test

dev-0/out.tsv

@@ -0,0 +1,471 @@
+od
+dawna
+i
+jeszcze
+,
+czy
+na
+niej
+;
+co
+on
+od
+.
+na
+co
+do
+głowy
+na
+jego
+i
+do
+panny
+do
+niego
+:
+po
+chwili
+o
+co
+?
+)
+:
+nie
+było
+mnie
+:
+nie
+ma
+do
+czasu
+i
+nie
+można
+do
+głowy
+do
+głowy
+i
+ja
+?
+.
+na
+drugi
+mnie
+nie
+był
+o
+,
+jak
+by
+od
+niej
+i
+o
+,
+jak
+nie
+można
+od
+niej
+,
+nie
+ma
+pan
+nie
+było
+i
+ja
+,
+co
+on
+jest
+na
+jego
+;
+ale
+co
+do
+domu
+?
+)
+?
+a
+jeżeli
+ja
+go
+o
+do
+niej
+;
+ale
+nie
+mogłem
+i
+na
+co
+?
+na
+niego
+nie
+ma
+.
+potem
+.
+ale
+co
+,
+i
+ja
+go
+do
+mnie
+na
+prawo
+.
+pan
+,
+o
+jego
+na
+nim
+?
+do
+panny
+;
+ale
+,
+co
+,
+ale
+co
+mi
+o
+pani
+nie
+był
+i
+nie
+można
+;
+bo
+pan
+i
+od
+mnie
+:
+potem
+nie
+było
+do
+nich
+:
+od
+mnie
+;
+o
+nim
+;
+co
+,
+a
+potem
+od
+,
+co
+mnie
+,
+nie
+jestem
+na
+prawo
+od
+.
+a
+co
+ja
+?
+i
+ja
+go
+od
+.
+ale
+i
+nawet
+od
+do
+paryża
+:
+po
+obiedzie
+;
+czy
+może
+.
+ja
+,
+ale
+,
+ale
+co
+,
+i
+o
+co
+do
+niego
+i
+jeszcze
+.
+nie
+jestem
+na
+niego
+nie
+do
+nas
+od
+niego
+nie
+ma
+;
+ale
+co
+,
+który
+,
+który
+mi
+,
+a
+jeżeli
+ja
+?
+o
+mnie
+.
+pan
+?
+nie
+chce
+na
+mnie
+:
+na
+drugim
+od
+powietrza
+do
+głowy
+;
+potem
+,
+nie
+mogłem
+nie
+był
+od
+niego
+nie
+,
+ani
+nie
+mam
+na
+nich
+i
+nie
+jest
+o
+,
+jak
+nie
+było
+i
+nie
+można
+?
+a
+nie
+ma
+na
+mnie
+na
+nich
+?
+i
+o
+ile
+nie
+było
+;
+a
+on
+na
+nich
+.
+na
+co
+.
+po
+francusku
+:
+ale
+co
+on
+?
+nie
+był
+,
+nie
+jest
+i
+po
+obiedzie
+o
+,
+co
+na
+niej
+od
+czasu
+,
+i
+ja
+o
+nim
+,
+a
+nie
+będzie
+do
+niej
+i
+ja
+?
+nie
+było
+na
+nich
+,
+który
+o
+mnie
+na
+mnie
+,
+na
+co
+od
+do
+panny
+do
+nich
+.
+a
+nie
+ma
+od
+mnie
+,
+o
+ile
+do
+nich
+nie
+do
+niego
+;
+a
+nie
+było
+i
+nie
+będzie
+i
+pan
+od
+dawna
+?
+po
+chwili
+!
+nawet
+mnie
+.
+i
+jeszcze
+mi
+nie
+na
+nim
+,
+który
+o
+co
+do
+niej
+i
+jeszcze
+od
+do
+nich
+.
+i
+o
+nim
+:
+i
+po
+do
+domu
+do
+niej
+od
+.
+ale
+i
+ja
+mu
+do
+niego
+na
+nim
+od
+.
+ja
+i
+na
+mnie
+:
+potem
+,
+ażeby
+o
+,
+a
+nawet
+o
+nim
+i
+nie
+ma
+od
+.
+pan
+do
+nich
+i
+na
+jego
+jego
+nie
+było
+do
+nas
+,
+co
+?
