main.py

import os
import torch
import copy
from torch.utils.data import DataLoader
import utils.DataProcessing as DP
import utils.LSTMClassifier as LSTMC
import torch.optim as optim
import torch.nn as nn
from torch.autograd import Variable

use_plot = True
use_save = True
if use_save:
    import pickle
    from datetime import datetime

DATA_DIR = 'data'
TRAIN_DIR = 'train_txt'
TEST_DIR = 'test_txt'
TRAIN_FILE = 'train_txt.txt'
TEST_FILE = 'test_txt.txt'
TRAIN_LABEL = 'train_label.txt'
TEST_LABEL = 'test_label.txt'

## parameter setting
epochs = 50
batch_size = 5
use_gpu = torch.cuda.is_available()
learning_rate = 0.01

def adjust_learning_rate(optimizer, epoch):
    lr = learning_rate * (0.1 ** (epoch // 10))
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr
    return optimizer

if __name__=='__main__':
    ### parameter setting
    embedding_dim = 100
    hidden_dim = 50
    sentence_len = 32
    train_file = os.path.join(DATA_DIR, TRAIN_FILE)
    test_file = os.path.join(DATA_DIR, TEST_FILE)
    fp_train = open(train_file, 'r')
    train_filenames = [os.path.join(TRAIN_DIR, line.strip()) for line in fp_train]
    filenames = copy.deepcopy(train_filenames)
    fp_train.close()
    fp_test = open(test_file, 'r')
    test_filenames = [os.path.join(TEST_DIR, line.strip()) for line in fp_test]
    fp_test.close()
    filenames.extend(test_filenames)

    corpus = DP.Corpus(DATA_DIR, filenames)
    nlabel = 8

    ### create model
    model = LSTMC.LSTMClassifier(embedding_dim=embedding_dim,hidden_dim=hidden_dim,
                           vocab_size=len(corpus.dictionary),label_size=nlabel, batch_size=batch_size, use_gpu=use_gpu)
    if use_gpu:
        model = model.cuda()
    ### data processing
    dtrain_set = DP.TxtDatasetProcessing(DATA_DIR, TRAIN_DIR, TRAIN_FILE, TRAIN_LABEL, sentence_len, corpus)

    train_loader = DataLoader(dtrain_set,
                          batch_size=batch_size,
                          shuffle=True,
                          num_workers=4
                         )
    dtest_set = DP.TxtDatasetProcessing(DATA_DIR, TEST_DIR, TEST_FILE, TEST_LABEL, sentence_len, corpus)

    test_loader = DataLoader(dtest_set,
                          batch_size=batch_size,
                          shuffle=False,
                          num_workers=4
                         )

    optimizer = optim.SGD(model.parameters(), lr=learning_rate)
    loss_function = nn.CrossEntropyLoss()
    train_loss_ = []
    test_loss_ = []
    train_acc_ = []
    test_acc_ = []
    ### training procedure
    for epoch in range(epochs):
        optimizer = adjust_learning_rate(optimizer, epoch)

        ## training epoch
        total_acc = 0.0
        total_loss = 0.0
        total = 0.0
        for iter, traindata in enumerate(train_loader):
            train_inputs, train_labels = traindata
            train_labels = torch.squeeze(train_labels)

            if use_gpu:
                train_inputs, train_labels = Variable(train_inputs.cuda()), train_labels.cuda()
            else: train_inputs = Variable(train_inputs)

            model.zero_grad()
            model.batch_size = len(train_labels)
            model.hidden = model.init_hidden()
            output = model(train_inputs.t())

            loss = loss_function(output, Variable(train_labels))
            loss.backward()
            optimizer.step()

            # calc training acc
            _, predicted = torch.max(output.data, 1)
            total_acc += (predicted == train_labels).sum()
            total += len(train_labels)
            total_loss += loss.data[0]

        train_loss_.append(total_loss / total)
        train_acc_.append(total_acc / total)
        ## testing epoch
        total_acc = 0.0
        total_loss = 0.0
        total = 0.0
        for iter, testdata in enumerate(test_loader):
            test_inputs, test_labels = testdata
            test_labels = torch.squeeze(test_labels)

            if use_gpu:
                test_inputs, test_labels = Variable(test_inputs.cuda()), test_labels.cuda()
            else: test_inputs = Variable(test_inputs)

            model.batch_size = len(test_labels)
            model.hidden = model.init_hidden()
            output = model(test_inputs.t())

            loss = loss_function(output, Variable(test_labels))

            # calc testing acc
            _, predicted = torch.max(output.data, 1)
            total_acc += (predicted == test_labels).sum()
            total += len(test_labels)
            total_loss += loss.data[0]
        test_loss_.append(total_loss / total)
        test_acc_.append(total_acc / total)

        print('[Epoch: %3d/%3d] Training Loss: %.3f, Testing Loss: %.3f, Training Acc: %.3f, Testing Acc: %.3f'
              % (epoch, epochs, train_loss_[epoch], test_loss_[epoch], train_acc_[epoch], test_acc_[epoch]))

    param = {}
    param['lr'] = learning_rate
    param['batch size'] = batch_size
    param['embedding dim'] = embedding_dim
    param['hidden dim'] = hidden_dim
    param['sentence len'] = sentence_len

    result = {}
    result['train loss'] = train_loss_
    result['test loss'] = test_loss_
    result['train acc'] = train_acc_
    result['test acc'] = test_acc_
    result['param'] = param

    if use_plot:
        import PlotFigure as PF
        PF.PlotFigure(result, use_save)
    if use_save:
        filename = 'log/LSTM_classifier_' + datetime.now().strftime("%d-%h-%m-%s") + '.pkl'
        result['filename'] = filename

        fp = open(filename, 'wb')
        pickle.dump(result, fp)
        fp.close()
        print('File %s is saved.' % filename)