Skip to content
This repository has been archived by the owner on Jul 8, 2024. It is now read-only.

Batching speed spike #20

Merged
merged 1 commit into from
May 17, 2024
Merged

Batching speed spike #20

merged 1 commit into from
May 17, 2024

Conversation

wpfl-dbt
Copy link
Collaborator

Quick spike to investigate the relative speed of batching. Embedded 100 items across 7 loops on each method using all-mpnet-base-v2.

  • Batched embedding took 21.2 s ± 2.34 s per loop
  • Serial embedding took 26.5 s ± 4.57 s per loop
  • Serial embedding where all-mpnet-base-v2 was reloaded each time took 2min 47s ± 8.26 s per loop

Conclusion:

  • Get the model in shared memory ASAP
  • Consider batching at some point in the future once we're at scale

Copy link
Contributor

@gecBurton gecBurton left a comment

Choose a reason for hiding this comment

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

really good analysis ❤️

@wpfl-dbt wpfl-dbt merged commit 4122645 into main May 17, 2024
2 checks passed
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants