Skip to content

Optimize eager mode evaluation for branchy graphs #593

@pranavm-nvidia

Description

@pranavm-nvidia

When eagerly evaluating different branches of a graph, we may want to optimize by saving the output of the common ancestor of both branches.
We will need to be careful to avoid evaluating too many tensors at once since this could cause OOMs (otherwise the best approach would have been to store values for all intermediate tensors).

Metadata

Metadata

Assignees

Labels

tripyPull request for the tripy project

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions