问题
I want to create a model instance and then save it so I can load and solve at a later time (the initialization takes quite long compared to the solving). When I tried this it gave me the following error.
with open('model.pickle', 'w') as f:
pickle.dump(instance, f)
AttributeError: Can't pickle local object 'Euphemia.init..obj_expression
The objective function is this:
def obj_expression(model):
curve = sum(model.x[area, hour, Type, index] * model.Q[area, hour, Type, index] *
( model.P1[area, hour, Type, index] + model.P0[area, hour, Type, index] ) / 2
for (area, hour, Type, index) in model.Curve )
bids = sum(model.y[area, index] * model.PB[area, index] *
sum( model.QB[area, index, hour] for (hour) in model.Hours )
for (area, index) in model.Bids )
return curve + bids
self.model.OBJ = pe.Objective(rule = obj_expression, sense = pe.maximize)
does anybody know how to save a concrete model?
回答1:
The solution is the cloudpickle module, regular pickle has problems pickling functions. An example:
import cloudpickle
with open('test.pkl', mode='wb') as file:
cloudpickle.dump(instance, file)
with open('test.pkl', mode='rb') as file:
instance = cloudpickle.load(file)
回答2:
It looks like pickle (and also cloudpickle) will not work when DerivativeVar is used.
from pyomo.environ import *
from pyomo.dae import *
import pickle
import dill
model = ConcreteModel()
model.x = ContinuousSet(initialize=(0., 1.))
model.y = Var(model.x, initialize=1.)
with open('model1.pickle', 'wb') as f:
pickle.dump(model, f)
model.dydx = DerivativeVar(model.y, wrt=model.x)
with open('model2.pickle', 'wb') as f:
try:
pickle.dump(model, f)
except TypeError:
dill.dump(model, f)
dill appears to work instead for pickling a weakref: Pickling weakref in Python
来源:https://stackoverflow.com/questions/50834258/how-to-save-pickle-a-model-instance-in-pyomo