Create SentencePieceTokenizer from options object #7403
Merged
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
Member
Author
Contributor
There was a problem hiding this comment.
PR Overview
This PR introduces the ability to create a SentencePieceTokenizer from an options object, enabling support for models defined by JSON or similar non-protobuf formats. Key changes include adding a new SentencePieceOptions class with configurable properties, updating existing model implementations and tests to support options-based creation, and adding new constructors and methods to handle JSON-based initialization.
Reviewed Changes
| File | Description |
|---|---|
| src/Microsoft.ML.Tokenizers/Model/SentencePieceOptions.cs | Adds a new options class to configure SentencePiece model properties. |
| test/Microsoft.ML.Tokenizers.Tests/UnigramTests.cs | Introduces tests ensuring the created tokenizer from JSON behaves as expected and validates shifted token IDs. |
| src/Microsoft.ML.Tokenizers/Model/SentencePieceUnigramModel.cs | Updates constructor and internal methods to support options-based initialization and renames internal counter for normalized text. |
| src/Microsoft.ML.Tokenizers/Model/SentencePieceBpeModel.cs | Adds a new constructor that validates unsupported normalization data and initializes the vocabulary based on options. |
| src/Microsoft.ML.Tokenizers/Model/SentencePieceBaseModel.cs | Introduces a new constructor from options and changes token ID properties from read-only to mutable. |
| src/Microsoft.ML.Tokenizers/Model/SentencePieceTokenizer.cs | Adds an overload of the Create method that instantiates a tokenizer using the options object. |
Copilot reviewed 7 out of 7 changed files in this pull request and generated 3 comments.
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## main #7403 +/- ##
==========================================
+ Coverage 68.93% 68.97% +0.03%
==========================================
Files 1480 1481 +1
Lines 273273 273666 +393
Branches 28234 28287 +53
==========================================
+ Hits 188393 188751 +358
- Misses 77510 77521 +11
- Partials 7370 7394 +24
Flags with carried forward coverage won't be shown. Click here to find out more.
|
ericstj
reviewed
Feb 27, 2025
michaelgsharp
approved these changes
Feb 27, 2025
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
3 participants
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.
We have extended support for the SentencePiece tokenizer, which traditionally uses a protobuf data file (
tokenizer.model). This update introduces the ability to create a tokenizer object by passing an options object containing tokenizer data such as vocabulary and normalization. With this change, users can initialize a tokenizer without loading a protobuf file, enabling them to source tokenizer data from formats like JSON files and other alternatives.