Add support for prompt lookup decoding during generation #3917
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Implements support for Prompt Lookup Decoding by exposing a new generation config parameter called
prompt_lookup_num_tokens
. Compatible with transformer version >= 4.37.0.In scenarios where the prompt is long and the output generated might re-use a lot of common ngrams, this can speedup token generation by near 2x - 2.4x. However, in scenarios where that may not be the case, such as open-ended questions, it leads to a 10% decrease in tokens per second.
Demo
https://drive.google.com/file/d/1E8qq8HnJBhL7GOuFDuMdih_GY1aAVwEC/view?usp=sharing
Script to Reproduce Demo