Python Modules

Public Python API surface for Continuum.

class continuum.LM(model_id: str, adapter=None)

Bases: object

Minimal language-model callable facade used by examples.

class continuum.Optimizer(program, metric, *, lr_tensor=0.1, lr_text=1.0, seed=0)

Bases: object

Unified optimizer facade for all Continuum parameter kinds.

fit(dataset, epochs=1)

step(batch)

class continuum.Param(kind: str, value: Any, metadata: dict[str, ~typing.Any]=<factory>)

Bases: object

Typed program parameter tracked by the Continuum optimizer.

static continuous(initial: float = 0.0, *, min_value=None, max_value=None, **metadata) → Param

static discrete(initial: Any = 0, *, choices=None, **metadata) → Param

static fewshot(k: int = 3) → Param

kind: str

static lora(rank: int = 8) → Param

metadata: dict[str, Any]

static tensor(shape=None, dtype='f32', initial=None, **metadata) → Param

static text(initial: str, **metadata) → Param

value: Any

continuum.program(fn)

Decorate a Python function as a Continuum program.

continuum.tool(fn)

Mark a Python callable as a tool (v0.1 no-op marker).

class continuum.frontend.optimizer.BayesOpt(rng: Random)

Bases: object

Simple random-search optimizer over bounded continuous values.

update(p, evaluate)

class continuum.frontend.optimizer.GEPAOpt(rng: Random)

Bases: object

Discrete choice optimizer for categorical parameters.

update(p, evaluate)

class continuum.frontend.optimizer.Optimizer(program, metric, *, lr_tensor=0.1, lr_text=1.0, seed=0)

Bases: object

Unified optimizer facade for all Continuum parameter kinds.

fit(dataset, epochs=1)

step(batch)

class continuum.frontend.optimizer.TensorOpt(lr: float, rng: Random)

Bases: object

Local optimizer for numeric tensor-like scalar parameters.

update(p, evaluate)

class continuum.frontend.optimizer.TextGradOpt(lr: float, rng: Random)

Bases: object

Text mutation optimizer used for prompt-style parameters.

update(p, evaluate)

class continuum.frontend.param.Param(kind: str, value: Any, metadata: dict[str, ~typing.Any]=<factory>)

Bases: object

Typed program parameter tracked by the Continuum optimizer.

static continuous(initial: float = 0.0, *, min_value=None, max_value=None, **metadata) → Param

static discrete(initial: Any = 0, *, choices=None, **metadata) → Param

static fewshot(k: int = 3) → Param

kind: str

static lora(rank: int = 8) → Param

metadata: dict[str, Any]

static tensor(shape=None, dtype='f32', initial=None, **metadata) → Param

static text(initial: str, **metadata) → Param

value: Any

class continuum.nn.module.Linear(in_features, out_features)

Bases: Module

Tiny reference linear layer used by examples and tests.

forward(x)

class continuum.nn.module.Module

Bases: object

Minimal nn.Module-like base class for parameter discovery.

forward(*args, **kwargs)

parameters()

continuum.programs.program.program(fn)

Mark a function as a Continuum program; first call traces it to CIR.