tatm (Transformer Assistive Testbed Module) is a Python library that provides a set of tools to assist in the development of transformer-based (and other architecture) models for foundation model research on on-premise clusters. It currently provides an interface for accessing and manipulating data, and in the future will include features for training models and evaluating models. The library is designed to be modular and extensible, allowing for easy integration of new models, datasets, and training frameworks.
-
Python 3.10 is the minimum supported version.
-
tatmdepends on pytorch, which depends on CUDA. It is recommended to pre-install both pytorch and CUDA prior to installingtatm. Instructions for installing pytorch can be found here.
To install the latest stable version of tatm from GitHub, run the following command:
pip install git+ssh://git@github.com/KempnerInstitute/tatm.git@mainTo install the latest development version of tatm from GitHub, run the following command:
pip install git+ssh://git@github.com/KempnerInstitute/tatm.git@devFor a specific past version of tatm, replace main or dev with the desired version number (i.e. v0.1.0).
The package is not yet available on PyPI. Stay tuned for updates!