def test_keras_style_two_separate_input_spaces(self):
# Define two input Spaces first. Independently (no container).
input_space_1 = IntBox(3, add_batch_rank=True)
input_space_2 = FloatBox(shape=(4,), add_batch_rank=True)
# One-hot flatten the int tensor.
flatten_layer_out = ReShape(flatten=True, flatten_categories=True)(input_space_1)
# Run the float tensor through two dense layers.
dense_1_out = DenseLayer(units=3, scope="d1")(input_space_2)
dense_2_out = DenseLayer(units=5, scope="d2")(dense_1_out)
# Concat everything.
cat_out = ConcatLayer()(flatten_layer_out, dense_2_out)
# Use the `outputs` arg to allow your network to trace back the data flow until the input space.
neural_net = NeuralNetwork(inputs=[input_space_1, input_space_2], outputs=cat_out)
test = ComponentTest(component=neural_net, input_spaces=dict(inputs=[input_space_1, input_space_2]))
var_dict = neural_net.variable_registry
w1_value = test.read_variable_values(var_dict["neural-network/d1/dense/kernel"])
b1_value = test.read_variable_values(var_dict["neural-network/d1/dense/bias"])
w2_value = test.read_variable_values(var_dict["neural-network/d2/dense/kernel"])
b2_value = test.read_variable_values(var_dict["neural-network/d2/dense/bias"])
# Batch of size=n.
input_ = [input_space_1.sample(4), input_space_2.sample(4)]
expected = np.concatenate([ # concat everything
one_hot(input_[0]), # int flattening
dense_layer(dense_layer(input_[1], w1_value, b1_value), w2_value, b2_value) # float -> 2 x dense
], axis=-1)
out = test.test(("call", input_), expected_outputs=expected)
test.terminate()
def test_specifiable_server(self):
action_space = IntBox(2)
state_space = FloatBox()
env_spec = dict(type="random_env",
state_space=state_space,
action_space=action_space,
deterministic=True)
# Create the server, but don't start it yet. This will be done fully automatically by the tf-Session.
specifiable_server = SpecifiableServer(
Environment, env_spec, dict(step_flow=[state_space, float, bool]),
"terminate")
# ret are ops now in the graph.
ret1 = specifiable_server.step_flow(action_space.sample())
ret2 = specifiable_server.step_flow(action_space.sample())
# Check all 3 outputs of the Env step (next state, reward, terminal).
self.assertEqual(ret1[0].shape, ())
self.assertEqual(ret1[0].dtype, convert_dtype("float32"))
self.assertEqual(ret1[1].shape, ())
self.assertEqual(ret1[1].dtype, convert_dtype("float32"))
self.assertEqual(ret1[2].shape, ())
self.assertEqual(ret1[2].dtype, convert_dtype("bool"))
self.assertEqual(ret2[0].shape, ())
self.assertEqual(ret2[0].dtype, convert_dtype("float32"))
self.assertEqual(ret2[1].shape, ())
self.assertEqual(ret2[1].dtype, convert_dtype("float32"))
self.assertEqual(ret2[2].shape, ())
self.assertEqual(ret2[2].dtype, convert_dtype("bool"))
# Start the session and run the op, then check its actual values.
with tf.train.SingularMonitoredSession(
hooks=[SpecifiableServerHook()]) as sess:
out1 = sess.run(ret1)
out2 = sess.run(ret2)
# next state
self.assertAlmostEqual(out1[0], 0.7713, places=4)
self.assertAlmostEqual(out2[0], 0.7488, places=4)
# reward
self.assertAlmostEqual(out1[1], 0.0208, places=4)
self.assertAlmostEqual(out2[1], 0.4985, places=4)
# terminal
self.assertTrue(out1[2] is np.bool_(False))
self.assertTrue(out2[2] is np.bool_(False))
