spltr 0.3.1

A simple PyTorch-based data loader and splitter.

What is it?

Spltr is a simple PyTorch-based data loader and splitter. It may be used to load i) arrays and ii) matrices or iii) Pandas DataFrames and iv) CSV files containing numerical data with subsequent split it into train, test (validation) subsets in the form of PyTorch DataLoader objects. The special emphesis was given to ease of usage and automation of many data-handling procedures.

Originally it was developed in order to speed up a data preparation stage for number of trivial ML tasks. Hope it may be useful for you as well.

Main Features

Spltr.process_x|y : converts loaded Input/Target data into PyTorch Tensors with ability to i) preview, ii) define tensor dtype, iii) set the desired device of returned tensor (CPU/GPU), iv) use selected rows/columns from Input/Target data sources or process a single data table (for CSV and Pandas DataFrame only).

Spltr.reshape_xy : reshapes subsets.

Spltr.split_data : splits data subsets into train, test (validation) PyTorch DataLoader objects.

Spltr.clean_space : optimizes memory by deleting unnecessary variables.

Installation

pip install spltr

License

OSI Approved :: MIT License

Documentation

https://github.com/maksymsur/Spltr

Dependencies

• torch >= 1.1.0

• numpy >= 1.16.4

• pandas >= 0.24.2

Example of usage

Hereunder we'll build a simple neural network and describe how Spltr may be used in the process.

Loading and reading an Iris dataset to be used as an example. The dataset may be found at: https://github.com/maksymsur/Spltr/blob/master/dataset/iris_num.csv

import pandas as pd

db = pd.read_csv('/your_path_here/iris_num.csv')
print(db.info())

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 150 entries, 0 to 149
Data columns (total 5 columns):
sepal.length    150 non-null float64
sepal.width     150 non-null float64
petal.length    150 non-null float64
petal.width     150 non-null float64
variety         150 non-null int64
dtypes: float64(4), int64(1)
memory usage: 5.9 KB
None

By building the network we'll try to predict the 'variety' column. This column identifies Iris species: Setosa = 0, Versicolor = 1 and Virginica = 2. Let's verify that it is comprised from the mentioned 3 unique values.

print(db.variety.unique())

[0 1 2]

Making X (Input) dataset excluding the 'variety' column.

base_db = db.iloc[:,:-1]
print(base_db.head())

   sepal.length  sepal.width  petal.length  petal.width
0           5.1          3.5           1.4          0.2
1           4.9          3.0           1.4          0.2
2           4.7          3.2           1.3          0.2
3           4.6          3.1           1.5          0.2
4           5.0          3.6           1.4          0.2

Making y (Target) dataset including i) the 'variety' column and ii) another column: 'petal.width'. The latter will be excluded by Spltr later just to demonstrate how easily we may modify loaded datasets.

target_db = db.iloc[:,-2:]
print(target_db.head())

   petal.width  variety
0          0.2        0
1          0.2        0
2          0.2        0
3          0.2        0
4          0.2        0

Loading necessary packages and initiating Spltr by including X,y datasets into it.

Note that for Spltr to work properly torch, numpy and pandas shall be installed and loaded.

import torch
import numpy as np
from spltr import Spltr

splt = Spltr(base_db,target_db)

Preprocessing X (Input data) by converting it into PyTorch Tensors. Also by indicating nrows={0:150} we show how easily rows may be selected for reading. I'm not going to apply any data normalization procedures just to avoid overloading of this example.

splt.process_x(preview=True, nrows={0:150})

[X]: The following DataFrame was loaded:

   sepal.length  sepal.width  petal.length  petal.width
0           5.1          3.5           1.4          0.2
1           4.9          3.0           1.4          0.2

[X]: Tensor of shape torch.Size([150, 4]) was created

Preprocessing y (Target dataset) by converting it into PyTorch Tensors as well.

At this stage we also indicate that only 2nd column from the y dataset will be used. Thus, by specifying usecols = 1 ('variety' index) we exclude the 'petal.width' column with index 0. Note, usecols may also work with strings and expression usecols = 'variety' is also viable.

splt.process_y(preview=True, usecols=1)

[y]: The following DataFrame was loaded:

0    0
1    0
Name: variety, dtype: int64

[y]: Tensor of shape torch.Size([150]) was created

Splitting the dataset into train - 30%, test - 20% which makes validation set equal to 50% (calculated automatically: val = 100% - train - test) and initializing permutation of the data.

splt.split_data(splits=(0.3,0.2), perm_seed=8)

[X,y]: The Data is splited into 3 datasets of length: Train 45, Test 30, Validation 75.
       DataLoaders with Batch Size 1 and Shufle False are created

Now let's clean unnecessary variables saved in the memory. That step may be especially useful if you are dealing with a huge datasets and don't want for X,y tensors to share the memory with X,y DataLoader objects.

splt.clean_space()

All variables are deleted. Only Train-Test(Validation) data is left

Setting up a very simple neural network. Pls note that the network architecture is comprised only to demonstrate how Spltr may be adopted. That's not an optimal way to solve Iris classification problem.

from torch import nn, optim
import torch.nn.functional as F

class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()

        self.input = nn.Linear(4,8)
        self.output = nn.Linear(8,3)

    def forward(self,x):

        x = F.softsign(self.input(x))
        x = self.output(x)
        return x

model = Net()
criterion = nn.CrossEntropyLoss(reduction='none')
optimizer = optim.RMSprop(model.parameters())

for n in range(10):
    train_loss = 0.0

    # Fitting a dataset for Training

    for train_data, train_target in splt.xy_train:
        model.zero_grad()
        train_result = model(train_data.float())
        loss = criterion(train_result, train_target)
        train_loss += loss.item()
        loss.backward()
        optimizer.step()

    test_correct = 0
    with torch.no_grad():

        # Fitting a dataset for Testing

        for test_data, test_target in splt.xy_test:
            test_result = model(test_data.float())
            _, predicted = torch.max(test_result, 1)
            test_correct += (predicted == test_target).sum().item()

    print(f'Epoch {n+1}. Train loss: ', round(train_loss/len(splt.xy_train), 5))
    print(f'Testing: Correctly identified samples {test_correct} out of {len(splt.xy_test)}', end='\n\n')

val_correct = 0

with torch.no_grad():

    # Fitting a dataset for Validation

    for val_data, val_target in splt.xy_val:
        val_result = model(val_data.float())
        _, val_predicted = torch.max(val_result, 1)
        val_correct += (val_predicted == val_target).sum().item()

print(f'VALIDATION: Correctly identified samples {val_correct} out of {len(splt.xy_val)}', end='\n\n')

Epoch 1. Train loss:  1.01279
Testing: Correctly identified samples 14 out of 30

Epoch 2. Train loss:  0.57477
Testing: Correctly identified samples 14 out of 30

Epoch 3. Train loss:  0.48212
Testing: Correctly identified samples 14 out of 30

Epoch 4. Train loss:  0.44074
Testing: Correctly identified samples 16 out of 30

Epoch 5. Train loss:  0.41144
Testing: Correctly identified samples 24 out of 30

Epoch 6. Train loss:  0.38251
Testing: Correctly identified samples 29 out of 30

Epoch 7. Train loss:  0.34853
Testing: Correctly identified samples 29 out of 30

Epoch 8. Train loss:  0.31831
Testing: Correctly identified samples 29 out of 30

Epoch 9. Train loss:  0.29182
Testing: Correctly identified samples 29 out of 30

Epoch 10. Train loss:  0.27387
Testing: Correctly identified samples 30 out of 30

VALIDATION: Correctly identified samples 74 out of 75

As you might have noticed, we obtained pretty good results even by using a very few samples (45) for training the network.