abstract class Llama::Sampler::Base

• Llama::Sampler::Base
• Reference
• Object

Overview

Base class for all sampling methods. Sampling is the process of selecting the next token during text generation. This is an abstract class.

Included Modules

• Llama::NativeResource

Direct Known Subclasses

• Llama::Sampler::AdaptiveP
• Llama::Sampler::Dist
• Llama::Sampler::Grammar
• Llama::Sampler::GrammarLazyPatterns
• Llama::Sampler::Greedy
• Llama::Sampler::Infill
• Llama::Sampler::MinP
• Llama::Sampler::Mirostat
• Llama::Sampler::MirostatV2
• Llama::Sampler::Penalties
• Llama::Sampler::Temp
• Llama::Sampler::TempExt
• Llama::Sampler::TopK
• Llama::Sampler::TopNSigma
• Llama::Sampler::TopP
• Llama::Sampler::Typical
• Llama::Sampler::Xtc

Defined in:

llama/sampler/base.cr

Constructors

• .new(handle : Pointer(LibLlama::LlamaSampler))

  Creates a new Sampler instance from a raw pointer.

Instance Method Summary

• #close : Nil

  Releases this sampler when it is not owned by a chain.

• #closed? : Bool

  Returns whether the sampler's native handle has been released.

• #free : Nil

  Releases the underlying C resources.

• #to_unsafe : Pointer(Llama::LibLlama::LlamaSampler)

  Returns the raw pointer to the underlying llama_sampler structure.

• #unsafe_handle! : Pointer(LibLlama::LlamaSampler)

  Returns a checked native handle for internal wrapper use.

Constructor Detail

Instance Method Detail