+)
+o

main.py

@@ -0,0 +1,242 @@
+import numpy as np
+import torch
+from tqdm.notebook import tqdm
+from sklearn.model_selection import train_test_split
+import nltk
+from nltk.tokenize import word_tokenize
+
+nltk.download('punkt')
+
+device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+with open("train/train.tsv", "r") as lalka_path:
+    lines = lalka_path.readlines()
+    
+train, test = train_test_split(lines, test_size = 0.2)
+
+with open("train/train_train.tsv", "w") as out_train_file:
+    for i in train:
+        out_train_file.write(i)
+        
+with open("train/train_test.tsv", "w") as out_test_file:
+    for i in test:
+        out_test_file.write(i)
+
+lalka_path_train= 'train/train_train.tsv'
+lalka_path_valid= 'train/train_test.tsv'
+
+corpora_train = open(lalka_path_train).read()
+
+corpora_train_tokenized = list(word_tokenize(corpora_train))
+corpora_train_tokenized = [token.lower() for token in corpora_train_tokenized]
+
+vocab_itos = sorted(set(corpora_train_tokenized))
+
+vocab_itos = vocab_itos[:15005]
+vocab_itos[15001] = "<UNK>"
+vocab_itos[15002] = "<BOS>"
+vocab_itos[15003] = "<EOS>"
+vocab_itos[15004] = "<PAD>"
+
+vocab_stoi = dict()
+
+for i, token in enumerate(vocab_itos):
+    vocab_stoi[token] = i
+
+NGRAMS = 5
+
+def get_token_id(dataset):
+    token_ids = [vocab_stoi['<PAD>']] * (NGRAMS-1) + [vocab_stoi['<BOS>']]
+    for token in dataset:
+        try:
+            token_ids.append(vocab_stoi[token])
+        except KeyError:
+            token_ids.append(vocab_stoi['<UNK>'])
+    token_ids.append(vocab_stoi['<EOS>'])
+    
+    return token_ids
+
+train_ids = get_token_id(corpora_train_tokenized)
+
+def get_samples(dataset):
+    samples = []
+    for i in range(len(dataset)-NGRAMS):
+        samples.append(dataset[i:i+NGRAMS])
+        
+    return samples
+
+train_ids = get_samples(train_ids)
+
+train_ids = torch.tensor(train_ids, device = device)
+
+corpora_valid = open(lalka_path_valid).read()
+
+corpora_valid_tokenized = list(word_tokenize(corpora_valid))
+corpora_valid_tokenized = [token.lower() for token in corpora_valid_tokenized]
+
+valid_ids = get_token_id(corpora_valid_tokenized)
+
+valid_ids = torch.tensor(get_samples(valid_ids), dtype = torch.long, device = device)
+
+class GRU(torch.nn.Module):
+
+    def __init__(self):
+        super(GRU, self).__init__()
+        self.emb = torch.nn.Embedding(len(vocab_itos),100)
+        self.rec = torch.nn.GRU(100, 256, 1, batch_first = True)
+        self.fc1 = torch.nn.Linear( 256 ,len(vocab_itos))
+        #self.dropout = torch.nn.Dropout(0.5)
+
+    def forward(self, x):
+        emb = self.emb(x)
+        #emb = self.dropout(emb)
+        output, h_n = self.rec(emb)
+        hidden = h_n.squeeze(0)
+        out = self.fc1(hidden)
+        #out = self.dropout(out)
+        return out
+
+lm = GRU().to(device)
+
+criterion = torch.nn.CrossEntropyLoss()
+
+optimizer = torch.optim.Adam(lm.parameters(),lr=0.0001)
+
+BATCH_SIZE = 128
+EPOCHS = 15
+
+def get_ppl(dataset_ids):
+
+    lm.eval()
+
+    batches = 0
+    loss_sum = 0
+    acc_score = 0
+
+    for i in range(0, len(dataset_ids) - BATCH_SIZE + 1, BATCH_SIZE):
+        X = dataset_ids[i : i + BATCH_SIZE, :NGRAMS - 1]
+        Y = dataset_ids[i : i + BATCH_SIZE, NGRAMS - 1]
+        
+        predictions = lm(X)
+        
+        loss = criterion(predictions,Y)
+        loss_sum += loss.item()
+        batches += 1
+
+    return np.exp(loss_sum / batches)
+
