Get Started¶
This guide will help you install batchtensor and verify your installation.
Prerequisites¶
batchtensor requires:
- Python 3.10 or later
- PyTorch 2.4 or later
- A compatible operating system (Linux or macOS)
It is highly recommended to install in a virtual environment to keep your system in order.
Installing with uv pip (recommended)¶
The following command installs the latest version of the library:
uv pip install batchtensor
To make the package as slim as possible, only the packages required to use batchtensor are
installed.
It is possible to install all the optional dependencies by running the following command:
uv pip install 'batchtensor[all]'
Installing from source¶
To install batchtensor from source, you can follow the steps below.
Prerequisites¶
The project uses uv for dependency management.
Please refer to the uv documentation for installation instructions.
You can verify the installation by running:
uv --version
Clone the repository¶
git clone git@github.com:durandtibo/batchtensor.git
cd batchtensor
Create a virtual environment¶
It is recommended to create a Python 3.10+ virtual environment.
You can create a virtual environment with uv:
inv create-venv
Alternatively, you can use the Makefile shortcut which also installs all dependencies:
make setup-venv
source .venv/bin/activate
Install dependencies¶
Install all dependencies using uv:
inv install
To install with documentation dependencies:
inv install --docs-deps
Verify the installation¶
Run the test suite to verify everything is working:
inv unit-test --cov
Next Steps¶
After installation, explore the documentation:
- Tensor Operations Guide: Learn about single tensor operations
- Nested Operations Guide: Learn about nested structure operations
- Utils Guide: Learn about utility functions
- API Reference: Browse the complete API
Quick Example¶
Here's a simple example to verify your installation:
>>> import torch
>>> from batchtensor.nested import slice_along_batch
>>> batch = {
... "features": torch.tensor([[1, 2], [3, 4], [5, 6]]),
... "labels": torch.tensor([0, 1, 2]),
... }
>>> # Take the first 2 samples
>>> slice_along_batch(batch, stop=2)
{'features': tensor([[1, 2], [3, 4]]), 'labels': tensor([0, 1])}
If this runs without errors, your installation is successful!
Troubleshooting¶
Import Errors¶
If you encounter import errors, ensure that:
- You're using Python 3.10 or later
- PyTorch is properly installed
- Your virtual environment is activated (if using one)
PyTorch Installation¶
If PyTorch is not installed, install it following the official PyTorch installation guide.