-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathtransfer_learning.py
324 lines (270 loc) · 17.2 KB
/
transfer_learning.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
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
import matplotlib
from logparser.anomaly_injector import transfer_train_log
#matplotlib.use('Agg')
from settings import settings
from wordembeddings.transform_glove import merge_templates
import logparser.Drain.Drain_demo as drain
import wordembeddings.transform_bert as transform_bert
from loganaliser.regression import Regression
from loganaliser.multiclass import Multiclass
from wordembeddings.bert_finetuning import finetune
from shared_functions import calculate_precision_and_plot, get_cosine_distance, inject_anomalies, get_labels_from_corpus, \
pre_process_log_events, get_nearest_neighbour_embedding_label_mapping
import os
from wordembeddings.visualisation import write_to_tsv_files_bert_sentences
from shared_functions import get_embeddings
def experiment(epochs=0,
mode="multiclass",
anomaly_type='random_lines',
anomaly_amount=1,
clip=1.0,
attention=False,
prediction_only=False,
option='UtahUtahTransfer', seq_len=7, n_layers=1, n_hidden_units=128, batch_size=64, finetuning=False,
embeddings_model='bert', experiment='x', label_encoder=None):
cwd = os.getcwd() + "/"
print("############\n STARTING\n Epochs:{}, Mode:{}, Attention:{}, Anomaly Type:{}"
.format(epochs, mode, attention, anomaly_type))
no_anomaly = True if anomaly_type == "no_anomaly" else False
if finetuning:
results_dir = settings[option]["dataset_2"]["results_dir"] + "_finetune/"
else:
results_dir = settings[option]["dataset_2"]["results_dir"] + "/"
results_dir_experiment = "{}_{}_epochs_{}_seq_len_{}_anomaly_type_{}_{}_hidden_{}_layers_{}_clip_{}_experiment_{}/".format(
results_dir + mode, embeddings_model, epochs, seq_len, anomaly_type, anomaly_amount, n_hidden_units, n_layers, clip, experiment)
train_ds_1 = settings[option]["dataset_1"]["raw_normal"]
train_ds_2 = settings[option]["dataset_2"]["raw_normal"]
test_ds_2 = settings[option]["dataset_2"]["raw_anomaly"]
raw_dir_1 = settings[option]["dataset_1"]["raw_dir"]
raw_dir_2 = settings[option]["dataset_2"]["raw_dir"]
parsed_dir_1 = settings[option]["dataset_1"]["parsed_dir"]
parsed_dir_2 = settings[option]["dataset_2"]["parsed_dir"]
embeddings_dir_1 = settings[option]["dataset_1"]["embeddings_dir"]
embeddings_dir_2 = settings[option]["dataset_2"]["embeddings_dir"]
logtype_1 = settings[option]["dataset_1"]["logtype"]
logtype_2 = settings[option]["dataset_2"]["logtype"]
train_instance_information_1 = settings[option]["dataset_1"]['instance_information_file_normal']
train_instance_information_2 = settings[option]["dataset_2"]['instance_information_file_normal']
# for binary
BINARY_train_instance_information_injected_1 = settings[option]["dataset_1"]['instance_information_file_normal'] + train_ds_1 + "_" + anomaly_type + "_" + str(anomaly_amount)
BINARY_train_instance_information_injected_2 = settings[option]["dataset_2"]['instance_information_file_normal'] + train_ds_2 + "_" + anomaly_type + "_" + str(anomaly_amount)
test_instance_information_2 = settings[option]["dataset_2"]['instance_information_file_anomalies_pre_inject']
test_instance_information_injected_2 = settings[option]["dataset_2"]['instance_information_file_anomalies_injected'] + test_ds_2 + "_" + anomaly_type + "_" + str(anomaly_amount)
