Fix bug in tokenizer created with gguf metadata

[guoqingbao]

This PR address the bug in gguf tokenizer (tokenizer built from gguf metadata) related to #1436.

Tested cases:

Qwen3 Arch

target/debug/mistralrs-server -i gguf -m /home/data -f DeepSeek-R1-0528-Qwen3-8B-Q2_K.gguf

Qwen2 Arch

target/debug/mistralrs-server -i gguf -m /home/data -f qwen2.5-7b-instruct-q2_k.gguf

LLama/Mistral Arch

target/debug/mistralrs-server -i gguf -m /home/data/Devstral-Small-2505 -f Devstral-Small-2505-Q2_K.gguf

Summary by CodeRabbit

• Refactor

  • Simplified special token handling by adding them directly during tokenizer setup and enhanced support for token type metadata.
  • Improved BPE tokenizer processing with explicit regex-based splitting and consistent configuration of tokenization components.
  • Removed external tokenizer override for specific model architectures based on metadata.

• Bug Fixes

  • Enhanced recognition and management of special tokens beyond the standard bos, eos, and unk types.

[coderabbitai]

Walkthrough

The changes refactor GGUF tokenizer conversion by integrating special token handling directly into tokenizer construction, removing an intermediate helper and related struct field. Token type metadata is now used to identify additional special tokens. Additionally, logic for overriding the tokenizer for "qwen3" models using external repositories is removed, simplifying the pipeline.

Changes

File(s)	Change Summary
mistralrs-core/src/gguf/gguf_tokenizer.rs	Refactored GGUF tokenizer conversion: removed add_special_tokens helper, updated struct and function signatures, added token type metadata handling, and unified special token logic.
mistralrs-core/src/pipeline/gguf.rs	Removed logic for fetching and overriding the tokenizer from external repositories for "qwen3" architectures.

Sequence Diagram(s)

sequenceDiagram
    participant GGUFFile as GGUF File
    participant GGUFTokenizer as GGUF Tokenizer Converter
    participant HFTokenizer as HF Tokenizer

    GGUFFile->>GGUFTokenizer: Provide tokenizer data and metadata
    GGUFTokenizer->>GGUFTokenizer: Extract special tokens (bos, eos, unk, and by token type)
    GGUFTokenizer->>HFTokenizer: Construct tokenizer with special tokens integrated

Loading

Possibly related PRs

• Temporary fix for qwen3 gguf tokenizer #1433: Introduced the "qwen3" tokenizer override logic that this PR removes, reversing the external tokenizer replacement behavior.

Suggested reviewers

• EricLBuehler

Poem

In the warren where token bunnies hop,
Special tokens now land right on top.
No more fetching from distant lands,
The pipeline flows by local hands.
With metadata clear and code made neat,
This tokenizer’s change is quite a treat! 🐇✨

---

📜 Recent review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 9108af2 and dd0961b.

📒 Files selected for processing (1)

• mistralrs-core/src/gguf/gguf_tokenizer.rs (7 hunks)

⏰ Context from checks skipped due to timeout of 90000ms (8)

• GitHub Check: Check (windows-latest, stable)
• GitHub Check: Test Suite (ubuntu-latest, stable)
• GitHub Check: Clippy
• GitHub Check: Check (macOS-latest, stable)
• GitHub Check: Check (ubuntu-latest, stable)
• GitHub Check: Docs
• GitHub Check: Test Suite (macOS-latest, stable)
• GitHub Check: Test Suite (windows-latest, stable)

🔇 Additional comments (3)

mistralrs-core/src/gguf/gguf_tokenizer.rs (3)

101-112: LGTM! Efficient special token identification using token types.

The logic correctly identifies special tokens by checking token types != 1 and adds them appropriately. The implementation is clean and handles the case where token_types length matches tokens length properly.

---

190-194: LGTM! Direct special token addition simplifies the logic.

