Platforms to show: All Mac Windows Linux Cross-Platform

Llama

Classes for using llama cpp in Xojo to run local LLMs.

• Classes
  • class LlamaBackendMBS
    • method Constructor
    • method Devices as LlamaDeviceMBS()
    • property Features as Dictionary
    • property Name as String
  • class LlamaBatchMBS
    • method Constructor
    • method Constructor(tokens as Int32, embd as Int32, n_seq_max as Int32)
    • method Constructor(tokens() as Int32)
    • method Destructor
    • method Tokens as Int32()
    • property embd as Ptr
    • property Handle as Integer
    • property logits as Ptr
    • property n_tokens as Int32
  • class LlamaChatMessageMBS
    • method Constructor
    • method Constructor(Role as String, Content as String)
    • method Destructor
    • property Content as String
    • property Role as String
  • class LlamaContextMBS
    • method Ask(Sampler as LlamaSamplerMBS, prompt as String, MaxTokens as Integer = 100) as String
    • method ClearMemory
    • method Constructor
    • method Constructor(Model as LlamaModelMBS, Parameters as LlamaContextParametersMBS = nil)
    • method Decode(batch as LlamaBatchMBS) as Integer
    • method Destructor
    • method Embeddings as Ptr
    • method Embeddings(index as Integer) as Ptr
    • method EmbeddingsSeq(index as Integer) as Ptr
    • method Encode(batch as LlamaBatchMBS) as Integer
    • method Logits as Ptr
    • method Logits(index as Integer) as Ptr
    • method MemoryBreakdownPrint
    • method Performance as LlamaPerformanceContextDataMBS
    • method PerformancePrint
    • method PerformanceReset
    • method SetCausalAttention(causalAttn as boolean)
    • method SetEmbeddings(embeddings as Boolean)
    • method SetThreads(Threads as Integer, ThreadsBatch as Integer)
    • method SetWarmup(warmup as boolean)
    • method Synchronize
    • property Handle as Integer
    • property Model as LlamaModelMBS
    • property n_batch as UInt32
    • property n_ctx as UInt32
    • property n_ctx_seq as UInt32
    • property n_ctx_used as UInt32
    • property n_seq_max as UInt32
    • property n_threads as Integer
    • property n_threads_batch as Integer
    • property n_ubatch as UInt32
    • property pooling_type as Integer
  • class LlamaContextParametersMBS
    • method Constructor
    • method Destructor
    • shared method FlashAttnTypeName(FlashAttnType as Integer) as String
    • property attention_type as Integer
    • property defrag_thold as Single
    • property embeddings as Boolean
    • property flash_attn_type as Integer
    • property kv_unified as Boolean
    • property no_perf as Boolean
    • property n_batch as UInt32
    • property n_ctx as UInt32
    • property n_seq_max as UInt32
    • property n_threads as Int32
    • property n_threads_batch as Int32
    • property n_ubatch as UInt32
    • property offload_kqv as Boolean
    • property op_offload as Boolean
    • property pooling_type as Integer
    • property rope_freq_base as Single
    • property rope_freq_scale as Single
    • property rope_scaling_type as Integer
    • property swa_full as Boolean
    • property type_k as Integer
    • property type_v as Integer
    • property yarn_attn_factor as Single
    • property yarn_beta_fast as Single
    • property yarn_beta_slow as Single
    • property yarn_ext_factor as Single
    • property yarn_orig_ctx as UInt32
  • class LlamaDeviceMBS
    • method Constructor
    • property Description as String
    • property DeviceID as String
    • property MemoryFree as Integer
    • property MemoryTotal as Integer
    • property Name as String
    • property Type as Integer
    • property TypeString as String
  • class LlamaExceptionMBS
  • class LlamaModelMBS
    • method ChatTemplate(Name as String) as String
    • method ClassifierOutputLabel(index as integer) as String
    • method Constructor
    • method Constructor(File as FolderItem, Parameters as LlamaModelParametersMBS = nil)
    • method Constructor(Path as String, Parameters as LlamaModelParametersMBS = nil)
    • method DecoderStartToken as Int32
    • method Destructor
    • method MetaKey(Index as Integer) as String
    • shared method MetaKeyName(Key as Integer) as String
    • method MetaValue(Index as Integer) as String
    • method MetaValue(Key as String) as String
    • method SaveToFile(File as FolderItem)
    • method SaveToFile(Path as String)
    • property ClassifierOutputs as Integer
    • property ctx_train as Integer
    • property Description as String
    • property embd as Integer
    • property embd_inp as Integer
    • property embd_out as Integer
    • property Handle as Integer
    • property HasDecoder as Boolean
    • property HasEncoder as Boolean
    • property head as Integer
    • property head_kv as Integer
    • property IsDiffusion as Boolean
    • property IsHybrid as Boolean
    • property IsRecurrent as Boolean
    • property layer as Integer
    • property Meta as Dictionary
    • property MetaCount as Integer
    • property ParameterCount as Integer
    • property RopeType as Integer
    • property Size as Integer
    • property swa as Integer
    • property Vocab as LlamaVocabMBS
  • class LlamaModelParametersMBS
    • method Constructor
    • method Destructor
    • property check_tensors as Boolean
    • property main_gpu as Int32
    • property no_alloc as Boolean
    • property no_host as Boolean
    • property n_gpu_layers as Int32
    • property split_mode as Integer
    • property use_direct_io as Boolean
    • property use_extra_bufts as Boolean
