def test_supported_fingerprinting(cls, monkeypatch):
# patch so we can instantiate various solvers without the proper libraries
monkeypatch.setitem(sys.modules, "scipy", Mock())
monkeypatch.setitem(sys.modules, "scipy.optimize", Mock())
monkeypatch.setitem(sys.modules, "sklearn", Mock())
monkeypatch.setitem(sys.modules, "sklearn.linear_model", Mock())
monkeypatch.setitem(sys.modules, "sklearn.utils", Mock())
monkeypatch.setitem(sys.modules, "sklearn.utils.extmath", Mock())
obj = cls()
assert Fingerprint.supports(obj)
# check fingerprint is created without error and is a valid sha1 hash
fp = str(Fingerprint(obj))
assert len(fp) == 40 and set(fp).issubset("0123456789abcdef")
def test_supported_fingerprinting(cls, monkeypatch):
# patch so we can instantiate various solvers without the proper libraries
monkeypatch.setitem(sys.modules, "scipy", Mock())
monkeypatch.setitem(sys.modules, "scipy.optimize", Mock())
monkeypatch.setitem(sys.modules, "sklearn", Mock())
monkeypatch.setitem(sys.modules, "sklearn.linear_model", Mock())
monkeypatch.setitem(sys.modules, "sklearn.utils", Mock())
monkeypatch.setitem(sys.modules, "sklearn.utils.extmath", Mock())
args = []
if issubclass(cls, nengo.neurons.RatesToSpikesNeuronType):
# spiking types require a `base_type` argument, so provide one
args.append(nengo.neurons.RectifiedLinear())
obj = cls(*args)
assert Fingerprint.supports(obj)
# check fingerprint is created without error and is a valid sha1 hash
fp = str(Fingerprint(obj))
assert len(fp) == 40 and set(fp).issubset("0123456789abcdef")
def tune(self, dt, model, ens):
from nengo_bio.internal.multi_compartment_lif_sim import GaussianSource
from nengo_bio.internal.model_weights import tune_two_comp_model_weights
# Fetch the maximum rates for which we need to determine the neuron
# parameters
max_rates = model.params[ens].max_rates
if isinstance(ens.max_rates, Uniform):
min_max_rate = ens.max_rates.low
max_max_rate = ens.max_rates.high
else:
min_max_rate = np.min(max_rates)
max_max_rate = np.max(max_rates)
# Get the parameter hash
params_hash = str(Fingerprint([self, dt]))
# Function running a single neuron simulation
sim_class = self._compile(dt)
def run_single_sim(idx, out, in_exc, in_inh):
xs = np.asarray((in_exc, in_inh), order='C', dtype=np.float64)
# TODO: Get these parameters from the connection
sim_class().run_single_with_gaussian_sources(
out, [
GaussianSource(
seed=4902 + idx,
mu=in_exc * 0.95,
sigma=in_exc * 0.5,
tau=5e-3,
offs=0.0),
GaussianSource(
seed=5821 + 7 * idx,
mu=in_inh * 0.95,
sigma=in_inh * 0.5,
tau=5e-3,
offs=0.0),
])
return tune_two_comp_model_weights(
dt=dt,
max_rates=max_rates,
min_max_rate=min_max_rate,
max_max_rate=max_max_rate,
run_single_sim=run_single_sim,
estimate_input_range=self._estimate_input_range,
filter_input=self._filter_input,
lif_rate_inv=self._lif_rate_inv,
params_hash=params_hash)
def test_fails_for_lambda_expression():
with pytest.raises(FingerprintError):
Fingerprint(lambda x: x)
def test_unsupported_fingerprinting(obj):
with pytest.raises(FingerprintError):
Fingerprint(obj)
def test_fingerprinting(reference, equal, different):
assert str(Fingerprint(reference)) == str(Fingerprint(equal))
assert str(Fingerprint(reference)) != str(Fingerprint(different))
1.,
1.,
))
# Normalise ws[1] = 1
ws = ws / ws[1]
return ws
def _params_den(self):
return multi_compartment_lif_parameters.DendriticParameters.\
make_two_comp_lif(
C_som=self.C_som,
C_den=self.C_den,
g_leak_som=self.g_leak_som,
g_leak_den=self.g_leak_den,
g_couple=self.g_couple,
E_rev_leak=self.E_rev_leak,
E_rev_exc=self.E_rev_exc,
E_rev_inh=self.E_rev_inh
)
# Alias for TwoCompLIF emphasizing the use of conductance based synapses
TwoCompLIFCond = TwoCompLIF
# Whitelist the neuron types
Fingerprint.whitelist(LIF)
Fingerprint.whitelist(LIFCond)
Fingerprint.whitelist(TwoCompLIF)
def test_fails_for_lambda_expression():
with pytest.raises(FingerprintError, match="cannot be fingerprinted"):
Fingerprint(lambda x: x)
def test_unsupported_fingerprinting(obj):
with pytest.raises(FingerprintError, match="cannot be fingerprinted"):
Fingerprint(obj)
def test_fails_for_lambda_expression():
with pytest.raises((ValueError, AttributeError)):
Fingerprint(lambda x: x)