The refactoring to add special tokens directly within the unigram tokenizer function is cleaner than the previous helper approach. The loop through bos, eos, and unk tokens is concise and correct.

---

260-267: LGTM! Clean special token handling for BPE tokenizer.

The direct addition of bos, eos, and conditional unk tokens is well-implemented. The null check for unk token is appropriate since it's optional.

✨ Finishing Touches

• 📝 Generate Docstrings

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

• Review comments: Directly reply to a review comment made by CodeRabbit. Example:
  • I pushed a fix in commit <commit_id>, please review it.
  • Explain this complex logic.
  • Open a follow-up GitHub issue for this discussion.
• Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
  • @coderabbitai explain this code block.
  • @coderabbitai modularize this function.
• PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
  • @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
  • @coderabbitai read src/utils.ts and explain its main purpose.
  • @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
  • @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

• @coderabbitai pause to pause the reviews on a PR.
• @coderabbitai resume to resume the paused reviews.
• @coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
• @coderabbitai full review to do a full review from scratch and review all the files again.
• @coderabbitai summary to regenerate the summary of the PR.
• @coderabbitai generate docstrings to generate docstrings for this PR.
• @coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
• @coderabbitai resolve resolve all the CodeRabbit review comments.
• @coderabbitai configuration to show the current CodeRabbit configuration for the repository.
• @coderabbitai help to get help.

Other keywords and placeholders

• Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
• Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
• Add @coderabbitai anywhere in the PR title to generate the title automatically.

CodeRabbit Configuration File (.coderabbit.yaml)

• You can programmatically configure CodeRabbit by adding a .coderabbit.yaml file to the root of your repository.
• Please see the configuration documentation for more information.
• If your editor has YAML language server enabled, you can add the path at the top of this file to enable auto-completion and validation: # yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json

Documentation and Community

• Visit our Documentation for detailed information on how to use CodeRabbit.
• Join our Discord Community to get help, request features, and share feedback.
• Follow us on X/Twitter for updates and announcements.

[github-actions]

Code Metrics Report

===============================================================================
 Language            Files        Lines         Code     Comments       Blanks
===============================================================================
 C Header                3           62           53            0            9
 CSS                     1          473          408           14           51
 Dockerfile              1           42           23           10            9
 HTML                    1           73           61            4            8
 JavaScript              7         1248          936          174          138
 JSON                   14          123          122            0            1
 Makefile                1            6            5            0            1
 Python                 88         4164         3518          161          485
 Shell                   1           63           26           18           19
 Plain Text              3         3723            0         2413         1310
 TOML                   21          769          707           10           52
 YAML                    2           21           19            2            0
-------------------------------------------------------------------------------
 Jupyter Notebooks       3            0            0            0            0
 |- Markdown             2           77           32           31           14
 |- Python               2          205          178            1           26
 (Total)                            282          210           32           40
-------------------------------------------------------------------------------
 Markdown               60         5243            0         4008         1235
 |- BASH                11          123          117            2            4
 |- JSON                 2           42           42            0            0
 |- Python               7          121          109            0           12
 |- Rust                22          757          634            1          122
 |- TOML                 2           75           63            0           12
 (Total)                           6361          965         4011         1385
-------------------------------------------------------------------------------
 Rust                  377       132640       118009         2921        11710
 |- Markdown           176         2996           25         2662          309
 (Total)                         135636       118034         5583        12019
===============================================================================
 Total                 583       148650       123887         9735        15028
===============================================================================

[coderabbitai]

Actionable comments posted: 2

🧹 Nitpick comments (2)

mistralrs-core/src/gguf/gguf_tokenizer.rs (2)

101-102: Document the token type convention.

The code assumes token type 1 represents normal tokens and other values represent special tokens. This convention should be documented for clarity:

-    //token type other than 1 treated as special token
+    // Token type convention: 1 = normal token, other values = special tokens
+    // This follows the GGUF specification for token type metadata

---

234-238: Consider documenting the regex pattern source.

