class
Llama::Sampler::TopP
- Llama::Sampler::TopP
- Llama::Sampler::Base
- Reference
- Object
Overview
Top-P (nucleus) sampler
The Top-P sampler (also known as nucleus sampling) keeps tokens whose cumulative probability exceeds the threshold P. This helps to filter out the long tail of low-probability tokens.
Example:
sampler = Llama::Sampler::TopP.new(0.95, 1) # Keep tokens until 95% probability is coveredDefined in:
llama/sampler/top_p.crConstructors
-
.new(p : Float32, min_keep : Int32 = 1)
Creates a new Top-P 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(p : Float32, min_keep : Int32 = 1)
#
Creates a new Top-P sampler
Parameters:
- p: The cumulative probability threshold (0.0 to 1.0)
- min_keep: Minimum number of tokens to keep
Raises:
- Llama::Error if the sampler cannot be created