+history_ppl_train  = []
+history_ppl_valid  = []
+
+for epoch in range(EPOCHS):
+    
+    batches = 0
+    loss_sum = 0
+    acc_score = 0
+    lm.train()
+
+    for i in tqdm(range(0, len(train_ids)-BATCH_SIZE + 1, BATCH_SIZE)):
+        X = train_ids[i : i + BATCH_SIZE, :NGRAMS - 1]
+        Y = train_ids[i : i + BATCH_SIZE, NGRAMS - 1]
+        
+        predictions = lm(X)
+        loss = criterion(predictions,Y)
+          
+        optimizer.zero_grad()
+        loss.backward()
+        optimizer.step()
+        
+        loss_sum += loss.item()
+        batches += 1
+        
+    ppl_train = get_ppl(train_ids)
+    ppl_valid = get_ppl(valid_ids)
+    
+    history_ppl_train.append(ppl_train)
+    history_ppl_valid.append(ppl_valid)
+    
+    print('epoch: ', epoch)
+    print('train ppl: ', ppl_train)
+    print('valid ppl: ', ppl_valid)
+    print()
+
+tokenized = list(word_tokenize('Gości innych nie widział oprócz spółleśników'))
+tokenized = [token.lower() for token in tokenized]
+
+ids = []
+
+for word in tokenized:
+    if word in vocab_stoi:
+        ids.append(vocab_stoi[word])
+    else:
+        ids.append(vocab_stoi['<UNK>'])
+
+lm.eval()
+
+ids = torch.tensor(ids, dtype = torch.long, device = device)
+
+preds= lm(ids.unsqueeze(0))
+
+vocab_itos[torch.argmax(torch.softmax(preds, 1), 1).item()]
+
+tokenized = list(word_tokenize('Lalka'))
+tokenized = [token.lower() for token in tokenized]
+
+ids = []
+
+for word in tokenized:
+    if word in vocab_stoi:
+        ids.append(vocab_stoi[word])
+    else:
+        ids.append(vocab_stoi['<UNK>'])
+ids = torch.tensor([ids], dtype = torch.long, device = device)
+
+candidates_number = 10
+
+for i in range(30):
+    preds= lm(ids)
+    
+    candidates = torch.topk(torch.softmax(preds,1),candidates_number)[1][0].cpu().numpy()
+    candidate = 15001
+    
+    while candidate > 15000:
+        candidate = candidates[np.random.randint(candidates_number)]
+        
+    print(vocab_itos[candidate])
+    ids =  torch.cat((ids, torch.tensor([[candidate]], device = device)), 1)
+
+with open("dev-0/in.tsv", "r") as dev_path:
+    nr_of_dev_lines = len(dev_path.readlines())
+    
+with open("test-A/in.tsv", "r") as test_a_path:
+    nr_of_test_a_lines = len(test_a_path.readlines())
+    
+with open("dev-0/out.tsv", "w") as out_dev_file:
+    for i in range(nr_of_dev_lines):
+        preds= lm(ids)
+        
+        candidates = torch.topk(torch.softmax(preds,1),candidates_number)[1][0].cpu().numpy()
+        candidate = 15001
+        
+        while candidate > 15000:
+            candidate = candidates[np.random.randint(candidates_number)]
+            
+        print(vocab_itos[candidate])
+        
+        ids =  torch.cat((ids, torch.tensor([[candidate]], device = device)), 1)
+        out_dev_file.write(vocab_itos[candidate] + '\n')
+        
+with open("test-A/out.tsv", "w") as out_test_file:
+    for i in range(nr_of_dev_lines):
+        preds= lm(ids)
+        
+        candidates = torch.topk(torch.softmax(preds,1),candidates_number)[1][0].cpu().numpy()
+        candidate = 15001
+        
+        while candidate > 15000:
+            candidate = candidates[np.random.randint(candidates_number)]
+            
+        print(vocab_itos[candidate])
+        
+        ids =  torch.cat((ids, torch.tensor([[candidate]], device = device)), 1)
+        out_test_file.write(vocab_itos[candidate]  + '\n')

test-A/out.tsv

@@ -0,0 +1,471 @@
+,
+ale
+nie
+mam
+.
