Adding optimizer: ExcitationSolve #433

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open

S-Erik wants to merge 41 commits into tequilahub:devel from S-Erik:excsolve

.github/workflows/ci_basic.yml

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -16,7 +16,7 @@ jobs:
  
        runs-on: ubuntu-latest

        strategy:

          matrix:

            python-version: ['3.9', '3.10']

            python-version: ['3.10', '3.11', '3.12']

        steps:

        - uses: actions/checkout@v2

    @@ -30,6 +30,7 @@ jobs:
  
            pip install -r requirements.txt

            # test default simulator against in-house simulator

            pip install git+https://github.com/Mikel-Ma/spex@devel

            pip install excitationsolve

        - name: Lint with flake8

          run: |

            pip install flake8

.github/workflows/ci_basic_autograd.yml

-Original file line number
+Diff line change
@@ Expand Up / @@ -16,7 +16,7 @@ jobs: @@
         runs-on: ubuntu-latest
         strategy:
           matrix:
-            python-version: ['3.9', '3.10']
+            python-version: ['3.11']
         steps:
         - uses: actions/checkout@v2
@@ Expand Down @@

.github/workflows/ci_chemistry_psi4.yml

-Original file line number
+Diff line change
@@ Expand Up / @@ -22,7 +22,7 @@ jobs: @@
             with:
               miniconda-version: 'latest'
               auto-activate-base: false
-              python-version: '3.10'
+              python-version: '3.11'
           - name: Initialize Conda for the shell
             run: |
@@ Expand Down @@

.github/workflows/ci_chemistry_pyscf.yml

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -16,7 +16,7 @@ jobs:
  
        runs-on: ubuntu-latest

        strategy:

          matrix:

            python-version: [3.9]

            python-version: ['3.10']

        steps:

        - uses: actions/checkout@v2

    @@ -39,7 +39,7 @@ jobs:
  
            python -m pip install -r requirements.txt

            python -m pip install -e .

            python -m pip install pyscf

            python -m pip install 'h5py <= 3.1'

            #python -m pip install 'h5py <= 3.1'

            cd tests

            ls

            pytest test_chemistry.py test_TrotErr.py  -m "not dependencies"

.github/workflows/ci_conda_madness.yml

-Original file line number
+Diff line change
@@ Expand Up / @@ -16,7 +16,7 @@ jobs: @@
         runs-on: ubuntu-latest
         strategy:
           matrix:
-            python-version: [3.9]
+            python-version: [3.11]
         steps:
         - uses: actions/checkout@v2
@@ Expand Down @@

.github/workflows/ci_ml.yml

-Original file line number
+Diff line change
@@ Expand Up / @@ -16,7 +16,7 @@ jobs: @@
         runs-on: ubuntu-latest
         strategy:
           matrix:
-            python-version: [3.9]
+            python-version: [3.11]
         steps:
         - uses: actions/checkout@v2
@@ Expand Down @@

.github/workflows/ci_pyquil.yml

This file was deleted.

README.md

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -9,15 +9,16 @@ Tequila can execute the underlying quantum expectation values on state of the ar
  
    # Getting Started

    Get started with our collection of  

    - *[Tutorials](https://tequilahub.github.io/tequila-tutorials/)*  

    - *[Tutorials](https://tequilahub.github.io/tequila-tutorials/)*

    - *[Documentation](https://tequilahub.github.io/tequila-tutorials/docs/sphinx/)*  

    Further sources:

    - [overview article](https://arxiv.org/abs/2011.03057)   

    - [tequila in a nutshell](https://kottmanj.github.io/tequila-in-a-nutshell/#/)  

    - [talks and slides](https://kottmanj.github.io/talks_and_material/)  

    # Installation

    Recommended Python version is 3.10 (3.11).   

    Recommended Python version is 3.11 (3.10 or 3.12 shouldn't be a problem either).   

    Tequila supports linux, osx and windows. However, not all optional dependencies (especially chemistry) are supported on windows.  

    ## Install from PyPi

    @@ -295,12 +296,7 @@ pip install pyscf
  
    Works similar as Psi4. Classical methods are also integrated in the madness interface allowing to use them in a basis-set-free representation.

    # Documentation

    You can build the documentation by navigating to `docs` and entering `make html`.

