Fix BPE merge parsing to handle HuggingFace tokenizer.json format #159
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
meta-cla
bot
added
the
CLA Signed
mergennachin
force-pushed
the
fix-bpe-merge-parsing
branch
from
e2e854e to
20b0ba9
Compare
larryliu0820
approved these changes
Dec 8, 2025
|
@mergennachin has imported this pull request. If you are a Meta employee, you can view this in D88675986. |
Summary:
The BPE merge parsing code incorrectly assumed merges were arrays of
two elements (["a", "b"]), but HuggingFace tokenizer.json uses
space-separated strings ("a b") as the standard format.
This fix:
- Adds support for legacy string format: "token1 token2" (standard HF format)
- Keeps support for tuple array format: ["token1", "token2"] (for tokens with spaces)
- Skips #version header lines (matching HuggingFace Rust tokenizers behavior)
The implementation follows the HuggingFace Rust tokenizers library
(huggingface/tokenizers) which handles both formats in
tokenizers/src/models/bpe/serialization.rs.
Added tests for both merge formats to verify correct parsing.
Test Plan:
```
mkdir build-test
cd build-test
cmake ../test
cmake --build . --target test_hf_tokenizer
cd ..
RESOURCES_PATH="test/resources" ./build-test/test_hf_tokenizer
```
Reviewed By: larryliu0820
Differential Revision: D88675986
Pulled By: mergennachin
facebook-github-bot
force-pushed
the
fix-bpe-merge-parsing
branch
from
20b0ba9 to
6c37dc8
Compare
|
@mergennachin has exported this pull request. If you are a Meta employee, you can view the originating Diff in D88675986. |
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Outdated suggestions cannot be applied.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.
The BPE merge parsing code incorrectly assumed merges were arrays of
two elements (["a", "b"]), but HuggingFace tokenizer.json uses
space-separated strings ("a b") as the standard format.
This fix:
The implementation follows the HuggingFace Rust tokenizers library
(huggingface/tokenizers) which handles both formats in
tokenizers/src/models/bpe/serialization.rs.
Added tests for both merge formats to verify correct parsing.
Test Plan: