Skip to content

Conversation

d4l3k
Copy link
Member

@d4l3k d4l3k commented Dec 7, 2024

This adds an E2E integration test that actually runs a training loop with the Manager and Lighthouse pieces.

This only tests the happy path currently and no recovery. Follow up PRs will add failure recovery tests.

To support instrumenting Python tests this adds bindings for Lighthouse.

Test plan:

tristanr@devvm5553 ~/torchft (d4l3k/integ_tests)> pytest torchft/manager_integ_test.py
==================================== test session starts ====================================
platform linux -- Python 3.10.14, pytest-8.3.3, pluggy-1.5.0
rootdir: /home/tristanr/torchft
configfile: pytest.ini
plugins: anyio-4.6.2.post1, typeguard-2.13.3
collected 1 item                                                                            

torchft/manager_integ_test.py .                                                       [100%]

===================================== 1 passed in 2.73s =====================================

The test is pretty fast ~2.73s but likely could be even faster with some tuning. We run 5 steps.

@facebook-github-bot facebook-github-bot added the CLA Signed This label is managed by the Meta Open Source bot. label Dec 7, 2024
@d4l3k d4l3k merged commit 7b93da7 into main Dec 7, 2024
6 checks passed
@d4l3k d4l3k deleted the d4l3k/integ_tests branch December 7, 2024 02:34
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants