def start(self):
"""Main training loop."""
for i in range(CFG.num_iterations):
print("Iteration", i + 1)
training_data = [] # list to store self play states, pis and vs
for j in range(CFG.num_games):
print("Start Training Self-Play Game", j + 1)
game = self.game.clone() # Create a fresh clone for each game.
self.play_game(game, training_data)
# Save the current neural network model.
self.net.save_model()
# Load the recently saved model into the evaluator network.
self.eval_net.load_model()
# Train the network using self play values.
self.net.train(training_data)
# Initialize MonteCarloTreeSearch objects for both networks.
current_mcts = MonteCarloTreeSearch(self.net)
eval_mcts = MonteCarloTreeSearch(self.eval_net)
evaluator = Evaluate(current_mcts=current_mcts, eval_mcts=eval_mcts,
game=self.game)
wins, losses = evaluator.evaluate()
print("wins:", wins)
print("losses:", losses)
num_games = wins + losses
if num_games == 0:
win_rate = 0
else:
win_rate = wins / num_games
print("win rate:", win_rate)
if win_rate > CFG.eval_win_rate:
# Save current model as the best model.
print("New model saved as best model.")
self.net.save_model("best_model")
else:
print("New model discarded and previous model loaded.")
# Discard current model and use previous best model.
self.net.load_model()
def evaluateVSM(targeEventFile, collFolder,k,relevTh,vsmClassifierFileName,topK):
'''
docs = []
try:
classifierFile = open(vsmClassifierFileName,"rb")
classifier = pickle.load(classifierFile)
classifierFile.close()
except:
f = open(targeEventFile,'r')
for url in f:
url = url.strip()
d = Document(url)
if d:
docs.append(d)
f.close()
docsTF = []
for d in docs:
wordsFreq = getFreq(d.getWords())
docsTF.append(wordsFreq)
classifier = VSMClassifier(docsTF,relevTh)
evalres = []
for j in range(k):
fn = collFolder+str(j)+'.txt'
f = codecs.open(fn, encoding='utf-8')
ftext = f.read()
r = classifier.calculate_score(ftext)[0]
evalres.append(r)
f.close()
'''
evaluator = Evaluate()
evaluator.buildVSMClassifier(targeEventFile,vsmClassifierFileName,relevTh,topK)
collFiles = []
for j in range(k):
fn = collFolder+str(j)+'.txt'
f = codecs.open(fn, encoding='utf-8')
ftext = f.read()
o = myObj()
o.text = ftext
collFiles.append(o)
res = evaluator.evaluateFC(collFiles)
#f = open(collFolder+'evaluationRes_VSM.txt','w')
#f.write('\n'.join([str(r) for r in res]))
#f.close()
#print sum(res)
return res
def _mock_evaluate(self):
"""
Create a mock `Evaluate` class, with all methods that access the
influxdb database stubbed out.
Returns:
Evaluate: An instance of the Evaluate class.
"""
evaluate = Evaluate(self.PIT, self.TP, self.VERSION,
self._get_tmp_dir(), False, False)
evaluate.query_influx = MagicMock()
evaluate.query_influx = InfluxDBMock.query_influx
return evaluate
def main(args):
if args.save_path is not None and not os.path.exists(args.save_path):
os.makedirs(args.save_path)
summary_writer = tf.summary.FileWriter(os.path.join(args.save_path, 'log'))
global_steps_counter = itertools.count() # thread-safe
global_net = Net(S_DIM, A_DIM, 'global', args)
num_workers = args.threads
workers = []
# create workers
for i in range(1, num_workers + 1):
worker_summary_writer = summary_writer if i == 0 else None
worker = Worker(i, make_env(args), global_steps_counter,
worker_summary_writer, args)
workers.append(worker)
saver = tf.train.Saver(max_to_keep=5)
with tf.Session() as sess:
coord = tf.train.Coordinator()
if args.model_path is not None:
print('Loading model...\n')
ckpt = tf.train.get_checkpoint_state(args.model_path)
saver.restore(sess, ckpt.model_checkpoint_path)
else:
print('Initializing a new model...\n')
sess.run(tf.global_variables_initializer())
print_params_nums()
# Start work process for each worker in a seperated thread
worker_threads = []
for worker in workers:
t = threading.Thread(target=lambda: worker.run(sess, coord, saver))
t.start()
time.sleep(0.5)
worker_threads.append(t)
if args.eval_every > 0:
evaluator = Evaluate(
global_net, summary_writer, global_steps_counter, args)
evaluate_thread = threading.Thread(
target=lambda: evaluator.run(sess, coord))
evaluate_thread.start()
coord.join(worker_threads)
def evaluateClassifier(classifierFile,cf,k):
evaluator = Evaluate()
evaluator.buildClassifier("posFile","negFolder",classifierFile)
collFiles = []
for j in range(k):
fn = cf+str(j)+'.txt'
f = codecs.open(fn, encoding='utf-8')
ftext = f.read()
o = myObj()
o.text = ftext
collFiles.append(o)
res = evaluator.evaluateFC(collFiles)
f = open(cf+'evaluationRes_Classf.txt','w')
f.write('\n'.join([str(r) for r in res]))
f.close()
print sum(res)
def __call__(self):
all_counts = defaultdict(dict)
gold = sorted(Reader(open(self.gold)))
for path in self.systems:
system = sorted(Reader(open(path)))
for match, per_doc, overall in Evaluate.count_all(system, gold):
all_counts[match][path] = (per_doc, overall)
results = [{'sys1': sys1, 'sys2': sys2,
'match': match,
'stats': self.significance(match_counts[sys1], match_counts[sys2])}
for sys1, sys2 in itertools.combinations(self.systems, 2)
for match, match_counts in sorted(all_counts.iteritems(),
key=lambda (k, v): MATCHES.index(k))]
return self.fmt(results, self.metrics)
def compute_scores(raw_data_dir=FLAGS.raw_data, data_dir=FLAGS.data_dir,
dataset=FLAGS.dataset, save_recommendation=FLAGS.saverec,
train_dir=FLAGS.train_dir, test=FLAGS.test):
from evaluate import Evaluation as Evaluate
evaluation = Evaluate(raw_data_dir, test=test)
R = recommend(evaluation.get_uids(), data_dir=data_dir)
evaluation.eval_on(R)
scores_self, scores_ex = evaluation.get_scores()
mylog("====evaluation scores (NDCG, RECALL, PRECISION, MAP) @ 2,5,10,20,30====")
mylog("METRIC_FORMAT (self): {}".format(scores_self))
mylog("METRIC_FORMAT (ex ): {}".format(scores_ex))
if save_recommendation:
name_inds = os.path.join(train_dir, "indices.npy")
np.save(name_inds, rec)
def walk_proximity(self,
trained=True,
num_walks=100,
walk_length=40,
workers=5):
if trained:
return np.loadtxt(self.walk_structure_embedding)
walk_structure = utils.walk_proximity(self.graph.adj,
num_walks,
walk_length,
workers=workers)
print('游走已完成...')
loss = Evaluate(10).loss()
auto_encoder = SparseAE(self.args, walk_structure, loss,
self.walk_structure_embedding)
embedding = auto_encoder.train(parallel=False)
return embedding
def main():
url = "https://race.netkeiba.com/?pid=race_old&id=n201908050411"
html = requests.get(url)
soup = BeautifulSoup(html.content, 'lxml')
race_name, distance = Get_Race_Info(soup)
print(race_name)
link_list, horse_list = Get_Link_List(soup)
#print(link_list)
for link_url, horse_name in zip(link_list, horse_list):
df = Scraping(link_url)
print(horse_name)
#print(df)
ave_list = Evaluate(df, distance)
print(ave_list)
def main():
prog = "python -m allennlp.run"
subcommand_overrides = {}
# pylint: disable=dangerous-default-value
parser = argparse.ArgumentParser(description="Run AllenNLP",
usage='%(prog)s',
prog=prog)
print(parser)
subparsers = parser.add_subparsers(title='Commands', metavar='')
subcommands = {
# Default commands
"train": Train(),
"evaluate": Evaluate(),
"evaluate_mlqa": Evaluate_MLQA(),
"make-vocab": MakeVocab(),
"fine-tune": FineTune(),
# Superseded by overrides
**subcommand_overrides
}
for name, subcommand in subcommands.items():
subparser = subcommand.add_subparser(name, subparsers)
subparser.add_argument('--include-package',
type=str,
action='append',
default=[],
help='additional packages to include')
args = parser.parse_args()
# If a subparser is triggered, it adds its work as `args.func`.
# So if no such attribute has been added, no subparser was triggered,
# so give the user some help.
if 'func' in dir(args):
# Import any additional modules needed (to register custom classes).
for package_name in args.include_package:
import_submodules(package_name)
args.func(args)
else:
parser.print_help()
def main(out_dir, input_file, input_plus, input_minus, fa_file, keep_temp,
window, name, model, rst, threshold, penality, DB_file):
if not os.path.exists(out_dir):
os.makedirs(out_dir)
out_dir = out_dir + '/' + name
####Generate sliding windlows
Generate_windows(out_dir, input_file, input_plus, input_minus, fa_file,
keep_temp, window, name)
data_dir = out_dir + '/data'
data_files = glob.glob(data_dir + "/*")
for data in data_files:
if 'wig' in data:
continue
baseName = data.split('/')[-1]
Evaluate(model, out_dir, rst, window, baseName, keep_temp)
Scan_Forward(baseName, threshold, penality, out_dir)
Scan_Backward(baseName, threshold, penality, out_dir)
if (keep_temp != 'yes'):
predict_file = out_dir + '/predict/' + baseName + '.txt'
os.system('rm %s' % predict_file)
Postprocess(DB_file, baseName, threshold, penality, out_dir)
if (keep_temp != 'yes'):
forward_file = out_dir + "/maxSum/%s.forward.%d.%d.txt" % (
baseName, threshold, penality)
backward_file = out_dir + "/maxSum/%s.backward.%d.%d.txt" % (
baseName, threshold, penality)
os.system('rm %s %s' % (forward_file, backward_file))
out_file = '%s/%s.predicted.txt' % (out_dir, name)
ww = open(out_file, 'w')
ww.write('predicted_pasid\tdb_diff\tdb_pasid\tscore\n')
ww.close()
os.system('cat %s/maxSum/*bidirection* >>%s' % (out_dir, out_file))
if (keep_temp != 'yes'):
os.system('rm -rf %s/data %s/predict %s/maxSum' %
(out_dir, out_dir, out_dir))
print("Job Done!")
def __call__(self):
all_counts = defaultdict(dict)
gold = sorted(Reader(open(self.gold)))
for path in self.systems:
system = sorted(Reader(open(path)))
for match, per_doc, overall in Evaluate.count_all(system, gold):
all_counts[match][path] = (per_doc, overall)
results = [
{
'sys1': sys1,
'sys2': sys2,
'match': match,
'stats': self.significance(match_counts[sys1],
match_counts[sys2])
} for sys1, sys2 in itertools.combinations(self.systems, 2)
for match, match_counts in sorted(
all_counts.iteritems(), key=lambda (k, v): MATCHES.index(k))
]
return self.fmt(results, self.metrics)
def _train(self, fold_n, x_trn, y_trn, x_val, y_val, num_class=2):
# 初始化模型
model, emb = self.create_model(num_class)
opt = Adam(0.01)
model.compile(optimizer=opt, loss=categorical_crossentropy)
patient, best_score = 0, 0
best_embedding = None
for epoch in range(2000):
generator = utils.batch_iter(x_trn, self.batch_size)
for index in generator:
if self.types == 'classes':
model.train_on_batch([x_trn[index]],
np.eye(num_class)[y_trn[index]])
if self.types == 'link':
vi, vj = x_trn[index][:, 0], x_trn[index][:, 1]
model.train_on_batch(
[vi, vj],
np.eye(num_class)[y_trn[index].reshape(-1).astype(
int)])
if self.types == 'classes':
y_val_pred = np.argmax(model.predict([x_val]), -1)
micro, macro = Evaluate.f1(y_val, y_val_pred)
print('fold_{}:,{},{}'.format(fold_n, micro, macro))
score = micro + macro
if self.types == 'link':
y_val_pred = np.argmax(
model.predict([x_val[:, 0], x_val[:, 1]]), -1)
score = roc_auc_score(y_val, y_val_pred)
print('fold_{}:,{},{}'.format(fold_n, score, best_score))
if score > best_score:
patient = 0
best_score = score
best_embedding = emb.get_weights()[0]
patient += 1
if patient >= 50:
break
return best_embedding
def __call__(self,
number_of_iterations=2,
learning_rate=0.005,
embedding_size=300,
hidden_size=100,
batch_size=100):
print("Starting 'Image Retrieval' in 'GRU' mode with '" +
self.difficulty + "' data")
self.model_full_path = self.model_path + "/" + self.model_name + "_" + self.timestamp + "_" + str(
learning_rate) + "_" + str(embedding_size) + ".pty"
self.output_file_name = self.output_path + "/" + self.model_name + "_" + self.timestamp + "_" + str(
learning_rate) + "_" + str(embedding_size) + ".csv"
self.number_of_iterations = number_of_iterations
self.learning_rate = learning_rate
self.embedding_size = embedding_size
self.hidden_size = hidden_size
self.batch_size = batch_size
self.model = GRU(self.nwords, self.embedding_size,
self.image_feature_size, self.output_vector_size,
self.hidden_size, self.batch_size)
self.criterion = nn.CrossEntropyLoss()
self.evaluate = Evaluate(self.model, self.img_features, self.minibatch,
self.preprocess, self.image_feature_size,
self.output_vector_size)
print(self.model)
self.optimizer = optim.Adam(self.model.parameters(),
lr=self.learning_rate)
self.train_loss_values = []
self.magic()
self.save_model()
self.save_data()
def test(RL):
env = envR(show=False)
path, cost, density, num_find_target, opt_cost = [], [], [], 0, []
evaluate = Evaluate(rows=10, cols=10)
train = False
succ = 0
print("****************************************************")
for episode in range(100):
pre_maps = env.reset()
step = 0
evaluate.set_start(start_pos=env.agent)
evaluate.set_goals(real_pos=env.maze.food_pos[0],
fake_pos=env.maze.food_pos[1])
# print("****************************************************")
# print("EPISODE ", episode)
# start_test = time.time()
for step in range(100):
action = RL.choose_action(str(pre_maps), train)
reward, done, action_ = env.step(action)
path.append(action_)
step += 1
if done:
succ += 1
cost, density, num_find_target, opt_cost = evaluation(
evaluate, cost, density, num_find_target, opt_cost, path)
path = []
break
pre_maps = env.get_maps()
print('This is ', episode, 'cost:', step, 'succ', succ)
print('average cost:', np.mean(cost), ' average density:',
np.mean(density), ' deceptive extent:', num_find_target / succ)
print('optimal cost:', np.mean(opt_cost))
print()
class Train():
def __init__(self, difficulty):
self.data_path = "../data"
self.model_path = "../models"
self.output_path = "../outputs"
self.difficulty = difficulty
self.timestamp = str(int(time.time()))
self.model_name = "regression_" + self.difficulty
self.data = Data(difficulty=self.difficulty, data_path=self.data_path)
(self.img_features, self.w2i, self.i2w, self.nwords, self.UNK,
self.PAD) = self.data()
self.train = list(self.data.get_train_data())
self.dev = list(self.data.get_validation_data())
self.test = list(self.data.get_test_data())
self.image_feature_size = 2048
self.output_vector_size = 10
def __call__(self,
number_of_iterations=2,
learning_rate=0.005,
embedding_size=300):
print("Starting 'Image Retrieval' in 'Regression' mode with '" +
self.difficulty + "' data")
self.model_full_path = self.model_path + "/" + self.model_name + "_" + self.timestamp + "_" + str(
learning_rate) + "_" + str(embedding_size) + ".pty"
self.output_file_name = self.output_path + "/" + self.model_name + "_" + self.timestamp + "_" + str(
learning_rate) + "_" + str(embedding_size) + ".csv"
self.number_of_iterations = number_of_iterations
self.learning_rate = learning_rate
self.embedding_size = embedding_size
self.model = Regression(self.nwords, self.embedding_size,
self.image_feature_size,
self.output_vector_size)
self.criterion = nn.MSELoss()
self.evaluate = Evaluate(self.model, self.img_features, self.minibatch,
self.preprocess, self.image_feature_size)
print(self.model)
self.optimizer = optim.Adam(self.model.parameters(),
lr=self.learning_rate)
self.train_loss_values = []
self.dev_loss_values = []
self.test_loss_values = []
self.magic()
self.save_model()
self.save_data()
def minibatch(self, data, batch_size=50):
for i in range(0, len(data), batch_size):
yield data[i:i + batch_size]
def preprocess(self, batch):
"""Helper function for functional batches"""
correct_indexes = [observation[2] for observation in batch]
img_ids = [observation[1] for observation in batch]
text_features = [observation[0] for observation in batch]
#Add Padding to max len of sentence in batch
max_length = max(map(len, text_features))
text_features = [
txt + [self.PAD] * (max_length - len(txt)) for txt in text_features
]
#return in "stacked" format
return text_features, img_ids, correct_indexes
def magic(self):
for ITER in range(self.number_of_iterations):
random.shuffle(self.train)
train_loss = 0.0
start = time.time()
for iteration, batch in enumerate(self.minibatch(self.train)):
#Outputs matrices of batch size
text_features, h5_ids, correct_index = self.preprocess(batch)
lookup_text_tensor = Variable(torch.LongTensor([text_features
])).squeeze()
target = np.empty([len(batch), self.image_feature_size])
for obs, img_ids in enumerate(h5_ids):
target[obs] = self.img_features[img_ids[
correct_index[obs]]]
target = Variable(
torch.from_numpy(target).type(torch.FloatTensor))
#Run model and calculate loss
prediction = self.model(lookup_text_tensor)
loss = self.criterion(prediction, target)
train_loss += loss.data[0]
self.optimizer.zero_grad()
self.model.zero_grad()
loss.backward()
self.optimizer.step()
#if iteration % verbosity_interval == 0:
# print("ITERATION %r: %r: train loss/sent=%.4f, time=%.2fs" % (ITER+1, iteration, train_loss/(iteration + 1), time.time() - start))
print(
"ITERATION %r: train loss/sent=%.4f, time=%.2fs" %
(ITER + 1, train_loss / len(self.train), time.time() - start))
#print("Score on training", evaluate(train))
#print("Score on development", evaluate(dev))
self.train_loss_values.append(train_loss / len(self.train))
self.dev_loss_values.append(self.evaluate.calculate_loss(self.dev))
self.test_loss_values.append(
self.evaluate.calculate_loss(self.test))
def save_model(self):
#Save model
torch.save(self.model, self.model_full_path)
print("Saved model has test score", self.evaluate(self.test))
def plot(self):
plt.plot(self.train_loss_values, label="Train loss")
plt.plot(self.dev_loss_values, label="Validation loss")
plt.plot(self.test_loss_values, label="Test loss")
plt.legend(loc='best')
plt.xlabel("Epochs")
plt.ylabel("Loss")
plt.title(self.model_name +
" - has loss with lr = %.4f, embedding size = %r" %
(self.learning_rate, self.embedding_size))
plt.show()
def save_data(self):
file = open(self.output_file_name, "w")
file.write(", ".join(map(str, self.train_loss_values)))
file.write("\n")
file.write(", ".join(map(str, self.dev_loss_values)))
file.write("\n")
file.write(", ".join(map(str, self.test_loss_values)))
file.write("\n")
file.write(str(self.evaluate(self.dev)))
file.write("\n")
file.close()
class DataHandler:
evaluator = Evaluate() # 클래스 변수 : 클래스의 모든 인스턴스들이 공유하는 변수 (연산기)
# 객체합성 (선생님의 캡슐화 범위)
# class method : 전역함수처럼 사용
@classmethod # 인스턴스 메소드로 정의해도 문제는 없다
def GetRawdataInDic(cls, filename):
rawdata = {}
with open(filename, 'rb') as f:
while 1:
try:
data = pickle.load(f)
except EOFError:
break
rawdata.update(data)
return rawdata
def __init__(self, filename, clsname):
self.rawdata = DataHandler.GetRawdataInDic(filename)
self.clsname = clsname
self.cache = {} # 연산한 값을 저장해 두는 저장소
#(필요할 때 연산하되 이미 연산된 값이면 연산 없이 저장된 값을 반환)
def get_scores(self): # cache기법 사용
if 'scores' not in self.cache:
self.cache['scores'] = list(self.rawdata.values())
return self.cache.get('scores')
def get_average(self): # cache 사용
if 'average' not in self.cache:
self.cache['average'] = self.evaluator.average(self.get_scores())
return self.cache.get('average')
def get_variance(self): # cache 사용
if 'variance' not in self.cache:
vari = round(
self.evaluator.variance(self.get_scores(), self.get_average()),
1)
self.cache['variance'] = vari
return self.cache.get('variance')
def get_standard_deviation(self):
if 'standard_deviation' not in self.cache:
std_dev = round(math.sqrt(self.get_variance()), 1)
self.cache['standard_deviation'] = std_dev
return self.cache.get('standard_deviation')
def GetEvaluation(self):
print('*' * 50)
print("%s 반 성적 분석 결과" % self.clsname)
print("{0}반의 평균은 {1}점이고 분산은 {2}이며,따라서 표준편차는{3}이다".\
format(self.clsname, self.get_average(), self.get_variance()\
, self.get_standard_deviation()))
print('*' * 50)
print("%s 반 종합 평가" % self.clsname)
print('*' * 50)
self.evaluateClass()
def evaluateClass(self):
avrg = self.get_average()
std_dev = self.get_standard_deviation()
if avrg < 50 and std_dev > 20:
print("성적이 너무 저조하고 학생들의 실력 차이가 너무 크다.")
elif avrg > 50 and std_dev > 20:
print("성적은 평균이상이지만 학생들 실력 차이가 크다. 주의 요망!")
elif avrg < 50 and std_dev < 20:
print("학생들간 실력차는 나지 않으나 성적이 너무 저조하다. 주의 요망!")
elif avrg > 50 and std_dev < 20:
print("성적도 평균 이상이고 학생들의 실력차도 크지 않다.")
# def who_is_highest(self):
# h_score= max(list(self.rawdata.values()))
# for k, v in self.rawdata.items():
# if v == h_score:
# return k
# def get_highest_score(self):
# return max(list(self.rawdata.values()))
# 선생님 코드
def who_ist_highest(self):
if 'highest' not in self.cache:
self.cache['highest'] = reduce(
lambda a, b: a
if self.rawdata.get(a) > self.rawdata.get(b) else b,
self.rawdata.keys())
return self.cache.get('highest')
def get_highest_score(self):
return self.rawdata[self.who_ist_highest()]
def who_is_lowest(self):
if 'lowest' not in self.cache:
self.cache['lowest'] = reduce(
lambda a, b: a
if self.rawdata.get(a) < self.rawdata.get(b) else b,
self.rawdata.keys())
return self.cache.get('lowest')
def get_lowest_score(self):
return self.rawdata[self.who_is_lowest()]
if fit:
print("Fit tokenizer...")
tokenizer = text.Tokenizer(nb_words=vocab_size)
tokenizer.fit_on_texts(text_generator())
print("Save tokenizer...")
f = open(tokenizer_fname, "wb")
cPickle.dump(tokenizer, f, protocol=cPickle.HIGHEST_PROTOCOL)
f.close()
else:
print('Load tokenizer...')
f = open(tokenizer_fname, "rb")
tokenizer = cPickle.load(f)
f.close()
evaluator = Evaluate(tokenizer, words, context, average_scores)
sampling_table = sequence.make_sampling_table(vocab_size)
for e in range(nb_epoch):
print('-'*40)
print('Epoch', e)
print('-'*40)
progbar = Progbar(tokenizer.document_count)
samples_seen = 0
losses = []
batch_loss = []
for i, seq in enumerate(tokenizer.texts_to_sequences_generator(text_generator())):
# get skipgram couples for one text in the dataset
couples, labels = skipgrams_l2c_fast(seq, vocab_size, num_senses =num_senses, window_size=4, negative_samples=1., sampling_table=sampling_table)
if couples:
#posFiles = ['pos-FSU.txt','pos-Hagupit.txt','pos-LAFire.txt','pos-AirAsia.txt']
posFiles = ['pos-FSU.txt','pos-Hagupit.txt','pos-AirAsia.txt','pos-sydneyseige.txt','pos-Charlie.txt']
#negFolder = 'neg'
negFiles = ['neg-FSU.txt','neg-Hagupit.txt','neg-AirAsia.txt','neg-sydneyseige.txt','neg-Charlie.txt']
'''
seedsFiles=['seedsURLs_z_501.txt','seedsURLs_z_540.txt']
#posFiles = ['pos-FSU.txt','pos-Hagupit.txt','pos-AirAsia.txt']
#negFiles = ['neg-FSU.txt','neg-Hagupit.txt','neg-AirAsia.txt']
posFiles = ['pos-Charlie.txt','pos-sydneyseige.txt']
negFiles = ['neg-Charlie.txt','neg-sydneyseige.txt']
'''
evaluator = Evaluate()
#for i in range(3):
noK = 10
th = 0.75
i=3
posFile = posFiles[i]
negFile = negFiles[i]
#modelFile = modelFile +"-"+str(i)+".txt"
#classifierFileName = 'classifier'+posFile.split(".")[0].split('-')[1]+".p"
vsmClassifierFileName = 'classifierVSM-'+posFile.split(".")[0].split('-')[1]+".p"
#evaluator.buildClassifier(posFile,negFolder,classifierFileName)
#evaluator.buildClassifier(posFile,negFile,classifierFileName)
evaluator.buildVSMClassifier(posFile, vsmClassifierFileName,th,noK)
v = 0
def doEvaluate():
eva = Evaluate()
eva.eval()
def main():
args = parse_args()
model_dir = args.model_dir
"""LOAD CONFIG FILE"""
config_files = glob.glob(os.path.join(model_dir, '*.ini'))
assert len(config_files) == 1, 'Put only one config file in the directory'
config_file = config_files[0]
config = configparser.ConfigParser()
config.read(config_file)
"""LOGGER"""
logger = getLogger(__name__)
logger.setLevel(logging.INFO)
formatter = logging.Formatter('[%(asctime)s] %(message)s')
sh = logging.StreamHandler()
sh.setLevel(logging.INFO)
sh.setFormatter(formatter)
logger.addHandler(sh)
log_file = model_dir + 'log.txt'
fh = logging.FileHandler(log_file)
fh.setLevel(logging.INFO)
fh.setFormatter(formatter)
logger.addHandler(fh)
logger.info('[Training start] logging to {}'.format(log_file))
"""PARAMATER"""
embed_size = int(config['Parameter']['embed_size'])
hidden_size = int(config['Parameter']['hidden_size'])
dropout_ratio = float(config['Parameter']['dropout'])
weight_decay = float(config['Parameter']['weight_decay'])
gradclip = float(config['Parameter']['gradclip'])
vocab_type = config['Parameter']['vocab_type']
vocab_size = int(config['Parameter']['vocab_size'])
"""TRINING DETAIL"""
gpu_id = args.gpu
n_epoch = args.epoch
batch_size = args.batch
interval = args.interval
"""DATASET"""
train_src_file = config['Dataset']['train_src_file']
train_trg_file = config['Dataset']['train_trg_file']
valid_src_file = config['Dataset']['valid_src_file']
valid_trg_file = config['Dataset']['valid_trg_file']
test_src_file = config['Dataset']['test_src_file']
correct_txt_file = config['Dataset']['correct_txt_file']
train_data_size = dataset.data_size(train_trg_file)
valid_data_size = dataset.data_size(valid_trg_file)
logger.info('train size: {0}, valid size: {1}'.format(train_data_size, valid_data_size))
if vocab_type == 'normal':
init_vocab = {'<unk>': 0, '<s>': 1, '</s>': 2, '<eod>': 3}
vocab = dataset.VocabNormal()
vocab.make_vocab(train_src_file, train_trg_file, init_vocab, vocab_size, freq=0)
dataset.save_pickle(model_dir + 'src_vocab.pkl', vocab.src_vocab)
dataset.save_pickle(model_dir + 'trg_vocab.pkl', vocab.trg_vocab)
sos = vocab.src_vocab['<s>']
eos = vocab.src_vocab['</s>']
eod = vocab.src_vocab['<eod>']
elif vocab_type == 'subword':
vocab = dataset.VocabSubword()
if os.path.isfile(model_dir + 'src_vocab.sub.model') and os.path.isfile(model_dir + 'trg_vocab.sub.model'):
vocab.load_vocab(model_dir + 'src_vocab.sub.model', model_dir + 'trg_vocab.sub.model')
else:
vocab.make_vocab(train_trg_file + '.sub', train_trg_file + '.sub', model_dir, vocab_size)
sos = vocab.src_vocab.PieceToId('<s>')
eos = vocab.src_vocab.PieceToId('</s>')
eod = vocab.src_vocab.PieceToId('<eod>')
src_vocab_size = len(vocab.src_vocab)
trg_vocab_size = len(vocab.trg_vocab)
logger.info('src_vocab size: {}, trg_vocab size: {}'.format(src_vocab_size, trg_vocab_size))
train_iter = iterator.Iterator(train_src_file, train_trg_file, batch_size, sort=True, shuffle=True)
# train_iter = iterator.Iterator(train_src_file, train_trg_file, batch_size, sort=False, shuffle=False)
valid_iter = iterator.Iterator(valid_src_file, valid_trg_file, batch_size, sort=False, shuffle=False)
evaluater = Evaluate(correct_txt_file)
test_iter = iterator.Iterator(test_src_file, test_src_file, batch_size, sort=False, shuffle=False)
"""MODEL"""
model = HiSeq2SeqModel(
WordEnc(src_vocab_size, embed_size, hidden_size, dropout_ratio),
WordDec(trg_vocab_size, embed_size, hidden_size, dropout_ratio),
SentEnc(hidden_size, dropout_ratio),
SentDec(hidden_size, dropout_ratio),
sos, eos, eod)
"""OPTIMIZER"""
optimizer = chainer.optimizers.Adam()
optimizer.setup(model)
optimizer.add_hook(chainer.optimizer.GradientClipping(gradclip))
optimizer.add_hook(chainer.optimizer.WeightDecay(weight_decay))
"""GPU"""
if gpu_id >= 0:
logger.info('Use GPU')
chainer.cuda.get_device_from_id(gpu_id).use()
model.to_gpu()
"""TRAIN"""
sum_loss = 0
loss_dic = {}
for epoch in range(1, n_epoch + 1):
for i, batch in enumerate(train_iter.generate(), start=1):
print(batch)
exit()
batch = vocab.convert2label(batch)
data = converter.convert(batch, gpu_id)
loss = optimizer.target(*data)
sum_loss += loss.data
optimizer.target.cleargrads()
loss.backward()
optimizer.update()
if i % interval == 0:
logger.info('E{} ## iteration:{}, loss:{}'.format(epoch, i, sum_loss))
sum_loss = 0
chainer.serializers.save_npz(model_dir + 'model_epoch_{}.npz'.format(epoch), model)
# chainer.serializers.save_npz(model_dir + 'optimizer_epoch{0}.npz'.format(epoch), optimizer)
"""EVALUATE"""
valid_loss = 0
for batch in valid_iter.generate():
batch = vocab.convert2label(batch)
data = converter.convert(batch, gpu_id)
with chainer.no_backprop_mode(), chainer.using_config('train', False):
valid_loss += optimizer.target(*data).data
logger.info('E{} ## val loss:{}'.format(epoch, valid_loss))
loss_dic[epoch] = valid_loss
