def train():
model_version, model_name = get_latest_model()
logger.info("Training on gathered game data, initializing from {}".format(
model_name))
new_model_name = generate(model_version + 1)
logger.info("New model will be {}".format(new_model_name))
save_file = os.path.join(PATHS.MODELS_DIR, new_model_name)
try:
logger.info("Getting tf_records")
tf_records = sorted(
gfile.Glob(os.path.join(PATHS.TRAINING_CHUNK_DIR,
'*.tfrecord.zz')))
tf_records = tf_records[-1 *
(GLOBAL_PARAMETER_STORE.WINDOW_SIZE //
GLOBAL_PARAMETER_STORE.EXAMPLES_PER_RECORD):]
print("Training from:", tf_records[0], "to", tf_records[-1])
with timer("Training"):
network.train(PATHS.ESTIMATOR_WORKING_DIR, tf_records,
model_version + 1)
network.export_latest_checkpoint_model(PATHS.ESTIMATOR_WORKING_DIR,
save_file)
except:
logger.info("Got an error training")
logging.exception("Train error")
def train_network(dataPath, matPath, networkPath, epochSize, learningRate,
hiddenSize, batchSize):
try:
m_data = sio.loadmat(matPath)
im_data = m_data['images']
l_data = np.squeeze(m_data['labels'])
except:
print("\nPreprocessing train data\n")
training_data = create_training_data(dataPath)
imgs = []
lbs = []
for featureVector, label in training_data:
imgs.append(featureVector)
lbs.append(label)
im_data = np.squeeze(imgs).transpose()
l_data = np.asarray(lbs)
sio.savemat(matPath, {'images': im_data, 'labels': l_data})
trainset = network.ProcessedDataset(im_data, l_data)
trainloader = torch.utils.data.DataLoader(trainset,
batch_size=batchSize,
shuffle=True,
num_workers=0)
net = network.Feedforward(hiddenSize)
network.train(net, trainloader, networkPath, learningRate, epochSize)
def train_network():
"""
Train a neural net using the pipeline.
"""
import network
network.train()
input("Press Enter to continue...")
def run(arg, f1, f2):
if arg == '--train':
agent = objects.Agent(14, 2)
fight_data = construct.fights('../xml_data/schedule.xml')
write_file(fight_data, TRAIN_FILE_NAME)
network.train(agent)
elif arg == '--predict':
agent = objects.Agent(14, 2)
fight_data = construct.fights('../xml_data/schedule.xml')
write_file(fight_data, FIGHT_FILE_NAME)
network.predict(agent, f1.replace(" ", ""), f2.replace(" ", ""))
print "Arguments Processed!"
def train_nets(inputs, outputs, training_rate, epochs, batch_size, outer_min,
random_limit, layers, activations, d_activations, cost, d_cost,
num_nets):
minimum = 100
minnet = 3
while (minimum > outer_min): #or random_limit>100001):
sum = 0
#while minimum > outer_min:
print('building networks')
print("Random Limit :", random_limit)
build_networks(layers, activations, d_activations, cost, d_cost,
num_nets, random_limit)
for network in networks:
output = network.train(inputs, outputs, training_rate, epochs,
batch_size, False)
print('finished a network')
print("output error =", output)
# Everything below subjected to changes
sum += output
if output < minimum:
minimum = output
minnet = network
#if minimum < outer_min:
# return minnet
print("Couldn't find anything")
avg = sum / num_nets
print(minimum, avg, random_limit)
networks.clear()
random_limit *= 10
return minnet
def train_net(self, comb):
self.txt_resultsText.clear()
if (self.dataPath is not None):
print("Loading Data...")
if (comb and len(self.app_combs) > 0):
train_data, train_labels, test_data, test_labels = data.data_init_comb(
self.dataPath,
train_comb=self.app_combs,
test_comb=self.app_combs)
_, _, self.Net = network.train(train_data, train_labels,
test_data, test_labels)
if (not comb):
train_data, train_labels, test_data, test_labels = data.data_init_measured(
self.dataPath)
_, _, self.Net = network.train(train_data, train_labels,
test_data, test_labels)
self.btn_disaggComb.setEnabled(True)
def train_network():
print('BEGIN_TRAINING')
images = normalize_images_255(read_images(TRAIN_IMAGE_FILE))
labels = read_labels(TRAIN_LABEL_FILE)
network = Network([784, 16, 16, 10])
print('TRAINING')
train(network, images, labels, 0.5, 20)
training_values = []
for i in range(len(images)):
classify_outputs = classify_network(network, images[i])
output_value = single_network_output(classify_outputs)
training_values.append(output_value)
accuracy = calculate_network_accuracy(training_values, labels)
print('TRAIN_COMPLETE')
print('Accuracy')
print(accuracy)
return network
def base_experiment():
n_train, n_valid = 55000, 5000
num_pruning_iter = 15 # How many iterative pruning steps to perform
print(
"Running a base experiment with input_size:{}, hidden_size:{}, num_classes:{}, "
"batch_size:{}, num_epochs:{}, num_pruning_iter:{}, pruning_rates:{}".
