def train_multitask_model(model, dataset, config_params, use_elmo):
name = model._name
train_x, train_y = dataset.get('train_x'), dataset.get('train_y')
output_size = dataset.get('output_size')
mask_builder = dataset.get('mask_builder')
tokenizer = dataset.get('tokenizer')
dom_y = dataset.get("dom_labeled_list")
lex_y = dataset.get("lex_labeled_list")
del dataset
log_dir = os.path.join("logs", "fit_generator_multitask",
datetime.datetime.now().strftime("%Y%m%d-%H%M%S"))
logger = TensorBoard(log_dir)
check_dir = os.path.join("checkpoint_multitask", f'{name}.hdf5')
model_chkpt = ModelCheckpoint(filepath=check_dir,
monitor="loss",
mode="min",
save_best_only=True,
save_weights_only=True,
verbose=True)
early_stopping = EarlyStopping(monitor="loss", patience=3)
reduce_lr = ReduceLROnPlateau(monitor='loss',
factor=0.2,
patience=3,
min_lr=1e-6)
epochs = int(config_params['epochs'])
batch_size = int(config_params['batch_size'])
cbks = [logger, model_chkpt, early_stopping, reduce_lr]
resources_path = os.path.join(os.getcwd(), 'resources')
try:
history = model.fit_generator(
multitask_train_generator(train_x,
train_y,
dom_y,
lex_y,
batch_size,
output_size,
use_elmo,
mask_builder,
tokenizer,
use_bert=False),
verbose=1,
epochs=epochs,
steps_per_epoch=np.ceil(len(train_x) / batch_size),
callbacks=cbks)
history_path = os.path.join(resources_path, f'{name}_history.pkl')
save_pickle(history_path, history.history)
plot_history(history, os.path.join(resources_path, f'{name}_history'))
model.save(os.path.join(resources_path, f'{name}_model.h5'))
model.save_weights(os.path.join(resources_path, f'{name}_weights.h5'))
return history
except KeyboardInterrupt:
model.save(os.path.join(resources_path, f'{name}.h5'))
model.save_weights(os.path.join(resources_path, f'{name}_weights.h5'))
def train_tagger(corpus_name, corpus): """ Train the taggers and saves them Args: corpus_name: name of the corpus used to create the tagger corpus: corpus for creating the tagger """ #List of n-gram taggers names complete_names = [corpus_name + '_' + x for x in N_GRAM_NAMES] # Training UnigramTagger tagger1 = UnigramTagger(corpus) utilities.save_pickle(tagger1, complete_names[0], TAGGER_EXTENSION, TAGGER_PATH) print "UnigramTagger trained with", corpus_name # Training BigramTagger tagger2 = BigramTagger(corpus) utilities.save_pickle(tagger2, complete_names[1], TAGGER_EXTENSION, TAGGER_PATH) print "BigramTagger trained with", corpus_name # Training TrigramTagger tagger3 = TrigramTagger(corpus) utilities.save_pickle(tagger3, complete_names[2], TAGGER_EXTENSION, TAGGER_PATH) print "TrigramTagger trained with", corpus_name
print(f'train_y shape is: {train_x.shape}')
# x.shape = [number of samples, max characters/sentence] = [31_553 , 256]
train_y = torch.LongTensor(train_dataset.train_y)
print(f'train_y shape is: {train_y.shape}')
# y.shape = [number of samples, max characters/sentence] = [31_553 , 256]
print('\n========== Validation Dataset ==========')
dev_file_path = os.path.join(DATA_PATH, 'en.wiki.sentences.dev')
dev_gold_file_path = os.path.join(DATA_PATH, 'en.wiki.gold.dev')
dev_dataset = WikiDataset(dev_file_path, dev_gold_file_path)
dev_dataset.vectorize_data()
dev_x = torch.tensor(dev_dataset.train_x)
print(f'dev_x shape is: {dev_x.shape}')
# x.shape = [number of samples, max characters/sentence] = [3_994 , 256]
dev_y = torch.tensor(dev_dataset.train_y)
print(f'dev_y shape is: {dev_y.shape}')
# y.shape = [number of samples, max characters/sentence] = [3_994 , 256]
char2idx_path_save = os.path.join(RESOURCES_PATH, 'char2idx.pkl')
save_pickle(char2idx_path_save, train_dataset.char2idx)
idx2char_path_save = os.path.join(RESOURCES_PATH, 'idx2char.pkl')
save_pickle(idx2char_path_save, train_dataset.idx2char)
label2idx_path_save = os.path.join(RESOURCES_PATH, 'label2idx.pkl')
save_pickle(label2idx_path_save, train_dataset.label2idx)
idx2label_path_save = os.path.join(RESOURCES_PATH, 'idx2label.pkl')