anomalies_injected_dir_2 = parsed_dir_2 + "anomalies_injected/"
anomaly_indeces_dir_2 = parsed_dir_2 + "anomalies_injected/anomaly_indeces/"
# corpus files produced by Drain
corpus_train_1 = cwd + parsed_dir_1 + train_ds_1 + '_corpus'
corpus_train_2 = cwd + parsed_dir_2 + train_ds_2 + '_corpus'
corpus_test_2 = cwd + parsed_dir_2 + test_ds_2 + '_corpus'
# bert vectors as pickle files
embeddings_train_1 = cwd + embeddings_dir_1 + train_ds_1 + '.pickle'
embeddings_train_2 = cwd + embeddings_dir_2 + train_ds_2 + '.pickle'
embeddings_test_2 = cwd + embeddings_dir_2 + test_ds_2 + '.pickle'
if finetuning:
finetuning_model_dir = "wordembeddings/finetuning-models/" + train_ds_1
lstm_model_save_path = cwd + 'loganaliser/saved_models/' + train_ds_1 + '_with_finetune' + '_lstm.pth'
else:
finetuning_model_dir = "bert-base-uncased"
lstm_model_save_path = cwd + 'loganaliser/saved_models/' + train_ds_1 + "_" + experiment + '_lstm.pth'
# take corpus parsed by drain, inject anomalies in this file
corpus_test_injected = cwd + anomalies_injected_dir_2 + test_ds_2 + "_" + anomaly_type
BINARY_corpus_train_injected = cwd + anomalies_injected_dir_2 + train_ds_1 + "_" + anomaly_type
BINARY_train_anomaly_indeces = cwd + results_dir_experiment + "train_anomaly_labels.txt"
test_anomaly_indeces = cwd + results_dir_experiment + "test_anomaly_labels.txt"
os.makedirs(results_dir, exist_ok=True)
os.makedirs(results_dir_experiment, exist_ok=True)
os.makedirs(raw_dir_1, exist_ok=True)
os.makedirs(raw_dir_2, exist_ok=True)
os.makedirs(parsed_dir_1, exist_ok=True)
os.makedirs(parsed_dir_2, exist_ok=True)
os.makedirs(embeddings_dir_1, exist_ok=True)
os.makedirs(embeddings_dir_2, exist_ok=True)
os.makedirs(anomalies_injected_dir_2, exist_ok=True)
os.makedirs(anomaly_indeces_dir_2, exist_ok=True)
### DRAIN PARSING
#if not os.path.exists(corpus_train_1) or not os.path.exists(corpus_test_2):
drain.execute(directory=raw_dir_1, file=train_ds_1, output=parsed_dir_1, logtype=logtype_1)
drain.execute(directory=raw_dir_2, file=train_ds_2, output=parsed_dir_2, logtype=logtype_2)
drain.execute(directory=raw_dir_2, file=test_ds_2, output=parsed_dir_2, logtype=logtype_2)
pre_process_log_events(corpus_test_2, corpus_train_1, corpus_train_2)
# manipulate train ds 1
if option != "UtahSashoTransfer":
transfer_train_log(corpus_train_1, corpus_train_1)
### INJECT ANOMALIES in test ds
test_ds_anomaly_lines, test_ds_lines_before_injection, train_ds_lines_after_injection = \
inject_anomalies(anomaly_type=anomaly_type,
corpus_input=corpus_test_2,
corpus_output=corpus_test_injected,
anomaly_indices_output_path=test_anomaly_indeces,
instance_information_in=test_instance_information_2,
instance_information_out=test_instance_information_injected_2,
anomaly_amount=anomaly_amount,
results_dir=results_dir_experiment)
### if in binary mode, inject anomalies also in train ds
if mode == "binary":
train_ds_anomaly_lines, train_ds_lines_before_injection, train_ds_lines_after_injection = \
inject_anomalies(
anomaly_type=anomaly_type,
corpus_input=corpus_train_1,
corpus_output=BINARY_corpus_train_injected,
anomaly_indices_output_path=BINARY_train_anomaly_indeces,
instance_information_in=train_instance_information_1,
instance_information_out=BINARY_train_instance_information_injected_1,
anomaly_amount=anomaly_amount,
