Distribute Lite's code with PyTorch Lightning
See original GitHub issueProposed refactor
Remove this line https://github.com/Lightning-AI/lightning/blob/5025c3ec162b7a1af978bfb19424fd1bb2683012/requirements/pytorch/base.txt#L6
and distribute the Lite source code together with PL when it’s installed
Motivation
When you do
PACKAGE_NAME=pytorch pip install -e .
to create an editable installation of PL. Lite gets installed from PyPI:
$ pip freeze | grep lightning
lightning-lite==1.8.0.post1
-e git+https://github.com/Lightning-AI/lightning.git@5025c3ec162b7a1af978bfb19424fd1bb2683012#egg=pytorch_lightning
...
This means that local changes to Lite are not reflected when developing locally
One downside of this is that if you also do pip install lightning_lite
, you might have conflicting lite installations. The PyPI version will take precedence.
If you enjoy Lightning, check out our other projects! ⚡
-
Metrics: Machine learning metrics for distributed, scalable PyTorch applications.
-
Lite: enables pure PyTorch users to scale their existing code on any kind of device while retaining full control over their own loops and optimization logic.
-
Flash: The fastest way to get a Lightning baseline! A collection of tasks for fast prototyping, baselining, fine-tuning, and solving problems with deep learning.
-
Bolts: Pretrained SOTA Deep Learning models, callbacks, and more for research and production with PyTorch Lightning and PyTorch.
-
Lightning Transformers: Flexible interface for high-performance research using SOTA Transformers leveraging PyTorch Lightning, Transformers, and Hydra.
Issue Analytics
- State:
- Created a year ago
- Comments:6 (6 by maintainers)
That happens because the docker image used already contains PyPI’s
lighting_lite
installation. And given the issue described above:The PyPI version is used over the version packaged inside PL.
Let’s give it a day to see if the rebuilt docker images do not have it installed which would make this issue disappear.
That’s weird! I will have a look. Thanks for reporting