Skip to content

reduce vram/ram usage for vl in calibration phase #1214

@wenhuach21

Description

@wenhuach21

The forward pass should stop at the last required block, rather than forwarding through the entire model.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions