ml_switcheroo.semantics.schema

Pydantic Schemas for Semantic Knowledge Base.

This module defines the data structure of the JSON mapping files (k_array_api.json, etc.) and the Framework Configuration blocks.

Updated to support: - PluginTraits: Capability flags for decoupling plugins from framework identifiers. - Rich Argument Definitions in std_args with Typed Defaults (Any). - Conditional Dispatch Rules. - PatternDef: Graph fusion patterns.

Classes

PluginTraits	Defines capability flags for a framework to guide plugin logic.
StructuralTraits	Defines structural patterns for a framework to guide generic rewriters.
FrameworkAlias	Configuration for global module aliasing.
FrameworkTraits	Configuration for a framework's structural behavior.
SemanticsFile	Schema representing an entire semantics JSON file.

Module Contents

class ml_switcheroo.semantics.schema.PluginTraits(/, **data: Any)

Bases: pydantic.BaseModel

Defines capability flags for a framework to guide plugin logic.

model_config

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

has_numpy_compatible_arrays: bool = None

requires_explicit_rng: bool = None

requires_functional_state: bool = None

requires_functional_control_flow: bool = None

enforce_purity_analysis: bool = None

strict_materialization_method: str | None = None

class ml_switcheroo.semantics.schema.StructuralTraits(/, **data: Any)

Bases: pydantic.BaseModel

Defines structural patterns for a framework to guide generic rewriters.

model_config

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

module_base: str | None = None

forward_method: str | None = None

known_inference_methods: Set[str] = None

functional_execution_method: str | None = None

requires_super_init: bool = None

init_method_name: str | None = None

inject_magic_args: List[Tuple[str, str | None]] = None

strip_magic_args: List[str] = None

auto_strip_magic_args: bool = None

lifecycle_strip_methods: List[str] = None

lifecycle_warn_methods: List[str] = None

impurity_methods: List[str] = None

implicit_method_roots: List[str] = None

jit_static_args: List[str] = None

class ml_switcheroo.semantics.schema.FrameworkAlias(/, **data: Any)

Bases: pydantic.BaseModel

Configuration for global module aliasing.

module: str

name: str

class ml_switcheroo.semantics.schema.FrameworkTraits(/, **data: Any)

Bases: pydantic.BaseModel

Configuration for a framework’s structural behavior.

model_config

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

extends: str | None = None

alias: FrameworkAlias | None = None

stateful_call: Dict[str, str] | None = None

tiers: List[ml_switcheroo.enums.SemanticTier] | None = None

traits: StructuralTraits | None = None

plugin_traits: PluginTraits | None = None

class ml_switcheroo.semantics.schema.SemanticsFile(/, **data: Any)

Bases: pydantic.BaseModel

Schema representing an entire semantics JSON file.

model_config

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

frameworks: Dict[str, FrameworkTraits] | None = None

imports: Dict[str, Any] | None = None

patterns: List[ml_switcheroo.core.dsl.PatternDef] | None = None