def update(t):
# Get eta for this timestep (integer divide isn't too slow)
eta_i = min(t / steps_per_eta_val, eta_i_max)
eta = eta_vals[eta_i]
eta_text.set_text("$\eta=$%g" % eta)
# All particles take one step
model.step(state, L, N, v0, R, eta, t)
t = t + 1
# Update animation
particles.set_data(state[:, 0], state[:, 1])
leaders.set_data(state[range(N_ldr), 0], state[range(N_ldr), 1])
return particles, leaders, box, eta_text
def update(t):
# Get eta for this timestep
eta_i = min(t / steps_per_eta_val, eta_i_max)
eta = eta_vals[eta_i]
eta_text.set_text("$\eta=$%g" % eta)
# All particles take one step
model.step(state, L, N, v0, R, eta, t)
V = model.order_parameter(state, V_coeff)
V_series.append(V)
t = t + 1
# Update particle animation
particles.set_data(state[:, 0], state[:, 1])
leaders.set_data(state[range(N_ldr), 0], state[range(N_ldr), 1])
# Take average of 'window' data points every 'plot_step' steps
if t > window and t % plot_step == 0:
# Do averages
V_window = np.array(V_series[-window:])
V_mean = np.mean(V_window)
Vsq_mean = np.mean(V_window * V_window)
chi = chi_coeff * (Vsq_mean - V_mean * V_mean)
# Update data lists
V_mean_series.append(V_mean)
chi_series.append(chi)
t_series.append(t * dt)
# Update animated plots
V_plot.set_data(t_series, V_mean_series)
chi_plot.set_data(t_series, chi_series)
# Update limits if necessary
tmax = ax2.get_xlim()[1]
chimax = np.max(chi_series)
if t > tmax or chi > chimax:
ax2.set_xlim(0, 2 * tmax)
ax3.set_xlim(0, 2 * tmax)
ax3.set_ylim(0, 1.1 * chimax)
ax2.figure.canvas.draw()
ax3.figure.canvas.draw()
return particles, leaders, box, eta_text, V_plot, chi_plot
def decode():
with tf.Session() as sess:
# Create model and load parameters.
model = create_model(sess, True)
model.batch_size = batch_size # We decode one sentence at a time.
# Load vocabularies.
vocab_path = os.path.join(vocab_dir, "vocab")
vocab_list, vocab_dict = data_util.initialize_vocabulary(vocab_path)
# Decode from standard input.
document = data_util.basic_tokenizer(input_dir)
while document:
# Get token-ids for the input sentence.
token_ids = data_util.sentence_to_token_ids(document, vocab_dict)
# Which bucket does it belong to?
bucket_id = len(buckets) - 1
for i, bucket in enumerate(buckets):
if bucket[0] >= len(token_ids):
bucket_id = i
break
else:
logging.warning("Document truncated: %s", document)
# Get a 1-element batch to feed the sentence to the model.
encoder_inputs, decoder_inputs, target_weights = model.get_batch(
{bucket_id: [(token_ids, [])]}, bucket_id)
# Get output logits for the sentence.
_, _, output_logits = model.step(sess, encoder_inputs,
decoder_inputs, target_weights,
bucket_id, True)
# This is a greedy decoder - outputs are just argmaxes of output_logits.
outputs = [
int(np.argmax(logit, axis=1)) for logit in output_logits
]
