# ClumsyGrad
[](https://badge.fury.io/py/clumsygrad)
[](https://clumsygrad.readthedocs.io/en/latest/)
**ClumsyGrad** is a minimal Python library for automatic differentiation, built on top of NumPy. The `Tensor` class has support for creating and expanding a computation graph dynamically with each operation.
## Features
- **Dynamic Computational Graphs**: Graphs are created on the fly.
- **Automatic Differentiation**: Compute gradients automatically using the chain rule.
- **Basic Tensor Operations**: Supports addition, subtraction, multiplication, matrix multiplication, power, exp, etc.
## Installation
You can install ClumsyGrad using pip:
```shell
pip install clumsygrad
```
## Basics
Here's a brief overview of how to use the library:
### Creating Tensors
```python
from clumsygrad.tensor import Tensor, TensorType
import numpy as np
# Create a tensor from a list (defaults to TensorType.INPUT)
a = Tensor([1.0, 2.0, 3.0])
print(a)
# Output: Tensor(id=0, shape=(3,), tensor_type=INPUT, grad_fn=None, requires_grad=False)
# Create a tensor that requires gradients (e.g., a parameter)
b = Tensor([[4.0], [5.0], [6.0]], tensor_type=TensorType.PARAMETER)
print(b)
# Output: Tensor(id=1, shape=(3, 1), tensor_type=PARAMETER, grad_fn=None, requires_grad=True)
# NumPy arrays can also be directly passed to the Tensor object
```
### Performing Operations
```python
from clumsygrad.tensor import Tensor, TensorType
# Define some tensors
x = Tensor([2.0, 3.0], tensor_type=TensorType.PARAMETER)
y = Tensor([4.0, 5.0], tensor_type=TensorType.PARAMETER)
s = Tensor(10.0, tensor_type=TensorType.PARAMETER) # A scalar tensor
# Addition
z_add = x + y
print(f"x + y = {z_add.data}")
# Output: x + y = [6. 7.]
# Element-wise multiplication
z_mul = x * y
print(f"x * y = {z_mul.data}")
# Output: x * y = [ 8. 15.]
# Scalar multiplication
z_scalar_mul = x * s # or x * 10.0
print(f"x * s = {z_scalar_mul.data}")
# Output: x * s = [20. 30.]
# Power
z_pow = x ** 2
print(f"x ** 2 = {z_pow.data}")
# Output: x ** 2 = [4. 9.]
# Matrix multiplication
mat_a = Tensor([[1, 2], [3, 4]], tensor_type=TensorType.PARAMETER)
mat_b = Tensor([[5, 6], [7, 8]], tensor_type=TensorType.PARAMETER)
mat_c = mat_a @ mat_b
print(f"mat_a @ mat_b = \n{mat_c.data}")
# Output: mat_a @ mat_b =
# [[19. 22.]
# [43. 50.]]
```
### Automatic Differentiation (Backpropagation)
For the function `L = sum(a * b + c)`. The gradient of L wrt. a, b and c is calculated as:
```python
from clumsygrad.tensor import Tensor
from clumsygrad.types import TensorType
import numpy as np
# Define input tensors that require gradients
a = Tensor([2.0, 3.0], tensor_type=TensorType.PARAMETER)
b = Tensor([4.0, 1.0], tensor_type=TensorType.PARAMETER)
c = Tensor([-1.0, 2.0], tensor_type=TensorType.PARAMETER)
# Define the computation
# x = a * b => x = [8.0, 3.0]
# y = x + c => y = [7.0, 5.0]
# L = sum(y) => L = 12.0
x = a * b
y = x + c
L = sum(y)
print(f"L = {L.data}")
# Output: L = 12.0
# Perform backpropagation
L.backward()
# Access the gradients
print(f"Gradient of L with respect to a (dL/da): {a.grad}")
# Output: Gradient of L with respect to a (dL/da): [4. 1.]
# (dL/da_i = b_i)
print(f"Gradient of L with respect to b (dL/db): {b.grad}")
# Output: Gradient of L with respect to b (dL/db): [2. 3.]
# (dL/db_i = a_i)
print(f"Gradient of L with respect to c (dL/dc): {c.grad}")
# Output: Gradient of L with respect to c (dL/dc): [1. 1.]
# (dL/dc_i = 1)
```
## Contributing
Contributions are welcome! If you'd like to contribute, please feel free to fork the repository, make your changes, and submit a pull request. You can also open an issue if you find a bug.
## License
This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.
## Documentation
For more detailed information, tutorials, and API reference, you can check out the [official documentation](https://clumsygrad.readthedocs.io/en/latest/).
Raw data
{
"_id": null,
"home_page": null,
"name": "clumsygrad",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.9",
"maintainer_email": "Sayan Gupta <sayangupta540@gmail.com>",
"keywords": "autograd, automatic-differentiation, machine-learning",
"author": null,
"author_email": "Sayan Gupta <sayangupta540@gmail.com>",
"download_url": "https://files.pythonhosted.org/packages/ff/a3/2d344c17112fc5c58644cae52962c0fe115d33db91195c030ebf789d2b77/clumsygrad-0.1.2.tar.gz",
"platform": null,
"description": "# ClumsyGrad\r\n\r\n[](https://badge.fury.io/py/clumsygrad)\r\n[](https://clumsygrad.readthedocs.io/en/latest/)\r\n\r\n**ClumsyGrad** is a minimal Python library for automatic differentiation, built on top of NumPy. The `Tensor` class has support for creating and expanding a computation graph dynamically with each operation.\r\n\r\n## Features\r\n\r\n- **Dynamic Computational Graphs**: Graphs are created on the fly.\r\n- **Automatic Differentiation**: Compute gradients automatically using the chain rule.\r\n- **Basic Tensor Operations**: Supports addition, subtraction, multiplication, matrix multiplication, power, exp, etc.\r\n\r\n## Installation\r\n\r\nYou can install ClumsyGrad using pip:\r\n\r\n```shell\r\npip install clumsygrad\r\n```\r\n\r\n## Basics\r\n\r\nHere's a brief overview of how to use the library:\r\n\r\n### Creating Tensors\r\n\r\n```python\r\nfrom clumsygrad.tensor import Tensor, TensorType\r\nimport numpy as np\r\n\r\n# Create a tensor from a list (defaults to TensorType.INPUT)\r\na = Tensor([1.0, 2.0, 3.0])\r\nprint(a)\r\n# Output: Tensor(id=0, shape=(3,), tensor_type=INPUT, grad_fn=None, requires_grad=False)\r\n\r\n# Create a tensor that requires gradients (e.g., a parameter)\r\nb = Tensor([[4.0], [5.0], [6.0]], tensor_type=TensorType.PARAMETER)\r\nprint(b)\r\n# Output: Tensor(id=1, shape=(3, 1), tensor_type=PARAMETER, grad_fn=None, requires_grad=True)\r\n\r\n# NumPy arrays can also be directly passed to the Tensor object\r\n```\r\n\r\n### Performing Operations\r\n\r\n```python\r\nfrom clumsygrad.tensor import Tensor, TensorType\r\n\r\n# Define some tensors\r\nx = Tensor([2.0, 3.0], tensor_type=TensorType.PARAMETER)\r\ny = Tensor([4.0, 5.0], tensor_type=TensorType.PARAMETER)\r\ns = Tensor(10.0, tensor_type=TensorType.PARAMETER) # A scalar tensor\r\n\r\n# Addition\r\nz_add = x + y\r\nprint(f\"x + y = {z_add.data}\")\r\n# Output: x + y = [6. 7.]\r\n\r\n# Element-wise multiplication\r\nz_mul = x * y\r\nprint(f\"x * y = {z_mul.data}\")\r\n# Output: x * y = [ 8. 15.]\r\n\r\n# Scalar multiplication\r\nz_scalar_mul = x * s # or x * 10.0\r\nprint(f\"x * s = {z_scalar_mul.data}\")\r\n# Output: x * s = [20. 30.]\r\n\r\n# Power\r\nz_pow = x ** 2\r\nprint(f\"x ** 2 = {z_pow.data}\")\r\n# Output: x ** 2 = [4. 9.]\r\n\r\n# Matrix multiplication\r\nmat_a = Tensor([[1, 2], [3, 4]], tensor_type=TensorType.PARAMETER)\r\nmat_b = Tensor([[5, 6], [7, 8]], tensor_type=TensorType.PARAMETER)\r\nmat_c = mat_a @ mat_b\r\nprint(f\"mat_a @ mat_b = \\n{mat_c.data}\")\r\n# Output: mat_a @ mat_b =\r\n# [[19. 22.]\r\n# [43. 50.]]\r\n```\r\n\r\n### Automatic Differentiation (Backpropagation)\r\n\r\nFor the function `L = sum(a * b + c)`. The gradient of L wrt. a, b and c is calculated as:\r\n\r\n```python\r\nfrom clumsygrad.tensor import Tensor\r\nfrom clumsygrad.types import TensorType\r\nimport numpy as np\r\n\r\n# Define input tensors that require gradients\r\na = Tensor([2.0, 3.0], tensor_type=TensorType.PARAMETER)\r\nb = Tensor([4.0, 1.0], tensor_type=TensorType.PARAMETER)\r\nc = Tensor([-1.0, 2.0], tensor_type=TensorType.PARAMETER)\r\n\r\n# Define the computation\r\n# x = a * b => x = [8.0, 3.0]\r\n# y = x + c => y = [7.0, 5.0]\r\n# L = sum(y) => L = 12.0\r\nx = a * b\r\ny = x + c\r\nL = sum(y)\r\n\r\nprint(f\"L = {L.data}\")\r\n# Output: L = 12.0\r\n\r\n# Perform backpropagation\r\nL.backward()\r\n\r\n# Access the gradients\r\nprint(f\"Gradient of L with respect to a (dL/da): {a.grad}\")\r\n# Output: Gradient of L with respect to a (dL/da): [4. 1.]\r\n# (dL/da_i = b_i)\r\n\r\nprint(f\"Gradient of L with respect to b (dL/db): {b.grad}\")\r\n# Output: Gradient of L with respect to b (dL/db): [2. 3.]\r\n# (dL/db_i = a_i)\r\n\r\nprint(f\"Gradient of L with respect to c (dL/dc): {c.grad}\")\r\n# Output: Gradient of L with respect to c (dL/dc): [1. 1.]\r\n# (dL/dc_i = 1)\r\n```\r\n\r\n## Contributing\r\n\r\nContributions are welcome! If you'd like to contribute, please feel free to fork the repository, make your changes, and submit a pull request. You can also open an issue if you find a bug.\r\n\r\n## License\r\n\r\nThis project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.\r\n\r\n## Documentation\r\n\r\nFor more detailed information, tutorials, and API reference, you can check out the [official documentation](https://clumsygrad.readthedocs.io/en/latest/).\r\n",
"bugtrack_url": null,
"license": null,
"summary": "A minimal python library for automatic differentiation, built on top of NumPy.",
"version": "0.1.2",
"project_urls": {
"Homepage": "https://github.com/Sayan-001/ClumsyGrad",
"Issues": "https://github.com/Sayan-001/ClumsyGrad/issues"
},
"split_keywords": [
"autograd",
" automatic-differentiation",
" machine-learning"
],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "668eb5ed645a8ddcf5c9b972b0657500eb5f03adfd22221df574e2f4f840d206",
"md5": "cb7edb9babc4cd1d34886b9942ae20bd",
"sha256": "e353da9a0bfb8f6084bdf922c6788aa62bd5e6db816263ee6e3e8a8e3d473d15"
},
"downloads": -1,
"filename": "clumsygrad-0.1.2-py3-none-any.whl",
"has_sig": false,
"md5_digest": "cb7edb9babc4cd1d34886b9942ae20bd",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.9",
"size": 15241,
"upload_time": "2025-07-08T18:22:55",
"upload_time_iso_8601": "2025-07-08T18:22:55.142269Z",
"url": "https://files.pythonhosted.org/packages/66/8e/b5ed645a8ddcf5c9b972b0657500eb5f03adfd22221df574e2f4f840d206/clumsygrad-0.1.2-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "ffa32d344c17112fc5c58644cae52962c0fe115d33db91195c030ebf789d2b77",
"md5": "92590a9fdf19773be3520039f1da7505",
"sha256": "5757e42a40639cffb4e91882e6599ada856b84c84b117d16186b1fa84e035dc0"
},
"downloads": -1,
"filename": "clumsygrad-0.1.2.tar.gz",
"has_sig": false,
"md5_digest": "92590a9fdf19773be3520039f1da7505",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.9",
"size": 20265,
"upload_time": "2025-07-08T18:22:56",
"upload_time_iso_8601": "2025-07-08T18:22:56.292603Z",
"url": "https://files.pythonhosted.org/packages/ff/a3/2d344c17112fc5c58644cae52962c0fe115d33db91195c030ebf789d2b77/clumsygrad-0.1.2.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-07-08 18:22:56",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "Sayan-001",
"github_project": "ClumsyGrad",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "clumsygrad"
}
JSON
