| Name | puretorch JSON |
| Version |
1.1.1
JSON |
| download |
| home_page | https://github.com/Dristro/PureTorch |
| Summary | Custom implementation of a Neural Network library using numpy. Now supporting autograd. |
| upload_time | 2025-08-17 19:12:26 |
| maintainer | None |
| docs_url | None |
| author | Dhruv |
| requires_python | >=3.11 |
| license | None |
| keywords |
|
| VCS |
 |
| bugtrack_url |
|
| requirements |
No requirements were recorded.
|
| Travis-CI |
No Travis.
|
| coveralls test coverage |
No coveralls.
|
# PureTorch
**PureTorch** is a NumPy-based implementation of PyTorch. It aims to replicate the functionality and user experience of PyTorch while offering a transparent and lightweight codebase for educational and experimental purposes.
---
# Table of Contents
| Section | Description |
|----------------------|------------------------------------------|
| [Version](#version-info) | Current version and features |
| [Purpose](#purpose) | The what and why of this repository |
| [Structure](#structure) | File structure of various modules |
| [Upcoming Features](#upcoming-features) | Features under development |
| [Setup](#setup) | Instructions for local setup |
| [Report Bugs](#report-bugs) | How to report issues or contribute |
---
# Version Info
**Current Version**: **1.1.0+dev**
### Migration Guide
If you're upgrading from `v0.x.x` to `v1.x.x-dev`, please refer to the [Migration Guide](MIGRATIONGUIDE.md) for detailed instructions on updating your codebase.
The guide covers:
- Breaking changes in the API.
- Updated module structures.
- Examples of how to transition your code.
Click [here](MIGRATIONGUIDE.md) to view the Migration Guide.
<br></br>
### New Features
1. **`puretorch.Tensor`**:
- autograd.tensor wrapper, tracks gradients for efficient back-propagation.
- back-propagation is easier and more convenient.
2. **`puretorch.nn`**:
- neural network modules like: `linear`, `sequential`, more.
3. **puretorch.optim**:
- optimizers like: `SGD`, other (will add more soon).
4. **autograd**:
- autograd is now supported, allowing users to build custom functions and give full control over gradnients.
<br></br>
### Deprecations
1. The **puretorch.layers.x** system is replaced with a simpler, PyTorch-like API.
2. Temporarily removed activations, losses, and optimizers. These will return with updated functionality, supporting Tensors.
<br></br>
> *Runs on CPU only. Might add GPU support later.*
---
# Purpose
Raw implementation of PyTorch-like library using NumPy.
The structure and essence of torch remains the same, but its implemented using NumPy.
---
# Structure
<u><b>New file structure</b></u>:
- puretorch
- nn
- linear
- Perceptron (will be deprecated in v1.1.0)
- sequential
- optim
- optimizer
- sgd
- tensor (autograd.tensor, modified for better nn compatibility)
- autograd
- context
- engine (not in use now. Abstractions of tensor-ops will be added here, along with higher level tensor logic)
- function
- ops
- tensor
Will be adding other layers, activations, losses and optimizers.
---
# Upcoming features
These are the features that im working on, and will soon be a part of PureTorch.
1. More optimizers (like sgd w/ momentum, adam, etc)
2. Loss functions
3. Model summary (like torchinfo.summary())
---
# Setup
> **Note**:
this for setting-up the "dev" branch locally.\
> **For stable installation**: go to the "main" branch's setup guide
### Prerequisites
- Python 3.8 or higher
- Git installed on your system
### Installation Steps
1. **Install the development branch:**
```bash
pip install PureTorch
```
2. **Verify installation in Python**
```Python
import puretorch
print(puretorch.__version__)
```
3. **Verify installation in the terminal**
```bash
python3 -c "import puretorch; print(puretorch.__version__)"
```
If the version indicated is: **1.1.0**, then the package was installed correctly.\
If not, try reinstalling the package (or) check if you installed the stable (vs) development package.
---
# Report Bugs
If you encounter issues or have suggestions, please open an issue via the [GitHub Issues tab](https://github.com/Dristro/PureTorch/issues).
For those interested in contributing:
- Fork this repository.
- Make your changes.
- Open a pull request with a detailed description of your changes.
*Test cases will be added soon to help verify contributions and new features.*
---
Raw data
{
"_id": null,
"home_page": "https://github.com/Dristro/PureTorch",
"name": "puretorch",
"maintainer": null,
"docs_url": null,
"requires_python": ">=3.11",
"maintainer_email": null,
"keywords": null,
"author": "Dhruv",
"author_email": null,
"download_url": "https://files.pythonhosted.org/packages/d9/a9/4033871b4ca1cb7d8af740200c97365cf96a8eca230f4aa185fee86a489b/puretorch-1.1.1.tar.gz",
"platform": null,
"description": "# PureTorch\n\n**PureTorch** is a NumPy-based implementation of PyTorch. It aims to replicate the functionality and user experience of PyTorch while offering a transparent and lightweight codebase for educational and experimental purposes.\n\n---\n\n# Table of Contents\n| Section | Description |\n|----------------------|------------------------------------------|\n| [Version](#version-info) | Current version and features |\n| [Purpose](#purpose) | The what and why of this repository |\n| [Structure](#structure) | File structure of various modules |\n| [Upcoming Features](#upcoming-features) | Features under development |\n| [Setup](#setup) | Instructions for local setup |\n| [Report Bugs](#report-bugs) | How to report issues or contribute |\n\n\n---\n\n# Version Info\n\n**Current Version**: **1.1.0+dev**\n### Migration Guide\nIf you're upgrading from `v0.x.x` to `v1.x.x-dev`, please refer to the [Migration Guide](MIGRATIONGUIDE.md) for detailed instructions on updating your codebase.\n\nThe guide covers:\n- Breaking changes in the API.\n- Updated module structures.\n- Examples of how to transition your code.\n\nClick [here](MIGRATIONGUIDE.md) to view the Migration Guide.\n<br></br>\n\n### New Features\n1. **`puretorch.Tensor`**:\n - autograd.tensor wrapper, tracks gradients for efficient back-propagation.\n - back-propagation is easier and more convenient.\n2. **`puretorch.nn`**:\n - neural network modules like: `linear`, `sequential`, more.\n3. **puretorch.optim**:\n - optimizers like: `SGD`, other (will add more soon).\n4. **autograd**:\n - autograd is now supported, allowing users to build custom functions and give full control over gradnients.\n<br></br>\n\n### Deprecations\n1. The **puretorch.layers.x** system is replaced with a simpler, PyTorch-like API.\n2. Temporarily removed activations, losses, and optimizers. These will return with updated functionality, supporting Tensors.\n<br></br>\n\n> *Runs on CPU only. Might add GPU support later.*\n\n---\n\n# Purpose\nRaw implementation of PyTorch-like library using NumPy.\nThe structure and essence of torch remains the same, but its implemented using NumPy.\n\n---\n\n# Structure\n<u><b>New file structure</b></u>:\n- puretorch\n - nn\n - linear\n - Perceptron (will be deprecated in v1.1.0)\n - sequential\n - optim\n - optimizer\n - sgd\n - tensor (autograd.tensor, modified for better nn compatibility)\n- autograd\n - context\n - engine (not in use now. Abstractions of tensor-ops will be added here, along with higher level tensor logic)\n - function\n - ops\n - tensor\n\nWill be adding other layers, activations, losses and optimizers.