The regex pattern appears to be the standard GPT-2 tokenization pattern. Consider adding a comment to document its origin:

+    // GPT-2 standard regex pattern for tokenization
+    // Handles contractions, words, numbers, and special characters
     let split = Split::new(
         SplitPattern::Regex("(?i:'s|'t|'re|'ve|'m|'ll|'d)|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+".to_string()),

📜 Review details

Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between 3d1b29b and 9108af2.

📒 Files selected for processing (2)

• mistralrs-core/src/gguf/gguf_tokenizer.rs (7 hunks)
• mistralrs-core/src/pipeline/gguf.rs (1 hunks)

⏰ Context from checks skipped due to timeout of 90000ms (8)

• GitHub Check: Docs
• GitHub Check: Clippy
• GitHub Check: Check (ubuntu-latest, stable)
• GitHub Check: Check (windows-latest, stable)
• GitHub Check: Test Suite (windows-latest, stable)
• GitHub Check: Test Suite (ubuntu-latest, stable)
• GitHub Check: Check (macOS-latest, stable)
• GitHub Check: Test Suite (macOS-latest, stable)

🔇 Additional comments (2)

mistralrs-core/src/pipeline/gguf.rs (1)

381-382: Good simplification - tokenizer no longer needs to be mutable.

The removal of the qwen3-specific tokenizer override logic simplifies the pipeline and reduces external dependencies. This aligns well with the improved GGUF tokenizer conversion that now handles special tokens properly.

mistralrs-core/src/gguf/gguf_tokenizer.rs (1)

147-194: Clean refactoring of unigram tokenizer special token handling.

The direct addition of special tokens simplifies the code and removes the need for the intermediate collection pattern.

[coderabbitai]

💡 Verification agent

🧩 Analysis chain

Verify ByteLevel configuration parameters.

All ByteLevel components (pre-tokenizer, decoder, post-processor) are configured with (false, false, false) parameters. Please verify these settings match the expected behavior for GGUF BPE tokenizers:

• add_prefix_space: false
• trim_offsets: false
• use_regex: false

---

🌐 Web query:

What are the correct ByteLevel parameters for GPT-2 style BPE tokenizers in the tokenizers library?

💡 Result:

For GPT-2 style BPE tokenizers in the tokenizers library (from Hugging Face), the typical ByteLevel parameters used are as follows:

• add_prefix_space=True
• use_regex=True
• trim_offsets=True

When you instantiate the ByteLevel pre-tokenizer, it looks like this:

from tokenizers.pre_tokenizers import ByteLevel

pre_tokenizer = ByteLevel(add_prefix_space=True, use_regex=True, trim_offsets=True)

Explanation:

• add_prefix_space=True adds a leading space before tokenizing, which helps with consistent tokenization especially for the first word.
• use_regex=True enables using regex-based splitting for tokens.
• trim_offsets=True trims offsets for cleaner span alignment, which is useful when mapping tokens back to original text.

This setup mirrors the original GPT-2 tokenizer's behavior using ByteLevel BPE.

If you need the full example or more details, let me know!

---

Update ByteLevel parameters to GPT-2 defaults

The ByteLevel pre-tokenizer, decoder, and post-processor are currently all configured with (false, false, false), but GPT-2–style BPE in Hugging Face’s tokenizers library uses:

• add_prefix_space = true
• use_regex = true
• trim_offsets = true

Please update in mistralrs-core/src/gguf/gguf_tokenizer.rs (around lines 240–257):

- PreTokenizerWrapper::ByteLevel(ByteLevel::new(false, false, false)),
+ PreTokenizerWrapper::ByteLevel(ByteLevel::new(true, true, true)),

- decoder: ByteLevel::new(false, false, false),
+ decoder: ByteLevel::new(true, true, true),

- post_processor: ByteLevel::new(false, false, false),
+ post_processor: ByteLevel::new(true, true, true),

This ensures consistent GPT-2–style tokenization, decoding, and post-processing.

