common : compute average token length from vocabulary #17632
+25
−1
Conversation
This file contains hidden or 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
Replace hardcoded token length (8) with dynamic computation based on actual vocabulary. Sample up to 1000 tokens to determine average length, cache result in static variable for one-time cost. Implementation: - Add compute_avg_token_length() helper function - Sample evenly across vocabulary (max 1000 tokens or 10%) - Use static caching to compute only once - Fallback to 8 if computation fails Benefits: - Adapts to any vocabulary automatically - Improves memory allocation accuracy (±25% depending on model) - No runtime overhead after initial computation - Backward compatible with existing models
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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.
Replace hardcoded token length (8 bytes) in
common_sampler_prev_str()with dynamic computation based on actual vocabulary.Problem
common/sampling.cppassumes 8 bytes per token when pre-allocating memory:This is inaccurate for many models - some average 6 bytes (Phi-3), others 10 bytes (Command-R).
Solution
Sample up to 1000 tokens from vocabulary to compute average length. Cache result in static variable for zero runtime overhead.
Testing
Tested with multiple vocabularies:
Builds cleanly, no performance regression.