class
Llama::Sampler::MirostatV2
- Llama::Sampler::MirostatV2
- Llama::Sampler::Base
- Reference
- Object
Overview
Mirostat sampler (version 2)
The Mirostat V2 sampler is an improved version of the Mirostat algorithm that requires fewer parameters and is more efficient. It dynamically adjusts sampling to maintain a target entropy level.
Based on the paper: https://arxiv.org/abs/2007.14966
Example:
sampler = Llama::Sampler::MirostatV2.new(42, 5.0, 0.1)Defined in:
llama/sampler/mirostat_v2.crConstructors
-
.new(seed : UInt32, tau : Float32, eta : Float32)
Creates a new Mirostat V2 sampler
Instance methods inherited from class Llama::Sampler::Base
to_unsafe : Pointer(Llama::LibLlama::LlamaSampler)
to_unsafe
Constructor methods inherited from class Llama::Sampler::Base
new(handle : Pointer(LibLlama::LlamaSampler))
new
Constructor Detail
def self.new(seed : UInt32, tau : Float32, eta : Float32)
#
Creates a new Mirostat V2 sampler
Parameters:
- seed: Random seed
- tau: Target entropy (5.0 is a good default)
- eta: Learning rate (0.1 is a good default)
Raises:
- Llama::Error if the sampler cannot be created