Skip to content

Conversation

@daniellepintz
Copy link
Contributor

Summary:
Add a config to test weight sync for Qwen3 30B MoE model
Note: the test will fail with this config until we update the torchtitan version.

Differential Revision: D89833611

@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Dec 27, 2025
@meta-codesync
Copy link

meta-codesync bot commented Dec 27, 2025

@daniellepintz has exported this pull request. If you are a Meta employee, you can view the originating Diff in D89833611.

daniellepintz added a commit that referenced this pull request Dec 27, 2025
Summary: A new stable version of torchtitan was just released (v0.2.1) corresponding to the commit here. The new version includes a [fix](pytorch/torchtitan@fb54971) to a bug which will fix our weight sync test for Qwen 30B MoE (#682)

Differential Revision: D89833771
daniellepintz added a commit that referenced this pull request Dec 27, 2025
Summary:
Pull Request resolved: #683

A new stable version of torchtitan was just released (v0.2.1) corresponding to the commit here. The new version includes a [fix](pytorch/torchtitan@fb54971) to a bug which will fix our weight sync test for Qwen 30B MoE (#682)

Differential Revision: D89833771
@JenniferWang
Copy link
Contributor

JenniferWang commented Dec 29, 2025

This test is used for two purposes

  • Correctness check for onboarding new model and therefore the test can run on any model config
  • Correctness check for changing infra code (e.g. Generator) and therefore it's a test and shall use a small model config.

In general,

  1. Please do not introduce new test fixture that fails the test.
  2. Do not introduce new test fixture unless it tests new aspect of our infra.

Also, this test is not configured to run continuously yet ... so it's not used to test infra regression :(

Copy link
Contributor

@allenwang28 allenwang28 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Review automatically exported from Phabricator review in Meta.

daniellepintz added a commit that referenced this pull request Jan 16, 2026
Summary:
Pull Request resolved: #682

Add a config to test weight sync for Qwen3 30B MoE model
Note: the test will fail with this config until we update the torchtitan version.

Reviewed By: allenwang28

Differential Revision: D89833611
daniellepintz added a commit that referenced this pull request Jan 17, 2026
Summary:
Pull Request resolved: #682

Add a config to test weight sync for Qwen3 30B MoE model
Note: the test will fail with this config until we update the torchtitan version.

Reviewed By: allenwang28

Differential Revision: D89833611
facebook-github-bot pushed a commit that referenced this pull request Jan 17, 2026
Summary:

Add a config to test weight sync for Qwen3 30B MoE model
Note: the test will fail with this config until we update the torchtitan version.

Reviewed By: allenwang28

Differential Revision: D89833611
facebook-github-bot pushed a commit that referenced this pull request Jan 17, 2026
Summary:

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
daniellepintz added a commit that referenced this pull request Jan 19, 2026
Summary:
Pull Request resolved: #682

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
daniellepintz added a commit that referenced this pull request Jan 19, 2026
Summary:
Pull Request resolved: #682

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
daniellepintz added a commit that referenced this pull request Jan 21, 2026
Summary:
Pull Request resolved: #682

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
facebook-github-bot pushed a commit that referenced this pull request Jan 21, 2026
Summary:

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
daniellepintz added a commit that referenced this pull request Jan 23, 2026
Summary:
Pull Request resolved: #682

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
Summary:
Pull Request resolved: #682

- Add a config to test weight sync for Qwen3 30B MoE model
- Add the config to run in the CI on every commit to main
- Fix docstring in test_policy_update.py

Reviewed By: allenwang28

Differential Revision: D89833611
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. fb-exported meta-exported

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants