def predict():
"""Predict unseen images"""
"""Step 0: load data and trained model"""
mnist = input_data.read_data_sets("./data/", one_hot=True)
checkpoint_dir = sys.argv[1]
"""Step 1: build the rnn model"""
x = tf.placeholder("float", [None, n_steps, n_input])
y = tf.placeholder("float", [None, n_classes])
weights = tf.Variable(tf.random_normal([n_hidden, n_classes]), name='weights')
biases = tf.Variable(tf.random_normal([n_classes]), name='biases')
pred = rnn_model(x, weights, biases)
correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
"""Step 2: predict new images with the trained model"""
with tf.Session() as sess:
sess.run(tf.initialize_all_variables())
"""Step 2.0: load the trained model"""
checkpoint_file = tf.train.latest_checkpoint(checkpoint_dir + 'checkpoints')
print('Loaded the trained model: {}'.format(checkpoint_file))
saver = tf.train.Saver()
saver.restore(sess, checkpoint_file)
"""Step 2.1: predict new data"""
test_len = 500
test_data = mnist.test.images[:test_len].reshape((-1, n_steps, n_input))
test_label = mnist.test.labels[:test_len]
print("Testing Accuracy:", sess.run(accuracy, feed_dict={x: test_data, y: test_label}))
def test_conv_net(conv_net):
test_x = tf.placeholder(tf.float32, [None, 32, 32, 3])
test_k = tf.placeholder(tf.float32)
logits_out = conv_net(test_x, test_k)
assert logits_out.get_shape().as_list() == [None, 10],\
'Incorrect Model Output. Found {}'.format(logits_out.get_shape().as_list())
print('Neural Network Built!')
def test_flatten(flatten):
test_x = tf.placeholder(tf.float32, [None, 10, 30, 6])
flat_out = flatten(test_x)
assert flat_out.get_shape().as_list() == [None, 10*30*6],\
'Incorrect Shape. Found {} shape'.format(flat_out.get_shape().as_list())
_print_success_message()
def q_train(make_obs_ph_n,
act_space_n,
q_index,
q_func,
optimizer,
grad_norm_clipping=None,
local_q_func=False,
scope="trainer",
reuse=None,
num_units=64):
with tf.variable_scope(scope, reuse=reuse):
# create distribtuions
act_pdtype_n = [make_pdtype(act_space) for act_space in act_space_n]
# set up placeholders
obs_ph_n = make_obs_ph_n
act_ph_n = [
act_pdtype_n[i].sample_placeholder([None], name="action" + str(i))
for i in range(len(act_space_n))
]
target_ph = tf.placeholder(tf.float32, [None], name="target")
q_input = tf.concat(obs_ph_n + act_ph_n, 1)
if local_q_func:
q_input = tf.concat([obs_ph_n[q_index], act_ph_n[q_index]], 1)
q = q_func(q_input, 1, scope="q_func", num_units=num_units)[:, 0]
q_func_vars = U.scope_vars(U.absolute_scope_name("q_func"))
q_loss = tf.reduce_mean(tf.square(q - target_ph))
# viscosity solution to Bellman differential equation in place of an initial condition
q_reg = tf.reduce_mean(tf.square(q))
loss = q_loss #+ 1e-3 * q_reg
optimize_expr = U.minimize_and_clip(optimizer, loss, q_func_vars,
grad_norm_clipping)
# Create callable functions
train = U.function(inputs=obs_ph_n + act_ph_n + [target_ph],
outputs=loss,
updates=[optimize_expr])
q_values = U.function(obs_ph_n + act_ph_n, q)
# target network
target_q = q_func(q_input,
1,
scope="target_q_func",
num_units=num_units)[:, 0]
target_q_func_vars = U.scope_vars(
U.absolute_scope_name("target_q_func"))
update_target_q = make_update_exp(q_func_vars, target_q_func_vars)
target_q_values = U.function(obs_ph_n + act_ph_n, target_q)
return train, update_target_q, {
'q_values': q_values,
'target_q_values': target_q_values
}
def test_output(output):
test_x = tf.placeholder(tf.float32, [None, 128])
test_num_outputs = 40
output_out = output(test_x, test_num_outputs)
assert output_out.get_shape().as_list() == [None, 40],\
'Incorrect Shape. Found {} shape'.format(output_out.get_shape().as_list())
_print_success_message()
def test_fully_conn(fully_conn):
test_x = tf.placeholder(tf.float32, [None, 128])
test_num_outputs = 40
fc_out = fully_conn(test_x, test_num_outputs)
assert fc_out.get_shape().as_list() == [None, 40],\
'Incorrect Shape. Found {} shape'.format(fc_out.get_shape().as_list())
_print_success_message()
def test_con_pool(conv2d_maxpool):
test_x = tf.placeholder(tf.float32, [None, 32, 32, 5])
test_num_outputs = 10
test_con_k = (2, 2)
test_con_s = (4, 4)
test_pool_k = (2, 2)
test_pool_s = (2, 2)
conv2d_maxpool_out = conv2d_maxpool(test_x, test_num_outputs, test_con_k, test_con_s, test_pool_k, test_pool_s)
assert conv2d_maxpool_out.get_shape().as_list() == [None, 4, 4, 10],\
'Incorrect Shape. Found {} shape'.format(conv2d_maxpool_out.get_shape().as_list())
_print_success_message()