    Open the documentation with a browser over like `firefox docs/build/html/index.html`

    Note that you will need some additional python packages like `sphinx` and `mr2` that are not explicitly listed in the requirements.txt

    You can also visit our prebuild online [documentation](https://tequilahub.github.io/tequila/)

    that will correspond to the github master branch

    see [here](https://tequilahub.github.io/tequila-tutorials/docs/sphinx/)  

    # How to contribute

    If you find any bugs or inconveniences in `tequila` please don't be shy and let us know.

requirements.txt

-Original file line number
+Diff line change
@@ Expand Up / @@ -10,6 +10,7 @@ pytest @@
     openfermion ~= 1.0 # can not be smaller than 1.0
     #cmake # needed by qulacs, can be removed otherwise, now in qulacs requirements
     qulacs # default simulator (best integration), remove if the installation gives you trouble and just install one of the other supported backend. Version restriction only for noise models, otherwise the new version is fine
+    excitationsolve
     #optional quantum backends
     #cirq >= 0.9.2 #
@@ Expand Down @@

src/tequila/grouping/fermionic_functions.py

-Original file line number
+Diff line change
@@ Expand Up / @@ -589,8 +589,12 @@ def find_index(basis_state): @@
         index = 0
         n_qubits = len(basis_state)
         for j in range(n_qubits):
-            index += int(basis_state[j]) * 2 ** (n_qubits - j - 1)
+            try:
+                index += int(basis_state[j]) * 2 ** (n_qubits - j - 1)
+            except Exception as E:
+                print("basis_state=", basis_state)
+                print(f"j={j}")
+                print(f"basis_state[j]={basis_state[j]}")
         return index
@@ Expand Down @@

src/tequila/optimizers/__init__.py

            
                      Original file line number
                      Diff line number
                      Diff line change
                  
    @@ -1,8 +1,10 @@
  
    from tequila.optimizers.optimizer_base import OptimizerHistory, Optimizer, TequilaOptimizerException, OptimizerResults

    from tequila.optimizers.optimizer_scipy import OptimizerSciPy

    from tequila.optimizers.optimizer_gd import OptimizerGD

    from tequila.optimizers.optimizer_excsolve import OptimizerExcitationSolve

    from tequila.optimizers.optimizer_scipy import minimize as minimize_scipy

    from tequila.optimizers.optimizer_gd import minimize as minimize_gd

    from tequila.optimizers.optimizer_excsolve import minimize as minimize_excsolve

    from tequila.simulators.simulator_api import simulate

    from dataclasses import dataclass

    @@ -18,13 +20,14 @@ class _Optimizers:
  
        methods: list = None

    SUPPORTED_OPTIMIZERS = ["scipy", "gpyopt", "gd"]

    SUPPORTED_OPTIMIZERS = ["scipy", "gpyopt", "gd", "excitationsolve"]

    INSTALLED_OPTIMIZERS = {}

    INSTALLED_OPTIMIZERS["scipy"] = _Optimizers(

        cls=OptimizerSciPy, minimize=minimize_scipy, methods=OptimizerSciPy.available_methods()

    )

    INSTALLED_OPTIMIZERS["gd"] = _Optimizers(cls=OptimizerGD, minimize=minimize_gd, methods=OptimizerGD.available_methods())

    has_gpyopt = False

    try:

        from tequila.optimizers.optimizer_gpyopt import OptimizerGPyOpt

    @@ -37,6 +40,18 @@ class _Optimizers:
  
    except ImportError:

        has_gpyopt = False

    has_excsolve = False

    try:

        from tequila.optimizers.optimizer_excsolve import OptimizerExcitationSolve

        from tequila.optimizers.optimizer_excsolve import minimize as minimize_excsolve

        INSTALLED_OPTIMIZERS["excitationsolve"] = _Optimizers(

            cls=OptimizerExcitationSolve, minimize=minimize_excsolve, methods=OptimizerExcitationSolve.available_methods()

        )

        has_excsolve = True

    except ImportError:

        has_excsolve = False

    def show_available_optimizers(module=None):

        """

    @@ -138,5 +153,5 @@ def minimize(
  
                )

        raise TequilaOptimizerException(

            "Could not find optimization method {} in tequila optimizers. You might miss dependencies"

            f"Could not find optimization method {method} in tequila optimizers. You might miss dependencies"

        )

