-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathbind_learn.py
140 lines (119 loc) · 5.29 KB
/
bind_learn.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
# -*- coding: utf-8 -*-
"""
Created on Sun Nov 15 19:11:14 2020
@author: Bosec
"""
from torch.utils.data import DataLoader
import os
import numpy as np
import pandas as pd
from model import train_NN, predict, fit
from model_helper import ColumnarDataset
from sklearn import preprocessing
"""
def prepare_dataset(data, prep = prepare_texts_5net):
train_texts = data["text_a"].to_list()
train_y = data['label'].to_list()
train_matrix = prep(train_texts)
#del train_texts
#train_matrix = pd.DataFrame(train_matrix)
print(train_matrix)
train_dataset = ColumnarDataset(train_matrix, train_y)
return train_dataset, train_matrix.shape[1]
def fit_space(data, prep = prepare_texts_berts):
train_texts = data
train_y = [0] * len(data)
train_matrix = prep(train_texts)
#del train_texts
#train_matrix = pd.DataFrame(train_matrix)
print(train_matrix)
train_dataset = ColumnarDataset(train_matrix, train_y)
test_loader = DataLoader(train_dataset, batch_size=32, shuffle=True)
return test_loader
import torch
def train_nets(train_data, dev_data , test_data):
train_dataset, dims = prepare_dataset(train_data, prep = prepare_texts_berts)
valid_dataset, dims = prepare_dataset(dev_data, prep = prepare_texts_berts)
test_dataset, dims = prepare_dataset(test_data, prep = prepare_texts_berts)
for lr in [0.0001, 0.005, 0.001, 0.005, 0.01, 0.05, 0.1]:
for p in [0.1, 0.3, 0.5, 0.7]:
print(lr, p)
net = train_NN(train_dataset, valid_dataset, dims ,max_epochs=100, batch_size = 32, lr = lr, dropout = p)
predict(test_dataset, net)
torch.save(net.state_dict(), os.path.join("pickles","t2v_bert_lsa_"+str(lr)+"_"+str(p)+".pymodel"))
"""
import torch
def prepare_dataset(dataset, split):
prefix = dataset+"/"+split
train_matrix = prepare(prefix)
train_y = np.loadtxt(prefix+"/_ys.csv").astype(np.int64)
train_y = train_y.tolist()
train_dataset = ColumnarDataset(train_matrix, train_y)
return train_dataset, train_matrix.shape[1]
def train_nets(dataset):
train_dataset, dims = prepare_dataset(dataset, split = 'train')
valid_dataset, dims = prepare_dataset(dataset, split = 'dev')
test_dataset, dims = prepare_dataset(dataset, split = 'test')
for lr in [0.0001, 0.005, 0.001, 0.005, 0.01, 0.05, 0.1]:
for p in [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.9]:
for e_max in [12,13,14]:
net = train_NN(train_dataset, valid_dataset, dims, e_max = e_max, max_epochs=1000, batch_size = 32, lr = lr, dropout = p)
predict(test_dataset, net)
#torch.save(net.state_dict(), os.path.join("pickles","t2v_bert_lsa_"+str(lr)+"_"+str(p)+".pymodel"))
kg_types = ["complex", "transe", "rotate", "quate", "distmult", "simple"]
LM_types = ["distilbert-base-nli-mean-tokens","lsa","roberta-large-nli-stsb-mean-tokens","stat","xlm-r-large-en-ko-nli-ststb"]
def prepare(split):
dataz = None
first = True
for lm in LM_types:
path = split+"/"+lm+".csv"
df = pd.read_csv(path,header=None).to_numpy()
if first:
dataz = df
first = False
else:
dataz = np.hstack((dataz, df))
for kg in kg_types:
path = split+"/"+kg+".csv"
df = pd.read_csv(path,header=None).to_numpy()
dataz = np.hstack((dataz, df))
return dataz
def prepare_dataset_snd(dataset, split):
prefix = dataset+"/"+split
train_matrix = prepare(prefix)
train_y = np.loadtxt(prefix+"/_ys.csv").astype(np.int64)
train_y = train_y.tolist()
return train_matrix, train_y
#outs = prepare("pan2020/train")
#print(outs.shape)
#train_nets("pan2020")
#Prepare the train data
train_matrix = prepare_text(X, model)
parameters = {"loss":["hinge","log"],"penalty":["elasticnet"],"alpha":[0.01,0.001,0.0001,0.0005],"l1_ratio":[0.05,0.25,0.3,0.6,0.8,0.95],"power_t":[0.5,0.1,0.9]}
svc = SGDClassifier()
gs1 = GridSearchCV(svc, parameters, verbose = 1, n_jobs = 8,cv = 10, refit = True)
gs1.fit(train_matrix, final_y)
clf = gs1.best_estimator_
scores = cross_val_score(clf, train_matrix, final_y, n_jobs = 8, verbose = 1, cv=10, scoring='f1_macro')
acc_svm = scores.mean()
logging.info("TRAIN SGD 10fCV F1-score: %0.4f (+/- %0.2f)" % (scores.mean(), scores.std() * 2))
parameters = {"C":[0.1,1,10,25,50,100,500],"penalty":["l1","l2"]}
"""
lr_learner = LogisticRegression(max_iter = 100000, solver="saga")
gs = GridSearchCV(lr_learner, parameters, verbose = 1, n_jobs = 8,cv = 10, refit = True)
gs.fit(train_matrix, final_y)
clf = gs.best_estimator_
acc_lr = cross_val_score(clf, train_matrix, y, n_jobs = 8, verbose = 1, cv=10, scoring='f1_macro')
logging.info("TRAIN SGD 10fCV F1-score: %0.4f (+/- %0.4f)" % (scores.mean(), scores.std() * 2))
if acc_svm > acc_lr:
clf = clf1
"""
# Prepare output
#fitted = clf.fit(train_matrix, train_y)
with open(os.path.join(config.PICKLES_PATH, model + "_cv_clf.pkl"), "wb") as f:
pickle.dump(clf, f)
return
return clf
train_matrix, train_y = prepare_dataset_snd("pan2020", "train")
dev_matrix, dev_y = prepare_dataset_snd("pan2020", "dev")
test_matrix, test_y = prepare_dataset_snd("pan2020", "test")