def test_simple_stats(self):
net = collections.OrderedDict()
net['l_in'] = InputLayer((None, 784))
net['l_shape'] = ReshapeLayer(net['l_in'], (-1, 1, 28, 28))
net['l_conv'] = Conv2DLayer(net['l_shape'],
num_filters=3,
filter_size=3,
W=Constant(1.),
pad=1)
net['l_out'] = DenseLayer(net['l_conv'],
num_units=10,
W=Constant(1.),
nonlinearity=softmax)
no_epochs = 5
ws = lasagne_visualizer.weight_supervisor(net,
no_epochs,
mode='all_trainable')
ws.initialize_grid()
ws.accumulate_weight_stats()
self.assertEquals(ws.max_weights.values(), [[1.]] * 2)
self.assertEquals(ws.min_weights.values(), [[1.]] * 2)
self.assertItemsEqual(ws.mean_weights.values(), [[1.]] * 2)
self.assertItemsEqual(ws.err_band_lo_weights.values(), [[1.]] * 2)
self.assertItemsEqual(ws.err_band_hi_weights.values(), [[1.]] * 2)
def test_init_custom_exception(self):
net = self.net
no_epochs = 5
self.assertRaises(
Exception,
lasagne_visualizer.weight_supervisor(net, no_epochs,
mode='custom'))
def test_custom_weight_range_aspect_ratio_exception(self):
net = self.net
no_epochs = 5
custom_weight_range = {'l_out': [1., -1.]}
ws = lasagne_visualizer.weight_supervisor(
net,
no_epochs,
mode='all_trainable',
custom_weight_ranges=custom_weight_range)
self.assertRaises(Exception, ws.initialize_grid)
def test_custom_weight_ranges_initialization(self):
net = self.net
no_epochs = 5
custom_weight_range = {'l_out': [-5., 5.]}
ws = lasagne_visualizer.weight_supervisor(
net,
no_epochs,
mode='all_trainable',
custom_weight_ranges=custom_weight_range)
self.assertEquals(ws.weight_ranges['l_conv'], [-1., 1.])
self.assertEquals(ws.weight_ranges['l_out'], [-5., 5.])
def test_init_mode_custom(self):
net = self.net
no_epochs = 5
ws = lasagne_visualizer.weight_supervisor(net,
no_epochs,
mode='custom',
layer_names=['l_conv'])
self.assertItemsEqual(ws.layer_names, ['l_conv'])
self.assertItemsEqual(ws.max_weights.keys(), ['l_conv'])
self.assertItemsEqual(ws.min_weights.keys(), ['l_conv'])
self.assertItemsEqual(ws.mean_weights.keys(), ['l_conv'])
self.assertItemsEqual(ws.err_band_lo_weights.keys(), ['l_conv'])
self.assertItemsEqual(ws.err_band_hi_weights.keys(), ['l_conv'])
def test_weight_ranges_initialization(self):
net = self.net
no_epochs = 5
ws = lasagne_visualizer.weight_supervisor(net,
no_epochs,
mode='all_trainable')
self.assertEquals(ws.weight_ranges.values(), [[-1., 1.]] * 2)
self.assertItemsEqual(ws.layer_names, net.keys()[-2:])
self.assertItemsEqual(ws.max_weights.keys(), net.keys()[-2:])
self.assertItemsEqual(ws.min_weights.keys(), net.keys()[-2:])
self.assertItemsEqual(ws.mean_weights.keys(), net.keys()[-2:])
self.assertItemsEqual(ws.err_band_lo_weights.keys(), net.keys()[-2:])
self.assertItemsEqual(ws.err_band_hi_weights.keys(), net.keys()[-2:])
def test_number_of_matplotlib_objects(self):
net = self.net
no_epochs = 5
ws = lasagne_visualizer.weight_supervisor(net,
no_epochs,
mode='custom',
layer_names=['l_conv'])
ws.initialize_grid()
ws.accumulate_weight_stats()
ws.live_plot()
self.assertEquals(len(ws.axis[0].lines), 1)
self.assertEquals(len(ws.axis[0].collections), 3)
self.assertEquals(len(ws.axis[0].texts), 1)
def test_init_mode_currently_trainable(self):
net = self.net
no_epochs = 5
net['l_conv'].params[net['l_conv'].W].remove('trainable')
net['l_out'].params[net['l_out'].W].add('trainable')
ws = lasagne_visualizer.weight_supervisor(net,
no_epochs,
mode='currently_trainable')
self.assertItemsEqual(ws.layer_names, ['l_out'])
self.assertItemsEqual(ws.max_weights.keys(), ['l_out'])
self.assertItemsEqual(ws.min_weights.keys(), ['l_out'])
self.assertItemsEqual(ws.mean_weights.keys(), ['l_out'])
self.assertItemsEqual(ws.err_band_lo_weights.keys(), ['l_out'])
self.assertItemsEqual(ws.err_band_hi_weights.keys(), ['l_out'])