networkx v2 has been released recently, which broke my code with hyperopt.
Test script
from hyperopt import hp, base, tpe, Trials, rand
import numpy as np
import json
space = {
"optimizer": hp.choice("optimizer", [
{
"$value": "sgd",
"lr": 0.1,
"momentum": 0.5
},
{
"$value": "adadelta",
"lr": 1
}
])
}
suggestions = []
for i in range(0, 1):
trials = Trials()
seed = np.random.RandomState().randint(2 ** 31 - 1)
domain = base.Domain(None, space)
new_ids = [len(trials._dynamic_trials) + 1]
new_trials = rand.suggest(new_ids, domain, trials, seed)
suggestions.append(new_trials[0]['misc']['vals'])
print(json.dumps(suggestions))Error:
Traceback (most recent call last):
File "test.py", line 1, in <module>
from hyperopt import hp, base, tpe, Trials, rand
File "/usr/lib/python2.7/site-packages/hyperopt/__init__.py", line 21, in <module>
from .fmin import fmin
File "/usr/lib/python2.7/site-packages/hyperopt/fmin.py", line 20, in <module>
standard_library.install_aliases()
File "/usr/lib/python2.7/site-packages/future/standard_library/__init__.py", line 483, in install_aliases
import test
File "/app/aetros-trainer/test.py", line 26, in <module>
domain = base.Domain(None, space)
File "/usr/lib/python2.7/site-packages/hyperopt/base.py", line 786, in __init__
pyll.toposort(self.expr)
File "/usr/lib/python2.7/site-packages/hyperopt/pyll/base.py", line 717, in toposort
assert order[-1] == expr
TypeError: 'generator' object has no attribute '__getitem__'
Error is that in hyperopt/pyll/base.py+714: nx.topological_sort(G) returns a generator without __getitem__ method since version 2, but the assert line below requires that.
When I install networkx:1.11 everything works fine.
networkx v2 has been released recently, which broke my code with hyperopt.
Test script
Error:
Error is that in hyperopt/pyll/base.py+714:
nx.topological_sort(G)returns a generator without__getitem__method since version 2, but theassertline below requires that.When I install networkx:1.11 everything works fine.