1236return new TiktokenTokenizer( 1264=> new TiktokenTokenizer(vocabFilePath, preTokenizer, specialTokens, normalizer, cacheSize); 1284=> new TiktokenTokenizer(vocabStream, preTokenizer, specialTokens, normalizer, cacheSize); 1315return new TiktokenTokenizer(encoder, decoder, vocab, preTokenizer, specialTokens, normalizer, cacheSize); 1379return new TiktokenTokenizer(vocabStream,

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);

1897new RegexPreTokenizer(TiktokenTokenizer.P50kBaseRegex(), CodeGenTokenizer.CodeGenSpecialTokens),

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

1203private static TiktokenTokenizer CreateForModel( 1258public static TiktokenTokenizer Create( 1278public static TiktokenTokenizer Create( 1299public static async Task<TiktokenTokenizer> CreateAsync( 1331public static async Task<TiktokenTokenizer> CreateAsync( 1357public static TiktokenTokenizer CreateForModel( 1396public static async Task<TiktokenTokenizer> CreateForModelAsync( 1433public static TiktokenTokenizer CreateForModel(string modelName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1443public static TiktokenTokenizer CreateForEncoding(string encodingName, IReadOnlyDictionary<string, int>? extraSpecialTokens = null, Normalizer? normalizer = null) 1473throw 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);

31public static Tokenizer GPT4 { get; } = TiktokenTokenizer.CreateForModel("gpt-4", _specialTokens); 32public static Tokenizer GPT2 { get; } = TiktokenTokenizer.CreateForModel("gpt2"); 33public static Tokenizer P50kBase { get; } = TiktokenTokenizer.CreateForModel("text-davinci-003"); 34public static Tokenizer R50kBase { get; } = TiktokenTokenizer.CreateForModel("ada"); 35public static Tokenizer P50kEdit { get; } = TiktokenTokenizer.CreateForModel("text-davinci-edit-001"); 36public static Tokenizer GPT4o { get; } = TiktokenTokenizer.CreateForModel("gpt-4o"); 43Assert.True(GPT4 is TiktokenTokenizer); 44IReadOnlyDictionary<string, int>? specialTokens = (GPT4 as TiktokenTokenizer)!.SpecialTokens; 48string assemblyName = typeof(TiktokenTokenizer).Assembly.FullName!; 59Tokenizer tokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, GPT4.PreTokenizer, null, specialTokens); 64tokenizer = TiktokenTokenizer.Create(stream, GPT4.PreTokenizer, null, specialTokens); 68tokenizer = await TiktokenTokenizer.CreateAsync(tokenizerDataFileName, GPT4.PreTokenizer, normalizer: null, specialTokens); 73tokenizer = await TiktokenTokenizer.CreateAsync(stream, GPT4.PreTokenizer, normalizer: null, specialTokens); 79tokenizer = TiktokenTokenizer.CreateForModel("gpt-4", stream); 85tokenizer = await TiktokenTokenizer.CreateForModelAsync("gpt-3.5-turbo", stream); 89tokenizer = TiktokenTokenizer.CreateForModel("gpt-4"); 116TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 117TiktokenTokenizer externalTokenizer = TiktokenTokenizer.Create(tokenizerDataFileName, tokenizer.PreTokenizer, null, tiktoken.SpecialTokens); 141TestDecodingWithSpan((tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 158private void TestDecodingWithSpan(TiktokenTokenizer tokenizer, int[] ids, string expectedDecoded) 194TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 225TestDecodingWithSpan((gpt4Tokenizer as TiktokenTokenizer)!, encoded.ToArray(), text); 237TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 272TestDecodingWithSpan((GPT4 as TiktokenTokenizer)!, encoded.ToArray(), text); 298TestDecodingWithSpan((GPT4o as TiktokenTokenizer)!, encoded.ToArray(), text); 306TestDecodingWithSpan((GPT4o as TiktokenTokenizer)!, encoded.ToArray(), text); 334TestDecodingWithSpan((GPT2 as TiktokenTokenizer)!, encoded.ToArray(), text); 353TestDecodingWithSpan((P50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 372TestDecodingWithSpan((P50kEdit as TiktokenTokenizer)!, encoded.ToArray(), text); 391TestDecodingWithSpan((R50kBase as TiktokenTokenizer)!, encoded.ToArray(), text); 444Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(modelName); 445Assert.True(tokenizer is TiktokenTokenizer); 457Tokenizer tokenizer = TiktokenTokenizer.CreateForEncoding(encodingName); 458Assert.True(tokenizer is TiktokenTokenizer); 471Tokenizer tokenizer1 = TiktokenTokenizer.CreateForModel(modelName); 473Assert.True(tokenizer is TiktokenTokenizer); 474Assert.True(tokenizer1 is TiktokenTokenizer); 476TiktokenTokenizer tiktoken = (tokenizer as TiktokenTokenizer)!; 477TiktokenTokenizer tiktoken1 = (tokenizer1 as TiktokenTokenizer)!; 488Assert.Throws<ArgumentNullException>(() => TiktokenTokenizer.CreateForEncoding(null!)); 489Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("r50k_base_")); 490Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_base_")); 491Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("p50k_edit_")); 492Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("cl100k_base_")); 493Assert.Throws<ArgumentException>(() => TiktokenTokenizer.CreateForEncoding("o200k_base_")); 511Tokenizer tokenizer = TiktokenTokenizer.CreateForModel(name); 512Assert.True(tokenizer is TiktokenTokenizer); 516int entriesCount = GetEncoder((tokenizer as TiktokenTokenizer)!)!.Count; 733private static IReadOnlyDictionary<ReadOnlyMemory<byte>, int>? GetEncoder(TiktokenTokenizer tiktoken) 734=> typeof(TiktokenTokenizer).GetProperty("Encoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<ReadOnlyMemory<byte>, int>; 736private static IReadOnlyDictionary<int, ReadOnlyMemory<byte>>? GetDecoder(TiktokenTokenizer tiktoken) 737=> typeof(TiktokenTokenizer).GetProperty("Decoder", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<int, ReadOnlyMemory<byte>>; 739private static IReadOnlyDictionary<string, int>? GetVocabulary(TiktokenTokenizer tiktoken) 740=> typeof(TiktokenTokenizer).GetProperty("Vocabulary", BindingFlags.Instance | BindingFlags.NonPublic)?.GetValue(tiktoken) as IReadOnlyDictionary<string, int>;