-
Notifications
You must be signed in to change notification settings - Fork 2
/
drive_dg.py
347 lines (301 loc) · 17.4 KB
/
drive_dg.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
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
import os
os.environ['CUDA_VISIBLE_DEVICES'] = '0, 1,2, 3'
from utils.data_reading import load_data_for_expert, read_yaml_file, read_json_file, write_json_file
import argparse
from utils.evaluation import acc_compute, calculate_macro_f1
from tqdm import tqdm
from utils.components.dnf_layer import LogicTrainer
import torch
import random
import sklearn.tree as tree
from sklearn.model_selection import cross_val_score
from matplotlib import pyplot as plt
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score, f1_score
from sklearn.naive_bayes import GaussianNB
# from comet_ml import Experiment
# from comet_ml.integration.pytorch import log_model
from utils.components.dnf_layer import batch_generation, transform_org_to_logic
class Gen_Expert:
def __init__(self, s_domain, t_domains, mode, data_path, gq_file, sq_file, model_name, args):
self.source_domains = [s.strip() for s in s_domain.split("|")]
self.target_domains = [t.strip() for t in t_domains.split("|")]
# label rule, choice = {"binary", "multiple"}
self.mode = mode
self.source_dataset = []
self.target_sets = []
self.gq_file = gq_file
self.sq_file = sq_file
self.model_name = model_name
self.args = args
for s in self.source_domains:
data_path_ = os.path.join(data_path, s)
dataset, self.rule = load_data_for_expert(data_path=data_path_, dataset_name=s,
mode=self.mode, gq_file=self.gq_file, sq_file=self.sq_file, evo_file=None, evo_flag=False)
self.source_dataset.append(dataset)
for t in self.target_domains:
data_path_ = os.path.join(data_path, t)
dataset, self.rule = load_data_for_expert(data_path=data_path_, dataset_name=t,
mode=self.mode, gq_file=self.gq_file, sq_file=self.sq_file,evo_file=None, evo_flag=False )
self.target_sets.append(dataset)
self.save_path = args.save_path
self.trainer = None
def train_logic(self, num_conjuncts, n_out, configure, weight_init_type, args, exp=None):
predicate_set = {}
for a in configure:
predicate_set[a[0]] = a[1]
# configure = [('P1', 1), ('P2', 1), ('P3', 1), ('P4', 1), ('P5', 1), ('P6', 1), ('P7', 3)]d
# prepare train, val, test datasets'
train_logics_inputs = []
train_label_inputs = []
# load data of source domains as training data
for s_set in self.source_dataset:
gq = s_set["gq"]
logics_input, label_input = transform_org_to_logic(configure, s_set['train'], gq,
mask_flag=args.mask_flag)
train_logics_inputs = train_logics_inputs + logics_input
train_label_inputs = train_label_inputs + label_input
train_set = [train_logics_inputs, train_label_inputs]
val_logics_inputs = []
val_label_inputs = []
# load data of source domains as validation data
for s_set in self.source_dataset:
gq = s_set["gq"]
logics_input, label_input = transform_org_to_logic(configure, s_set['val'], gq,
mask_flag=args.mask_flag)
val_logics_inputs = val_logics_inputs + logics_input
val_label_inputs = val_label_inputs + label_input
val_loader = batch_generation(val_logics_inputs, val_label_inputs, self.mode, args.batchsize)
test_logics_inputs = []
test_label_inputs = []
# load data of target domains as test data
for t_set in self.target_sets:
gq = t_set["gq"]
logics_input, label_input = transform_org_to_logic(configure, t_set['test'], gq,
mask_flag=args.mask_flag)
test_logics_inputs = test_logics_inputs + logics_input
test_label_inputs = test_label_inputs + label_input
test_loader = batch_generation(test_logics_inputs, test_label_inputs, self.mode, args.batchsize)
# label_input = [transform_symbols_to_long(label_input[i:i + batchsize], label_mapping=Label_Mapping_Rule[mode]) for i
# in range(0, len(label_input), batchsize)]
# logics_input = [torch.tensor(logics_input[i:i + batchsize]) for i in range(0, len(logics_input), batchsize)]
# return [(logics_input[i], label_input[i]) for i in range(len(logics_input))]
print("length of train_set {}, length of val_loader {}, length of test_loader {}".format(len(train_set[0])//args.batchsize, len(val_loader), len(test_loader)))
args.n_steps_per_epoch = len(train_set[0])//args.batchsize