results_dir=results_dir_experiment)
train_ds_2_anomaly_lines, train_ds_2_lines_before_injection, train_ds_2_lines_after_injection = \
inject_anomalies(
anomaly_type=anomaly_type,
corpus_input=corpus_train_2,
corpus_output=BINARY_corpus_train_injected,
anomaly_indices_output_path=BINARY_train_anomaly_indeces,
instance_information_in=train_instance_information_2,
instance_information_out=BINARY_train_instance_information_injected_2,
anomaly_amount=anomaly_amount,
results_dir=results_dir_experiment)
# produce templates out of the corpuses that we have from the anomaly file
templates_train_1 = list(set(open(corpus_train_1, 'r').readlines()))
templates_train_2 = list(set(open(corpus_train_2, 'r').readlines()))
# merge_templates(templates_normal_1, templates_normal_2, merged_template_path=parsed_dir_1 + "_merged_templates_normal")
templates_test_anomalies_injected = list(set(open(corpus_test_injected, 'r').readlines()))
merged_templates = merge_templates(templates_train_1, templates_train_2, templates_test_anomalies_injected, merged_template_path=None)
merged_templates = list(merged_templates)
if finetuning:
if not os.path.exists(finetuning_model_dir):
finetune(templates=templates_train_1, output_dir=finetuning_model_dir)
sentence_to_embeddings_mapping = get_embeddings(embeddings_model, merged_templates)
write_to_tsv_files_bert_sentences(word_embeddings=sentence_to_embeddings_mapping,
tsv_file_vectors=results_dir_experiment + "visualisation/vectors.tsv",
tsv_file_sentences=results_dir_experiment + "visualisation/sentences.tsv")
embeddings_dim = list(sentence_to_embeddings_mapping.values())[0].size()[0]
if anomaly_type in ["insert_words", "remove_words", "replace_words"]:
get_cosine_distance(test_ds_lines_before_injection, train_ds_lines_after_injection, results_dir_experiment, sentence_to_embeddings_mapping)
# transform output of bert into numpy word embedding vectors
transform_bert.transform(sentence_embeddings=sentence_to_embeddings_mapping, logfile=corpus_train_1, outputfile=embeddings_train_1)
transform_bert.transform(sentence_embeddings=sentence_to_embeddings_mapping, logfile=corpus_train_2, outputfile=embeddings_train_2)
transform_bert.transform(sentence_embeddings=sentence_to_embeddings_mapping, logfile=corpus_test_injected, outputfile=embeddings_test_2)
############################
###### MULTICLASS
############################
if mode == "multiclass":
target_normal_labels, n_classes, normal_label_embeddings_map, _ = get_labels_from_corpus(normal_corpus=open(corpus_train_1, 'r').readlines(),
encoder_path=label_encoder,
templates=templates_train_1,
embeddings=sentence_to_embeddings_mapping)
sentences_ds2_label_ds1_mapping = get_nearest_neighbour_embedding_label_mapping(sentence_to_embeddings_mapping,
normal_label_embeddings_map,
templates_train_2 + templates_test_anomalies_injected)
# Use sentence_label_mapping to map every sentence from corpus to labels
ds_2_target_labels = [sentences_ds2_label_ds1_mapping.get(sentence) for sentence in open(corpus_train_2, 'r').readlines()]
lstm_ds_1 = Multiclass(n_features=n_classes,
n_input=embeddings_dim,
target_labels=target_normal_labels,
train_vectors=embeddings_train_1,
train_instance_information_file=train_instance_information_1,
test_vectors=None,
test_instance_information_file=None,
savemodelpath=lstm_model_save_path,
seq_length=seq_len,
num_epochs=epochs,