+ale
+o
+on
+,
+czy
+,
+co
+on
+do
+głowy
+.
+na
+jego
+.
+pan
+i
+od
+czasu
+do
+domu
+od
+dawna
+;
+ale
+co
+do
+pani
+i
+do
+nich
+nie
+było
+do
+domu
+;
+potem
+mu
+do
+głowy
+?
+ale
+o
+czym
+;
+potem
+do
+niego
+.
+pan
+i
+nawet
+o
+do
+domu
+:
+od
+niej
+nie
+.
+nawet
+od
+czasu
+od
+niego
+na
+niego
+do
+domu
+i
+jeszcze
+i
+nawet
+:
+ale
+na
+nią
+:
+czy
+na
+lewo
+,
+a
+na
+niego
+nie
+było
+.
+i
+o
+co
+na
+nią
+.
+na
+drugim
+do
+niej
+;
+lecz
+nawet
+do
+mnie
+;
+i
+on
+do
+mnie
+i
+nie
+do
+paryża
+:
+od
+nie
+jest
+i
+na
+myśl
+do
+paryża
+,
+co
+,
+bo
+i
+jeszcze
+,
+ażeby
+mu
+do
+pani
+:
+potem
+:
+na
+prawo
+i
+nie
+można
+na
+lewo
+,
+ażeby
+o
+czym
+:
+czy
+pan
+jest
+od
+nas
+od
+powietrza
+?
+i
+pan
+,
+ażeby
+na
+nim
+.
+ale
+o
+nim
+.
+na
+lewo
+od
+niego
+;
+i
+nie
+jest
+do
+niego
+nie
+do
+czasu
+na
+myśl
+,
+o
+na
+mnie
+?
+,
+nie
+ma
+i
+nie
+może
+o
+pani
+,
+ale
+jeszcze
+od
+nie
+było
+do
+paryża
+i
+o
+pani
+?
+i
+o
+mnie
+,
+nie
+mam
+na
+mnie
+,
+jak
+pan
+od
+mnie
+na
+myśl
+i
+nawet
+:
+ale
+na
+niego
+na
+prawo
+od
+czasu
+do
+nas
+do
+panny
+.
+ja
+po
+odejściu
+?
+,
+o
+mnie
+na
+nią
+:
+co
+nie
+może
+być
+i
+ja
+od
+niej
+?
+od
+nie
+było
+na
+nich
+i
+jego
+do
+domu
+nie
+do
+pani
+.
+na
+mnie
+,
+a
+on
+i
+nie
+,
+na
+nim
+?
+a
+potem
+i
+na
+nim
+.
+ja
+na
+nią
+?
+a
+,
+który
+na
+mnie
+;
+potem
+do
+mnie
+i
+od
+czasu
+od
+niego
+.
+a
+potem
+o
+nim
+do
+niego
+i
+na
+nim
+do
+mnie
+.
+a
+jeżeli
+pan
+,
+czy
+pan
+,
+o
+ile
+:
+o
+nie
+,
+nie
+było
+.
+a
+ja
+na
+nią
+na
+prawo
+od
+mnie
+:
+i
+jeszcze
+nie
+do
+mnie
+?
+a
+on
+o
+nim
+do
+domu
+?
+ale
+ja
+mu
+nie
+na
+prawo
+nie
+było
+na
+jego
+,
+i
+ja
+go
+i
+nie
+jest
+,
+ażeby
+na
+niego
+do
+pani
+.
+ale
+i
+cóż
+?
+i
+pan
+?
+,
+czy
+o
+mnie
+:
+na
+prawo
+i
+jeszcze
+,
+co
+do
+paryża
+nie
+na
+.
+nie
+jest
+na
+prawo
+?
+pan
+,
+i
+pan
+na
+nią
+.
+ale
+po
+mnie
+?
+do
+domu
+nie
+na
+niego
+do
+głowy
+;
+ale
+co
+mi
+o
+mnie
+.
+na
+prawo
+.
+a
+jeżeli
+go
+o
+nie
+na
+prawo
+od
+dawna
+do
+paryża
+nie
+.
+a
+nawet
+mnie
+?
+na
+prawo
+?
+na
+nią
+:
+na
+nim
+,
+co
+na
+jego
+nie
+ma
+;
+i