format(network.input_size, network.hidden_size, network.num_classes,
network.batch_size, network.num_epochs, num_pruning_iter,
network.pruning_rates))
train_loader, val_loader = dataset.init_data_mask_base_expt(
n_train, n_valid)
model = network.MultiLayerPerceptron().to(network.device)
model.apply(weights_init)
presets = {}
for name, param in model.state_dict().items():
presets[name] = param.clone()
model.presets = presets
early_stop_iteration = network.train(model, train_loader, val_loader)
test_accuracy_history = []
test_accuracy = network.test(model)
test_accuracy_history.append(test_accuracy)
early_stop_iteration_history = [early_stop_iteration]
for iter in range(num_pruning_iter):
# This is the percentage of weights remaining in the network after pruning
print(
"\tResults for pruning round {} with percentage of weights remaining {}"
.format(iter + 1, 100 * 0.8**(iter + 1)))
network.prune(model)
network.reset_params(model)
early_stop_iteration = network.train(model, train_loader, val_loader)
early_stop_iteration_history.append(early_stop_iteration)
test_accuracy = network.test(model)
test_accuracy_history.append(test_accuracy)
print('Test accuracy history {}'.format(test_accuracy_history))
print(
'Early stop iteration history {}'.format(early_stop_iteration_history))
def train():
"""Train network for a number of steps."""
with tf.Graph().as_default():
global_step = tf.Variable(0, trainable=False)
# Get images and labels for network.
images, labels = network.distorted_inputs()
# Build a Graph that computes the logits predictions from the
# inference model.
logits = network.inference(images)
# Calculate loss.
loss = network.loss(logits, labels)
# Build a Graph that trains the model with one batch of examples and
# updates the model parameters.
train_op = network.train(loss, global_step)
# Create a saver.
saver = tf.train.Saver(tf.all_variables())
# Build the summary operation based on the TF collection of Summaries.
#summary_op = tf.merge_all_summaries()
# Build an initialization operation to run below.
init = tf.initialize_all_variables()
# Start running operations on the Graph.
sess = tf.Session(config=tf.ConfigProto(
log_device_placement=FLAGS.log_device_placement))
sess.run(init)
# Start the queue runners.
tf.train.start_queue_runners(sess=sess)
#summary_writer = tf.train.SummaryWriter(FLAGS.train_dir,
# graph_def=sess.graph_def)
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 0:
num_examples_per_step = FLAGS.batch_input_size
examples_per_sec = num_examples_per_step / duration
sec_per_batch = float(duration)
format_str = (
'%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print(format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
def level(data, targets, layers, pattern, seed):
# ---------------------------------------------------------------------------------------------------
# set up dataset for nn
active = organiser.remove_non_pfam(data)
if pattern is None:
active = organiser.reduce_none(targets, active, 20, seed)
else:
active = organiser.get_current(targets, active, pattern)
inputs = data.dense_shape[1]
outputs = network.get_outputs(targets, active, pattern)
nodes = range(inputs, outputs, (inputs-outputs)/layers)
nn = network.build(inputs, outputs, layers, "sigmoid", nodes, "rmsprop", "mse", ["accuracy"])
network.train(nn, data, targets, active, 1000, 5, 0)
return nn
def train():
"""Train network for a number of steps."""
with tf.Graph().as_default():
global_step = tf.Variable(0, trainable=False)
# Get images and labels for network.
images, labels = network.distorted_inputs()
# Build a Graph that computes the logits predictions from the
# inference model.
logits = network.inference(images)
# Calculate loss.
loss = network.loss(logits, labels)
# Build a Graph that trains the model with one batch of examples and
# updates the model parameters.
train_op = network.train(loss, global_step)
# Create a saver.
saver = tf.train.Saver(tf.all_variables())
