Skip to content

Conversation

@wwwjn
Copy link
Contributor

@wwwjn wwwjn commented Jan 2, 2026

[ghstack-poisoned]
wwwjn added a commit that referenced this pull request Jan 2, 2026
ghstack-source-id: 557ecd0
Pull Request resolved: #2194
@meta-cla meta-cla bot added the CLA Signed This label is managed by the Meta Open Source bot. label Jan 2, 2026
@wwwjn wwwjn closed this Jan 2, 2026
wwwjn added a commit that referenced this pull request Jan 2, 2026
ghstack-source-id: 557ecd0
Pull Request resolved: #2194
@wwwjn wwwjn reopened this Jan 2, 2026
@wwwjn wwwjn changed the title refactor model registery [rl] refactor model registery Jan 2, 2026
wwwjn added a commit that referenced this pull request Jan 13, 2026
ghstack-source-id: 557ecd0
Pull Request resolved: #2194
f"(model_cls={train_spec.model_cls.__name__}, flavor={model_flavor})"
)
The plugin is loaded at runtime when needed (in generator.py), following
the pattern used in native_sampler.py.
Copy link
Contributor

Choose a reason for hiding this comment

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

where is native_sampler.py?


This function registers TorchTitan models (e.g., Qwen3) with vLLM's ModelRegistry.
It's called automatically when the unified module is imported, or can be called
explicitly for manual registration.
Copy link
Contributor

Choose a reason for hiding this comment

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

why do we need the explicit way -- explicitly calling it would require importing in the first place, which would have already register it?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/8gpu 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.

3 participants