save_pickle(idx2label_path_save, train_dataset.idx2label)
file_path = os.path.join(DATA_PATH, 'en.wiki.sentences.train')
gold_file_path = os.path.join(DATA_PATH, 'en.wiki.gold.train')
print('========== Training Dataset ==========')
train_dataset = WikiDataset(file_path, gold_file_path)
train_dataset.vectorize_data()
# train_x = torch.LongTensor(train_dataset.train_x)
# print(f'train_x shape is: {train_x.shape}')
# x.shape = [number of samples, max characters/sentence] = [31_553, 256]
# train_y = torch.LongTensor(train_dataset.train_y)
# print(f'train_y shape is: {train_y.shape}')
# y.shape = [number of samples, max characters/sentence] = [31_553, 256]
char2idx_path_save = os.path.join(RESOURCES_PATH, 'char2idx.pkl')
save_pickle(char2idx_path_save, train_dataset.char2idx)
label2idx_path_save = os.path.join(RESOURCES_PATH, 'label2idx.pkl')
save_pickle(label2idx_path_save, train_dataset.label2idx)
print('\n========== Validation Dataset ==========')
dev_file_path = os.path.join(DATA_PATH, 'en.wiki.sentences.dev')
dev_gold_file_path = os.path.join(DATA_PATH, 'en.wiki.gold.dev')
dev_dataset = WikiDataset(dev_file_path, dev_gold_file_path)
dev_dataset.char2idx = train_dataset.char2idx
dev_dataset.idx2char = train_dataset.idx2char
dev_dataset.label2idx = train_dataset.label2idx
dev_dataset.idx2label = train_dataset.idx2label
dev_dataset.vectorize_data()
# dev_x = torch.LongTensor(dev_dataset.train_x)
def train_classifier(name, x_train, y_train):
"""
Using the caracteristics and the labels it will train the classifier
and save it as a pickle file
Args:
name: name of the classifier
x_train: metrics to train the classifier
y_train: labels to train the classifier
Returns:
the classifier
"""
"""
Classifiers info: http://scikit-learn.org/stable/auto_examples/classification/plot_classifier_comparison.html
"""
timer = u.Timer()
print "\nTraining", name
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler(copy=True, with_mean=True, with_std=True)
from sklearn.feature_selection import SelectKBest
select = SelectKBest()
list_k = [
2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21
]
### This functions allows to try other classifiers
def declare_NB():
from sklearn.naive_bayes import GaussianNB
naive_bayes = GaussianNB()
steps = [("scaler", scaler), ('feature_selection', select),
('naive_bayes', naive_bayes)]
parameters = dict(feature_selection__k=list_k)
return steps, parameters
def declare_SVM():
from sklearn.svm import SVC
SVM = SVC()
steps = [("scaler", scaler), ('feature_selection', select),
('SVM', SVM)]
list_C = [1, 2, 3, 4, 5, 10, 100, 1000, 10000]
parameters = dict(feature_selection__k=list_k,
SVM__kernel=["rbf"],
SVM__C=list_C)
return steps, parameters, list_C
def declare_adaboost():
from sklearn.ensemble import AdaBoostClassifier
adaboost = AdaBoostClassifier()
steps = [("scaler", scaler), ('feature_selection', select),
('adaboost', adaboost)]
parameters = dict(feature_selection__k=list_k)
return steps, parameters
def declare_Decision_tree():
from sklearn import tree
decision_tree = tree.DecisionTreeClassifier()
steps = [("scaler", scaler), ('feature_selection', select),
('decision_tree', decision_tree)]
min_samp_list = [20, 15, 10, 8, 6, 4]
parameters = dict(feature_selection__k=list_k,
decision_tree__min_samples_split=min_samp_list)
return steps, parameters, min_samp_list
def declare_Random_forest():
from sklearn.ensemble import RandomForestClassifier
random_forest = RandomForestClassifier()
steps = [("scaler", scaler), ('feature_selection', select),
('random_forest', random_forest)]
min_samp_list = [20, 15, 10, 8, 6, 4]
parameters = dict(feature_selection__k=list_k,
random_forest__min_samples_split=min_samp_list)
return steps, parameters, min_samp_list
#Use the apropiate algorithm
if name == name_naive_bayes:
steps, parameters = declare_NB()
elif name == name_SVM:
steps, parameters, list_C = declare_SVM()
elif name == name_adaBoost:
steps, parameters = declare_adaboost()
elif name == name_decision_tree:
steps, parameters, min_samp_list = declare_Decision_tree()
elif name == name_random_forest:
steps, parameters, min_samp_list = declare_Random_forest()
from sklearn.cross_validation import ShuffleSplit
cv = ShuffleSplit(len(x_train), n_iter=10, test_size=0.1, random_state=0)
from sklearn.pipeline import Pipeline
pipeline = Pipeline(steps)
from sklearn.grid_search import GridSearchCV
#Scoring options:
# accuracy, f1_weighted, r2, average_precision
clf = GridSearchCV(pipeline, cv=cv,
param_grid=parameters) #, scoring="f1_weighted")
clf.fit(x_train, y_train)
def report_NB():
import data_analisis
list_mean = []
for param, mean_score, cv_scores in clf.grid_scores_:
list_mean.append(mean_score)
data_analisis.scatter_plot_from_lists(list_k,
list_mean,
"NB accuracy by K variables",
Algorithms_path,
xlabel="Num variables",
ylabel="Accuracy")
def report_Adaboost():
import data_analisis
list_mean = []
for param, mean_score, cv_scores in clf.grid_scores_:
list_mean.append(mean_score)
data_analisis.scatter_plot_from_lists(
list_k,
list_mean,
"Adaboost accuracy by K variables",
Algorithms_path,
xlabel="Num variables",
ylabel="Accuracy")
def report_SVM():
size_k = len(list_k)
size_c = len(list_C)
print "k=", size_k, "c=", size_c
matrix = [[0 for x in range(size_c)] for y in range(size_k)]
i = 0
j = 0
last_value_C = list_C[0]
for param, mean_score, cv_scores in clf.grid_scores_:
if param["SVM__C"] != last_value_C:
i += 1
j = 0
last_value_C = param["SVM__C"]
#print param, "mean=", mean_score, "i=", i, "j=", j
matrix[j][i] = mean_score
j += 1
import numpy as np
header = [""] + list_C
matrix = np.c_[list_k, matrix]
u.save_to_csv("SVM.csv", Algorithms_path, matrix, header)
u.change_decimal_separator("SVM.csv", Algorithms_path)
def report_decision_tree():
size_k = len(list_k)
size_min_samp = len(min_samp_list)
print "k=", size_k, "min_samples=", size_min_samp
matrix = [[0 for x in range(size_min_samp)] for y in range(size_k)]
i = 0
j = 0
last_value_min_samples = min_samp_list[0]
for param, mean_score, cv_scores in clf.grid_scores_:
if param[
"decision_tree__min_samples_split"] != last_value_min_samples:
i += 1
j = 0
last_value_min_samples = param[
"decision_tree__min_samples_split"]
#print param, "mean=", mean_score, "i=", i, "j=", j
matrix[j][i] = mean_score
j += 1
import numpy as np
header = [""] + min_samp_list
matrix = np.c_[list_k, matrix]
u.save_to_csv("DecisionTree.csv", Algorithms_path, matrix, header)
u.change_decimal_separator("DecisionTree.csv", Algorithms_path)
def report_random_forest():
size_k = len(list_k)
size_min_samp = len(min_samp_list)
print "k=", size_k, "min_samples=", size_min_samp
matrix = [[0 for x in range(size_min_samp)] for y in range(size_k)]
i = 0
j = 0
last_value_min_samples = list_k[0]
for param, mean_score, cv_scores in clf.grid_scores_:
if param["feature_selection__k"] != last_value_min_samples:
i += 1
j = 0
last_value_min_samples = param["feature_selection__k"]
#print param, "mean=", mean_score, "i=", i, "j=", j
matrix[i][j] = mean_score
j += 1
import numpy as np
header = [""] + min_samp_list
matrix = np.c_[list_k, matrix]
u.save_to_csv("RandomForest.csv", Algorithms_path, matrix, header)
u.change_decimal_separator("RandomForest.csv", Algorithms_path)
#Use the apropiate algorithm
if name == name_naive_bayes:
report_NB()
elif name == name_SVM:
report_SVM()
elif name == name_adaBoost:
report_Adaboost()
elif name == name_decision_tree:
report_decision_tree()
elif name == name_random_forest:
report_random_forest()
print "\n\nBest estimator", clf.best_estimator_
print "\n\nBest score", clf.best_score_
print "Trained in", timer.get_time()
from process_text import get_metrics_header
final_feature_indices = clf.best_estimator_.named_steps[
"feature_selection"].get_support(indices=True)
final_feature_list = [
get_metrics_header()[i] for i in final_feature_indices
]
print "Selected vars:", final_feature_list
u.save_pickle(clf, name, path=ML_path)
return clf
