OMMX adapter for Fixstars Amplify

This package provides an adapter for the Fixstars Amplify SDK from/to OMMX.

How to install

The ommx-fixstars-amplify-adapter can be installed from PyPI as follows:

pip install ommx-fixstars-amplify-adapter

Solve problems formulated in OMMX with Fixstars Amplify

The ommx-fixstars-amplify-adapter allows problems formulated in OMMX to be solved in Fixstars Amplify.

sequenceDiagram
    participant U as User
    participant A as Adapter
    participant P as Fixstars Amplify
    U->>A: ommx.v1.Instance
    A->>U: amplify.Model
    U->>P: amplify.Model and parameters for solvers;
    P->>P: Solve with Fixstars Amplify
    P->>U: amplify.Result
    U->>A: amplify.Result and variable_map
    A->>U: ommx:State

Loading

For example, the following problem formulated in OMMX can be solved using Fixstars Amplify.

import amplify
from ommx.v1 import Instance, DecisionVariable
from ommx_fixstars_amplify_adapter import instance_to_model, result_to_state

q_0 = DecisionVariable.binary(id=0, name="q_0")
q_1 = DecisionVariable.binary(id=1, name="q_1")

ommx_instance = Instance.from_components(
    decision_variables=[q_0, q_1],
    objective=q_0 * q_1 + q_0 - q_1 + 1,
    constraints=[q_0 + q_1 == 1],
    sense=Instance.MAXIMIZE,
)

model, variable_map = instance_to_model(ommx_instance)
client = amplify.FixstarsClient(token="***FIXSTARS AMPLIFY TOKEN***")
result = amplify.solve(model, client=client)
state = result_to_state(result, variable_map)
print(state)

Solve problems formulated in Fixstars Amplify SDK with other solvers

The ommx-fixstars-amplify-adapter allows problems formulated in Fixstars Amplify SDK to be solved in other solvers.

sequenceDiagram
    participant U as User
    participant A as Adapter
    participant O as Other OMMX toolchain
    U->>A: amplify.Model
    A->>U: ommx.v1.Instance
    U->>O: ommx.v1.Instance and parameters for other solver
    O->>O: Solve the instance with other solver using other adapter
    O->>U: ommx.v1.State

Loading

For example, the following mixed integer programming problem formulated in Fixstars Amplify SDK can be solved using PythonMIP.

import amplify
from ommx_fixstars_amplify_adapter import model_to_instance
from ommx_python_mip_adapter import instance_to_model, model_to_solution

UPPER = float("inf")
LOWER = 0.0

gen = amplify.VariableGenerator()
x = gen.scalar("Integer", bounds=(LOWER, UPPER), name="x")
y = gen.scalar("Real", bounds=(LOWER, UPPER), name="y")

model = amplify.Model()
model += -10 * x - y
model += amplify.less_equal(x, 1)
model += amplify.less_equal(20 * x + y, 100)

ommx_instance = model_to_instance(model)
model = instance_to_model(ommx_instance)
model.optimize()
state = model_to_solution(model, ommx_instance)
print(state)

Note

Currently, the model_to_instance function does not support Ising variables. Therefore, if your model contains Ising variables, you must convert them to binary variables. The following function can be used for this conversion.

import typing
import amplify

def ising_to_binary(
    model: amplify.Model
) -> typing.Tuple[amplify.Model, amplify.Result.ModelConversion.IntermediateMapping]:
    ising_to_binary_settings = {
        amplify.VariableType.Binary: amplify.Degree.HighOrder,
        amplify.VariableType.Ising: amplify.Degree.Zero,
        amplify.VariableType.Integer: amplify.Degree.HighOrder,
        amplify.VariableType.Real: amplify.Degree.HighOrder,
    }
    return model.to_intermediate_model(
        amplify.AcceptableDegrees(
            objective=ising_to_binary_settings,  # type: ignore
            equality_constraints=ising_to_binary_settings,  # type: ignore
            inequality_constraints=ising_to_binary_settings,  # type: ignore
        )
    )

For Developer

The packages required for development can be installed as follows:

pip install ".[dev]"

Use the following commands to test, lint and format.

python -m pytest
python -m ruff check
python -m ruff format