no_anomaly=no_anomaly,
results_dir=cwd + results_dir_experiment,
embeddings_model='bert',
n_layers=n_layers,
n_hidden_units=n_hidden_units,
batch_size=batch_size,
clip=clip,
top_k_label_mapping=None,
normal_label_embeddings_map=normal_label_embeddings_map,
lines_that_have_anomalies=test_ds_anomaly_lines,
corpus_of_log_containing_anomalies=corpus_test_injected,
transfer_learning=False,
attention=attention,
prediction_only=prediction_only,
transfer_learning_initial_training=True,
mode=mode,
sentence_to_embeddings_mapping=sentence_to_embeddings_mapping)
if not prediction_only:
lstm_ds_1.start_training()
lstm_ds_2 = Multiclass(n_features=n_classes,
n_input=embeddings_dim,
target_labels=ds_2_target_labels,
train_vectors=embeddings_train_2,
train_instance_information_file=train_instance_information_2,
test_vectors=embeddings_test_2,
test_instance_information_file=test_instance_information_injected_2,
savemodelpath=lstm_model_save_path,
seq_length=seq_len,
num_epochs=5,
no_anomaly=no_anomaly,
results_dir=cwd + results_dir_experiment,
embeddings_model='bert',
n_layers=n_layers,
n_hidden_units=n_hidden_units,
batch_size=batch_size,
clip=clip,
top_k_label_mapping=sentences_ds2_label_ds1_mapping,
normal_label_embeddings_map=normal_label_embeddings_map,
lines_that_have_anomalies=test_ds_anomaly_lines,
corpus_of_log_containing_anomalies=corpus_test_injected,
transfer_learning=True,
attention=attention,
prediction_only=prediction_only,
mode=mode,
sentence_to_embeddings_mapping=sentence_to_embeddings_mapping)
if not prediction_only:
lstm_ds_2.start_training()
############################
###### REGRESSION
############################
elif mode == "regression":
lstm_ds_1 = Regression(train_vectors=embeddings_train_1,
train_instance_information_file=train_instance_information_1,
test_vectors=embeddings_train_1,
test_instance_information_file=train_instance_information_1,
savemodelpath=lstm_model_save_path,
seq_length=seq_len,
num_epochs=epochs,
n_hidden_units=n_hidden_units,
n_layers=n_layers,
embeddings_model='bert',
no_anomaly=no_anomaly,
clip=clip,
results_dir=cwd + results_dir_experiment,
batch_size=batch_size,
lines_that_have_anomalies=test_ds_anomaly_lines,
n_input=embeddings_dim,
n_features=embeddings_dim,
transfer_learning=False,
attention=attention,
prediction_only=prediction_only)
if not prediction_only:
lstm_ds_1.start_training()
lstm_ds_2 = Regression(train_vectors=embeddings_train_2,
train_instance_information_file=train_instance_information_2,
test_vectors=embeddings_test_2,
test_instance_information_file=test_instance_information_injected_2,
savemodelpath=lstm_model_save_path,
seq_length=seq_len,
num_epochs=5,
n_hidden_units=n_hidden_units,
n_layers=n_layers,
embeddings_model='bert',
no_anomaly=no_anomaly,
clip=clip,
results_dir=cwd + results_dir_experiment,
batch_size=batch_size,
lines_that_have_anomalies=test_ds_anomaly_lines,
n_input=embeddings_dim,
n_features=embeddings_dim,
transfer_learning=True,
attention=attention,
prediction_only=prediction_only)
if not prediction_only:
lstm_ds_2.start_training()
f1, precision = lstm_ds_2.final_prediction()
calculate_precision_and_plot(results_dir_experiment, cwd, embeddings_model, epochs, seq_len, anomaly_type,
anomaly_amount, n_hidden_units, n_layers, clip, experiment, mode)
print("done.")
return f1, precision
if __name__ == '__main__':
experiment()