v0.11.1
Added use_bitmask flag to TokenEnforcerTokenizerData which causes the allowed tokens data to be torch tensor bitmask based, for smooth vLLM V1 integration. This is a breaking change if you directly use the TokenEnforcer.get_allowed_tokens() function as its return type has changed.