src/tequila/optimizers/optimizer_excsolve.py

-Original file line number
+Diff line change
@@ -0,0 +1,142 @@
+    import scipy
+    import numpy
+    import typing
+    import numbers
+    from excitationsolve import ExcitationSolveScipy
+    from tequila.objective import Objective
+    from tequila.objective.objective import assign_variable, Variable, format_variable_dictionary, format_variable_list
+    from .optimizer_base import Optimizer, OptimizerResults
+    from ._containers import _EvalContainer, _GradContainer, _HessContainer, _QngContainer
+    from .optimizer_scipy import SciPyResults
+    from tequila.utils.exceptions import TequilaException
+    from tequila.circuit.noise import NoiseModel
+    from tequila.tools.qng import get_qng_combos
+    from dataclasses import dataclass
+    class OptimizerExcitationSolve(Optimizer):
+        r"""The ExcitationSolve optimizer as a SciPy optimizer that can be given to the scipy.optimize.minimize function.
+        Usage:
+        ```python
+            excsolve_obj = ExcitationSolveScipy(maxiter=100, tol=1e-10, save_parameters=True)
+            optimizer = excsolve_obj.minimize
+            res = scipy.optimize.minimize(cost, params, method=optimizer)
+            energies = excsolve_obj.energies
+            counts = excsolve_obj.nfevs
+        ```
+        Note that this optimizer never needs to evaluate the ansatz circuit
+        at the (current) optimal parameters, unless the optimal parameters fall onto
+        the sample points used to reconstruct the energy function.
+        Therefore, when used with a qiskit VQE object, the energies transmitted
+        to a VQE callback function, do not seem to improve or converge. Nevertheless,
+        the determined optimal energy and parameters are still returned.
+        Args:
+            maxiter (int): Maximum number of VQE iterations (maximum number of times to optimize all parameters)
+            tol: Threshold of energy difference after subsequent VQE iterations defining convergence
+            num_samples (int, optional): Number of different parameter values at which to sample
+                the energy to reconstruct the energy function in one parameter.
+                Must be greater or equal to 5. Defaults to 5.
+            hf_energy (float | None, optional): The Hartree-Fock energy, i.e. the energy of the
+                system where all parameters in the circuit are zero. If none, this will be
+                calculated by evaluating the energy of the ansatz with all parameters set to zero.
+                If this energy is known from a prior classical calculation, e.g. a Hartree-Fock
+                calculation, one energy evaluation is saved. Defaults to None.
+            save_parameters (bool, optional): If True, params member variable contains
+                all optimal parameter values after each optimization step,
+                i.e. after optimizing each single parameter. Defaults to False.
+            param_scaling (float, optional): Factor used for rescaling the parameters. This ExcitationSolve optimizer
+                                                expects the parameters to be 2\pi periodic. For example, in Qiskit
+                                                the excitation parameters result in excitation operators being \pi periodic.
+                                                Therefore, we use a factor of 0.5 for qiskit, resulting in a Period of 2\pi.
+        """
+        @classmethod
+        def available_methods(cls):
+            """:return: All tested available methods"""
+            return ["excitationsolve"]
+        def __init__(
+            self, maxiter, tol=1e-12, num_samples=5, hf_energy=None, save_parameters=False, param_scaling=0.5, **kwargs
+        ):
+            if maxiter is None:
+                maxiter = 10
+            super().__init__(**kwargs)
+            self.opt = ExcitationSolveScipy(
+                maxiter=maxiter,
+                tol=tol,
+                num_samples=num_samples,
+                hf_energy=hf_energy,
+                save_parameters=save_parameters,
+                param_scaling=param_scaling,
+            )
+        def __call__(
+            self,
+            objective: Objective,
+            variables: typing.List[Variable],
+            initial_values: typing.Dict[Variable, numbers.Real] = None,
+            *args,
+            **kwargs,
+        ) -> SciPyResults:
+            objective = objective.contract()
+            infostring = "{:15} : {}\n".format("Method", "ExcitationSolve")
+            infostring += "{:15} : {} expectationvalues\n".format("Objective", objective.count_expectationvalues())
+            # if self.save_history and reset_history:
+            #     self.reset_history()
+            active_angles, passive_angles, variables = self.initialize_variables(objective, initial_values, variables)
+            # Transform the initial value directory into (ordered) arrays
+            param_keys, param_values = zip(*active_angles.items())
+            param_values = numpy.array(param_values)
+            # do the compilation here to avoid costly recompilation during the optimization
+            compiled_objective = self.compile_objective(objective=objective, *args, **kwargs)
+            E = _EvalContainer(
+                objective=compiled_objective,
+                param_keys=param_keys,
+                samples=self.samples,
+                passive_angles=passive_angles,
+                save_history=self.save_history,
+                print_level=self.print_level,
+            )
+            res = self.opt.minimize(E, param_values, *args, **kwargs)
+            if self.save_history:
+                self.history.energies = self.opt.energies
+                self.history.angles = self.opt.params
+                # self.history.gradients = self.opt.energies_shiftste
+            return SciPyResults(energy=res.fun, history=self.history, variables=res.x, scipy_result=res)
+    def minimize(
+        objective: Objective,
+        variables: typing.List[Variable],
+        initial_values: typing.Dict[Variable, numbers.Real] = None,
+        method: str = "excitationsolve",
+        maxiter: int = 10,
+        *args,
+        **kwargs,
+    ):
+        optimize = OptimizerExcitationSolve(
+            maxiter=maxiter,
+            save_parameters=True,
+            *args,
+            **kwargs,
+        )
+        return optimize(
+            objective=objective,
+            variables=variables,
+            initial_values=initial_values,
+            *args,
+            **kwargs,
+        )

src/tequila/version.py

-Original file line number
+Diff line change
@@ -1,2 +1,2 @@
-    __version__ = "1.9.10.dev"
+    __version__ = "1.9.10"
     __author__ = "Tequila Developers "

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adding optimizer: ExcitationSolve #433

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!

Uh oh!

Uh oh!

Adding optimizer: ExcitationSolve #433

Are you sure you want to change the base?

Uh oh!

Adding optimizer: ExcitationSolve #433

Uh oh!

Uh oh!

Diff view

Diff view

There are no files selected for viewing

Uh oh!

Uh oh!

Uh oh!