# Build the summary operation based on the TF collection of Summaries.
#summary_op = tf.merge_all_summaries()
# Build an initialization operation to run below.
init = tf.initialize_all_variables()
# Start running operations on the Graph.
sess = tf.Session(config=tf.ConfigProto(
log_device_placement=FLAGS.log_device_placement))
sess.run(init)
# Start the queue runners.
tf.train.start_queue_runners(sess=sess)
#summary_writer = tf.train.SummaryWriter(FLAGS.train_dir,
# graph_def=sess.graph_def)
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 0:
num_examples_per_step = FLAGS.batch_input_size
examples_per_sec = num_examples_per_step / duration
sec_per_batch = float(duration)
format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print (format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
def train():
with tf.Graph().as_default():
global_step = tf.Variable(0, trainable=False)
images, labels = network.distorted_inputs()
logits = network.inference(images)
loss = network.loss(logits, labels)
train_op = network.train(loss, global_step)
saver = tf.train.Saver(tf.all_variables())
summary_op = tf.merge_all_summaries()
init = tf.initialize_all_variables()
sess = tf.Session(config=tf.ConfigProto(
log_device_placement=FLAGS.log_device_placement))
sess.run(init)
tf.train.start_queue_runners(sess=sess)
summary_writer = tf.train.SummaryWriter(FLAGS.train_dir, sess.graph)
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 0:
num_examples_per_step = FLAGS.batch_input_size
examples_per_sec = num_examples_per_step / duration
sec_per_batch = float(duration)
format_str = (
'%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print(format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
if step % 100 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
if step % 1000 == 0 or (step + 1) == FLAGS.max_steps:
checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def reinforcement_learn(predictor, turns=1000):
game = CookieClickerGame(verbose=False)
game.cookies = 100
state = State(game)
# predictor = Predictor()
# predictor = LinearPredictor(len(state.get_state()), state.get_action_space())
for i in range(1000):
# print(game.str_basic())
# print(game)
# pred = predictor.predict(state)
inputs_tensor = torch.tensor(state.get_state(), dtype=torch.float)
pred_tensor = predictor(inputs_tensor)
avail_preds = [
p * a for p, a in zip(pred_tensor.detach().numpy(),
state.get_action_availability())
]
# pred_idx = argmax(avail_preds)
pred_idx = random.choices(range(len(avail_preds)), avail_preds, k=1)[0]
action = state.prediction_to_action(pred_idx)
reward = state.perform_action(action)
# print(avail_preds, "->", pred_idx)
# print("reward:", reward)
desired_tensor = pred_tensor.clone()
alpha = 0.01
if reward > 0:
desired_tensor[pred_idx] += alpha
train(predictor, desired_tensor, pred_tensor)
elif reward < 0:
desired_tensor[pred_idx] -= alpha
train(predictor, desired_tensor, pred_tensor)
print(game.total_cookies, game.cpt)
def test_network():
wine_attributes = ["alch","malic","ash","alcash","mag","phen","flav","nfphens","proant","color","hue","dil","prol"]
columns = ["class","alch","malic","ash","alcash","mag","phen","flav","nfphens","proant","color","hue","dil","prol"]
"""
0) class
1) Alcohol
2) Malic acid
3) Ash
4) Alcalinity of ash
5) Magnesium
6) Total phenols
7) Flavanoids
8) Nonflavanoid phenols
9) Proanthocyanins
10)Color intensity
11)Hue
12)OD280/OD315 of diluted wines
13)Proline
"""
from Induction.IntGrad.integratedGradients import random_baseline_integrated_gradients, integrated_gradients
exit()
df = read_data_pd("../../Data/wine.csv",columns = columns)
df.columns = columns # Add columns to dataframe.
#Cov.columns = ["Sequence", "Start", "End", "Coverage"]
dataman = Datamanager.Datamanager(dataframe_train=df,classes=3,dataset="wine")
model = network.NN_3_25("wine",in_dim=13,out_dim=3)
print(model.input_type)
optimizer = optim.Adam(model.parameters(), lr=0.01,betas=(0.9,0.999),eps=1e-6)
#loss = network.RootMeanSquareLoss()
#loss = t_loss.L1Loss()
loss = t_loss.MSELoss()
network.train(model, dataman,validation=True, optimizer = optimizer,loss_function = loss, batch=20, iterations=50)
def split_data_experiment():
n_train1, n_train2 = 27500, 27500
n_valid1, n_valid2 = 2500, 2500
num_pruning_iter = 27 # How many iterative pruning steps to perform
print(
"Running a data split experiment with input_size:{}, hidden_size:{}, num_classes:{}, "
"batch_size:{}, num_epochs:{}, num_pruning_iter:{}, pruning_rates:{}".
