pytorch dataloader

def create_dataloader(self):
        print("creating data loaders")
        loaders = {}
        for s in self.data_params['sets']:
            if s == 'train':
                tranform = self.create_transform(self.data_params['mean'], self.data_params['std'],
                                                 new_size=self.data_params['resize'])
                dataset = datasets.ImageFolder(os.path.join(self.data_params['data_path'], 'training'), tranform)
                loaders[s] = torch.utils.data.DataLoader(dataset,
                                                         self.training_params['batch_size'] * self.exp_params[
                                                             'num_gpus'], shuffle=True)
            elif s == 'val':
                tranform = self.create_transform(self.data_params['mean'], self.data_params['std'],
                                                 new_size=self.data_params['resize'])
                dataset = datasets.ImageFolder(os.path.join(self.data_params['data_path'], 'val'), tranform)
                loaders[s] = torch.utils.data.DataLoader(dataset,
                                                         self.training_params['batch_size'] * self.exp_params[
                                                             'num_gpus'], shuffle=False)
        self.loaders = loaders

    def create_transform(self, mean=[0, 0, 0], std=[1, 1, 1], new_size=None):
        transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize(mean=mean, std=std)])
        if new_size:
            shape_transform = transforms.Resize(new_size)
            transform = transforms.Compose([shape_transform, transform])

        return transform