def setUp(self):
tree = ((0, 1), ((2, 3), 4))
self.link = links.BinaryHierarchicalSoftmax(3, tree)
self.link.cleargrads()
self.x = numpy.random.uniform(-1, 1, (2, 3)).astype(numpy.float32)
self.t = numpy.array([0, 2]).astype(numpy.int32)
self.gy = numpy.random.uniform(-1, 1, ()).astype(numpy.float32)
self.W = self.link.W.data.copy()
def __init__(self, n_vocab, n_units, tree, ratio=1., train=True):
super(HSMmodel, self).__init__(
embed=L.EmbedID(n_vocab, n_units),
l1=L.LSTM(n_units, n_units),
l2=L.LSTM(n_units, n_units),
l3=L.Linear(n_units, n_units),
# lossfun=SoftmaxCrossEntropyLoss(n_units, n_vocab),
lossfun=L.BinaryHierarchicalSoftmax(n_units, tree),
)
self.train = train
self.ratio = 1. - ratio
self.y = None
self.loss = None
self.accuracy = None
self.compute_accuracy = True
def __init__(self, params):
self.__dict__.update(params)
super(GRU_Baseline, self).__init__(
embed=L.EmbedID(self.n_vocab + 3, self.word_dim),
gru=L.GRU(n_units=self.n_units, n_inputs=self.word_dim),
# l2v = L.Linear(
# in_size = self.n_units,
# out_size = self.n_vocab + 2
# )
# blackout = L.BlackOut(
# in_size = self.n_units,
# counts = self.vocab_counts,
# sample_size = self.blackout_sample_size
# )
# bottleneck = L.Linear(
# in_size = self.n_units,
# out_size = 128
# ),
bh_softmax=L.BinaryHierarchicalSoftmax(in_size=self.n_units,
tree=self.huffman_tree))
def setUp(self):
self._config_user = chainer.using_config('dtype', self.dtype)
self._config_user.__enter__()
tree = ((0, 1), ((2, 3), 4))
self.link = links.BinaryHierarchicalSoftmax(3, tree)
self.link.cleargrads()
self.x = numpy.random.uniform(-1, 1, (2, 3)).astype(self.dtype)
self.t = numpy.array([0, 2]).astype(numpy.int32)
self.gy = numpy.random.uniform(-1, 1, ()).astype(self.dtype)
self.W = self.link.W.data.copy()
if self.dtype == numpy.float16:
self.check_sum_options = {'delta': 1e-3}
self.test_forward_options = {'atol': 0.005}
self.check_backward_options = {'dtype': numpy.float64}
else:
self.check_sum_options = {'delta': 1e-5}
self.test_forward_options = {}
self.check_backward_options = {}