016572015,
'filters_cnn': 277,
'dense_size': 98,
'dropout_rate': 0.4986233680859435},
{'batch_size': 30,
'lear_rate': 0.021880881947614603,
'lear_rate_decay': 0.014620662267840959,
'epochs': 23}],
# 500
[{'coef_reg_cnn': 0.00011826989851694623,
'coef_reg_den': 0.00057033663916566111,
'filters_cnn': 298,
'dense_size': 71,
'dropout_rate': 0.4026373274835373},
{'batch_size': 21,
'lear_rate': 0.025750585638000676,
'lear_rate_decay': 0.023253677502792103,
'epochs': 34}],
# 1000
[{'coef_reg_cnn': 0.00046255365614283103,
'coef_reg_den': 0.0014098076556438696,
'filters_cnn': 210,
'dense_size': 59,
'dropout_rate': 0.5557728960043049},
{'batch_size': 28,
'lear_rate': 0.024490853695736985,
'lear_rate_decay': 0.028121698403082398,
'epochs': 47}]]
print("___TO CALCULATE AVERAGE ACCURACY FOR PARAMETERS____")
f1_mean_scores = []
for p in range(NUM_OF_CALCS):
AverageRecognizer = IntentRecognizer(intents, fasttext_embedding_model=fasttext_model, n_splits=n_splits)
AverageRecognizer.init_network_parameters([params[n_size][0]])
AverageRecognizer.init_learning_parameters([params[n_size][1]])
AverageRecognizer.init_model(cnn_word_model, text_size, embedding_size, kernel_sizes, add_network_params=None)
AverageRecognizer.fit_model(train_requests_part, train_classes_part, to_use_kfold=False, verbose=True)
train_predictions = AverageRecognizer.predict(train_requests_part)
AverageRecognizer.report(np.vstack([train_classes_part[i] for i in range(n_splits)]),
np.vstack([train_predictions[i] for i in range(n_splits)]),
mode='TRAIN')
test_predictions = AverageRecognizer.predict(test_requests)
f1_scores = AverageRecognizer.report(np.vstack([test_classes[i] for i in range(n_splits)]),
np.vstack([test_predictions[i] for i in range(n_splits)]),
mode='TEST')[0]
f1_mean_scores.append(np.mean(f1_scores))
f1_mean_per_size.append(np.mean(f1_mean_scores))
f1_std_per_size.append(np.std(f1_mean_scores))
print("___MEAN-STD___:\n size: %d\t f1-mean: %f\tf1-std: %f" % (
train_size, f1_mean_per_size[n_size], f1_std_per_size[n_size]))
if AVERAGE_FOR_PARAMS:
for n_size, train_size in enumerate(train_sizes):
print("size: %d\t f1-mean: %f\tf1-std: %f" % (train_size, f1_mean_per_size[n_size], f1_std_per_size[n_size]))