def test_pyll_lockdown():
pf0 = model_params.pyll_param_func()
sample = pyll.stochastic.sample
s = sample(pf0, np.random.RandomState(123))
# -- this was done with commit
# f11026e235e57f85e9dd4f3a23a67dac2b33e8db
print s
assert s == {'preproc':
{'global_normalize': 0, 'crop': (0, 0, 250, 250), 'size':
(100, 100)}, 'slm': ((('lnorm', {'kwargs': {'inker_shape': (8.0, 8.0),
'outker_shape': (8.0, 8.0), 'remove_mean': 1, 'threshold':
0.11155144268807565, 'stretch': 0.10625644175250697}}),),
(('fbcorr', {'initialize': {'n_filters': 16.0, 'filter_shape':
(8.0, 8.0), 'generate': ('random:uniform', {'rseed': 1})},
'kwargs': {}}), ('lpool', {'kwargs': {'ker_shape': (4.0, 4.0),
'order': 2.1093609092062242, 'stride': 2}}), ('lnorm',
{'kwargs': {'inker_shape': (7.0, 7.0), 'outker_shape':
(7.0, 7.0), 'remove_mean': 0, 'threshold':
0.87658837006354096, 'stretch':
0.47462712500820414}})), (('fbcorr',
{'initialize': {'n_filters': 16.0,
'filter_shape': (8.0, 8.0), 'generate':
('random:uniform', {'rseed': 12})},
'kwargs': {}}), ('lpool', {'kwargs':
{'ker_shape': (6.0, 6.0), 'order':
5.2378168706681931, 'stride':
2}}), ('lnorm', {'kwargs':
{'inker_shape': (8.0, 8.0),
'outker_shape': (8.0,
8.0), 'remove_mean':
0, 'threshold':
3.8851479053533247,
'stretch':
0.22825379389892816}})))}
def test_pyll_param_func_valid():
# -- test that the pyll_param_func returns interchangable things
# and that those things deliver samples predictably
pf0 = model_params.pyll_param_func()
pf1 = model_params.pyll_param_func()
sample = pyll.stochastic.sample
for seed in range(20):
s00 = sample(pf0, np.random.RandomState(seed))
s10 = sample(pf1, np.random.RandomState(seed))
s01 = sample(pf0, np.random.RandomState(seed + 1))
s11 = sample(pf1, np.random.RandomState(seed + 1))
assert s00 == s10
assert s01 == s11
assert s00 != s01
