Skip to main content

Python class

SamplingConfig

SamplingConfig

class max.pipelines.SamplingConfig(*, config_file=None, section_name=None, in_dtype=float32, out_dtype=float32, enable_structured_output=False, enable_variable_logits=False, enable_penalties=False, enable_min_tokens=False)

source

Bases: ConfigFileModel

Configuration for the sampling stage of token generation.

Parameters:

  • config_file (str | None)
  • section_name (str | None)
  • in_dtype (Annotated[DType, BeforeValidator(func=~max.pipelines.lib.sampling.sampling_config._coerce_dtype, json_schema_input_type=PydanticUndefined)])
  • out_dtype (Annotated[DType, BeforeValidator(func=~max.pipelines.lib.sampling.sampling_config._coerce_dtype, json_schema_input_type=PydanticUndefined)])
  • enable_structured_output (bool)
  • enable_variable_logits (bool)
  • enable_penalties (bool)
  • enable_min_tokens (bool)

enable_min_tokens

enable_min_tokens: bool

source

enable_penalties

enable_penalties: bool

source

enable_structured_output

enable_structured_output: bool

source

enable_variable_logits

enable_variable_logits: bool

source

from_generation_config_sampling_defaults()

classmethod from_generation_config_sampling_defaults(sampling_params_defaults, **kwargs)

source

Creates a SamplingConfig from generation config defaults and kwargs.

Inspects the provided defaults to determine if penalty-related or min-tokens-related fields are set to non-default values; if so, enables the corresponding flags in the result unless already set in kwargs.

Parameters:

  • sampling_params_defaults (SamplingParamsGenerationConfigDefaults) – The generation config defaults containing explicit values for sampling parameters.
  • **kwargs – Additional keyword arguments to override or supplement the config.

Returns:

A new SamplingConfig instance with the appropriate fields set.

Return type:

SamplingConfig

in_dtype

in_dtype: CoercedDType

source

model_config

model_config: ClassVar[ConfigDict] = {'extra': 'forbid', 'strict': False}

source

Configuration for the model, should be a dictionary conforming to [ConfigDict][pydantic.config.ConfigDict].

model_post_init()

model_post_init(context, /)

source

This function is meant to behave like a BaseModel method to initialise private attributes.

It takes context as an argument since that’s what pydantic-core passes when calling it.

Parameters:

  • self (BaseModel) – The BaseModel instance.
  • context (Any) – The context.

Return type:

None

out_dtype

out_dtype: CoercedDType

source