Class LocalTokenizerProcessor
java.lang.Object
uno.anahata.asi.gemini.tokenizer.LocalTokenizerProcessor
-
Field Summary
Fields -
Constructor Summary
ConstructorsConstructorDescriptionLocalTokenizerProcessor(com.google.genai.proto.SentencepieceModel.ModelProto model) -
Method Summary
Modifier and TypeMethodDescriptionprivate voidaddNewCandidate(List<Symbol> symbols, PriorityQueue<MergeCandidate> pq, int left, int right) private intprivate booleanisByteId(int id) private booleanisMergeCandidateValid(List<Symbol> symbols, MergeCandidate symbol) private StringmaybeMerge(String a, String b) private Stringprivate StringreplaceSentencePieceSeparator(String pieceText) private intsymbolToID(Symbol symbol)
-
Field Details
-
log
-
userDefined
-
uTrie
-
pieces
-
reserved
-
byte2Token
-
idToByte
-
model
private com.google.genai.proto.SentencepieceModel.ModelProto model -
unkID
private int unkID
-
-
Constructor Details
-
LocalTokenizerProcessor
public LocalTokenizerProcessor(com.google.genai.proto.SentencepieceModel.ModelProto model)
-
-
Method Details
-
encode
- Throws:
IllegalStateException
-
decodeIds
-
normalize
-
addNewCandidate
private void addNewCandidate(List<Symbol> symbols, PriorityQueue<MergeCandidate> pq, int left, int right) -
maybeMerge
-
isMergeCandidateValid
-
symbolToID
-
isByteId
private boolean isByteId(int id) -
replaceSentencePieceSeparator
-
convertHexValue
-

