def test_network_step_no_mix_forward(self):
spec = blur.NetworkSpec(use_forward_activations_for_synapse_update=False)
genome = genome_util.convert_genome_to_tf_variables(
genome_util.create_backprop_genome(num_species=1))
initializer = lambda params: 2 * tf.ones(params.shape, dtype=tf.float32)
data = random_dataset()
state = blur_meta.init_first_state(
genome,
synapse_initializer=initializer,
data=data,
hidden_layers=[256, 128])
input_fn = data.make_one_shot_iterator().get_next
data_support_fn, _ = blur_meta.episode_data_fn_split(input_fn)
blur.network_step(
state,
genome,
data_support_fn,
data.make_one_shot_iterator().get_next,
network_spec=spec,
env=blur_env.tf_env)
g = tf.get_default_graph()
synapse_pre = g.get_operation_by_name(
'step/backward/synapse_update/hebbian_pre').inputs[0]
synapse_post = g.get_operation_by_name(
'step/backward/synapse_update/hebbian_post').inputs[0]
self.assertIn('backward', synapse_pre.name)
self.assertIn('backward', synapse_post.name)
self.assertNotIn('forward', synapse_pre.name)
self.assertNotIn('forward', synapse_post.name)
def test_verify_gradient_match_jp(self):
tf.reset_default_graph()
tf.disable_v2_behavior()
num_in = 10
num_out = 15
inp, out, ow = random_dense(num_in, num_out)
env = blur_env.jp_env
pre, post, synapse, _, network_spec = get_blur_state(env, inp, out, ow)
genome = genome_util.create_backprop_genome(num_species=1)
update1, update2 = blur.get_synaptic_update(
pre,
post,
synapse=synapse,
input_transform_gn=genome.neuron.transform,
update_type=synapse_util.UpdateType.BOTH,
env=env)
hebbian_update = blur.get_hebbian_update(
pre,
post,
genome.synapse.transform,
global_spec=network_spec,
env=env)
grad_weights, grad_image, y = tf_gradients(inp, out, ow)
np.set_printoptions(precision=4, linewidth=200)
with tf.Session():
verify_equal(update1, update2, hebbian_update, grad_weights.eval(),
grad_image.eval(), y.eval(), num_in, num_out)
def get_blur_state(env, inp, out, ow):
pre = np.concatenate([inp, np.zeros_like(inp)],
axis=-1).astype(blur_env.NP_FLOATING_TYPE)
post = np.concatenate([np.zeros_like(out), out],
axis=-1).astype(blur_env.NP_FLOATING_TYPE)
ww = ow.astype(blur_env.NP_FLOATING_TYPE)
ww = ww[Ellipsis, None]
synapse = synapse_util.combine_in_out_synapses(
ww, synapse_util.transpose_synapse(ww, env), env=env)
synapse = synapse_util.sync_states_synapse(synapse, env, num_states=2)
genome = genome_util.convert_genome_to_tf_variables(
genome_util.create_backprop_genome(num_species=1))
network_spec = blur.NetworkSpec()
network_spec.symmetric_synapses = True
network_spec.batch_average = False
network_spec.backward_update = 'multiplicative_second_state'
return pre, post, synapse, genome, network_spec