Source code for openjij.model.chimera_model

# Copyright 2019 Jij Inc.

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at

#     http://www.apache.org/licenses/LICENSE-2.0

# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import openjij
from openjij.model import make_BinaryQuadraticModel, make_BinaryQuadraticModel_from_JSON, BinaryQuadraticModel
import cxxjij as cj


[docs]def make_ChimeraModel(linear, quadratic): """ChimeraModel factory. Returns: generated ChimeraModel class """ class ChimeraModel(make_BinaryQuadraticModel(linear, quadratic)): """Binary quadnratic model dealing with chimera graph This model deal with chimera graph. ChimeraModel provide methods to verify whether a given interaction graph matches a Chimera graph and to convert it to cxxjij.graph.Chimera. Examples:: # This interactions satisfy chimera topology. >>> Q={(0, 4): -1, (4, 12): -1} >>> chimera_model = ChimeraModel(Q, unit_num_L=2) # make >>> chimera_self.validate_chimera() """ def __init__(self, linear=None, quadratic=None, offset=0.0, var_type=openjij.SPIN, unit_num_L=None, model=None, gpu=False): self.gpu = gpu if model: super().__init__(model.linear, model.quadratic, model.offset, model.var_type, gpu=gpu) else: super().__init__(linear, quadratic, offset, var_type, gpu=gpu) if not unit_num_L: raise ValueError( 'Input unit_num_L which is the length of the side of the two-dimensional grid where chimera unit cells are arranged.') self.unit_num_L = unit_num_L # check the type of indices is valid. self.coordinate = self._validate_indices(self.indices) # _chimera_index: 1-D index i,L -> chimera coordinate x,y,z # _to_index: chimera coordinate x,y,z,L -> 1-D index i if self.coordinate == 'index': self._chimera_index = lambda i, L: self.chimera_coordinate(i, L) self._to_index = lambda x, y, z, L: self.to_index(x, y, z, L) elif self.coordinate == 'chimera coordinate': self._chimera_index = lambda i, L: i self._to_index = lambda x, y, z, L: self.to_index(x, y, z, L) def _validate_indices(self, indices): """ Check if the type of indices is valid. Args: indices (list(int) or list(tuple)) Return: the type of indices ('chimera coordinate or index') """ if isinstance(indices[0], int): return 'index' elif isinstance(indices[0], (tuple, list)): if len(indices[0]) == len(indices[-1]) == 3: return 'chimera coordinate' raise ValueError( 'In the chimera graph, index should be int or tuple or list.') def validate_chimera(self): """ Check if the Chimera connectivity is valid. Chimera coordinate: r, c, z One dimension coordinate: i Relation: i = 8Lr + 8c + z Chimera unit cell (column reprezentation) 0 - 4 1 - 5 2 - 6 3 - 7 """ # check chimera interaction for (i, j) in self.quadratic.keys(): r_i, c_i, z_i = self._chimera_index(i, self.unit_num_L) # list up indices which can connect i adj_list = [] if z_i >= 4: # part of right side of a Chimera unit cell (in the column representation). if c_i > 0: adj_list.append(self._to_index( r_i, c_i-1, z_i, self.unit_num_L)) if c_i < self.unit_num_L-1: adj_list.append(self._to_index( r_i, c_i+1, z_i, self.unit_num_L)) adj_list += [self._to_index(r_i, c_i, z, self.unit_num_L) for z in range(0, 4)] else: # part of left side of a Chimera unit cell (in the column representation). if r_i > 0: adj_list.append(self._to_index( r_i-1, c_i, z_i, self.unit_num_L)) if r_i < self.unit_num_L-1: adj_list.append(self._to_index( r_i+1, c_i, z_i, self.unit_num_L)) adj_list += [self._to_index(r_i, c_i, z, self.unit_num_L) for z in range(4, 8)] connect_i = j if isinstance( j, int) else self._to_index(*j, self.unit_num_L) if connect_i not in adj_list: incomp_part = 'The connectable nodes of {} are {}, not {}.'.format( i, adj_list, j) raise ValueError( 'Problem graph incompatible with chimera graph.\n' + incomp_part) return False return True def to_index(self, r, c, i, unit_num_L): """ Chimera coordinate: r, c, i One dimension coordinate: i Relation: i = 8*L*r + 8*c + i Args: r (int): Row index of 2-D Chimera grid. c (int): Column index of 2-D Chimera grid. i (int): index in Chimera unit cell. unit_num_L (int): Row and Column length of 2-D Chimera grid. """ return 8*unit_num_L*r + 8*c + i def chimera_coordinate(self, i, unit_num_L): """Convert 1-d index to chimera corrdinate Args: i (int): 1-D index 0~L*L*8-1 unit_num_L (int): number of chimera grid size Returns: (int, int, int): chimera corrdinate """ z_i = i % 8 c_i = (i % (8 * unit_num_L) - z_i)/8 r_i = (i - 8*c_i - z_i) / (8 * unit_num_L) return int(r_i), int(c_i), int(z_i) def get_cxxjij_ising_graph(self): """Get cxxjij.graph.Chimera type instance Args: i (int): 1-D index 0~L*L*8-1 unit_num_L (int): number of chimera grid size Returns: object (cxxjij.graph.Chimera) """ chimera_L = self.unit_num_L if not self.validate_chimera(): raise