GitHub - EzraEdelman/parallel_training: heavily parallelized hyper parameter sweeps, even on one gpu

If you just want to use the code as is, use train.py

Feel free to ignore testing.ipynb and nnx entirely.

Required packages: Jax, Equinox, tqdm

Name		Name	Last commit message	Last commit date
Latest commit History 6 Commits
datasets		datasets
models		models
.gitignore		.gitignore
.python-version		.python-version
LICENSE		LICENSE
README.md		README.md
eq_parallel.py		eq_parallel.py
main.py		main.py
parity_transformer_results.png		parity_transformer_results.png
pyproject.toml		pyproject.toml
temp.py		temp.py
testing.ipynb		testing.ipynb
testing_wrapper.ipynb		testing_wrapper.ipynb
train.py		train.py
uv.lock		uv.lock

Provide feedback