1322return new TiktokenTokenizer( 1350=> new TiktokenTokenizer(vocabFilePath, preTokenizer, specialTokens, normalizer, cacheSize); 1370=> new TiktokenTokenizer(vocabStream, preTokenizer, specialTokens, normalizer, cacheSize); 1401return new TiktokenTokenizer(encoder, decoder, vocab, preTokenizer, specialTokens, normalizer, cacheSize); 1465return new TiktokenTokenizer(vocabStream,

21private static readonly Tokenizer _gpt4oTokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

12private static readonly Tokenizer _tokenizer = TiktokenTokenizer.CreateForModel("gpt-4");

39HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4"))); 71HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 13 }); 96HeaderChunker lessThanContext = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 5 }); 99HeaderChunker sameAsContext = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 6 }); 119HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 30 }); 136HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 37 }); 146HeaderChunker chunker = new(new(TiktokenTokenizer.CreateForModel("gpt-4")) { MaxTokensPerChunk = 100 }); 175Tokenizer tokenizer = TiktokenTokenizer.CreateForModel("gpt-4");

16var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

17var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o"); 25var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

16var tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

26Tokenizer tokenizer = TiktokenTokenizer.CreateForModel("gpt-4o");

226private static IngestionChunker<string> CreateChunker() => new HeaderChunker(new(TiktokenTokenizer.CreateForModel("gpt-4")));

43/// Create <see cref="TiktokenTokenizer"/> from tokenizer model file. 47public static TiktokenTokenizer FromPretrained( 53return TiktokenTokenizer.Create(File.OpenRead(modelFilePath), preTokenizer, normalizer: null, specialTokens: _specialTokens);

68var tokenizer = LlamaTokenizerHelper.FromPretrained(modelWeightFolder);

36var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 39var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

81public static ICausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM> LoadModel(string weightFolder, string checkPointName = "model.safetensors.index.json") 91var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 94var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

37var tokenizer = LlamaTokenizerHelper.FromPretrained(originalWeightFolder); 40var pipeline = new CausalLMPipeline<TiktokenTokenizer, LlamaForCausalLM>(tokenizer, model, device);

1879new RegexPreTokenizer(TiktokenTokenizer.P50kBaseRegex(), CodeGenTokenizer.CodeGenSpecialTokens),

117vocabStream, mergesStream, new RegexPreTokenizer(TiktokenTokenizer.P50kBaseRegex(), CodeGenTokenizer.CodeGenSpecialTokens), normalizer: null,

1289private static TiktokenTokenizer CreateForModel( 1344public static TiktokenTokenizer Create( 1364public static TiktokenTokenizer Create( 1385public static async Task<TiktokenTokenizer> CreateAsync( 1417public static async Task<TiktokenTokenizer> CreateAsync( 1443public static TiktokenTokenizer CreateForModel( 1482public static async Task<TiktokenTokenizer> CreateForModelAsync( 1519public static TiktokenTokenizer CreateForModel(string modelName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1529public static TiktokenTokenizer CreateForEncoding(string encodingName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1563throw new ArgumentException($"The encoding name '{encodingName}' is not supported. The only supported encoding names are: {TiktokenTokenizer.Cl100kBaseEncodingName}, {TiktokenTokenizer.P50kBaseEncodingName}, {TiktokenTokenizer.P50kEditEncodingName}, and {TiktokenTokenizer.R50kBaseEncodingName}.", nameof(encodingName));

32return SplitText(text, TiktokenTokenizer.P50kBaseRegex()); 47return SplitText(text, TiktokenTokenizer.P50kBaseRegex());

30var exception = Record.Exception(() => TiktokenTokenizer.CreateForModel(modelName)); 53TiktokenTokenizer externalTokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, preTokenizer: null, normalizer: null);

32public static Tokenizer GPT4 { get; } = TiktokenTokenizer.CreateForModel("gpt-4", _specialTokens); 33public static Tokenizer GPT2 { get; } = TiktokenTokenizer.CreateForModel("gpt2"); 34public static Tokenizer P50kBase { get; } = TiktokenTokenizer.CreateForModel("text-davinci-003"); 35public static Tokenizer R50kBase { get; } = TiktokenTokenizer.CreateForModel("ada"); 36public static Tokenizer P50kEdit { get; } = TiktokenTokenizer.CreateForModel("text-davinci-edit-001"); 37public static Tokenizer GPT4o { get; } = TiktokenTokenizer.CreateForModel("gpt-4o"); 38public static Tokenizer GPT5 { get; } = TiktokenTokenizer.CreateForModel("gpt-5"); 39public static Tokenizer GPT5_1 { get; } = TiktokenTokenizer.CreateForModel("gpt-5.1"); 40public static Tokenizer GPT5_2 { get; } = TiktokenTokenizer.CreateForModel("gpt-5.2"); 41public static Tokenizer Phi4 { get; } = TiktokenTokenizer.CreateForModel("phi-4"); 42public static TiktokenTokenizer GptOss { get; } = TiktokenTokenizer.CreateForModel("gpt-oss-20b"); 50Assert.True(GPT4 is TiktokenTokenizer); 51IReadOnlyDictionary<string, int>? specialTokens = (GPT4 as TiktokenTokenizer)!.SpecialTokens; 55string assemblyName = typeof(TiktokenTokenizer).Assembly.FullName!; 66Tokenizer tokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, GPT4.PreTokenizer, null, specialTokens); 71tokenizer = TiktokenTokenizer.Create(stream, GPT4.PreTokenizer, null, specialTokens); 75tokenizer = await TiktokenTokenizer.CreateAsync(tokenizerDataFileName, GPT4.PreTokenizer, normalizer: null, specialTokens); 80tokenizer = await TiktokenTokenizer.CreateAsync(stream, GPT4.PreTokenizer, normalizer: null, specialTokens); 86tokenizer = TiktokenTokenizer.CreateForModel("gpt-4", stream); 92tokenizer = await TiktokenTokenizer.CreateForModelAsync("gpt-3.5-turbo", stream); 96tokenizer = TiktokenTokenizer.CreateForModel("gpt-4"); 123TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 124TiktokenTokenizer externalTokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, tokenizer.PreTokenizer, null, tiktoken.SpecialTokens); 148TestDecodingWithSpan((tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 165private void TestDecodingWithSpan(TiktokenTokenizer tokenizer, int[] ids, string expectedDecoded) 201TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 232TestDecodingWithSpan((gpt4Tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 244TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 279TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 291foreach (TiktokenTokenizer tokenizer in new[] { GPT4o, GptOss, GPT5, GPT5_1, GPT5_2 }) 344TestDecodingWithSpan((GPT2 as TiktokenTokenizer)!, encoded.ToArray(), text); 363TestDecodingWithSpan((P50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 382TestDecodingWithSpan((P50kEdit as TiktokenTokenizer)!, encoded.ToArray(), text); 401TestDecodingWithSpan((R50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 483Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(modelName); 484Assert.True(tokenizer is TiktokenTokenizer); 497Tokenizer tokenizer = TiktokenTokenizer.CreateForEncoding(encodingName); 498Assert.True(tokenizer is TiktokenTokenizer); 512Tokenizer tokenizer1 = TiktokenTokenizer.CreateForModel(modelName); 514Assert.True(tokenizer is TiktokenTokenizer); 515Assert.True(tokenizer1 is TiktokenTokenizer); 517TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 518TiktokenTokenizer tiktoken1 = (tokenizer1 as TiktokenTokenizer)!; 529Assert.Throws<ArgumentNullException>(() => TiktokenTokenizer.CreateForEncoding(null!)); 530Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("r50k_base_")); 531Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_base_")); 532Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_edit_")); 533Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("cl100k_base_")); 534Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("o200k_base_")); 535Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("o200k_harmony_")); 561Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(name); 562Assert.True(tokenizer is TiktokenTokenizer); 566int entriesCount = GetEncoder((tokenizer as TiktokenTokenizer)!)!.Count; 843private static IReadOnlyDictionary<ReadOnlyMemory<byte>, int>? GetEncoder(TiktokenTokenizer tiktoken) 844=> typeof(TiktokenTokenizer).GetProperty("Encoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<ReadOnlyMemory<byte>, int>; 846private static IReadOnlyDictionary<int, ReadOnlyMemory<byte>>? GetDecoder(TiktokenTokenizer tiktoken) 847=> typeof(TiktokenTokenizer).GetProperty("Decoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<int, ReadOnlyMemory<byte>>; 849private static IReadOnlyDictionary<string, int>? GetVocabulary(TiktokenTokenizer tiktoken) 850=> typeof(TiktokenTokenizer).GetProperty("Vocabulary", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<string, int>;