training.py

# -*- coding: utf-8 -*-
"""
Created on Sat Sep 14 17:50:00 2019

@author: HSU, CHIH-CHAO
"""

#try to use nn for crossentropy

import torch
import torch.nn.functional as F

#%% Training the Model
def train(m, device, train_itr, optimizer, epoch, max_epoch):
    m.train()
    corrects, train_loss = 0.0,0
    for batch in train_itr:
        text, target = batch.text, batch.label
        text = torch.transpose(text,0, 1)
        target.data.sub_(1)
        text, target = text.to(device), target.to(device)
        optimizer.zero_grad()
        logit = m(text)
        
        loss = F.cross_entropy(logit, target)
        loss.backward()
        optimizer.step()
        
        train_loss+= loss.item()
        result = torch.max(logit,1)[1]
        corrects += (result.view(target.size()).data == target.data).sum()
    
    size = len(train_itr.dataset)
    train_loss /= size 
    accuracy = 100.0 * corrects/size
  
    return train_loss, accuracy
    
def valid(m, device, test_itr):
    m.eval()
    corrects, test_loss = 0.0,0
    for batch in test_itr:
        text, target = batch.text, batch.label
        text = torch.transpose(text,0, 1)
        target.data.sub_(1)
        text, target = text.to(device), target.to(device)
        
        logit = m(text)
        loss = F.cross_entropy(logit, target)

        
        test_loss += loss.item()
        result = torch.max(logit,1)[1]
        corrects += (result.view(target.size()).data == target.data).sum()
    
    size = len(test_itr.dataset)
    test_loss /= size 
    accuracy = 100.0 * corrects/size
    
    return test_loss, accuracy