Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Adds rule of thumb for determining embeddings size #2050

Merged
merged 1 commit into from
May 23, 2022

Conversation

w4nderlust
Copy link
Collaborator

Also changes the default behavior for dense category embed encoder. This has repercussions on documentation, where we should document the rule of thumb for determining size if unspecified: min(512, round(1.6 * x**0.56))

@w4nderlust w4nderlust requested review from justinxzhao, ShreyaR and dantreiman and removed request for justinxzhao May 21, 2022 18:50
@github-actions
Copy link

Unit Test Results

       6 files  ±0         6 suites  ±0   1h 48m 51s ⏱️ + 1m 13s
2 787 tests ±0  2 752 ✔️ ±0    35 💤 ±0  0 ±0 
8 361 runs  ±0  8 252 ✔️ ±0  109 💤 ±0  0 ±0 

Results for commit c1500ab. ± Comparison against base commit 9590ceb.

Copy link
Contributor

@jppgks jppgks left a comment

Choose a reason for hiding this comment

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

Do we have a reference for the rule of thumb?

@w4nderlust
Copy link
Collaborator Author

Do we have a reference for the rule of thumb?

Just some empirical use: https://forums.fast.ai/t/size-of-embedding-for-categorical-variables/42608/2?u=knesgood

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.

None yet

3 participants