\n\n---\n\n# Upcoming features\nThese are the features that im working on, and will soon be a part of PureTorch.\n1. More optimizers (like sgd w/ momentum, adam, etc)\n2. Loss functions\n3. Model summary (like torchinfo.summary())\n\n---\n\n# Setup\n> **Note**:\nthis for setting-up the \"dev\" branch locally.\\\n> **For stable installation**: go to the \"main\" branch's setup guide\n\n### Prerequisites\n- Python 3.8 or higher\n- Git installed on your system\n\n\n### Installation Steps\n1. **Install the development branch:**\n```bash\npip install PureTorch\n```\n2. **Verify installation in Python**\n```Python\nimport puretorch\nprint(puretorch.__version__)\n```\n3. **Verify installation in the terminal**\n```bash\npython3 -c \"import puretorch; print(puretorch.__version__)\"\n```\n\nIf the version indicated is: **1.1.0**, then the package was installed correctly.\\\nIf not, try reinstalling the package (or) check if you installed the stable (vs) development package.\n\n---\n\n# Report Bugs\nIf you encounter issues or have suggestions, please open an issue via the [GitHub Issues tab](https://github.com/Dristro/PureTorch/issues). \n\nFor those interested in contributing:\n- Fork this repository.\n- Make your changes.\n- Open a pull request with a detailed description of your changes.\n\n*Test cases will be added soon to help verify contributions and new features.*\n\n---\n",
"bugtrack_url": null,
"license": null,
"summary": "Custom implementation of a Neural Network library using numpy. Now supporting autograd.",
"version": "1.1.1",
"project_urls": {
"Homepage": "https://github.com/Dristro/PureTorch"
},
"split_keywords": [],
"urls": [
{
"comment_text": null,
"digests": {
"blake2b_256": "2e79faafa15eccb2b5c7b5ec6bd1a5f5a61cfe89d7541f3383dd61fe64249aa3",
"md5": "0b93073bdf07017c4aa898a469edf564",
"sha256": "62ab08a647a7d97e49c166d704a06622fa13cc15beb579df19a81e9353c5d995"
},
"downloads": -1,
"filename": "puretorch-1.1.1-py3-none-any.whl",
"has_sig": false,
"md5_digest": "0b93073bdf07017c4aa898a469edf564",
"packagetype": "bdist_wheel",
"python_version": "py3",
"requires_python": ">=3.11",
"size": 32360,
"upload_time": "2025-08-17T19:12:25",
"upload_time_iso_8601": "2025-08-17T19:12:25.326429Z",
"url": "https://files.pythonhosted.org/packages/2e/79/faafa15eccb2b5c7b5ec6bd1a5f5a61cfe89d7541f3383dd61fe64249aa3/puretorch-1.1.1-py3-none-any.whl",
"yanked": false,
"yanked_reason": null
},
{
"comment_text": null,
"digests": {
"blake2b_256": "d9a94033871b4ca1cb7d8af740200c97365cf96a8eca230f4aa185fee86a489b",
"md5": "23573e7fb8e3c29a9dd8cbad3feb33ce",
"sha256": "feeb487dd26c96029344f96cafd21326625c2f33bcb46ac054b9891a5a831501"
},
"downloads": -1,
"filename": "puretorch-1.1.1.tar.gz",
"has_sig": false,
"md5_digest": "23573e7fb8e3c29a9dd8cbad3feb33ce",
"packagetype": "sdist",
"python_version": "source",
"requires_python": ">=3.11",
"size": 26853,
"upload_time": "2025-08-17T19:12:26",
"upload_time_iso_8601": "2025-08-17T19:12:26.692930Z",
"url": "https://files.pythonhosted.org/packages/d9/a9/4033871b4ca1cb7d8af740200c97365cf96a8eca230f4aa185fee86a489b/puretorch-1.1.1.tar.gz",
"yanked": false,
"yanked_reason": null
}
],
"upload_time": "2025-08-17 19:12:26",
"github": true,
"gitlab": false,
"bitbucket": false,
"codeberg": false,
"github_user": "Dristro",
"github_project": "PureTorch",
"travis_ci": false,
"coveralls": false,
"github_actions": false,
"lcname": "puretorch"
}
