Skip to content

README.md

Latest commit

 

History

History
130 lines (87 loc) · 4.62 KB

README.md

File metadata and controls

130 lines (87 loc) · 4.62 KB

Test Suite for Tensor Operations

This test suite validates the tensor operations implemented in babytorch by comparing the results with PyTorch. The tests cover various operations, including addition, multiplication, broadcasting, squeezing and activation functions.

Required Libraries:

The following libraries must be installed in your environment babytorch, torch, numpy, and unittest.

Addition & Multiplication

This file contains tests for scalar, vector, and matrix addition and multiplication operations. It ensures that the results and gradients obtained from babytorch match those from PyTorch.

To run the tests, you'll need:

  • Python 3.x
  • babytorch: A lightweight deep learning framework
  • PyTorch: A popular deep learning framework`

Running the Tests

To run the tests, simply execute the following command in your terminal:

cd tests/tensor_operations/math_operations
python add_mul_tests.py

The following tests will be executed and the results will be displayed in the console:

  1. Addition Tests:

    • Scalar Addition
    • Vector Addition
    • Matrix Addition
    • Scalar-Vector Addition
    • Scalar-Matrix Addition
    • Vector-Matrix Addition
    • Batched Matrix Addition
    • Batched Matrix Addition with Broadcasting

  2. Multiplication Tests:

    • Scalar Multiplication
    • Vector Multiplication
    • Matrix Multiplication
    • Matrix-Matrix Multiplication with Different Shapes
    • Matrix-Vector Multiplication with Different Shapes
    • Batched Matrix Multiplication with Broadcasting

Tensor Squeeze and Unsqueeze

To run the tests, simply execute the following command in your terminal:

cd tests/tensor_operations/math_operations
python squeeze_unsqueeze_tests.py

The following tests will be executed and the results will be displayed in the console:

  1. Squeeze Tests:

    • Vector Squeeze

  2. Unsqueeze Tests:

    • Scalar Unsqueeze
    • Vector Unsqueeze

Unary Operations

This file contains tests for various activation functions such as ReLU, Tanh, and Exponential implemented in babytorch. The tests validate the correctness of the implementations by comparing the results and gradients obtained from babytorch with those from PyTorch.

Running the Tests

To run the tests for activation functions, execute the following command in your terminal:

cd tests/tensor_operations/activation_functions
python unary_operations_tests.py

The following tests will be executed and the results will be displayed in the console:

  1. ReLU Tests:

    • Scalar ReLU
    • Vector ReLU
    • Matrix ReLU

  2. Tanh Tests:

    • Scalar Tanh
    • Vector Tanh
    • Matrix Tanh

  3. Exponential Tests:

    • Scalar Exp
    • Vector Exp
    • Matrix Exp

Iterablility & Subscribtion

This file contains tests for the iterable and subscriptable behavior of tensors in the babytorch library. The tests ensure that tensors can be iterated over and accessed using subscript notation, similar to NumPy arrays.

Running the Tests

To run the tests for iterable and subscriptable behavior, execute the following command in your terminal:

cd tests/tensor_operations/iterability_subscription
python iterability_subscription_tests.py

The following tests will be executed and the results will be displayed in the console:

  1. Iterable Tests:

    • Test that a tensor can be iterated over and returns the expected values.

  2. Subscriptable Get Tests:

    • Test that a specific element of a tensor can be accessed using subscript notation, and it returns the expected value.

  3. Subscriptable Set Tests:

    • Test that a specific element of a tensor can be modified using subscript notation, and the modification is reflected in the tensor.

  4. Multi-dimensional Tests:

    • Test that multi-dimensional tensors can be accessed using multiple subscripts, and the expected values are returned.

Conclusion

This test suite serves as a comprehensive validation tool for the tensor operations and activation functions implemented in babytorch. By comparing the results with PyTorch, it ensures the correctness and consistency of the framework's functionalities. Whether you're developing new features for babytorch or using it in your deep learning projects, these tests provide confidence in the reliability of the library.

Feel free to contribute to babytorch by adding new tests or improving existing ones. Your contributions play a vital role in enhancing the robustness and performance of the framework.

Thank you for using babytorch and happy testing!