def lstm(X, encoder):
layers = [
(InputLayer, {}),
(EmbeddingLayer, {
'output_size': 32
}),
(LSTMLayer, {
'grad_clipping': 5.,
'num_units': 100
}),
(LSTMLayer, {
'grad_clipping': 5.,
'num_units': 100
}),
(RNNDenseLayer, {
'nonlinearity': identity
}),
]
rnn = RNN(
layers=layers,
encoder=encoder,
verbose=1,
improvement_threshold=1.,
updater=partial(rmsprop, learning_rate=1e-2),
)
rnn.initialize()
return rnn
def lstm(X, encoder):
layers = [
(InputLayer, {}),
(EmbeddingLayer, {'output_size': 32}),
(LSTMLayer, {'grad_clipping': 5., 'num_units': 100}),
(LSTMLayer, {'grad_clipping': 5., 'num_units': 100}),
(RNNDenseLayer, {'nonlinearity': identity}),
]
rnn = RNN(
layers=layers,
encoder=encoder,
verbose=1,
improvement_threshold=1.,
updater=partial(rmsprop, learning_rate=1e-2),
)
rnn.initialize()
return rnn
def clf(self, encoder):
clf = RNN(encoder=encoder, **rnn_kwargs())
clf.initialize()
clf.fit(self.X, num_epochs=1)
return clf
def clf(self, encoder):
clf = RNN(encoder=encoder, **rnn_kwargs())
clf.initialize()
clf.fit(self.X, num_epochs=1)
return clf
def test_init_raises_when_patience_and_no_eval(self):
with pytest.raises(ValueError):
RNN(layers=[], encoder=Mock(), patience=3, eval_size=0)
