def load(self):
# Data Transformations
transform = transforms.Compose([
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
# Dataset and Creating Train/Test Split
train_set = torchvision.datasets.CIFAR10(root='./data',
train=True,
download=True,
transform=transform)
test_set = torchvision.datasets.CIFAR10(root='./data',
train=False,
download=True,
transform=transform)
# Dataloader Arguments & Test/Train Dataloaders
dataloader_args = dict(shuffle=True,
batch_size=self.batch_size_cuda,
num_workers=self.num_workers,
pin_memory=True) if has_cuda() else dict(
shuffle=True,
batch_size=self.batch_size_cpu)
self.train_loader = torch.utils.data.DataLoader(
train_set, **dataloader_args)
self.test_loader = torch.utils.data.DataLoader(test_set,
**dataloader_args)
def load(self):
# Get Train and Test Data
train_set, test_set = self._dataset()
# Dataloader Arguments & Test/Train Dataloaders
dataloader_args = dict(shuffle=True, batch_size=self.batch_size_cpu)
if has_cuda():
dataloader_args.update(batch_size=self.batch_size_cuda,
num_workers=self.num_workers,
pin_memory=True)
self.train_loader = torch.utils.data.DataLoader(
train_set, **dataloader_args)
self.test_loader = torch.utils.data.DataLoader(test_set,
**dataloader_args)
import os
import pdb
import sys
import torch.utils.data
from torchvision import datasets, transforms
from utils.has_cuda import *
device = has_cuda()
root_path = "C:/Users/gajanana_ganjigatti/Documents/Gaju_data/Quest/eva4/S12/tiny-imagenet-200/tiny-imagenet-200"
dataloader_args = dict(batch_size=512, num_workers=2,
pin_memory=True) if device else dict(batch_size=4)
imagenet_traindir = os.path.join(root_path, 'train')
imagenet_valdir = os.path.join(root_path, 'val')
imagenet_testdir = os.path.join(root_path, 'test')
imagenet_mean, imagenet_std = (0.485, 0.456, 0.406), (0.229, 0.224, 0.225)
def tinyImgNet_dataloader(split='train'):
if split == 'train':
data_transform = datasets.ImageFolder(
imagenet_traindir,
transform=transforms.Compose([
transforms.Pad(padding=1, padding_mode="edge"),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(20),
transforms.RandomCrop(size=(64, 64), padding=4),
# transforms.RandomErasing(scale=(0.16, 0.16), ratio=(1, 1)),
transforms.ToTensor(),