# If there is an EOS symbol in outputs, cut them at that point.
if data_util.EOS_ID in outputs:
outputs = outputs[:outputs.index(data_util.EOS_ID)]
summary = data_util.token_ids_to_sentence(outputs, vocab_list)
print(' '.join(summary))
def nlu_interface(nlu_inputs, sess, model):
# nlu_output = {'nlu_result':{'intent':'', 'slots':[]}}
# CONFIRM_LIST, REJECT_LIST = get_y_n()
nlu_inputs = nlu_inputs.strip().replace(' ', '')
assert type(nlu_inputs) == str
inputs = cut_sentence(nlu_inputs)
id_inputs = data_utils.nlu_input_to_token_ids(inputs, input_vocab_path,
data_utils.tab_tokenizer)
_inputs, _labels, _sequence_length = model.get_one([[id_inputs, [0]]], 0)
# pdb.set_trace()
_, step_loss, logits, summary = model.step(sess, _inputs, _labels,
_sequence_length, True)
label_logit = logits['label'][0]
predict_label_ids = np.argmax(label_logit, 1)[:_sequence_length[0]]
predict_label = [rev_label_vocab[x] for x in predict_label_ids]
nlu_output = '\t'.join(predict_label)
return nlu_output
# ------------------ #
# Initialise model #
# ------------------ #
box1, box2, glob = model.initialise()
# --------------------------- #
# Integrate forward in time #
# --------------------------- #
for n in range(nt):
# Update fluxes, moisture, circulation using current temperatures
model.update(n, box1, box2, glob)
# Step temperatures forward
model.step(n, box1, box2, glob)
# Print every 'n_print' steps
if n % n_print == 0:
years, months, days = model.simulation_time(n)
print "Simulation time: %dy, %dm, %dd" %(years, months, days)
# Save every n_save steps (exluding step 0)
if (n+1) % n_save == 0:
years, months, days = model.simulation_time(n)
"Saving time series at simulation time: %d years, %d months, %d days" %(years, months, days)
model.save(n+1, box1, box2, glob)
# Update fluxes, moisture, circulation for final timestep
model.update(nt, box1, box2, glob)
import model
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import colors
import matplotlib.animation as animation
from agent import House
model = model.BurglaryModel(0, 128, 128, 5, .06, 5.6, 0.1, 0.019, 1, 5, 5, 0.3)
# Press the green button in the gutter to run the script.
if __name__ == '__main__':
for i in range(1000):
model.step()
print(i)
crime_counts = np.zeros((model.grid.width, model.grid.height))
for cell in model.grid.coord_iter():
content, x, y = cell
crimes = 0
for row in content:
if isinstance(row, House):
crimes = row.att_t
crime_counts[x][y] = crimes
norm = colors.Normalize(vmin=0.1, vmax=(model.theta * 0.75))
plt.imshow(crime_counts,
interpolation='nearest',
cmap=plt.get_cmap('seismic'),
norm=norm)
plt.colorbar()
# Print current repeat number
print " %d/%d" % (rep + 1, repeats)
# Create a brand new initial state
state = model.init(L, N, v0, R, eta)
# Initialise V, t and set V_coeff
V = 0
t = 0
# Loop over steps until order parameter hits threshold
while V < 0.98:
# All particles take one step
model.step(state, L, N, v0, R, eta, t)
V = model.order_parameter(state, V_coeff)
t += 1
# Save steps to burn in
nt_burn_repeats[rep] = t
# End loop over repeats
# Compute mean and standard deviation of burn-in time
nt_burn_mean[iN] = np.mean(nt_burn_repeats)
nt_burn_stderr[iN] = np.std(nt_burn_repeats) / np.sqrt(repeats)
iN += 1
# End loop over particle numbers
def train():
from_train = None
to_train = None
from_dev = None
to_dev = None
with tf.Session() as sess:
# Create model.
print("Creating %d layers of %d units." % (num_layers, size))
model = create_model(sess, False)
# Read data into buckets and compute their sizes.
print("Reading development and training data (limit: %d)." %
max_train_data_size)
dev_set = read_data(from_dev, to_dev)
train_set = read_data(from_train, to_train, max_train_data_size)
train_bucket_sizes = [len(train_set[b]) for b in range(len(buckets))]
train_total_size = float(sum(train_bucket_sizes))
# A bucket scale is a list of increasing numbers from 0 to 1 that we'll use
# to select a bucket. Length of [scale[i], scale[i+1]] is proportional to
# the size if i-th training bucket, as used later.
train_buckets_scale = [
sum(train_bucket_sizes[:i + 1]) / train_total_size
for i in range(len(train_bucket_sizes))
]
# This is the training loop.
step_time, loss = 0.0, 0.0
current_step = 0
previous_losses = []
while True:
# Choose a bucket according to data distribution. We pick a random number
# in [0, 1] and use the corresponding interval in train_buckets_scale.
random_number_01 = np.random.random_sample()
bucket_id = min([
i for i in range(len(train_buckets_scale))
if train_buckets_scale[i] > random_number_01
])
# Get a batch and make a step.
start_time = time.time()
encoder_inputs, decoder_inputs, target_weights = model.get_batch(
train_set, bucket_id)
_, step_loss, _ = model.step(sess, encoder_inputs, decoder_inputs,
target_weights, bucket_id, False)
