Skip to main content

Crate fastokens

Crate fastokens 

Source

Re-exports§

pub use self::added_tokens::AddedTokenInfo;
pub use self::added_tokens::AddedTokens;
pub use self::json_structs::AddedTokenConfig;
pub use self::json_structs::DecoderConfig;
pub use self::json_structs::DecoderKind;
pub use self::json_structs::ModelConfig;
pub use self::json_structs::ModelKind;
pub use self::json_structs::NormalizerConfig;
pub use self::json_structs::NormalizerKind;
pub use self::json_structs::PostProcessorConfig;
pub use self::json_structs::PostProcessorKind;
pub use self::json_structs::PreTokenizerConfig;
pub use self::json_structs::PreTokenizerKind;
pub use self::json_structs::TokenizerConfig;
pub use self::json_structs::TokenizerJson;
pub use self::models::Model;
pub use self::normalizers::Nfc;
pub use self::normalizers::Normalizer;
pub use self::normalizers::Replace;
pub use self::post_processors::PostProcessor;
pub use self::pre_tokenizers::ByteLevel;
pub use self::pre_tokenizers::Pcre2Limits;
pub use self::pre_tokenizers::PreTokenizer;
pub use self::pre_tokenizers::Split;
pub use self::pre_tokenizers::SplitBehavior;
pub use self::tiktoken::CL100K_BASE_PATTERN;
pub use self::tiktoken::KIMI_PATTERN;
pub use self::tiktoken::KIMI_RESERVED_SPECIAL_TOKENS;
pub use self::tiktoken::O200K_BASE_PATTERN;
pub use self::tiktoken::TiktokenConfig;
pub use self::tiktoken::TiktokenFamily;

Modules§

added_tokens
decoders
json_structs
models
normalizers
post_processors
pre_tokenized
pre_tokenizers
tiktoken
Support for the tiktoken model format (OpenAI’s .tiktoken / tiktoken.model files, e.g. cl100k_base, o200k_base, or the tiktoken.model shipped by models such as Kimi-K2).

Structs§

DecodeStream
Stateful incremental decoder.
EncodeSegment
One piece of input for Tokenizer::encode_segments.
Tokenizer
An LLM tokenizer backed by tokenizer.json.
TokenizerOptions
Options applied while constructing a Tokenizer.

Enums§

Error
Errors that can occur when constructing a Tokenizer.

Functions§

decode_stream_step
Advance an incremental decode stream by one or more token IDs.