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	// example:
	// "type": "ByteLevel",
	// "add_prefix_space": false,
	// "trim_offsets": false,
	// "use_regex": false
	let pre_tokenizer = Sequence::new(vec![
	PreTokenizerWrapper::Split(split),
	PreTokenizerWrapper::ByteLevel(ByteLevel::new(false, false, false)),
	]);
	tokenizer.with_pre_tokenizer(Some(pre_tokenizer));
	tokenizer.with_decoder(Some(decoders::byte_level::ByteLevel::new(
	false, false, false,
	)));
	tokenizer.with_post_processor(Some(processors::byte_level::ByteLevel::new(
	false, false, false,
	)));
	// example:
	// "type": "ByteLevel",
	// "add_prefix_space": false,
	// "trim_offsets": false,
	// "use_regex": false
	let pre_tokenizer = Sequence::new(vec![
	PreTokenizerWrapper::Split(split),
	PreTokenizerWrapper::ByteLevel(ByteLevel::new(true, true, true)),
	]);
	tokenizer.with_pre_tokenizer(Some(pre_tokenizer));
	tokenizer.with_decoder(Some(decoders::byte_level::ByteLevel::new(
	true, true, true,
	)));
	tokenizer.with_post_processor(Some(processors::byte_level::ByteLevel::new(
	true, true, true,
	)));

🤖 Prompt for AI Agents

In mistralrs-core/src/gguf/gguf_tokenizer.rs around lines 240 to 257, the
ByteLevel pre-tokenizer, decoder, and post-processor are all instantiated with
parameters set to false (false, false, false). Update these parameters to match
the GPT-2 style defaults by setting add_prefix_space, use_regex, and
trim_offsets all to true. This means changing the ByteLevel::new calls to
ByteLevel::new(true, true, true) for the pre-tokenizer, decoder, and
post-processor to ensure consistent GPT-2 style tokenization behavior.

[coderabbitai]

⚠️ Potential issue

Potential panic from chained unwraps in token type extraction.

The code uses multiple unwrap() calls that could panic if the metadata format is unexpected. Consider using proper error handling:

-        let vtypes: &Vec<Value> = md_get("tokenizer.ggml.token_type")
-            .unwrap()
-            .to_vec()
-            .unwrap();
-        let v: Vec<i32> = vtypes.iter().map(|v| v.to_i32().unwrap()).collect();
+        let vtypes = md_get("tokenizer.ggml.token_type")?
+            .to_vec()?;
+        let v: Vec<i32> = vtypes
+            .iter()
+            .map(|v| v.to_i32())
+            .collect::<Result<Vec<_>, _>>()?;

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	let vtypes: &Vec<Value> = md_get("tokenizer.ggml.token_type")
	.unwrap()
	.to_vec()
	.unwrap();
	let v: Vec<i32> = vtypes.iter().map(|v| v.to_i32().unwrap()).collect();
	token_types.extend(v);
	// Replace chained unwraps with ?-based error handling
	- let vtypes: &Vec<Value> = md_get("tokenizer.ggml.token_type")
	- .unwrap()
	- .to_vec()
	- .unwrap();
	- let v: Vec<i32> = vtypes.iter().map(|v| v.to_i32().unwrap()).collect();
	+ let vtypes = md_get("tokenizer.ggml.token_type")?
	+ .to_vec()?;
	+ let v: Vec<i32> = vtypes
	+ .iter()
	+ .map(|v| v.to_i32())
	+ .collect::<Result<Vec<_>, _>>()?;
	token_types.extend(v);

🤖 Prompt for AI Agents

In mistralrs-core/src/gguf/gguf_tokenizer.rs around lines 83 to 88, the code
uses chained unwrap() calls on metadata extraction that can cause panics if the
expected data is missing or malformed. Replace these unwraps with proper error
handling by checking the presence and validity of the metadata at each step,
using match or if let constructs to handle errors gracefully and return or
propagate meaningful error messages instead of panicking.