Skip to content

Cache and gen#13

Merged
neelnanda-io merged 12 commits into
mainfrom
cache_and_gen
Oct 4, 2022
Merged

Cache and gen#13
neelnanda-io merged 12 commits into
mainfrom
cache_and_gen

Conversation

@anshradh

Copy link
Copy Markdown
Collaborator

This has working implementations of sampling (with various modifications to logits, i.e. temperature, top-k, top-p, etc.) and greedy decoding, both with and without caching past keys and values (cache offers 3-4x speedup). These are also all batched. Will put up a PR for beam search next

@anshradh
anshradh marked this pull request as ready for review September 16, 2022 21:11
@anshradh
anshradh marked this pull request as draft September 16, 2022 21:20
@anshradh
anshradh marked this pull request as ready for review September 16, 2022 22:03
@neelnanda-io

Copy link
Copy Markdown
Collaborator

Thanks and sorry for taking so long on this! Did a bunch of refactoring and integration into my new changes, and just merged in :)

@neelnanda-io
neelnanda-io merged commit c50b003 into main Oct 4, 2022
@anshradh

anshradh commented Oct 4, 2022

Copy link
Copy Markdown
Collaborator Author

No worries and sorry for not keeping it more up-to-date!

arthurdupe pushed a commit to ArthurConmy/TransformerLens that referenced this pull request Jul 30, 2023
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.

2 participants