    • property use_mlock as Boolean
    • property use_mmap as Boolean
    • property vocab_only as Boolean
    • property Yield as Boolean
    • event Progress(progress as Single) as Boolean
  • class LlamaModelQuantizeParametersMBS
    • method Constructor
    • method Destructor
    • property allow_requantize as Boolean
    • property ftype as Integer
    • property keep_split as Boolean
    • property nthread as Int32
    • property only_copy as Boolean
    • property output_tensor_type as Integer
    • property pure as Boolean
    • property quantize_output_tensor as Boolean
    • property token_embedding_type as Integer
  • class LlamaPerformanceContextDataMBS
    • method Constructor
    • method Destructor
    • property n_eval as Int32
    • property n_p_eval as Int32
    • property n_reused as Int32
    • property t_eval_ms as Double
    • property t_load_ms as Double
    • property t_p_eval_ms as Double
    • property t_start_ms as Double
  • class LlamaPerformanceSamplerDataMBS
    • method Constructor
    • method Destructor
    • property n_sample as Int32
    • property t_sample_ms as Double
  • class LlamaSamplerChainParametersMBS
    • method Constructor
    • method Destructor
    • property no_perf as Boolean
  • class LlamaSamplerMBS
    • method Accept(token as Int32)
    • method AddToChain(sampler as LlamaSamplerMBS)
    • method Chain(Index as Integer) as LlamaSamplerMBS
    • method Clone as LlamaSamplerMBS
    • method Constructor
    • method Constructor(parameters as LlamaSamplerChainParametersMBS)
    • method Destructor
    • shared method InitDist(seed as UInt32) as LlamaSamplerMBS
    • shared method InitGreedy as LlamaSamplerMBS
    • shared method InitInfill(vocab as LlamaVocabMBS) as LlamaSamplerMBS
    • shared method InitMinP(p as Single, minKeep as UInt64) as LlamaSamplerMBS
    • shared method InitTemp(p as Single) as LlamaSamplerMBS
    • shared method InitTempExt(t as Single, delta as Single, exponent as Single) as LlamaSamplerMBS
    • shared method InitTopK(k as Int32) as LlamaSamplerMBS
    • shared method InitTopNSigma(n as Single) as LlamaSamplerMBS
    • shared method InitTopP(p as Single, minKeep as UInt64) as LlamaSamplerMBS
    • shared method InitTypical(p as Single, minKeep as UInt64) as LlamaSamplerMBS
    • shared method InitXTC(p as Single, t as Single, minKeep as UInt64, seed as UInt32) as LlamaSamplerMBS
    • method Performance as LlamaPerformanceSamplerDataMBS
    • method PerformancePrint
    • method PerformanceReset
    • method RemoveFromChain(Index as Integer)
    • method Reset
    • method Sample(context as LlamaContextMBS, Index as Integer) as Int32
    • property ChainLength as Integer
    • property Handle as Integer
    • property Name as String
    • property Seed as UInt32
  • class LlamaVocabMBS
    • method Constructor
    • method Destructor
    • method Detokenize(tokens() as Int32, removeSpecial as boolean, unparseSpecial as Boolean) as String
    • method isControl(Token as Int32) as Boolean
    • method isEOG(Token as Int32) as Boolean
    • method Text(Token as Int32) as String
    • method TokenAttributes(Token as Int32) as Integer
    • method Tokenize(text as String, AddSpecial as Boolean = true, ParseSpecial as Boolean = true) as Int32()
    • method TokenScore(Token as Int32) as Single
    • method TokenToPiece(Token as Int32, special as boolean = true) as String
    • property bos as Int32
    • property eot as Int32
    • property Handle as Integer
    • property nl as Int32
    • property n_tokens as Integer
    • property pad as Int32
    • property sep as Int32
    • property type as Integer
• Modules
  • module LlamaMBS
    • method ApplyChatTemplate(Template as String, ChatMessages() as LlamaChatMessageMBS, AddAssistentTokens as boolean) as String
    • method BackendInit
    • method BackendLoad(Path as String) as Boolean
    • method BackendLoadAll
    • method Backends as LlamaBackendMBS()
    • method ChatBuiltInTemplates as String()
    • method Loaded as Boolean
    • method LoadErrorMessage as String
    • method LoadLibrary(File as FolderItem) as boolean
    • method LoadLibrary(Path as String) as boolean
    • method PrintSystemInfo
    • method SetDebugLogger
    • method SetLogger(Logger as LlamaLogMBS)
    • property MaxDevices as UInt64
    • property MaxParallelSequences as UInt64
    • property MaxTensorBufTOverrides as UInt64
    • property SupportsGPUOffload as Boolean
    • property SupportsMLock as Boolean
    • property SupportsMMap as Boolean
    • property SupportsRPC as Boolean
    • property TimeUS as Int64
    • delegate LlamaLogMBS(Level as Integer, Message as String)
• Blog Entries
  • Use Llama.cpp in Xojo
  • MBS Xojo Plugins, version 26.1pr3

Release notes

• Version 26.1
  • Updated LlamaMBS module to use newer headers.
  • Added no_alloc and use_direct_io properties to LlamaModelParametersMBS class.
  • Added MaxTensorBufTOverrides property to LlamaMBS module.
  • Added n_ctx_seq property to LlamaContextMBS class.
  • Added embd_out and embd_inp to LlamaModelMBS class.
  • Added MetaKeyName function and related constants to LlamaModelMBS class.
  • Added LlamaExceptionMBS class.
  • Fixed a memory leak in Tokenize and Detokenize functions in LlamaVocabMBS class.
  • Added SetDebugLogger method to LlamaMBS module.
  • Added LlamaDeviceMBS and LlamaBackendMBS classes.
  • Added Backends method to LlamaMBS module.
• Version 25.5
  • Added LlamaMBS module for using Llama.cpp in Xojo.

The items on this page are in the following plugins: MBS Tools Plugin.