def testNNMultipleInputs(self):
nn = blocks_std.NN(10, bias=blocks_std.Bias(0), act=tf.tanh)
x = [tf.placeholder(dtype=tf.float32, shape=[5, 7]),
tf.placeholder(dtype=tf.float32, shape=[5, 3]),
tf.placeholder(dtype=tf.float32, shape=[5, 5])]
y = nn(*x)
xs = self.CheckNN(y, nn, 'Tanh')
self.assertEqual(len(x), len(xs))
for u, v in zip(x, xs):
self.assertIs(u, v)
def __init__(self,
depth,
bias=LSTMBiasInit,
initializer=block_util.RsqrtInitializer(),
name=None):
super(LSTM, self).__init__([depth], name)
with self._BlockScope():
self._depth = depth
self._nn = blocks_std.NN(
4 * depth, bias=bias, act=None, initializer=initializer)
self._hidden_linear = blocks_std.Linear(
4 * depth, initializer=initializer)
def testNNWithBiasWithAct(self): nn = blocks_std.NN(10, bias=blocks_std.Bias(0), act=tf.square) x = tf.placeholder(dtype=tf.float32, shape=[5, 7]) y = nn(x) self.assertIs(x, self.CheckNN(y, nn, 'Square')[0])
def testNNWithoutBiasWithAct(self): nn = blocks_std.NN(10, act=tf.nn.relu, bias=None) x = tf.placeholder(dtype=tf.float32, shape=[5, 7]) y = nn(x) self.assertIs(x, self.CheckNN(y, nn, 'Relu')[0])
def testNNWithoutActWithoutBias(self): nn = blocks_std.NN(10, act=None, bias=None) x = tf.placeholder(dtype=tf.float32, shape=[5, 7]) y = nn(x) self.assertIs(x, self.CheckNN(y, nn)[0])