'''
epoch = 1 forward and backward pass of ALL training samples
batch_size = number of training samples in one forward & backward pass
number of = iterations = number of passes, each pass using [batch_size] number of sampeles
e.g. 100 samples, batch_size = 20 --> 100/20 = 5 iterations for 1 epoch
'''
import torch
import torchvision
from torch.utils.data import Dataset, DataLoader
import numpy as np
import math
class WineDataset(Dataset):
def __init__(self):
# data loading
xy = np.loadtxt('./practic_torch/data/wine.csv', delimiter=',', dtype= np.float32, skiprows = 1)
self.x = torch.from_numpy(xy[:, 1:])
self.y = torch.from_numpy(xy[:, [0]]) # n_samples, 1
self.n_samples = xy.shape[0]
def __getitem__(self, index):
return self.x[index], self.y[index]
def __len__(self):
return self.n_samples
dataset = WineDataset()
dataloader = DataLoader(dataset=dataset, batch_size=4, shuffle = True, num_workers=2)
# training loop
num_epochs = 2
total_samples = len(dataset)
print(total_samples)
n_iterations = math.ceil(total_samples / 4)
print(n_iterations)
for epoch in range(num_epochs):
for i, (inputs, labels) in enumerate(dataloader):
# forwar bacward. update
if (i+1) % 5 == 0:
print(f'epoch {epoch+1} / {num_epochs}, step {i+1}/{n_iterations}, inputs {inputs.shape}')
