Add tokenizer.json at the repository root for consumers that load the fast tokenizer serialization directly, including Infinity v3.
Converted from the original vocabulary and tokenizer configuration using AutoTokenizer with use_fast=True. The resulting JSON is identical to the existing onnx/tokenizer.json at this revision. No weights or tokenizer configuration are changed.
Validation: 156 comparisons of token IDs, offsets, token type IDs and attention masks covering casing, Unicode, sentence pairs, padding and left/right truncation. Infinity 3.0.0 GPU NER inference and all aggregation modes passed using this tokenizer.