"""TEST"""
output = []
for batch in test_iter.generate():
# batch: (articlesのリスト, abstracts_sosのリスト, abstracts_eosのリスト)タプル
batch = vocab.convert2label(batch)
data = converter.convert(batch, gpu_id)
"""
out: [(sent, attn), (sent, attn), ...] <-バッチサイズ
sent: decodeされた文のリスト
attn: 各文のdecode時のattentionのリスト
"""
with chainer.no_backprop_mode(), chainer.using_config('train', False):
out = model.generate(data[0], data[3])
output.extend(out)
res_decode = []
res_attn = []
for o in output:
sent, attn = o
sentence = dataset.to_list(sent)
sentence = dataset.eod_truncate(sentence, eod)
sent_num = len(sentence)
sentence = [dataset.eos_truncate(s, eos) for s in sentence]
sentence = [vocab.label2word(s) for s in sentence]
sentence = dataset.join_sentences(sentence)
res_decode.append(sentence)
attn = np.sum(np.array(attn[:sent_num]), axis=0) / sent_num
res_attn.append(attn)
rank_list = evaluater.rank(res_attn)
single = evaluater.single(rank_list)
multiple = evaluater.multiple(rank_list)
logger.info('E{} ## precision'.format(epoch))
logger.info('single: {} | {}'.format(single[0], single[1]))
logger.info('multi : {} | {}'.format(multiple[0], multiple[1]))
with open(model_dir + 'model_epoch_{}.hypo'.format(epoch), 'w')as f:
[f.write(r + '\n') for r in res_decode]
with open(model_dir + 'model_epoch_{}.attn'.format(epoch), 'w')as f:
[f.write('{}\n'.format(r)) for r in res_attn]
with open(model_dir + 'model_epoch_{}.prec'.format(epoch), 'w')as f:
f.write('single\n')
f.write(single[0] + '\n')
f.write(single[1] + '\n')
f.write('multiple\n')
f.write(multiple[0] + '\n')
f.write(multiple[1] + '\n')
"""MODEL SAVE"""
best_epoch = min(loss_dic, key=(lambda x: loss_dic[x]))
logger.info('best_epoch:{0}'.format(best_epoch))
chainer.serializers.save_npz(model_dir + 'best_model.npz', model)
def test(test_filename, time_now, title, **kwargs):
# 超参数设置
network_name = get_value_or_default(kwargs, 'network', default='LSTM')
affect = get_value_or_default(kwargs, 'affect', default=30)
filename = test_filename
column = get_value_or_default(kwargs, 'column', default='ClPr')
index_col = 'TrdDt'
batch_size = 1
plot_name = get_value_or_default(kwargs, 'plot_name', default=['fig1', ])
# 加载数据
data = Action.generate_df(
filename,
column,
index_col,
affect
)
data_loader = DataLoader(data['dataset'], batch_size=batch_size, shuffle=False)
net = torch.load('save/{}.pt'.format(network_name))
predict = list()
for tx, ty in data_loader:
output = net(tx.reshape(1, batch_size, affect))
output = output.reshape(1).detach()
predict.append(float(output) * data['std'] + data['mean'])
plt1 = Plot(1, time_now, network_name)
plt1.plot(data['index'], data['real_data'][affect:], 'real data')
plt1.plot(data['index'], predict, 'predict data')
plt1.title(title, zh=True)
plt1.xylabel('Datetime', 'price')
plt1.save(plot_name[0])
# Plot.show()
Plot.cla()
evaluator = Evaluate(title, data['real_data'][affect:], predict)
logger = Logger('test.log')
basic_info = 'tested {}.'.format(network_name)
logger.set_log(basic_info,
t=time_now,
filename=filename,
column=column,
affect_days=affect,
network=net,
plot_name=plot_name,
MSELoss=evaluator.MSELoss(),
DA=evaluator.DA(),
Theil=evaluator.Theil_U(),
L1Loss=evaluator.L1Loss(),
Customize=evaluator.customize(),
title=title,
MAPE=evaluator.MAPE(),
R=evaluator.R()
)
f_out = open('log/{}.txt'.format(title), 'w')
print('{} = {}'.format('time', time_now),
'{} = {}'.format('MSELoss', evaluator.MSELoss()),
'{} = {}'.format('DA', evaluator.DA()),
'{} = {}'.format('Theil_U', evaluator.Theil_U()),
'{} = {}'.format('L1Loss', evaluator.L1Loss()),
'{} = {}'.format('MAPE', evaluator.MAPE()),
'{} = {}'.format('R', evaluator.R()),
file=f_out,
sep='\n')
f_out.close()
return evaluator
def evaluate():
eval = Evaluate(self.arch, self.params, self.train_dir)
#!/usr/bin/env python
from evaluate import Evaluate
if __name__ == '__main__':
ev = Evaluate(timesteps=2000)
exp = './'
filename = exp + '/output_truth.txt'
truth = ev.get_data(filename)
filename = exp + '/output_bckgd.txt'
analy = ev.get_data(filename)
# print('file :',filename)
# print('truth : ',truth.shape,' analysis ',analy.shape)
ev.plot_state(truth, analy)
def main(train_file_to_use, test_file_to_use, test_type, features_combination_list, lamda, comp):
# for perm in itertools.combinations(features_combination_list_sub, 4):
# features_combination_list.append(list(perm))
# start all combination of features
for features_combination in features_combination_list:
print('{}: Start creating MEMM for features : {}'.format(time.asctime(time.localtime(time.time())),
features_combination))
logging.info('{}: Start creating MEMM for features : {}'.format(time.asctime(time.localtime(time.time())),
features_combination))
train_start_time = time.time()
memm_class = MEMM(directory, train_file_to_use, features_combination)
logging.info('{}: Finish MEMM for features : {}'.format(time.asctime(time.localtime(time.time())),
features_combination))
print('{}: Finish MEMM for features : {}'.format(time.asctime(time.localtime(time.time())),
features_combination))
print('{}: Start gradient for features : {} and lambda: {}'.
format(time.asctime(time.localtime(time.time())), features_combination, lamda))
logging.info('{}: Start gradient for features : {} and lambda: {}'.
format(time.asctime(time.localtime(time.time())), features_combination, lamda))
gradient_class = Gradient(model=memm_class, lambda_value=lamda)
gradient_result = gradient_class.gradient_descent()
train_run_time = (time.time() - train_start_time) / 60.0
print('{}: Finish gradient for features : {} and lambda: {}. run time: {}'.
format(time.asctime(time.localtime(time.time())), features_combination, lamda, train_run_time))
logging.info('{}: Finish gradient for features : {} and lambda: {}. run time: {}'.
format(time.asctime(time.localtime(time.time())), features_combination, lamda, train_run_time))
weights = gradient_result.x
# np.savetxt(gradient_file, weights, delimiter=",")
viterbi_start_time = time.time()
print('{}: Start viterbi'.format((time.asctime(time.localtime(time.time())))))
viterbi_class = viterbi(memm_class, data_file=test_file_to_use, w=weights)
viterbi_result = viterbi_class.viterbi_all_data
viterbi_run_time = (time.time() - viterbi_start_time) / 60.0
print('{}: Finish viterbi. run time: {}'.format((time.asctime(time.localtime(time.time()))), viterbi_run_time))
logging.info('{}: Finish viterbi. run time: {}'.format((time.asctime(time.localtime(time.time()))),
viterbi_run_time))
write_file_name = datetime.now().strftime(directory + 'file_results/result_MEMM_most_common_tags_' + test_type +
'%d_%m_%Y_%H_%M.wtag')
confusion_file_name = datetime.now().strftime(directory + 'confusion_files/CM_MEMM_most_common_tags_' + test_type +
'%d_%m_%Y_%H_%M.xls')
evaluate_class = Evaluate(memm_class, test_file_to_use, viterbi_result, write_file_name,
confusion_file_name, comp=comp)
if not comp:
word_results_dictionary = evaluate_class.run()
if comp:
evaluate_class.write_result_doc()
logging.info('{}: The model hyper parameters: \n lambda:{} \n test file: {} \n train file: {}'
.format(time.asctime(time.localtime(time.time())), lamda, test_file_to_use, train_file_to_use))
logging.info('{}: Related results files are: \n {} \n {}'.
format(time.asctime(time.localtime(time.time())), write_file_name, confusion_file_name))
# print(word_results_dictionary)
summary_file_name = '{0}analysis/summary_{1}_{2.day}_{2.month}_{2.year}_{2.hour}_{2.minute}.csv' \
.format(directory, test_type, datetime.now())
evaluate_class.create_summary_file(lamda, features_combination, test_file_to_use, train_file_to_use,
summary_file_name, gradient_class.file_name, comp)
logging.info('{}: Following Evaluation results for features {}'.
