Skip to content

Conversation

@yaoyu-33
Copy link
Contributor

Description

Adds documentation to the stream_adapter_weights.py example script explaining the verification feature.

Changes

  • Added a "Verification" section to the module docstring that explains:
    • The verification is enabled by default (disable with --no-verify)
    • What the verification does: checks that manually merging adapter weights (lora_A, lora_B) onto base weights using the LoRA formula produces the same result as exporting with merge_adapter_weights=True
    • This ensures the adapter streaming and merge-back paths are mathematically consistent

@copy-pr-bot
Copy link

copy-pr-bot bot commented Jan 23, 2026

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

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.

3 participants