Skip to content

Transforms

author: Juma Shafara date: "2024-09-04" title: Transform Practice keywords: [Training Two Parameter, Mini-Batch Gradient Decent, Training Two Parameter Mini-Batch Gradient Decent] description: In this lab, you will review how to make a prediction in several different ways by using PyTorch.

Photo by DATAIDEA

Transforms

import dataidea_science as ds
import torch
from torch.utils.data import Dataset
from torch.utils.data import DataLoader

boston_ = ds.loadDataset('boston')

Custom Dataset

class BostonDataset(Dataset): 

    def __init__(self, transform=None):
        # define our dataset
        self.data = boston_
        self.x = self.data.drop('MEDV', axis=1).values
        self.y = self.data.MEDV.values
        self.samples = self.data.shape[0]
        self.transform = transform

    def __getitem__(self, index):
        # access samples
        sample = (self.x[index], self.y[index])

        if self.transform:
            sample = self.transform(sample)

        return sample

    def __len__(self):
        # len(dataset)
        return self.samples

boston_dataset = BostonDataset()

row_1 = boston_dataset[1]
print('Row 1 Features:', row_1[0])
print('Row 1 Outcome:', row_1[1])

length_ = len(boston_dataset)
print('Total Samples: ', length_)

Row 1 Features: [2.7310e-02 0.0000e+00 7.0700e+00 0.0000e+00 4.6900e-01 6.4210e+00
 7.8900e+01 4.9671e+00 2.0000e+00 2.4200e+02 1.7800e+01 3.9690e+02
 9.1400e+00]
Row 1 Outcome: 21.6
Total Samples:  506

DataLoader

boston_dataloader = DataLoader(dataset=boston_dataset,
                               batch_size=3,
                               shuffle=True,
                               num_workers=2)

for batch_no, (x, y) in enumerate(boston_dataloader):
    print(f'Batch: {batch_no}:')
    print(f'Data: {x}')
    print(f'Labels: {y}')

    if batch_no == 0:
        break

Batch: 0:
Data: tensor([[9.7617e-01, 0.0000e+00, 2.1890e+01, 0.0000e+00, 6.2400e-01, 5.7570e+00,
         9.8400e+01, 2.3460e+00, 4.0000e+00, 4.3700e+02, 2.1200e+01, 2.6276e+02,
         1.7310e+01],
        [2.9090e-01, 0.0000e+00, 2.1890e+01, 0.0000e+00, 6.2400e-01, 6.1740e+00,
         9.3600e+01, 1.6119e+00, 4.0000e+00, 4.3700e+02, 2.1200e+01, 3.8808e+02,
         2.4160e+01],
        [5.5007e-01, 2.0000e+01, 3.9700e+00, 0.0000e+00, 6.4700e-01, 7.2060e+00,
         9.1600e+01, 1.9301e+00, 5.0000e+00, 2.6400e+02, 1.3000e+01, 3.8789e+02,
         8.1000e+00]], dtype=torch.float64)
Labels: tensor([15.6000, 14.0000, 36.5000], dtype=torch.float64)

Transformer

class TensorTransformer:

    def __init__(self, dtype=torch.float32):
        self.dtype = dtype

    def __call__(self, sample):
        x_tensor = torch.tensor(data=sample[0], dtype=self.dtype)
        y_tensor = torch.tensor(data=sample[1], dtype=self.dtype)
        return x_tensor, y_tensor

boston_dataset = BostonDataset(transform=TensorTransformer())

row_1 = boston_dataset[1]
print('Row 1 Features:', row_1[0])
print('Row 1 Outcome:', row_1[1])

Row 1 Features: tensor([2.7310e-02, 0.0000e+00, 7.0700e+00, 0.0000e+00, 4.6900e-01, 6.4210e+00,
        7.8900e+01, 4.9671e+00, 2.0000e+00, 2.4200e+02, 1.7800e+01, 3.9690e+02,
        9.1400e+00])
Row 1 Outcome: tensor(21.6000)