Tokenizer

Microsoft.ML.Tokenizers adapters and factories.

FunctionDescription
fromBert configCreate a BERT tokenizer. Encoding adds [CLS]/[SEP] when AddSpecialTokens is true.
fromBpe configCreate a BPE tokenizer.
fromFunctions encode decodeCreate a tokenizer from encode and decode functions.
fromSentencePiece configCreate a SentencePiece tokenizer.
fromTiktoken configCreate a Tiktoken tokenizer.
fromWordPiece configCreate a WordPiece tokenizer.
inner tokenizerReturn the wrapped Microsoft.ML.Tokenizers instance.
Tokenizers created with fromFunctions have no backend; this call fails.
wrap innerWrap a Microsoft.ML.Tokenizers tokenizer.

fromBert

fromBert config

Create a BERT tokenizer. Encoding adds [CLS]/[SEP] when AddSpecialTokens is true.

Parameters

  • config : BertConfig

Returns Tokenizer


fromBpe

fromBpe config

Create a BPE tokenizer.

Parameters

  • config : BpeConfig

Returns Tokenizer


fromFunctions

fromFunctions encode decode

Create a tokenizer from encode and decode functions.

Parameters

  • encode : string -> int64 list
  • decode : int64 list -> string

Returns Tokenizer


fromSentencePiece

fromSentencePiece config

Create a SentencePiece tokenizer.

Parameters

  • config : SentencePieceConfig

Returns Tokenizer


fromTiktoken

fromTiktoken config

Create a Tiktoken tokenizer.

Parameters

  • config : TiktokenConfig

Returns Tokenizer


fromWordPiece

fromWordPiece config

Create a WordPiece tokenizer.

Parameters

  • config : WordPieceConfig

Returns Tokenizer


inner

inner tokenizer

Return the wrapped Microsoft.ML.Tokenizers instance. Tokenizers created with fromFunctions have no backend; this call fails.

Parameters

  • tokenizer : Tokenizer

Returns Tokenizer


wrap

wrap inner

Wrap a Microsoft.ML.Tokenizers tokenizer.

Parameters

  • inner : Tokenizer

Returns Tokenizer