# initialize the training class
if args.type_of_logic_model == "tree":
clf = DecisionTreeClassifier(random_state=0, max_depth=5, max_leaf_nodes=10, min_weight_fraction_leaf=0.01)
# clf = GaussianNB()
ind_list = [i for i in range(len(train_set[0]))]
random.shuffle(ind_list)
# may be change tos shuffle per epoch
train_logics_inputs = [train_set[0][i] for i in ind_list]
train_label_inputs = [train_set[1][i] for i in ind_list]
train_loader = batch_generation(train_logics_inputs, train_label_inputs, self.mode, args.batchsize)
train_data = torch.cat([tmp[0] for tmp in train_loader], dim=0).numpy()
train_label = torch.cat([tmp[1] for tmp in train_loader]).numpy()
clf.fit(train_data, train_label)
t_data = torch.cat([tmp[0] for tmp in test_loader], dim=0).numpy()
t_label = torch.cat([tmp[1] for tmp in test_loader]).numpy()
p_label = clf.predict(t_data)
# Compute accuracy
accuracy = accuracy_score(t_label, p_label)
# Compute macro-F1 score
macro_f1 = f1_score(t_label, p_label, average='macro')
# plt.figure(dpi=500)
# tree.plot_tree(clf)
# plt.show()
print(accuracy, macro_f1)
return accuracy
if args.type_of_logic_model == "bayes":
clf = GaussianNB()
# clf = GaussianNB()
ind_list = [i for i in range(len(train_set[0]))]
random.shuffle(ind_list)
# may be change tos shuffle per epoch
train_logics_inputs = [train_set[0][i] for i in ind_list]
train_label_inputs = [train_set[1][i] for i in ind_list]
train_loader = batch_generation(train_logics_inputs, train_label_inputs, self.mode, args.batchsize)
train_data = torch.cat([tmp[0] for tmp in train_loader], dim=0).numpy()
train_label = torch.cat([tmp[1] for tmp in train_loader]).numpy()
clf.fit(train_data, train_label)
t_data = torch.cat([tmp[0] for tmp in test_loader], dim=0).numpy()
t_label = torch.cat([tmp[1] for tmp in test_loader]).numpy()
p_label = clf.predict(t_data)
# Compute accuracy
accuracy = accuracy_score(t_label, p_label)
# Compute macro-F1 score
macro_f1 = f1_score(t_label, p_label, average='macro')
# plt.figure(dpi=500)
# tree.plot_tree(clf)
# plt.show()
print(accuracy, macro_f1)
return accuracy
else:
# train the logic model
trainer = LogicTrainer(num_conjuncts=num_conjuncts, n_out=n_out, delta=args.initial_delta, configure=configure,
weight_init_type=weight_init_type, device=self.args.device, args=args, exp=exp)
reported_test_metrics = trainer.train(train_set, val_loader, test_loader)
return reported_test_metrics
# train
# eval on val
# save model
# eval on test
# def eval
def parse_args():
parser = argparse.ArgumentParser()
# dataset args
parser.add_argument('--s_domains', default='GOSSIPCOP|POLITIFACT', type=str)
parser.add_argument('--t_domains', default='Constraint', type=str,
choices=["Constraint", "GOSSIPCOP", "POLITIFACT"])
parser.add_argument('--data_path', type=str, default='/home/liuhui/unify/data')
parser.add_argument('--mode', type=str, default='binary', choices=['binary', 'multiple'])
# choose fewer smale for testing
parser.add_argument('--num_eval_samples', default=5, type=int)
parser.add_argument('--shot_number', default=0, type=int)
parser.add_argument('--save_path', default="/reports.json", type=str)
parser.add_argument('--save_all_path', default='/home/liuhui/unify/data/', type=str)
parser.add_argument('--model_name', type=str, default="flan-t5-xl",
choices=["flan-t5-xxl", "flan-t5-xl", "flan-t5-large", "flan-t5-base", "flan-t5-small", "Llama-2-7b-chat-hf",
"Llama-2-13b-chat-hf", "gpt-3.5-turbo"])
parser.add_argument('--device', default="cuda", choices=["cuda", "cpu"])
parser.add_argument('--evi_flag', action="store_true")
parser.add_argument('--eval_mode', type=str, default='logics', choices=['logics', 'sampling'])
# the parameters of the logic model
parser.add_argument('--num_conjuncts', default=20, type=int)
parser.add_argument('--n_out', default=2, type=int, choices=[2, 6])
parser.add_argument('--delta', default=0.01, type=float)
parser.add_argument('--weight_init_type', default="normal", type=str, choices=["normal", "uniform"])
parser.add_argument('--mask_flag', default=-2, type=int, choices=[-2, 0])
parser.add_argument('--initial_delta', '-initial_delta', type=float, default=0.01,
help='initial delta.')
parser.add_argument('--delta_decay_delay', '-delta_decay_delay', type=int, default=1,
help='delta_decay_delay.')
parser.add_argument('--delta_decay_steps', '-delta_decay_steps', type=int, default=1,
help='delta_decay_steps.')
# 0.01 1.3 -> 25 0.1 1.1
parser.add_argument('--delta_decay_rate', '-delta_decay_rate', type=float, default=1.1,
help='delta_decay_rate.')
# the logic model type
parser.add_argument('--type_of_logic_model', default="mlp", type=str, choices=["logic", "mlp", "tree", "bayes"])
# the parameters of training the logic model， optimizer, schedule
parser.add_argument('--SGD', '-sgd', action='store_true', help='use optimizer')
parser.add_argument('--lr', default=1e-3, type=float, help='learning rate')
parser.add_argument('--weight_decay', '-wd', default=1e-4, type=float, help='weight decay')
parser.add_argument('--n_steps_per_epoch', default=1, type=int)
parser.add_argument('--scheduler', '-sch', type=str, default='StepLR', choices=['StepLR', 'MultiStepLR', 'CosLR'])
parser.add_argument('--step_size', '-stp', type=int, default=20, help='fixed step size for StepLR')
parser.add_argument('--n_epoch', type=int, default=30, help='the number of epochs')
parser.add_argument('--n_batch_step', type=int, default=50,
help='the number of batches per step for delta scheduler')
parser.add_argument('--batchsize', default=64, type=int)
parser.add_argument('--gqfile', default="flan-t5-large_False.json", type=str)
# save the model
parser.add_argument('--best_target_ckpoint', default="xx.pt", type=str)
parser.add_argument('--save_flag', action="store_true")
# the parameters of decision tree
parser.add_argument('--max_depth', default=6, type=int, help='max_depth of decision tree')
parser.add_argument('--max_leaf_nodes', default=30, type=int, help='max_leaf_nodes of decision tree')
parser.add_argument('--min_weight_fraction_leaf', default=0.01, type=float, help='min_weight_fraction_leaf of decision tree')
args = parser.parse_args()
return args
if __name__ == "__main__":
############################# eval by LLMs
args = parse_args()
# eval using zero-shot faln-t5 and llama2
# e = Expert(dataset_name=args.dataset_name, mode=args.mode, data_path=args.data_path,
# gq_file=None, sq_file=None, model_name="flan-t5-xl", args=args)
# e.eval_gq(model_name=args.model_name, device=args.device, evi_flag=args.evi_flag, mode=args.eval_mode)
############################# eval by Logic Model
if args.evi_flag:
gq_files = ["flan-t5-large_True.json", "flan-t5-xl_True.json", "flan-t5-xxl_True.json", "Llama-2-7b-chat-hf_True.json",
"Llama-2-13b-chat-hf_True.json"]
# gq_files = ["gpt-3.5-turbo_True.json"]
else:
gq_files = [
# "flan-t5-large_False.json",
"flan-t5-xl_False.json", "flan-t5-xxl_False.json",
"Llama-2-7b-chat-hf_False.json", "Llama-2-13b-chat-hf_False.json"
]
# ["flan-t5-large_True.json", "flan-t5-xl_True.json", "flan-t5-xxl_True.json",
# "Llama-2-7b-chat-hf_True.json", "Llama-2-13b-chat-hf_True.json ", "gpt-3.5-turbo_True.json"]
args.save_path = os.path.join(args.data_path, "dg", args.s_domains+args.t_domains+str(args.evi_flag)+".json")
conjuncts = [50]
if args.n_out == 2:
args.mode = 'binary'
else:
args.mode = 'multiple'
wds = [1e-4]
final_results_wd_con = {}
final_results = {}
