Coverage Status pipeline status Latest Release

Get YOKEd!

A general prototyping, training, and testing harness for pytorch used for models developed under the ArtIMis: Multi-physics/Multi-material Applications and ASC-PEM-EADA(Enabling Agile Design and Assessment) projects.

The YOKE module is divided into submodules, installed in a python environment:

• datasets/
• helpers/
• models/
• metrics/
• losses/
• utils/
• utils/training/
• utils/training/datastep/
• utils/training/epoch/
• lr_schedulers.py
• parellel_utils.py

Helper utilities and examples are under applications:

• harnesses
• makefilelists.py
• filelists
• normalization
• evaluation
• viewers

NOTE: Data for training is not housed within YOKE. The data locations are specified through command-line arguments passed to the programs in harnesses, evaluation, and viewers.

The Python environment is specified through the pyproject.toml file. YOKE is meant to be installed using flit in a minimal python environment.

Setup your base environment and activate it (we use conda):

>> conda create -n <yoke_env_name> python=3.9 flit
>> conda activate <yoke_env_name>
>> cd <yoke_repo_clone_directory>
>> flit install

WARNING

This install process does not guarantee that PyTorch is installed to utilize your GPUs. If you want to ensure that PyTorch is installed to make optimal use of your hardware we suggest manually installing torch prior to installing YOKE with flit.

An alternative to building your Python environment for a particular computing environment is to use containerization. This allows for portability and some degree of isolation from the host system. You can then develop a Yoke application on one system and easily move it to another for larger training runs etc.

Using the Docker runtime, build the container (this could be done as part of a CI pipeline) and drop in to an interative shell:

>> cd <yoke_repo_clone_directory>
>> docker buildx build -f Dockerfile . -t yoke
>> docker run -it yoke /bin/bash

When you want to push your application to a different machine or deploy to a service like Kubernetes, RunAI etc., you can run docker push or docker export along with any transfers of necessary training data.

To run the tests use...

>> pytest
>> pytest --cov
>> pytest --cov --cov-report term-missing

The ruff linter is used in YOKE to enforce coding and formatting standards. To run the linter do

>> ruff check
>> ruff check --preview

You can make ruff fix automatic standards using

>> ruff check --fix
>> ruff check --preview --fix

Use ruff to then check your code formatting and show you what would be adjusted, then fix formatting

>> ruff format --check --diff
>> ruff format

LANL O4863

© 2025. Triad National Security, LLC. All rights reserved.

This program was produced under U.S. Government contract 89233218CNA000001 for Los Alamos National Laboratory (LANL), which is operated by Triad National Security, LLC for the U.S. Department of Energy/National Nuclear Security Administration. All rights in the program are reserved by Triad National Security, LLC, and the U.S. Department of Energy/National Nuclear Security Administration. The Government is granted for itself and others acting on its behalf a nonexclusive, paid-up, irrevocable worldwide license in this material to reproduce, prepare. derivative works, distribute copies to the public, perform publicly and display publicly, and to permit others to do so.