ValueError("Problem graph incompatible with chimera graph.") _h, _J, _offset = self.to_ising() if self.gpu: chimera = cj.graph.ChimeraGPU(chimera_L, chimera_L) else: chimera = cj.graph.Chimera(chimera_L, chimera_L) for i, hi in _h.items(): r_i, c_i, zi = self._chimera_index(i, L=chimera_L) if not self._index_validate(i, chimera_L): raise ValueError( "Problem graph incompatible with chimera graph. Node {}.".format(i)) chimera[r_i, c_i, zi] = hi for (i, j), Jij in _J.items(): r_i, c_i, zi = self._chimera_index(i, L=chimera_L) r_j, c_j, zj = self._chimera_index(j, L=chimera_L) # validate connection error_msg = "In the {}*{} Chimera grid, ".format( chimera_L, chimera_L) error_msg += "there is no connection between node {} and node {}.".format( i, j) linear_vldt = self._index_validate( i, chimera_L) and self._index_validate(j, chimera_L) if not (linear_vldt and self._validate((r_i, c_i, zi), (r_j, c_j, zj), chimera_L)): raise ValueError( "Problem graph incompatible with chimera graph.\n" + error_msg) if r_i == r_j and c_i == c_j: # connection in Chimera unit cell if zj in [0, 4]: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.IN_0or4] = Jij elif zj in [1, 5]: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.IN_1or5] = Jij elif zj in [2, 6]: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.IN_2or6] = Jij else: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.IN_3or7] = Jij # connection between Chimera unit cells elif r_i - r_j == -1: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.PLUS_R] = Jij elif r_i - r_j == 1: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.MINUS_R] = Jij elif c_i - c_j == -1: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.PLUS_C] = Jij elif c_i - c_j == 1: chimera[r_i, c_i, zi, cj.graph.ChimeraDir.MINUS_C] = Jij return chimera def energy(self, sample, convert_sample=False): """calc energy of the BinaryQuadraticModel. Args: sample: samples convert_sample: if true, the type of sample is automatically converted to self.vartype. """ return super().energy(sample, sparse=True, convert_sample=convert_sample) def energies(self, samples_like, convert_sample=False): return super().energies(sample_like, sparse=True, convert_sample=convert_sample) def _validate(self, rcz1, rcz2, L): """ Check if the connectivity is valid. Args: rcz1 (int), rcz2(int), L(int) Returns: result (bool) """ r1, c1, z1 = rcz1 r2, c2, z2 = rcz2 left_side = [0, 1, 2, 3] right_side = [4, 5, 6, 7] if r1 == r2 and c1 == c2: if ((z1 in left_side) and (z2 in right_side)): return True elif ((z2 in left_side) and (z1 in right_side)): return True elif ((c1 == c2 and abs(r1 - r2) == 1) or (r1 == r2 and abs(c1 - c2) == 1)): return True return False def _index_validate(self, i, L): """ Check if the index is valid. Args: i(int), L(int) Returns: result (bool) """ if isinstance(i, tuple): two_d_bool = (i[0] < self.unit_num_L) and (i[1] < self.unit_num_L) return two_d_bool and (i[2] < 8) max_index = 8 * L * L return 0 <= i < max_index return ChimeraModel
[docs]def make_ChimeraModel_from_JSON(obj): """make ChimeraModel from JSON. Returns: corresponding ChimeraModel type """ label = obj['variable_labels'][0] if isinstance(label, list): #convert to tuple label = tuple(label) mock_linear = {label:1.0} return make_ChimeraModel(mock_linear, {})
[docs]def ChimeraModel(linear: dict=None, quadratic: dict=None, offset: float=0.0, var_type=openjij.SPIN, unit_num_L: int=None, model=None, gpu: bool=False): """generate ChimeraModel object. This model deal with chimera graph. ChimeraModel provide methods to verify whether a given interaction graph matches a Chimera graph and to convert it to cxxjij.graph.Chimera. Args: linear (dict): linear biases quadratic (dict): quadratic biases offset (float): offset var_type: vartype ('SPIN' or 'BINARY') unit_num_L (int): unit_num_L model (BinaryQuadraticModel): if model is not None, the object is initialized by model. gpu (bool): if true, this can be used for gpu samplers. Returns: generated ChimeraModel Examples: Example shows how to initialize ChimeraModel.:: # This interactions satisfy chimera topology. >>> Q={(0, 4): -1, (4, 12): -1} >>> chimera_model = ChimeraModel(Q, unit_num_L=2) # make >>> chimera_self.validate_chimera() """ Model = make_ChimeraModel(linear, quadratic) return Model(linear, quadratic, offset, var_type, unit_num_L, model, gpu)
#classmethods ChimeraModel.from_qubo = \ lambda Q, offset=0.0, **kwargs: make_ChimeraModel({}, Q).from_qubo(Q, offset, **kwargs) ChimeraModel.from_ising = \ lambda linear, quadratic, offset=0.0, **kwargs: make_ChimeraModel(linear, quadratic).from_ising(linear, quadratic, offset, **kwargs) ChimeraModel.from_serializable = \ lambda obj: make_ChimeraModel_from_JSON(obj).from_serializable(obj)