for wd in wds:
for conjunct in conjuncts:
args.num_conjuncts = conjunct
args.weight_decay = wd
exp_name_wd_con = '_'.join([args.s_domains, str(args.n_out), str(args.num_conjuncts), str(args.weight_decay)])
final_results_wd_con[exp_name_wd_con] = {}
final_results_wd_con[exp_name_wd_con]["reported_metrics"] = {}
avg_acc = []
for gq_file in gq_files:
args.gqfile = gq_file
exp_name = '_'.join([args.s_domains, str(args.n_out), str(args.num_conjuncts), str(args.weight_decay), args.gqfile])
# experiment.set_name(exp_name)
experiment = None
configure = [('P1', 1), ('P2', 1), ('P3', 1), ('P4', 1), ('P5', 1), ('P7', 3), ('P8', 1)]
e = Gen_Expert(s_domain=args.s_domains, t_domains=args.t_domains, mode=args.mode, data_path=args.data_path,
gq_file=args.gqfile, sq_file="sq.json", model_name=args.model_name, args=args)
reported_test_metrics = e.train_logic(args.num_conjuncts, args.n_out, configure=configure, weight_init_type=args.weight_init_type, args=args, exp=experiment)
final_results[exp_name] = reported_test_metrics
final_results_wd_con[exp_name_wd_con]["reported_metrics"][exp_name] = reported_test_metrics
avg_acc.append(reported_test_metrics["final_acc"])
final_results_wd_con[exp_name_wd_con]['avg_acc'] = sum(avg_acc)/len(avg_acc)
max_para = None
max_acc = 0
for key in final_results_wd_con.keys():
if max_acc<final_results_wd_con[key]['avg_acc']:
max_para = key
max_acc = final_results_wd_con[key]['avg_acc']
print(max_para)
print(max_acc)
print("#################################")
print(final_results_wd_con[max_para]["reported_metrics"])
#
write_json_file([final_results_wd_con, final_results] , args.save_path)
## for other decision models
# conjuncts = [10]
#
# if args.n_out == 2:
# args.mode = 'binary'
# else:
# args.mode = 'multiple'
# wds = [1e-4]
# final_results_wd_con = {}
# final_results = {}
# for wd in wds:
# for conjunct in conjuncts:
# args.num_conjuncts = conjunct
# args.weight_decay = wd
# exp_name_wd_con = '_'.join([args.s_domains, str(args.n_out), str(args.num_conjuncts), str(args.weight_decay)])
# final_results_wd_con[exp_name_wd_con] = {}
# final_results_wd_con[exp_name_wd_con]["reported_metrics"] = {}
# avg_acc = []
# for gq_file in gq_files:
# args.gqfile = gq_file
# exp_name = '_'.join([args.s_domains, str(args.n_out), str(args.num_conjuncts), str(args.weight_decay), args.gqfile])
# # experiment.set_name(exp_name)
# experiment = None
# configure = [('P1', 1), ('P2', 1), ('P3', 1), ('P4', 1), ('P5', 1), ('P7', 3), ('P8', 1)]
# e = Gen_Expert(s_domain=args.s_domains, t_domains=args.t_domains, mode=args.mode, data_path=args.data_path,
# gq_file=args.gqfile, sq_file="sq.json", model_name=args.model_name, args=args)
# reported_test_metrics = e.train_logic(args.num_conjuncts, args.n_out, configure=configure, weight_init_type=args.weight_init_type, args=args, exp=experiment)
# final_results[exp_name] = reported_test_metrics
# final_results_wd_con[exp_name_wd_con]["reported_metrics"][exp_name] = reported_test_metrics
# avg_acc.append(reported_test_metrics)
#
# final_results_wd_con[exp_name_wd_con]['avg_acc'] = sum(avg_acc)/len(avg_acc)
# max_para = None
# max_acc = 0
#
# print("#################################")
# print(final_results_wd_con[max_para]["reported_metrics"])
#