95 lines
2.3 KiB
Python
95 lines
2.3 KiB
Python
|
import gensim.downloader as gensim
|
||
|
|
import numpy as np
|
||
|
|
import pandas as pd
|
||
|
|
import torch
|
||
|
|
from nltk.tokenize import word_tokenize
|
||
|
|
|
||
|
|
|
||
|
|
class NeuralNetworkModel(torch.nn.Module):
|
||
|
|
def __init__(self):
|
||
|
|
super(NeuralNetworkModel, self).__init__()
|
||
|
|
self.1 = torch.nn.Linear(300, 300)
|
||
|
|
self.2 = torch.nn.Linear(300, 1)
|
||
|
|
|
||
|
|
def forward(self, x):
|
||
|
|
x = self.1(x)
|
||
|
|
x = torch.relu(x)
|
||
|
|
x = self.2(x)
|
||
|
|
x = torch.sigmoid(x)
|
||
|
|
return x
|
||
|
|
|
||
|
|
nm = NeuralNetworkModel()
|
||
|
|
dev_train = []
|
||
|
|
test_train = []
|
||
|
|
word2vec = gensim.load('word2vec-google-news-300')
|
||
|
|
|
||
|
|
|
||
|
|
np.mean([word2vec[word] for word in doc if word in word2vec] or [np.zeros(300)], axis=0)
|
||
|
|
|
||
|
|
|
||
|
|
def model_train():
|
||
|
|
|
||
|
|
train = pd.read_table('train/in.tsv.xz', compression='xz', sep='\t', quoting=3)
|
||
|
|
trainy = pd.read_table('train/expected.tsv', sep='\t', quoting=3)
|
||
|
|
trainy = trainy[0]
|
||
|
|
|
||
|
|
def model_prepare():
|
||
|
|
dev = pd.read_table('dev-0/in.tsv.xz', compression='xz', sep='\t', quoting=3)
|
||
|
|
test = pd.read_table('test-A/in.tsv.xz', compression='xz', sep='\t', quoting=3)
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
train = [word_tokenize(x) for x in train]
|
||
|
|
dev = [word_tokenize(x) for x in dev]
|
||
|
|
test = [word_tokenize(x) for x in test]
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
def word_2_voc():
|
||
|
|
train = [d2v(doc) for doc in train]
|
||
|
|
dev = [d2v(doc) for doc in dev]
|
||
|
|
test = [d2v(doc) for doc in test]
|
||
|
|
|
||
|
|
|
||
|
|
criterion = torch.nn.BCELoss()
|
||
|
|
optimizer = torch.optim.Adam(model.parameters())
|
||
|
|
|
||
|
|
print ("1")
|
||
|
|
|
||
|
|
for epoch in range(5):
|
||
|
|
model.train()
|
||
|
|
for i in range(0, y_train.shape[0], 5):
|
||
|
|
X = train[i:i + 5]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
y = trainy[i:i + 5]
|
||
|
|
y = torch.tensor(y.astype(np.float32).to_numpy()).reshape(-1, 1)
|
||
|
|
optimizer.zero_grad()
|
||
|
|
outputs = nm(X.float())
|
||
|
|
loss = criterion(outputs, y)
|
||
|
|
loss.backward()
|
||
|
|
optimizer.step()
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
print ("2")
|
||
|
|
|
||
|
|
with torch.no_grad():
|
||
|
|
for i in range(0, len(dev), 5):
|
||
|
|
X = dev[i:i + 5]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
outputs = nm(X.float())
|
||
|
|
y = (outputs > 0.5)
|
||
|
|
dev_train.extend(y)
|
||
|
|
|
||
|
|
for i in range(0, len(test), 5):
|
||
|
|
X = test[i:i + 5]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
outputs = nm(X.float())
|
||
|
|
y = (outputs >= 0.5)
|
||
|
|
testy.extend(y)
|
||
|
|
|
||
|
|
|
||
|
|
dev_train.to_csv(r'dev-0/out.tsv', sep='\t', index=False, header=False)
|
||
|
|
test_train.to_csv(r'test-A/out.tsv', sep='\t', index=False, header=False)
|
||
|
|
|