Prześlij pliki do ''

run.py

@@ -1,91 +1,105 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+import torch.optim as optim
+from torchvision import transforms
+import pickle
 import numpy as np
 import pandas as pd
-import torch
+from word2vec import Word2Vec
-import csv
-import lzma
-import gensim.downloader
-from nltk import word_tokenize
-
-#print('wczytanie danych')
-
-x_train = pd.read_table('train/in.tsv.xz', compression='xz', sep='\t', header=None, quoting=3)
-y_train = pd.read_table('train/expected.tsv', sep='\t', header=None, quoting=3)
-x_dev = pd.read_table('dev-0/in.tsv.xz', compression='xz', sep='\t', header=None, quoting=3)
-x_test = pd.read_table('test-A/in.tsv.xz', compression='xz', sep='\t', header=None, quoting=3)
-
-#print('inicjalizacja modelu')
-class NeuralNetworkModel(torch.nn.Module):
-    def __init__(self):
-        super(NeuralNetworkModel, self).__init__()
-        self.l01 = torch.nn.Linear(300, 300)
-        self.l02 = torch.nn.Linear(300, 1)
-
-    def forward(self, x):
-        x = self.l01(x)
-        x = torch.relu(x)
-        x = self.l02(x)
-        x = torch.sigmoid(x)
-        return x
 
 
-#print('przygotowanie danych')
+class FFN(nn.Module):
 
-x_train = x_train[0].str.lower()
+    def __init__(self, input_dim, output_dim, hidden1_size, hidden2_size, lr, epochs, batch_size):
-y_train = y_train[0]
+        super(FFN, self).__init__()
-x_dev = x_dev[0].str.lower()
+        self.path = 'model1.pickle'
-x_test = x_test[0].str.lower()
+        self.lr = lr
+        self.epochs = epochs
+        self.output_dim = output_dim
+        self.word2vec = Word2Vec()
+        self.word2vec.load()
+        self.batch_size = batch_size
+        self.input_dim = input_dim
+        self.fc1 = nn.Linear(batch_size, hidden1_size)
+        self.fc2 = nn.Linear(hidden1_size, hidden2_size)
+        self.fc3 = nn.Linear(hidden2_size, hidden2_size)
+        self.fc4 = nn.Linear(hidden2_size, hidden2_size)
+        self.fc5 = nn.Linear(hidden2_size, batch_size)
 
-x_train = [word_tokenize(x) for x in x_train]
+    def forward(self, data):
-x_dev = [word_tokenize(x) for x in x_dev]
+        data = F.relu(self.fc1(data))
-x_test = [word_tokenize(x) for x in x_test]
+        data = F.relu(self.fc2(data))
+        data = F.relu(self.fc3(data))
+        data = F.relu(self.fc4(data))
+        data = F.sigmoid(self.fc5(data))
+        return data
 
-word2vec = gensim.downloader.load('word2vec-google-news-300')
+    def serialize(self):
-x_train = [np.mean([word2vec[word] for word in content if word in word2vec] or [np.zeros(300)], axis=0) for content in x_train]
+        with open(self.path, 'wb') as file:
-x_dev = [np.mean([word2vec[word] for word in content if word in word2vec] or [np.zeros(300)], axis=0) for content in x_dev]
+            pickle.dump(self, file)
-x_test = [np.mean([word2vec[word] for word in content if word in word2vec] or [np.zeros(300)], axis=0) for content in x_test]
 
+    def load(self):
+        with open(self.path, 'rb') as file:
+            self = pickle.load(file)
 
-#print('trenowanie modelu')
+    def batch(self, iterable, n=1):
-model = NeuralNetworkModel()
+        l = len(iterable)
-BATCH_SIZE = 5
+        for ndx in range(0, l, n):
-criterion = torch.nn.BCELoss()
+            yield iterable[ndx:min(ndx + n, l)]
-optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
 
-for epoch in range(BATCH_SIZE):
+    """
-    model.train()
+    data is a tuple of embedding vector and a label of 0/1
-    for i in range(0, y_train.shape[0], BATCH_SIZE):
+    """
-        X = x_train[i:i + BATCH_SIZE]
+
-        X = torch.tensor(X)
+    def train(self, data, expected):
-        y = y_train[i:i + BATCH_SIZE]
+        self.zero_grad()
-        y = torch.tensor(y.astype(np.float32).to_numpy()).reshape(-1, 1)
+        criterion = torch.nn.BCELoss()
+        optimizer = optim.Adam(self.parameters(), lr=self.lr)
+        batch_size = self.batch_size
+        num_of_classes = self.output_dim
+        for epoch in range(self.epochs):
+            epoch_loss = 0.0
+            idx = 0
+            for i in range(0, int(len(data) / batch_size) * batch_size, batch_size):
+                inputs = data[i:i + batch_size]
+                labels = expected[i:i + batch_size]
                 optimizer.zero_grad()
-        outputs = model(X.float())
+                outputs = self.forward(torch.tensor(self.word2vec.list_of_sentences2vec(inputs)))
-        loss = criterion(outputs, y)
+                target = torch.tensor(labels.values).double()
+                loss = criterion(outputs.view(batch_size), target.view(-1, ))
                 loss.backward()
                 optimizer.step()
 
-#print('predykcja wynikow')
+                epoch_loss += loss.item()
-y_dev = []
+                if (idx % 1000 == 0):
-y_test = []
+                    print('epoch: {}, idx: {}, loss: {}'.format(epoch, idx, epoch_loss / 1000))
-model.eval()
+                    epoch_loss = 0
+                idx += 1
+        self.serialize()
 
-with torch.no_grad():
+    def test(self, data, expected, path):
-    for i in range(0, len(x_dev), BATCH_SIZE):
+        correct = 0
-        X = x_dev[i:i + BATCH_SIZE]
+        incorrect = 0
-        X = torch.tensor(X)
+        total = 0
-        outputs = model(X.float())
+        predictions = []
-        prediction = (outputs > 0.5)
+        batch_size = self.batch_size
-        y_dev += prediction.tolist()
+        for i in range(0, int(len(data) / batch_size) * batch_size, batch_size):
+            inputs = data[i:i + batch_size]
+            labels = expected[i:i + batch_size]
+            predicted = self.forward(torch.tensor(self.word2vec.list_of_sentences2vec(inputs)))
+            score = [1 if x > 0.5 else 0 for x in predicted]
 
-    for i in range(0, len(x_test), BATCH_SIZE):
+            for x, y in zip(score, labels):
-        X = x_test[i:i + BATCH_SIZE]
+                if (x == y):
-        X = torch.tensor(X)
+                    correct += 1
-        outputs = model(X.float())
+                else:
-        y = (outputs >= 0.5)
+                    incorrect += 1
-        y_test += prediction.tolist()
+            predictions.append(score)
 
-# print('eksportowanie do plików')
+        print(correct)
-y_dev = np.asarray(y_dev, dtype=np.int32)
+        print(incorrect)
-y_test = np.asarray(y_test, dtype=np.int32)
+        print(correct / (incorrect + correct))
-y_dev.tofile('./dev-0/out.tsv', sep='\n')
+        df = pd.DataFrame(np.asarray(predictions).reshape(int(len(data) / batch_size) * batch_size))
-y_test.tofile('./test-A/out.tsv', sep='\n')
+        df.reset_index(drop=True, inplace=True)
+        df.to_csv(path, sep="\t", index=False)