step_time += (time.time() - start_time) / steps_per_checkpoint
loss += step_loss / steps_per_checkpoint
current_step += 1
# Once in a while, we save checkpoint, print statistics, and run evals.
if current_step % steps_per_checkpoint == 0:
# Print statistics for the previous epoch.
perplexity = math.exp(
float(loss)) if loss < 300 else float("inf")
print(
"global step %d learning rate %.4f step-time %.2f perplexity "
"%.2f" %
(model.global_step.eval(), model.learning_rate.eval(),
step_time, perplexity))
# Decrease learning rate if no improvement was seen over last 3 times.
if len(previous_losses) > 2 and loss > max(
previous_losses[-3:]):
sess.run(model.learning_rate_decay_op)
previous_losses.append(loss)
# Save checkpoint and zero timer and loss.
checkpoint_path = os.path.join(train_dir, "fyp.ckpt")
model.saver.save(sess,
checkpoint_path,
global_step=model.global_step)
step_time, loss = 0.0, 0.0
# Run evals on development set and print their perplexity.
for bucket_id in range(len(buckets)):
if len(dev_set[bucket_id]) == 0:
print(" eval: empty bucket %d" % (bucket_id))
continue
encoder_inputs, decoder_inputs, target_weights = model.get_batch(
dev_set, bucket_id)
_, eval_loss, _ = model.step(sess, encoder_inputs,
decoder_inputs,
target_weights, bucket_id,
True)
eval_ppx = math.exp(
float(eval_loss)) if eval_loss < 300 else float("inf")
print(" eval: bucket %d perplexity %.2f" %
(bucket_id, eval_ppx))
sys.stdout.flush()
def main(_):
pp.pprint(FLAGS.__flags)
emb = None
try:
# pre-trained chars embedding
emb = np.load("./data/emb.npy")
chars = cPickle.load(open("./data/vocab.pkl", 'rb'))
vocab_size, emb_size = np.shape(emb)
data_loader = TextLoader('./data', FLAGS.batch_size, chars)
except Exception:
data_loader = TextLoader('./data', FLAGS.batch_size)
emb_size = FLAGS.emb_size
vocab_size = data_loader.vocab_size
model = DialogueModel(batch_size=FLAGS.batch_size,
max_seq_length=data_loader.seq_length,
vocab_size=vocab_size,
pad_token_id=0,
unk_token_id=UNK_ID,
emb_size=emb_size,
memory_size=FLAGS.memory_size,
keep_prob=FLAGS.keep_prob,
learning_rate=FLAGS.learning_rate,
grad_clip=FLAGS.grad_clip,
temperature=FLAGS.temperature,
infer=False)
summaries = tf.summary.merge_all()
init = tf.global_variables_initializer()
# save hyper-parameters
cPickle.dump(FLAGS.__flags, open(FLAGS.logdir + "/hyperparams.pkl", 'wb'))
checkpoint = FLAGS.checkpoint + '/model.ckpt'
count = 0
saver = tf.train.Saver()
with tf.Session() as sess:
summary_writer = tf.summary.FileWriter(FLAGS.logdir, sess.graph)
sess.run(init)
if len(glob(checkpoint + "*")) > 0:
saver.restore(sess, checkpoint)
print("Model restored!")
else:
# load embedding
if emb is not None:
sess.run([], {model.embedding: emb})
print("Fresh variables!")
current_step = 0
count = 0
for e in range(FLAGS.num_epochs):
data_loader.reset_batch_pointer()
state = None
# iterate by batch
for _ in range(data_loader.num_batches):
x, y, input_lengths, output_lengths = data_loader.next_batch()
if (current_step + 1) % 10 != 0:
res = model.step(sess, x, y, input_lengths, output_lengths,
state)
else:
res = model.step(sess, x, y, input_lengths, output_lengths,
state, summaries)
summary_writer.add_summary(res["summary_out"],
current_step)
loss = res["loss"]
perplexity = np.exp(loss)
count += 1
print("{0}/{1}({2}), perplexity {3}".format(
current_step + 1,
FLAGS.num_epochs * data_loader.num_batches, e,
perplexity))
state = res["final_state"]
if (current_step + 1) % 2000 == 0:
count = 0
summary_writer.flush()
save_path = saver.save(sess, checkpoint)
print("Model saved in file:", save_path)
current_step = tf.train.global_step(sess, model.global_step)
summary_writer.close()
save_path = saver.save(sess, checkpoint)
print("Model saved in file:", save_path)
