SamplerFactory class final
Builds a Sampler from declarative SamplerParams.
Default chain order, mirroring llama.cpp's recommended layout:
logit_bias → penalties → dry → top-n-sigma → top-k → typical-p → top-p → min-p → xtc → (dynamic-)temperature → grammar → infill → mirostat | adaptive-p | dist
Stages whose config is disabled are skipped. greedy short-circuits to
argmax sampling and ignores everything else.
Pass model when the chain may use grammar / DRY / infill / logit-bias /
mirostat-v1 — those samplers need the vocab or n_ctx_train. A
StateError is thrown if a stage is requested without the model.
Properties
- hashCode → int
-
The hash code for this object.
no setterinherited
- runtimeType → Type
-
A representation of the runtime type of the object.
no setterinherited
Methods
-
noSuchMethod(
Invocation invocation) → dynamic -
Invoked when a nonexistent method or property is accessed.
inherited
-
toString(
) → String -
A string representation of this object.
inherited
Operators
-
operator ==(
Object other) → bool -
The equality operator.
inherited
Static Methods
-
build(
SamplerParams params, {LlamaModel? model}) → Sampler