Skip to content

Conversation

@gagika
Copy link
Collaborator

@gagika gagika commented Jan 28, 2026

This PR extends the transfer_state_directly utility to support weight synchronization from scanned MaxText model (where layers are stacked in a single tensor) to unscanned MaxText + vLLM models (where layers are separate parameters).

Previously, transfer_state_directly only supported 1-to-1 mapping (Unscanned -> Unscanned). This change adds logic to detect and unroll scanned layers during the transfer process.

  • I have added all the necessary unit tests for my change.
  • I have verified that my change does not break existing code and all unit tests pass.
  • I have added all appropriate doc-strings/documentation.
  • My PR is based on the latest changes of the main branch (if unsure, rebase the code).
  • I have signed the Contributor License Agreement.
  • I have followed Contribution Guidelines.

…e_directly

Dynamically detecting scan dim + path caching.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants