Optimal algorithm for _encode_chunk(): 20% faster encoding, with 0.5% better COMPRESSION #84
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…f the chunk in terms of the number of tokens. It uses dynamic programming instead of BPE to encode the chunk. The compression is improved by 0.5%, and it is more than 20% faster.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.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.
This PR reimplements RegexTokenizer._encode_chunk() using dynamic programming to return a guaranteed optimal (minimum number of tokens) tokenization of a chunk.
After training, the vocabulary is fixed. During encoding, _encode_chunk() is called for each chunk to get a tokenization of it. Currently, _encode_chunk() uses something similar to the BPE algorithm used in the training (but with the fixed vocabulary) to determine the tokens. However, this approach is slow and does not guarantee an optimal tokenization of the chunk.
Tested on a 1MB text from wikitext_103, the encoding is more than 20% faster, and the compression is improved by about 0.5%.