format(time.asctime(time.localtime(time.time())), features_combination))
if not comp:
logging.info('{}: Evaluation results are: \n {} \n'.format(time.asctime(time.localtime(time.time())),
word_results_dictionary))
logging.info('-----------------------------------------------------------------------------------')
def load_model():
s = Summarizer()
e = Evaluate()
return s, e
Created on Tue Jun 16 17:57:09 2015
@author: Paco
"""
from utils import Utils
from evaluate import Evaluate
from metrics import Metrics
# Load data
u = Utils()
train_hard = u.load_matrix('data/data_train_difficile.mat')
#generate pairs
pairs_idx, pairs_label = u.generate_pairs(train_hard['label'], 1000, 0.1)
# Calculate distance
m = Metrics()
dist = m.braycurtis_dist(train_hard['X'], pairs_idx)
# Evaluate model
e = Evaluate()
e.evaluation(pairs_label,dist)
# display results
e.display_roc()
e.hard_score()
# Evaluate test dataset and save it
test_hard = u.load_matrix('data/data_test_difficile.mat')
dist_test = m.braycurtis_dist(test_hard['X'], test_hard['pairs'])
u.save_test(dist_test,filetxt='soumission_dur.txt')
def trainer(epochs, model, optimizer, scheduler, train_dataloader,
test_dataloader, batch_train, batch_test, device):
max_grad_norm = 1.0
train_loss_set = []
for e in trange(epochs, desc="Epoch"):
while gc.collect() > 0:
pass
# Training
# Set our model to training mode (as opposed to evaluation mode)
model.train()
# if e > 8:
# model.freeze_bert()
# Tracking variables
tr_loss = 0
nb_tr_examples, nb_tr_steps = 0, 0
# Train the data for one epoch
for step, batch in enumerate(train_dataloader):
# Add batch to GPU
batch = tuple(t.to(device) for t in batch)
# Unpack the inputs from our dataloader
b_input_ids, b_input_mask, b_adj, b_adj_mwe, b_labels, b_target_idx, _ = batch
# Clear out the gradients (by default they accumulate)
optimizer.zero_grad()
# Forward pass
### For BERT + GCN and MWE
loss = model(b_input_ids.to(device), adj=b_adj, adj_mwe=b_adj_mwe ,attention_mask=b_input_mask.to(device), \
labels=b_labels, batch=batch_train, target_token_idx=b_target_idx.to(device))
train_loss_set.append(loss.item())
# Backward pass
loss.backward(retain_graph=True)
torch.nn.utils.clip_grad_norm_(model.parameters(), max_grad_norm)
# Update parameters and take a step using the computed gradient
optimizer.step()
scheduler.step()
optimizer.zero_grad()
# Update tracking variables
tr_loss += loss.item()
nb_tr_examples += b_input_ids.size(0)
nb_tr_steps += 1
print("Train loss: {}".format(tr_loss / nb_tr_steps))
# Validation
# Put model in evaluation mode to evaluate loss on the validation set
model.eval()
all_preds = torch.FloatTensor()
all_labels = torch.LongTensor()
test_indices = torch.LongTensor()
# Evaluate data for one epoch
for batch in test_dataloader:
# Add batch to GPU
batch = tuple(t.to(device) for t in batch)
# Unpack the inputs from our dataloader
b_input_ids, b_input_mask, b_adj, b_adj_mwe, b_labels, b_target_idx, test_idx = batch
# Telling the model not to compute or store gradients, saving memory and speeding up validation
with torch.no_grad():
# Forward pass, calculate logit predictions
### For BERT + GCN and MWE
logits = model(b_input_ids.to(device), adj=b_adj, adj_mwe=b_adj_mwe, attention_mask=b_input_mask.to(device), \
batch=batch_test, target_token_idx=b_target_idx.to(device))
# Move logits and labels to CPU
logits = logits.detach().cpu()
label_ids = b_labels.cpu()
test_idx = test_idx.cpu()
all_preds = torch.cat([all_preds, logits])
all_labels = torch.cat([all_labels, label_ids])
test_indices = torch.cat([test_indices, test_idx])
scores = Evaluate(all_preds, all_labels)
print('scores.accuracy()={}\nscores.precision_recall_fscore()={}'.format(
scores.accuracy(), scores.precision_recall_fscore()))
return scores, all_preds, all_labels, test_indices
class Segmenter(object):
def __init__(self, hdfs_client, flags):
self.train_is_alive = False
self.hdfs_client = hdfs_client
self.flags = flags
self.data_utils = DataUtils()
def update_config(self):
config_path = os.path.join(self.flags.raw_data_path, 'config.json')
try:
with open(config_path, encoding='utf-8', mode='r') as data_file:
config_json = json.load(data_file)
if 'use_lstm' in config_json:
self.flags.use_lstm = config_json['use_lstm']
elif 'use_dynamic_rnn' in config_json:
self.flags.use_dynamic_rnn = config_json['use_dynamic_rnn']
elif 'use_bidirectional_rnn' in config_json:
self.flags.use_bidirectional_rnn = config_json[
'use_bidirectional_rnn']
elif 'vocab_drop_limit' in config_json:
self.flags.vocab_drop_limit = config_json[
'vocab_drop_limit']
elif 'batch_size' in config_json:
self.flags.batch_size = config_json['batch_size']
elif 'num_steps' in config_json:
self.flags.num_steps = config_json['num_steps']
elif 'num_layer' in config_json:
self.flags.num_layer = config_json['num_layer']
elif 'embedding_size' in config_json:
self.flags.embedding_size = config_json['embedding_size']
elif 'learning_rate' in config_json:
self.flags.learning_rate = config_json['learning_rate']
elif 'learning_rate_decay_factor' in config_json:
self.flags.learning_rate_decay_factor = config_json[
'learning_rate_decay_factor']
elif 'keep_prob' in config_json:
self.flags.keep_prob = config_json['keep_prob']
elif 'clip_norm' in config_json:
self.flags.clip_norm = config_json['clip_norm']
except:
raise Exception('ERROR: config.json content invalid')
def train(self):
self.hdfs_client.hdfs_download(
os.path.join(self.flags.input_path, 'train.txt'),
os.path.join(self.flags.datasets_path, 'train.txt'))
self.hdfs_client.hdfs_download(
os.path.join(self.flags.input_path, 'test.txt'),
os.path.join(self.flags.datasets_path, 'test.txt'))
self.data_utils.label_segment_file(
os.path.join(self.flags.datasets_path, 'train.txt'),
os.path.join(self.flags.datasets_path, 'label_train.txt'))
self.data_utils.label_segment_file(
os.path.join(self.flags.datasets_path, 'test.txt'),
os.path.join(self.flags.datasets_path, 'label_test.txt'))
self.data_utils.split_label_file(
os.path.join(self.flags.datasets_path, 'label_train.txt'),
os.path.join(self.flags.datasets_path, 'split_train.txt'))
self.data_utils.split_label_file(
os.path.join(self.flags.datasets_path, 'label_test.txt'),
os.path.join(self.flags.datasets_path, 'split_test.txt'))
words_vocab, labels_vocab = self.data_utils.create_vocabulary(
os.path.join(self.flags.datasets_path, 'split_train.txt'),
self.flags.vocab_path, self.flags.vocab_drop_limit)
train_word_ids_list, train_label_ids_list = self.data_utils.file_to_word_ids(
os.path.join(self.flags.datasets_path, 'split_train.txt'),
words_vocab, labels_vocab)
test_word_ids_list, test_label_ids_list = self.data_utils.file_to_word_ids(
os.path.join(self.flags.datasets_path, 'split_test.txt'),
words_vocab, labels_vocab)
tensorflow_utils = TensorflowUtils()
tensorflow_utils.create_record(
train_word_ids_list, train_label_ids_list,
os.path.join(self.flags.tfrecords_path, 'train.tfrecords'))
tensorflow_utils.create_record(
test_word_ids_list, test_label_ids_list,
os.path.join(self.flags.tfrecords_path, 'test.tfrecords'))
self.hdfs_client.hdfs_upload(
self.flags.vocab_path,
os.path.join(self.flags.output_path,
os.path.basename(self.flags.vocab_path)))
train = Train()
train.train()
def upload_tensorboard(self):
hdfs_tensorboard_path = os.path.join(
self.flags.output_path,
os.path.basename(os.path.normpath(self.flags.tensorboard_path)))
temp_hdfs_tensorboard_path = hdfs_tensorboard_path + '-temp'
self.hdfs_client.hdfs_upload(self.flags.tensorboard_path,
temp_hdfs_tensorboard_path)
self.hdfs_client.hdfs_delete(hdfs_tensorboard_path)
self.hdfs_client.hdfs_mv(temp_hdfs_tensorboard_path,
hdfs_tensorboard_path)
def log_monitor(self):
while (self.train_is_alive):
time.sleep(120)
self.upload_tensorboard()
def upload_model(self):
predict = Predict()
predict.saved_model_pb()
hdfs_checkpoint_path = os.path.join(
self.flags.output_path,
os.path.basename(os.path.normpath(self.flags.checkpoint_path)))
hdfs_saved_model_path = os.path.join(
self.flags.output_path,
os.path.basename(os.path.normpath(self.flags.saved_model_path)))
temp_hdfs_checkpoint_path = hdfs_checkpoint_path + '-temp'
temp_hdfs_saved_model_path = hdfs_saved_model_path + '-temp'
self.hdfs_client.hdfs_upload(self.flags.checkpoint_path,
temp_hdfs_checkpoint_path)
self.hdfs_client.hdfs_upload(self.flags.saved_model_path,
temp_hdfs_saved_model_path)
self.hdfs_client.hdfs_delete(hdfs_checkpoint_path)
self.hdfs_client.hdfs_delete(hdfs_saved_model_path)
self.hdfs_client.hdfs_mv(temp_hdfs_checkpoint_path,
hdfs_checkpoint_path)
self.hdfs_client.hdfs_mv(temp_hdfs_saved_model_path,
hdfs_saved_model_path)
def evaluate(self):
shutil.rmtree(self.flags.vocab_path)
shutil.rmtree(self.flags.checkpoint_path)
self.hdfs_client.hdfs_download(
os.path.join(self.flags.input_path,
os.path.basename(self.flags.vocab_path)),
self.flags.vocab_path)
self.hdfs_client.hdfs_download(
os.path.join(self.flags.input_path, 'test.txt'),
os.path.join(self.flags.datasets_path, 'test.txt'))
hdfs_checkpoint_path = os.path.join(
self.flags.input_path,
os.path.basename(self.flags.checkpoint_path))
self.hdfs_client.hdfs_download(hdfs_checkpoint_path,
self.flags.checkpoint_path)
self.data_utils.label_segment_file(
os.path.join(self.flags.datasets_path, 'test.txt'),
os.path.join(self.flags.datasets_path, 'label_test.txt'))
self.data_utils.split_label_file(
os.path.join(self.flags.datasets_path, 'label_test.txt'),
os.path.join(self.flags.datasets_path, 'split_test.txt'))
predict = Predict()
predict.file_predict(
os.path.join(self.flags.datasets_path, 'split_test.txt'),
os.path.join(self.flags.datasets_path, 'test_predict.txt'))
self.model_evaluate = Evaluate()
self.model_evaluate.evaluate(
os.path.join(self.flags.datasets_path, 'test_predict.txt'),
os.path.join(self.flags.datasets_path, 'test_evaluate.txt'))
self.hdfs_client.hdfs_delete(
os.path.join(self.flags.output_path, 'test_evaluate.txt'))
self.hdfs_client.hdfs_upload(
os.path.join(self.flags.datasets_path, 'test_evaluate.txt'),
os.path.join(self.flags.input_path, 'test_evaluate.txt'))
p2 = os.path.join(path, "a-" + file)
al = align.face_features(p, p2)
ev = utils.parse_evaluate(al, args.parsing_checkpoint, cuda=cuda)
p = os.path.join(path, "b-" + file)
cv2.imwrite(p, ev)
ev = 255 - utils.img_edge(ev)
p = os.path.join(path, "c-" + file)
cv2.imwrite(p, ev)
elif args.phase == "dataset":
dataset = FaceDataset(args, "test")
dataset.pre_process(cuda)
elif args.phase == "preview":
log.info("preview picture")
path = "../export/regular/model.jpg"
img = cv2.imread(path)
img2 = utils.parse_evaluate(img, args.parsing_checkpoint, cuda)
img3 = utils.img_edge(img2)
img3_ = ops.fill_grey(img3)
img4 = align.face_features(path)
log.info("{0} {1} {2} {3}".format(img.shape, img2.shape, img3_.shape,
img4.shape))
ops.merge_4image(img, img2, img3_, img4, show=True)
elif args.phase == "evaluate":
log.info("evaluation mode start")
evl = Evaluate(args, cuda=cuda)
img = cv2.imread(args.eval_image).astype(np.float32)
x_ = evl.itr_train(img)
evl.output(x_, img)
else:
log.error("not known phase %s", args.phase)
inp[el] = self.means[0, el]
# convert the input list to a numpy matric and normalise it
inp = np.matrix([inp])
inp = (inp - self.means) / self.std
# get a result and prediction using the logistic function
result = (self.coef * inp.T)[0, 0] + self.bias
prediction = 1.0 / (1 + np.exp(-result))
assert prediction <= 1 and prediction >= 0
return prediction
# # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # # #
if __name__ == "__main__":
m = LogReg()
predictions = m.make_prediction_set(
2000,
2016,
"2000-2016 (it9)",
testStart=2015,
slamsOnly=False,
trainName='LogReg_final',
inputProcessedFile="2000-2016 (it9)-processed")
e = Evaluate(predictions)
print('Final log reg model')
e.display_summary()
ftext.close()
f.close()
furl.close()
res = evaluator.evaluateFC(rp)
writeEvaluation(res,evalFilename)
print sum(res)
print len(res)
if __name__ == "__main__":
seedsFiles=['Output-CharlestonShooting.txt','seeds-Sandra.txt','Output-tunisiaHotelAttack.txt','Output-samesexmarriage.txt','Output-fifaArrests.txt','Output-boatCapsized.txt','Output-nepalEarthquake.txt','seeds_459.txt','seeds_474.txt','seedsURLs_z_534.txt','seedsURLs_z_501.txt','seedsURLs_z_540.txt']
posFiles = ['charlestonShootingPos.txt','evaluate-SandraBland.txt','pos-tunisiaHotelAttack.txt','pos-samesexmarriage.txt','Output-fifaArrests.txt','Output-boatCapsized.txt','Output-nepalEarthquake.txt','pos-FSU.txt','pos-Hagupit.txt','pos-AirAsia.txt','pos-sydneyseige.txt','pos-Charlie.txt']
negFiles = ['charlestonShootingNeg.txt','neg-FSU.txt','neg-Hagupit.txt','neg-AirAsia.txt','neg-sydneyseige.txt','neg-Charlie.txt']
modelFiles = ['Output-CharlestonShooting.txt','model-SandraBland.txt','model-tunisiaHotelAttack.txt','model-samesexmarriage.txt','model-CharlestonShooting.txt']
evaluator = Evaluate()
#for i in range(3):
pagesLimit = 300
seedsFiles=['seeds-Sandra.txt','Output-tunisiaHotelAttack.txt','Output-samesexmarriage.txt','Output-CharlestonShooting.txt','Output-fifaArrests.txt','Output-boatCapsized.txt','Output-nepalEarthquake.txt','seeds_459.txt','seeds_474.txt','seedsURLs_z_534.txt','seedsURLs_z_501.txt','seedsURLs_z_540.txt']
posFiles = ['evaluate-SandraBland.txt','pos-tunisiaHotelAttack.txt','pos-samesexmarriage.txt','pos-CharlestonShooting.txt','Output-fifaArrests.txt','Output-boatCapsized.txt','Output-nepalEarthquake.txt','pos-FSU.txt','pos-Hagupit.txt','pos-AirAsia.txt','pos-sydneyseige.txt','pos-Charlie.txt']
negFiles = ['neg-FSU.txt','neg-Hagupit.txt','neg-AirAsia.txt','neg-sydneyseige.txt','neg-Charlie.txt']
modelFiles = ['model-SandraBland.txt','model-tunisiaHotelAttack.txt','model-samesexmarriage.txt','model-CharlestonShooting.txt']
evaluator = Evaluate()
#for i in range(3):
pagesLimit = 100
noK = 5
pageTh = 0.2
urlsTh = 0
i=0
from board import Board, PieceStack, Turn, get_piece_text, EMPTY
from evaluate import Evaluate, WIN
from random import randint
from search import RootOfAlphaBetaSearch
piecestack = PieceStack()
turn = Turn()
board = Board()
evaluate = Evaluate()
def UserTurn(piecestack, board, piece):
board.show()
piecestack.show()
piecestack.TakePiece(piece)
print('Piece: {0}'.format(get_piece_text(piece)))
while True:
x, y = [
int(i) - 1 for i in raw_input(
"Enter x y coordinates to place piece: ").split()
]
if board.pieces[x][y] is EMPTY:
break
else:
print('Square is not empty. Try another one.')
def classifier_test():
# setting
dataset_para = '[email protected]@partition@selection'
# 选择使用哪些特征
feature_para = (1, 2, 3, 4)
# file directory
# feature_dir = dataset_dir + r'\feature1'
# 预处理：从原始数据yoochoose-data中提取出实验数据所需要部分数据（根据实验数据session进行提取）
# 输入1：（实验数据）dataset_dir\train\session_item.txt .\test\session_item.txt
# 输入2：（yoochoose-data）yoochoose_data_dir\yoochoose-clicks.dat .\yoochoose-buys.dat .\yoochoose-test.dat
# 输出：dataset_dir\yoochoose-selected\yoochoose-clicks-selected.dat .\yoochoose-buys-selected.dat .\yoochoose-test-selected.dat
dataset_dir = r'E:\ranking aggregation\dataset\yoochoose\Full' + '\\' + dataset_para
yoochoose_data_dir = r'E:\recsyschallenge2015\mycode\yoochoose-data'
# 输出路径
yoochoose_selected_dir = dataset_dir + r'\yoochoose-selected'
# 假如输出文件夹不存在，则创建文件夹
# if not os.path.exists(yoochoose_selected_dir):
# os.makedirs(yoochoose_selected_dir)
# Preprocess2.extract_data(dataset_dir, yoochoose_data_dir, yoochoose_selected_dir)
# 提取特征
# 输入：yoochoose selected data（及groundtruth）
# 输出：特征
feature_dir = r'E:\recsyschallenge2015\mycode\result-data'
# feature_dir = dataset_dir + r'\feature1'
# 假如输出文件夹不存在，则创建文件夹
if not os.path.exists(feature_dir):
os.makedirs(feature_dir)
print('feature_para:', feature_para)
Feature4.go(dataset_dir, feature_dir, feature_para)
# 读取特征
X_train, y_train = Input2.read_train(feature_dir)
X_test, y_test, test_dic_data, session_item_data, session_idx_dic = Input2.read_test(
dataset_dir, feature_dir)
groundtruth_path = dataset_dir + r'\test\session_item.txt'
# 模型部分
print('model: LogisticRegression')
model = LogisticRegression()
model.fit(X_train, y_train)
# print(model)
# make predictions
y_predict = model.predict(X_test)
# 结果评估
solution = Solution.generate(test_dic_data, y_predict)
Evaluate.go(solution, groundtruth_path)
# 模型部分
print('model: GaussianNB')
model = GaussianNB()
model.fit(X_train, y_train)
# print(model)
# make predictions
y_predict = model.predict(X_test)
# 结果评估
solution = Solution.generate(test_dic_data, y_predict)
Evaluate.go(solution, groundtruth_path)
# 模型部分
print('model: SVM')
model = SVC()
model.fit(X_train, y_train)
# print(model)
# make predictions
y_predict = model.predict(X_test)
# 结果评估
solution = Solution.generate(test_dic_data, y_predict)
Evaluate.go(solution, groundtruth_path)
class DataHandler:
#클래스 멤버: 연산기 하나
evaluator = Evaluate()
#class method : 전역함수처럼 쓸 수 있다
@classmethod
def GetRawdataInDic(cls, filename):
rawdata = {}
with open(filename, 'rb') as f:
while 1:
try:
data = pickle.load(f)
except EOFError:
break
rawdata.update(data)
return rawdata
def __init__(self, filename, clsname):
self.rawdata = DataHandler.GetRawdataInDic(filename)
self.clsname = clsname
#연산한 값을 저장해두는 저장소
#필요할 떄 연산하되, 이미 연산된 값이면 연산없이 저장된 값을 반환
self.cache = {}
def get_scores(self):
if 'scores' not in self.cache:
self.cache['scores'] = list(self.rawdata.values())
return self.cache.get('scores')
#cache
def get_average(self):
if 'average' not in self.cache:
self.cache['average'] = self.evaluator.average(self.get_scores())
return self.cache.get('average')
def get_variance(self):
if 'variace' not in self.cache:
vari = round(
self.evaluator.variance(self.get_scores(), self.get_average()))
self.cache['variance'] = vari
return self.cache.get('variance')
def get_standard_deviation(self):
if "standard_deviation" not in self.cache:
std_dev = round(math.sqrt(self.get_variance()), 1)
self.cache["standard_deviation"] = std_dev
return self.cache.get("standard_deviation")
def WhoIsHighest(self):
if 'highest' not in self.cache:
self.cache['highest'] = reduce(
lambda a, b: a
if self.rawdata.get(a) > self.rawdata.get(b) else b,
self.rawdata.keys())
return self.cache.get('highest')
def GetHighestScore(self):
return self.rawdata[self.WhoIsHighest()]
def WhoIsLowest(self):
if "lowest" not in self.cache:
self.cache['lowest'] = reduce(
lambda a, b: a
if self.rawdata.get(a) < self.rawdata.get(b) else b,
self.rawdata.keys())
return self.cache.get('lowest')
def GetLowestScore(self):
return self.rawdata[self.WhoIsLowest()]
def get_evaluation(self):
print('*' * 50)
print("%s 반 성적 분석 결과" % self.clsname)
print("{0}반의 평균은 {1}점이고 분산은 {2}이며,따라서 표준편차는{3}이다".\
format(self.clsname, self.get_average(), self.get_variance()\
, self.get_standard_deviation()))
print('*' * 50)
print("%s 반 종합 평가" % self.clsname)
print('*' * 50)
self.evaluateClass()
def evaluateclass(self):
avrg = self.get_average()
std_dev = self.get_standard_deviation()
if avrg < 50 and std_dev > 20:
print("성적이 너무 저조하고 학생들의 실력 차이가 너무 크다.")
elif avrg > 50 and std_dev > 20:
print("성적은 평균이상이지만 학생들 실력 차이가 크다. 주의 요망!")
elif avrg < 50 and std_dev < 20:
print("학생들간 실력차는 나지 않으나 성적이 너무 저조하다. 주의 요망!")
elif avrg > 50 and std_dev < 20:
print("성적도 평균 이상이고 학생들의 실력차도 크지 않다.")
def main():
# action space
actionSpace = [[10, 0], [7, 7], [0, 10], [-7, 7], [-10, 0], [-7, -7], [0, -10], [7, -7]]
numActionSpace = len(actionSpace)
# state space
numStateSpace = 4
xBoundary = [0, 360]
yBoundary = [0, 360]
checkBoundaryAndAdjust = ag.CheckBoundaryAndAdjust(xBoundary, yBoundary)
initSheepPositionMean = np.array([180, 180])
initWolfPositionMean = np.array([180, 180])
initSheepPositionNoise = np.array([120, 120])
initWolfPositionNoise = np.array([60, 60])
sheepPositionReset = ag.SheepPositionReset(initSheepPositionMean, initSheepPositionNoise, checkBoundaryAndAdjust)
wolfPositionReset = ag.WolfPositionReset(initWolfPositionMean, initWolfPositionNoise, checkBoundaryAndAdjust)
numOneAgentState = 2
positionIndex = [0, 1]
sheepPositionTransition = ag.SheepPositionTransition(numOneAgentState, positionIndex, checkBoundaryAndAdjust)
wolfPositionTransition = ag.WolfPositionTransition(numOneAgentState, positionIndex, checkBoundaryAndAdjust)
numAgent = 2
sheepId = 0
wolfId = 1
transitionFunction = env.TransitionFunction(sheepId, wolfId, sheepPositionReset, wolfPositionReset,
sheepPositionTransition, wolfPositionTransition)
minDistance = 15
isTerminal = env.IsTerminal(sheepId, wolfId, numOneAgentState, positionIndex, minDistance)
screen = pg.display.set_mode([xBoundary[1], yBoundary[1]])
screenColor = [255, 255, 255]
circleColorList = [[50, 255, 50], [50, 50, 50], [50, 50, 50], [50, 50, 50], [50, 50, 50], [50, 50, 50],
[50, 50, 50], [50, 50, 50], [50, 50, 50]]
circleSize = 8
saveImage = False
saveImageFile = 'image'
render = env.Render(numAgent, numOneAgentState, positionIndex, screen, screenColor, circleColorList, circleSize,
saveImage, saveImageFile)
aliveBouns = -1
deathPenalty = 20
rewardDecay = 0.99
rewardFunction = reward.TerminalPenalty(sheepId, wolfId, numOneAgentState, positionIndex, aliveBouns, deathPenalty, isTerminal)
accumulateRewards = PG.AccumulateRewards(rewardDecay, rewardFunction)
maxTimeStep = 150
sampleTrajectory = PG.SampleTrajectory(maxTimeStep, transitionFunction, isTerminal)
approximatePolicy = PG.ApproximatePolicy(actionSpace)
trainPG = PG.TrainTensorflow(actionSpace)
numTrajectory = 20
maxEpisode = 1000
# Generate models.
learningRate = 1e-4
hiddenNeuronNumbers = [128, 256, 512, 1024]
hiddenDepths = [2, 4, 8]
# hiddenNeuronNumbers = [128]
# hiddenDepths = [2]
generateModel = GeneratePolicyNet(numStateSpace, numActionSpace, learningRate)
models = {(n, d): generateModel(d, round(n / d)) for n, d in it.product(hiddenNeuronNumbers, hiddenDepths)}
print("Models generated")
# Train.
policyGradient = PG.PolicyGradient(numTrajectory, maxEpisode, render)
trainModel = lambda model: policyGradient(model, approximatePolicy,
sampleTrajectory,
accumulateRewards,
trainPG)
trainedModels = {key: trainModel(model) for key, model in models.items()}
print("Finished training")
# Evaluate
modelEvaluate = Evaluate(numTrajectory, approximatePolicy, sampleTrajectory, rewardFunction)
meanEpisodeRewards = {key: modelEvaluate(model) for key, model in trainedModels.items()}
print("Finished evaluating")
# print(meanEpisodeRewards)
# Visualize
independentVariableNames = ['NeuroTotalNumber', 'layerNumber']
draw(meanEpisodeRewards, independentVariableNames)
print("Finished visualizing", meanEpisodeRewards)
axs[0].grid()
axs[0].set_title('Loss')
axs[1].plot(history['Train_dice'], label='Train Dice')
axs[1].plot(history['Valid_dice'], label='Valid Dice')
axs[1].legend()
axs[1].grid()
axs[1].set_title('Dice')
plt.savefig('../output/loss_dice.png')
########################################################################
# Evaluate the network
# get all predictions of the validation set: maybe a memory error here.
if args.load_mod:
# load the best model
net.load_state_dict(torch.load(MODEL_FILE))
eva = Evaluate(net, device, validloader, args, isTest=False)
eva.search_parameter()
dice, dicPred, dicSubmit = eva.predict_dataloader()
# eva.plot_sampled_predict()
# evaluate the prediction
sout = '\nFinal Dice {:.3f}\n'.format(dice) +\
'==============Predict===============\n' + \
analyze_labels(pd.DataFrame(dicPred)) # +\
# '==============True===============\n' + \
# analyze_labels(stat_df_valid)
# print(sout)
# print2file(sout, LOG_FILE)
# print2file(' '.join(str(key)+':'+str(val) for key,val in eva.dicPara.items()), LOG_FILE)
# load swa model
def __init__(self, n):
self.data = DataManager('../data/train.csv','../data/test.csv', n)
self.fe = FeatureExtractor(self.data)
self.eval = Evaluate()
def playit(self):
point = 0
evaluation_ai = Evaluate(self.board, True)
aiMoves = evaluation_ai.checkPossibleMoves()
depth = 0
print(aiMoves[0][0][0])
import numpy as np
import json
gen_data = False
plot_fig = True
basename = 'rmse'
jsonfile = basename + '.json'
print('gen_data/plot_fig/filename = ', gen_data, plot_fig, basename)
mem_list = [6, 10]
loc_list = np.arange(5, 60, 5).tolist()
wgt_list = np.arange(0, 10, 1).tolist()
if gen_data:
ev = Evaluate()
truth = ev.get_data('output_truth.txt')
error = {}
rmsedata = {}
amedata = {}
for mem in mem_list:
rmseloc = {}
ameloc = {}
for loc in loc_list:
rmse = []
ame = []
for wgt in wgt_list:
filename = 'h3dw%2.2dl%2.2dm%2.2d/output_analy.txt' % (
wgt, loc, mem)
# analysis
optimizer = torch.optim.AdamW(model.parameters(), lr=args.lr)
print(model)
print(model_fname)
# # train the model
# for param in model.parameters():
# param.requires_grad = True
model.train()
train_loop()
# evaluate the model
if args.evaluate:
if args.evaluate_on_cpu:
device = "cpu"
model = model.to(device)
model.eval()
if args.train:
Evaluate(model, test_loader, outpath, args.target, device, args.n_epochs)
elif args.load:
Evaluate(model, test_loader, outpath, args.target, device, args.load_epoch)
## -----------------------------------------------------------
# # to retrieve a stored variable in pkl file
# import pickle
# with open('../../test_tmp_delphes/experiments/PFNet7_gen_ntrain_2_nepochs_3_batch_size_3_lr_0.0001/confusion_matrix_plots/cmT_normed_epoch_0.pkl', 'rb') as f: # Python 3: open(..., 'rb')
# a = pickle.load(f)
class StanceDetector:
def __init__(self, n):
self.data = DataManager('../data/train.csv','../data/test.csv', n)
self.fe = FeatureExtractor(self.data)
self.eval = Evaluate()
def buildBaseline(self, model):
print 'Training baseline',model
feats = ['words']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
for mode in ['simple','tfidf']:
if model=='bayes':
cl = MultinomialNB()
elif model=='svm':
cl = LinearSVC()
if mode=='tfidf':
cl = Pipeline([('tfidf', TfidfTransformer()),
('clf', cl), ])
clf = cl.fit(X, y)
y_pred = clf.predict(X_test)
print mode, accuracy_score(y_true, y_pred)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def buildSimple(self, model):
feats = ['topicVecs','words2vec']
print feats
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
for mode in ['simple']:#,'tfidf']:
if model=='bayes':
cl = MultinomialNB()
elif model=='svm':
# cl = LinearSVC()
cl = LinearSVC()
cl = GridSearchCV(cl, self.getGridSearchParams())
if mode=='tfidf':
cl = Pipeline([('tfidf', TfidfTransformer()),
('clf', cl), ])
clf = cl.fit(X, y)
# print cl.best_params_
y_pred = clf.predict(X_test)
print mode, accuracy_score(y_true, y_pred)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
#train in name means helper function
def trainSVC(self, feats, y_attribute, proba=False):
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
clf = SVC(probability=proba)
clf = clf.fit(X,y)
if proba:
y_proba = clf.predict_proba(X_test)
return clf, y_proba
else:
y_pr = clf.predict(X_test)
return clf, y_pr
def trainLinearSVC(self, feats, y_attribute, dec=False):
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
clf = LinearSVC()
clf = clf.fit(X,y)
if dec:
y_pr = clf.decision_function(X_test)
return clf, y_pr
else:
y_pr = clf.predict(X_test)
return clf, y_pr
#TODO: revisit
#check lable transform encodings of NONE, FAVOR, AGAINST
# def buildTopicStanceSeparate(self):
# feats = ['words']
# y_attribute = 'stance'
# X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
# #builds two separate for topic and stance
# topic_clf, y_topic_proba = self.trainLinearSVC(feats = ['words','lexiconsbyword'],y_attribute = 'topic',dec=True)
# #WRONR
# #WRONG
# #WRONG
# boost_factors = np.ones_like(y_true)
# #multiply by NONE (0) = 0
# #multiply by FAVOR (1) = 1
# #multiply by AGAINST (2) = 2
# #has index of class with max prob for each sample
# topic_preds = np.argmax(y_topic_proba,axis=1)
# for ind,s in enumerate(y_topic_proba):
# prob = y_topic_proba[ind][topic_preds[ind]]
# if prob < 0.4:
# boost_factors[ind] = 0 #corresponds to NONE
# stance_clf,stance_pred = self.trainLinearSVC(feats = ['words','lexiconsbyword','topic'],y_attribute = 'stance')
# # for i in range(0, len(stance_pred)):
# # if boost_factors[i] == 2:
# # stance_pred[i] = self.fe.labelenc.transform("NONE")
# #with numpy arrays now, above is equivalent to below , right?
# stance_pred = np.multiply(stance_pred, boost_factors)
# stance_pred_labels = self.fe.labelenc.inverse_transform(stance_pred)
# # print [(self.data.testLabels[i], stance_pred_labels[i]) for i in range(len(stance_pred))]
# score = accuracy_score(y_true, stance_pred)
# print score
# pprint(self.eval.computeFscores(self.data.testTweets, stance_pred_labels))
def buildTopicOnlyMultiple(self):
#one svm for each topic
feats = ['words2vec']
y_attribute = 'stance'
clf_topic = {}
for topic in list(self.fe.topicenc.classes_):
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute, topic)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute, topic)
clf = LinearSVC()
clf = clf.fit(X,y)
clf_topic[topic] = clf
print topic, clf.score(Xt,yt)
# not useful. still less than single SVM. but not as much as avg of above
# X_whole,y_whole = self.fe.getFeaturesMatrix('train',feats,y_attribute)
# Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
# newX = []
# newXt = []
# for topic in clf_topic:
# newX.append(clf_topic[topic].transform(X_whole))
# newXt.append(clf_topic[topic].transform(Xt))
# newX = np.concatenate(tuple(newX),axis=1)
# newXt = np.concatenate(tuple(newXt),axis=1)
# newclf = LinearSVC()
# newclf = newclf.fit(newX, y_whole)
# print newclf.score(newXt, yt)
def trainTopicSVM(self, topic):
feats = ['words2vec','clusteredLexicons','topic1hot']
y_attribute = 'stance'
X,y = self.fe.getFeaturesTopicNontopic('train',feats,y_attribute, topic=topic)
X_test,y_true = self.fe.getFeaturesTopicNontopic('test',feats,y_attribute, topic=topic)
clf = LinearSVC()
clf = GridSearchCV(clf,self.getGridSearchParams())
clf = clf.fit(X,y)
print clf.best_params_
print topic #,clf.score(X_test, y_true)
return clf
#WRITE
#WRITE
#WRITE
def buildTopicWise(self):
#separate SVC for each topic, tests on that class only first, then on all
topic_clf = {}
feats = ['words2vec','clusteredLexicons','topic1hot']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,y_attribute)
#X matrix for new classifier which uses this as train matrix
#has columns of each topic classifier's confidence function
# X_fx = []
# X_ftestx = []
preds = []
for topic in list(self.fe.topicenc.classes_):
topic_clf[topic] = self.trainTopicSVM(topic)
preds.append(topic_clf[topic].predict(X_test))
# X_fx.append(topic_clf[topic].decision_function(X))
# X_ftestx.append(topic_clf[topic].decision_function(X_test))
allpreds = np.vstack(tuple(preds))
topic1hot, temp = self.fe.getFeaturesMatrix('test',['topic1hot'],'stance')
# print allpreds.shape, topic1hot.T.shape
allpreds[allpreds==5] = 1
final_pred = np.multiply(topic1hot.T,allpreds)
prediction = np.sum(final_pred, axis=0).astype(int)
# X_fx = np.concatenate(tuple(X_fx), axis=1)
# X_ftestx = np.concatenate(tuple(X_ftestx), axis=1)
# clf = LinearSVC().fit(X_fx, y)
# y_pred = clf.predict(X_ftestx)
print accuracy_score(y_true, prediction)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(prediction)))
#GOOD 66%acc
#1.2 % increase with change topic to 1hot
def buildSVMWord2Vec(self):
feats = ['words2vec','topic1hot']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
clf = LinearSVC(C=0.01,penalty='l1',dual=False)
clf = clf.fit(X,y)
y_pred = clf.predict(Xt)
print clf.score(Xt, yt)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def buildSVMTrial(self):
feats = ['topic1hot','words2vec']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
clf = LinearSVC(C=0.001)
clf = clf.fit(X,y)
y_pred = clf.predict(Xt)
print clf.score(Xt, yt)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def buildTrial(self):
# feats = ['pos','words2vec','clusteredLexicons','topic1hot']
# 'givenSentiment','givenOpinion'
feats = ['words2vec','pos','clusteredLexicons','top1grams','top2grams']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
# clf = DecisionTreeClassifier()
# clf = LogisticRegression()
clf = LinearSVC(C=1, class_weight='balanced', penalty='l1',dual=False)
clf = clf.fit(X,y)
y_pred = clf.predict(Xt)
# print y_pred
print len(np.where(y_pred==0)[0]),len(np.where(y_pred==1)[0]),len(np.where(y_pred>1)[0])
print len(y_pred)
print 'training accuracy',clf.score(X, y)
print clf.score(Xt, yt)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def buildGithubSGDModel(self):
# feats = ['words2vec','topic1hot','pos']
y_attribute = 'stance'
dataset = self.fe.getDataset('train')
dataset2 = self.fe.getDataset('test')
y_train = self.fe.getY('train',dataset, y_attribute)
y_test = self.fe.getY('train',dataset2, y_attribute)
tfidf = TfidfVectorizer(ngram_range=(1, 2), max_df=1.0, min_df=1, binary=True, norm='l2', use_idf=True, smooth_idf=False, sublinear_tf=True, encoding='latin1')
X_train = tfidf.fit_transform(self.data.trainTweetsText)
X_test = tfidf.transform(self.data.testTweetsText)
tuned_parameters = {'alpha': [10 ** a for a in range(-12, 0)]}
clf = GridSearchCV(SGDClassifier(loss='hinge', penalty='elasticnet',l1_ratio=0.75, n_iter=10, shuffle=True, verbose=False, n_jobs=4, average=False)
, tuned_parameters, cv=10, scoring='f1_weighted')
clf.fit(X_train, y_train)
print clf.best_params_
print("Grid scores on development set:")
for params, mean_score, scores in clf.grid_scores_:
print "%0.3f (+/-%0.03f) for %r" % (mean_score, scores.std() * 2, params)
print classification_report(y_test, clf.predict(X_test))
print clf.score(X_test, y_test)
def getGridSearchParams(self):
param_grid = [
{'C': [0.001, 0.01, 0.1, 1], 'dual':[False, True],'class_weight':['balanced',None]}
]
return param_grid
def getGridSearchParamsForXGBoost(self):
param_grid = [
{'n_estimators':[10,20,30,40,50], 'max_depth': [1,2,3,4,5]}
]
def buildSVMWord2VecWithClusters(self):
#feats = ['topic1hot']
#feats = ['words2vec', 'top1grams', 'top2grams']
#feats = ['words2vec', 'top1grams']
#feats = ['words2vec', 'top2grams']
feats = ['words2vec', 'clusteredLexicons', 'topic1hot', 'pos']
#feats = ['words2vec','topic1hot', 'pos','clusteredLexicons', 'top2grams']
#feats = ['clusteredLexicons']
#feats = ['pos']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
print (X.shape)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
clf = LinearSVC(C=1,penalty='l1',dual=False)
clf = clf.fit(X,y)
y_pred = clf.predict(Xt)
# f = open('pred','w')
# for i in y_pred:
# #print type(i)
# f.write('{0}'.format(i))
# f.close()
accuracy = clf.score(Xt, yt)
# print clf.score(Xt, yt)
fscores = self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred))
# print type(fscores)
# print fscores
# pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
# print (accuracy, fscores['Macro'])
return (accuracy, fscores['Macro'])
def buildSVMWord2VecWithClustersGridSearch(self):
feats = ['words2vec','topic1hot','pos', 'clusteredLexicons']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
svmclf = LinearSVC(C=0.01,penalty='l1',dual=False)
clf = GridSearchCV(svmclf, self.getGridSearchParams())
clf = clf.fit(X,y)
print clf.best_params_
y_pred = clf.predict(Xt)
print clf.score(Xt, yt)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def trainStanceNone(self, feats):
# feats = ['words2vec','topic1hot','pos']
X,y = self.fe.getFeaturesStanceNone('train',feats)
Xt,yt = self.fe.getFeaturesStanceNone('test',feats)
svmclf = LinearSVC()
stance_none_clf = GridSearchCV(svmclf, self.getGridSearchParams()).fit(X, y)
# print stance_none_clf.score(Xt, yt)
pred = stance_none_clf.predict(Xt)
print classification_report(yt, pred)
return stance_none_clf
def trainFavorAgainst(self,feats):
# feats = ['words2vec','topic1hot','pos']
X,y = self.fe.getFeaturesFavorAgainst('train',feats)
Xt,yt = self.fe.getFeaturesFavorAgainst('test',feats)
svmclf = LinearSVC()
fav_agnst_clf = GridSearchCV(svmclf, self.getGridSearchParams()).fit(X, y)
pred = fav_agnst_clf.predict(Xt)
print classification_report(yt, pred)
# print fav_agnst_clf.score(Xt, yt)
return fav_agnst_clf
def buildModel2(self):
#one SVM for Stance/None and other for Favor/Against
feats = ['words2vec','topic1hot','pos']
print feats
stance_none_clf = self.trainStanceNone(feats)
fav_agnst_clf = self.trainFavorAgainst(feats)
X_test,y_true = self.fe.getFeaturesMatrix('test',feats,'stance')
st_pred = stance_none_clf.predict(X_test)
favaga_pred = fav_agnst_clf.predict(X_test)
for index,row in enumerate(st_pred):
if row==3:
st_pred[index] = favaga_pred[index]
print classification_report(y_true, st_pred)
print accuracy_score(y_true, st_pred)
# assert(stance_none_clf.classes_[1]==3) #stance(3)
# # >0 means this class - stance will be predicted
# # <0 means none is predicted
# confi = stance_none_clf.decision_function(X_test)
# # treat as confident about none if confi<-0.25:
# y_pred = fav_agnst_clf.predict(X_test)
# print accuracy_score(y_true, y_pred)
# pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
# threshold = -0.25
# confi_high = np.where(confi<threshold)[0]
# for loc in confi_high:
# y_pred[loc] = self.fe.labelenc.transform('NONE')
# print 'Boosted', accuracy_score(y_true, y_pred)
# print len(np.where(y_pred==0)[0]),len(np.where(y_pred==1)[0]), len(np.where(y_pred==2)[0]),
# pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
def get_proba_one(self, model, X):
predicted = model.predict_proba(X)
return predicted[:, 1]
def runXGBoostModel(self,model, model_name, X, target, X_test, crossOn):
print "Trying to fit model"
print X.shape, target.shape
model.fit(X, target)
print "Successfully fit model"
predicted = self.get_proba_one(model, X)
predicted_test = self.get_proba_one(model, X_test)
predicted_test = model.predict(X_test)
print predicted_test
return predicted_test
def word2VecXGBoost(self):
feats = ['words2vec','pos','clusteredLexicons', 'top1grams','top2grams', 'topic1hot' ]
#feats = ['words2vec']
#feats = ['clusteredLexicons']
#feats = ['pos']
y_attribute = 'stance'
X,y = self.fe.getFeaturesMatrix('train',feats,y_attribute)
print (X.shape)
Xt,yt = self.fe.getFeaturesMatrix('test',feats,y_attribute)
#clf = LinearSVC(C=0.01,penalty='l1',dual=False)
#clf = clf.fit(X,y)
#y_pred = clf.predict(Xt)
# f = open('pred','w')
# for i in y_pred:
# #print type(i)
# f.write('{0}'.format(i))
# f.close()
#print clf.score(Xt, yt)
#pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(y_pred)))
m2_xgb = xgb.XGBClassifier(n_estimators=10, nthread=-1, max_depth = 2 , seed=500)
#m2_xgb = GridSearchCV(m2_xgb, self.getGridSearchParamsForXGBoost())
print "Run Model"
y_pred = self.runXGBoostModel(m2_xgb, "m2_xgb_OS_ENN", X, y, Xt, True)
# print type(yt)
# print type(y_pred)
# print len(yt)
# print len(y_pred)
# print yt.shape
# print y_pred.shape
# print yt
# print y_pred
# print(m2_xgb)
print accuracy_score(yt, y_pred)
def buildModel3(self):
#feats = [['words2vec'],['pos'],['clusteredLexicons']]
feats = [['words2vec'],['pos'],['clusteredLexicons']]
y_attribute = 'stance'
y_pred = []
y_t = []
for f in feats:
X,y = self.fe.getFeaturesMatrix('train',f,y_attribute)
Xt,yt = self.fe.getFeaturesMatrix('test',f,y_attribute)
clf = SVC(C=1, probability=True)
clf = clf.fit(X,y)
train_transform = clf.predict_log_proba(X)
test_transform = clf.predict_log_proba(Xt)
# print 'Train transform ',train_transform.shape
# print 'Test transform ',test_transform.shape
y_pred.append(train_transform)
y_t.append(test_transform)
#y_pred_h = np.hstack(tuple(y_pred))
#y_t_h = np.hstack(tuple(y_t))
x = 0
for i in y_pred:
x += i
y_pred_h = x
x = 0
for i in y_t:
x += i
y_t_h = x
# print type(y_pred_h)
# print y_pred_h[0]
# print y_pred_h.shape
regr = linear_model.LogisticRegression()
regr.fit(y_pred_h, y)
final_pred = regr.predict(y_t_h)
print accuracy_score(final_pred, yt)
pprint(self.eval.computeFscores(self.data.testTweets, self.fe.labelenc.inverse_transform(final_pred)))