nn - bigrams

This commit is contained in:
adnovac 2022-04-29 13:01:36 +02:00
parent 27944ca2c8
commit 3161a6a902
3 changed files with 17971 additions and 17957 deletions

File diff suppressed because it is too large Load Diff

62
run.py
View File

@ -1,4 +1,15 @@
#%% # -*- coding: utf-8 -*-
"""Untitled0.ipynb
Automatically generated by Colaboratory.
Original file is located at
https://colab.research.google.com/drive/1vjpmLsNPjPLM1_5fBGbBYg-ZqdXQeGQH
"""
from google.colab import drive
drive.mount('/content/gdrive/')
# importy # importy
from torchtext.vocab import build_vocab_from_iterator from torchtext.vocab import build_vocab_from_iterator
from torch.utils.data import DataLoader from torch.utils.data import DataLoader
@ -11,7 +22,7 @@ from os.path import exists
vocab_size = 30000 vocab_size = 30000
embed_size = 150 embed_size = 150
#%%
# funkcje pomocnicze # funkcje pomocnicze
def clean(text): def clean(text):
text = str(text).strip().lower() text = str(text).strip().lower()
@ -32,7 +43,7 @@ def get_word_lines_from_data(d):
for line in d: for line in d:
yield get_words_from_line(line) yield get_words_from_line(line)
#%%
class Model(torch.nn.Module): class Model(torch.nn.Module):
def __init__(self, vocabulary_size, embedding_size): def __init__(self, vocabulary_size, embedding_size):
super(Model, self).__init__() super(Model, self).__init__()
@ -45,7 +56,7 @@ class Model(torch.nn.Module):
def forward(self, x): def forward(self, x):
return self.model(x) return self.model(x)
#%%
class Trigrams(torch.utils.data.IterableDataset): class Trigrams(torch.utils.data.IterableDataset):
def __init__(self, data, vocabulary_size): def __init__(self, data, vocabulary_size):
self.vocab = build_vocab_from_iterator( self.vocab = build_vocab_from_iterator(
@ -69,45 +80,48 @@ class Trigrams(torch.utils.data.IterableDataset):
(self.vocab[t] for t in itertools.chain.from_iterable(get_word_lines_from_data(self.data)))) (self.vocab[t] for t in itertools.chain.from_iterable(get_word_lines_from_data(self.data))))
#%%
# ładowanie danych treningowych # ładowanie danych treningowych
train_in = pd.read_csv("train/in.tsv.xz", sep='\t', header=None, encoding="UTF-8", on_bad_lines="skip", quoting=csv.QUOTE_NONE, nrows=300000)[[6, 7]] train_in = pd.read_csv("gdrive/MyDrive/train/in.tsv.xz", sep='\t', header=None, encoding="UTF-8", on_bad_lines="skip", quoting=csv.QUOTE_NONE, nrows=300000)[[6, 7]]
train_expected = pd.read_csv("train/expected.tsv", sep='\t', header=None, encoding="UTF-8", on_bad_lines="skip", quoting=csv.QUOTE_NONE, nrows=300000) train_expected = pd.read_csv("gdrive/MyDrive/train/expected.tsv", sep='\t', header=None, encoding="UTF-8", on_bad_lines="skip", quoting=csv.QUOTE_NONE, nrows=300000)
train_data = pd.concat([train_in, train_expected], axis=1) train_data = pd.concat([train_in, train_expected], axis=1)
train_data = train_data[6] + train_data[0] + train_data[7] train_data = train_data[6] + train_data[0] + train_data[7]
train_data = train_data.apply(clean) train_data = train_data.apply(clean)
train_dataset = Trigrams(train_data, vocab_size) train_dataset = Trigrams(train_data, vocab_size)
#%%
# trenowanie/wczytywanie modelu # trenowanie/wczytywanie modelu
device = 'cuda' if torch.cuda.is_available() else 'cpu' device = 'cuda' if torch.cuda.is_available() else 'cpu'
model = Model(vocab_size, embed_size).to(device) model = Model(vocab_size, embed_size).to(device)
print(device)
if(not exists('model1.bin')): if(not exists('model1.bin')):
data = DataLoader(train_dataset, batch_size=200) data = DataLoader(train_dataset, batch_size=8000)
optimizer = torch.optim.Adam(model.parameters()) optimizer = torch.optim.Adam(model.parameters())
criterion = torch.nn.NLLLoss() criterion = torch.nn.NLLLoss()
model.train() model.train()
step = 0 step = 0
for x, y in data: for i in range(2):
x = x.to(device) print(f"EPOCH {i}=========================")
y = y.to(device) for x, y in data:
optimizer.zero_grad() x = x.to(device)
ypredicted = model(x) y = y.to(device)
loss = criterion(torch.log(ypredicted), y) optimizer.zero_grad()
if step % 100 == 0: ypredicted = model(x)
print(step, loss) loss = criterion(torch.log(ypredicted), y)
step += 1 if step % 100 == 0:
loss.backward() print(step, loss)
optimizer.step() step += 1
loss.backward()
optimizer.step()
torch.save(model.state_dict(), 'model1.bin') torch.save(model.state_dict(), 'model1.bin')
else: else:
model.load_state_dict(torch.load('model1.bin')) model.load_state_dict(torch.load('model1.bin'))
#%%
vocab = train_dataset.vocab vocab = train_dataset.vocab
import nltk
nltk.download('punkt')
def predict(tokens): def predict(tokens):
ixs = torch.tensor(vocab.forward(tokens)).to(device) ixs = torch.tensor(vocab.forward(tokens)).to(device)
out = model(ixs) out = model(ixs)
@ -134,11 +148,11 @@ def predict_file(result_path, data):
f.write(result + "\n") f.write(result + "\n")
print(result) print(result)
#%%
dev_data = pd.read_csv("dev-0/in.tsv.xz", sep='\t', header=None, quoting=csv.QUOTE_NONE)[6] dev_data = pd.read_csv("gdrive/MyDrive/dev-0/in.tsv.xz", sep='\t', header=None, quoting=csv.QUOTE_NONE)[6]
dev_data = dev_data.apply(clean) dev_data = dev_data.apply(clean)
predict_file("dev-0/out.tsv", dev_data) predict_file("dev-0/out.tsv", dev_data)
test_data = pd.read_csv("test-A/in.tsv.xz", sep='\t', header=None, quoting=csv.QUOTE_NONE)[6] test_data = pd.read_csv("gdrive/MyDrive/test-A/in.tsv.xz", sep='\t', header=None, quoting=csv.QUOTE_NONE)[6]
test_data = test_data.apply(clean) test_data = test_data.apply(clean)
predict_file("test-A/out.tsv", test_data) predict_file("test-A/out.tsv", test_data)

File diff suppressed because it is too large Load Diff