Idiomatic Structural Handling for Special Tokens via Macro-Driven Dispatch in Token Library Struct#260
Closed
sempervictus wants to merge 7 commits into
Closed
Idiomatic Structural Handling for Special Tokens via Macro-Driven Dispatch in Token Library Struct#260sempervictus wants to merge 7 commits into
sempervictus wants to merge 7 commits into
Conversation
This implements the full llguidance integration enabling grammar-constrained inference for structured outputs, tool calling, and custom constraints. Architecture: - TopLevelGrammar serialized via rmp_serde across RPC boundaries - Grammar flows: Server → params.grammar → Runner → GuidanceState → Matcher - Inline correction via logits masking during sampling - Post-process correction via rollback on validation failure Key components: - params.grammar field in SamplingParams for RPC serialization GuidanceState - GuidanceState::new() with Matcher state management - GuidanceState::reset() for proper state cleanup - Rollback counter (MAX_ROLLBACK_ATTEMPTS=3) preventing infinite loops - guidance_failed/guidance_mismatch sets cleared on rollback - Vocab size validation in build_llg_factory() - Lark grammar generation from tools via build_tool_call_lark_grammar() CLI flags: - --enable-tool-grammar: Auto-build LLG grammar from MCP tools - --allow-constraint-api: Accept client-provided structured_outputs/response_format
Expand SpecialTokens usage to cover EOS uses across the codebase to include the chat template. This gates access to the EOS tokens through a single common API providing an interdiction point to add or remove them as needed per-model or family as requried.
- Replace manual EOS token extraction logic with centralized SpecialTokens::new() and idiomatic accessors - Eliminate EosTokenId enum and related complex serialization logic in favor of direct Vec<u32> - Update all callers to use SpecialTokens for tool start/end token IDs - Remove stop_token_ids from SamplingParams and related logic (now handled via SpecialTokens) - Simplify tokenizer config by replacing EosTokenEntry with Option<String> - Add comprehensive SpecialTokens API with category-based accessors, ID/string sets, and search methods
sempervictus
force-pushed
the
specialtokens/pr
branch
from
a908fa2 to
64d9397
Compare
Closed
added 3 commits
March 9, 2026 13:21
Improve the binary example to be a handy extractor for models which developers can use to update special_tokens.rs quickly. Add tags extracted from Qwen3.5 0.8B
Narrow the Common category search specifically to find string dups of actually special tokens (handle "aftermarket" models/merges). Add and test Llama4 and Qwen3.5 MoE
4 tasks
Contributor
Author
|
replaced by #262 |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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.
@guoqingbao the pain of magic strings not passing through code agent tools is over, this should put that mess to bed once and for all. Quite handy for dynamic grammar generation and quashes the presumption of "there is only one EOS" since q35 has several and one of the weird artifacts we see with things like DeepResearch is actually the re-trained model throwing out string token representation of
<|im_start|>due to some SFT quirk when it should have been using the properly masked<\0xFF|im_start|>AddedToken (which iirc also has.special == true).I cant easily drop this in without #232 but that's all set for merge IMO and actually did a bunch of the work for this PR (the logic/structure/macro stuff is all me but it did a nice job locating callers and splicing in the remedy).