visiontext

Utilities for deep learning on multimodal data.

• jupyter notebooks / jupyter lab / ipython
• matplotlib
• pandas
• webdataset / tar
• pytorch

Install

Requires python>=3.10, requires pytorch to be installed already, see https://pytorch.org/

pip install visiontext

Full build

Additionally requires libjpeg-turbo and sqlite

pip install visiontext[full]

Dev install

Clone repository and cd into, then:

pip install pytest pytest-cov pylint black[jupyter]
pylint visiontext
pylint tests

# full build
pip install -e .[full]
python -m pytest --cov

# minimal build
pip install -e .
python -m pytest --cov -m "not full"

Changelog

• 0.23.1: Drop python 3.8 support since PyTorch dropped it. Set minimum python version as 3.10.
  • PyTorch disabled the install via conda, change builds to install PyTorch CPU via pip.
• 0.10.1: Test with python 3.12
• 0.8.1: Set minimum python version to 3.8 since PyTorch requires it