format(network.input_size, network.hidden_size, network.num_classes,
network.batch_size, network.num_epochs, num_pruning_iter,
network.pruning_rates))
train_loader1, val_loader1, train_loader2, val_loader2 = \
dataset.init_data_mask_split_data_expt(n_train1, n_valid1, n_train2, n_valid2)
model = network.MultiLayerPerceptron().to(network.device)
model.apply(weights_init)
presets = {}
for name, param in model.state_dict().items():
presets[name] = param.clone()
model.presets = presets
network.train(model, train_loader1, val_loader1)
test_accuracy1 = network.test(model)
test_accuracy_history1 = [test_accuracy1]
test_accuracy_history2 = [test_accuracy1]
for iter in range(num_pruning_iter):
# This is the percentage of weights remaining in the network after pruning
print(
"Results for pruning round {} with percentage of weights remaining {}"
.format(iter + 1, 100 * 0.8**(iter + 1)))
# prune model after training with first half of data
network.prune(model)
# Reset, retrain on second half of data - perform testing
network.reset_params(model)
network.train(model, train_loader2, val_loader1)
test_accuracy2 = network.test(model)
test_accuracy_history2.append(test_accuracy2)
# reset the model, retrain with first half of data - then perform testing
network.reset_params(model)
network.train(model, train_loader1, val_loader1)
test_accuracy1 = network.test(model)
test_accuracy_history1.append(test_accuracy1)
print(
'Test accuracy history after re-training with first half of the training dataset {}'
.format(test_accuracy_history1))
print(
'Test accuracy history after re-training with second half of the training dataset {}'
.format(test_accuracy_history2))
visualize.plot_test_accuracy_coarse(test_accuracy_history1,
test_accuracy_history2)
def train():
with tf.Graph().as_default():
global_step = tf.Variable(0, trainable=False)
# 获得图片数据和对应的标签batch
float_image, label = tfrecord.train_data_read(tfrecord_path=FLAGS.train_data)
images, labels = tfrecord.create_batch(float_image,label,count_num=FLAGS.train_num)
logits = network.inference(images)
# 误差计算
loss = network.loss(logits, labels)
# 模型训练
train_op = network.train(loss, global_step)
# 存储模型
saver = tf.train.Saver(tf.global_variables())
# 存储所有操作
summary_op = tf.summary.merge_all()
# 初始化所有变量.
init = tf.initialize_all_variables()
# 开始计算流图
sess = tf.Session(config=tf.ConfigProto(
log_device_placement=FLAGS.log_device_placement))
sess.run(init)
# 队列开始
tf.train.start_queue_runners(sess=sess)
summary_writer = tf.summary.FileWriter(FLAGS.train_dir,
graph_def=sess.graph_def)
for step in xrange(FLAGS.max_steps):
start_time = time.time()
_, loss_value = sess.run([train_op, loss])
duration = time.time() - start_time
assert not np.isnan(loss_value), 'Model diverged with loss = NaN'
if step % 10 == 0:
num_examples_per_step = FLAGS.batch_size
examples_per_sec = num_examples_per_step / duration
sec_per_batch = float(duration)
format_str = ('%s: step %d, loss = %.2f (%.1f examples/sec; %.3f '
'sec/batch)')
print (format_str % (datetime.now(), step, loss_value,
examples_per_sec, sec_per_batch))
if step % 50 == 0:
summary_str = sess.run(summary_op)
summary_writer.add_summary(summary_str, step)
# 保存模型检查点.
if step % 1000 == 0 or (step + 1) == FLAGS.max_steps:
checkpoint_path = os.path.join(FLAGS.train_dir, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=step)
def train(network, train_images, train_labels, batch_size, hm_epochs):
for epoch in range(hm_epochs):
batch_count = round(len(train_images) / batch_size)
c_average = 0
for batch in range(batch_count):
c = network.train(
train_images[batch * batch_size:batch * batch_size +
batch_size],
train_labels[batch * batch_size:batch * batch_size +
batch_size])
c_average += c * len(train_labels[batch * batch_size:batch *
batch_size + batch_size])
print('Batch {} out of {} finished'.format(batch + 1, batch_count))
print('Epoch {} completet out of {} cost:\t{c}'.format(
epoch + 1, hm_epochs, c_average / len(train_images)))
def main():
print("Extracting Features...")
Training_DS, Testing_DS = network.getDataSet()
print("Features Extracted")
print("Building Network...")
net = network.build(10)
print("Network Built")
print("Training...")
trainer = network.train(net, Training_DS, epoch=3000)
print("Trained")
print("Testing..")
test(net, Testing_DS, trainer)
return net
def train_nets(inputs, outputs, training_rate, epochs, batch_size, outer_min,
random_limit, layers, activations, d_activations, cost, d_cost,
num_nets):
minimum = 100
minnet = 3
while minimum > outer_min:
build_networks(layers, activations, d_activations, cost, d_cost,
num_nets, random_limit)
for network in networks:
output = network.train(inputs, outputs, training_rate, epochs,
batch_size, False)
if output < minimum:
minimum = output
minnet = network
print("Couldn't find anything")
print(minimum, random_limit)
random_limit *= 10
networks.clear()
return minnet
def train(model):
torch.manual_seed(2)
landmarks = face_key_point.load_face_landmark_data(path=os.path.join(
"data", "training.csv"),
batch_size=32)
train_d = landmarks.data["train"]
valid_d = landmarks.data["valid"]
loss_fn = torch.nn.MSELoss()
optimizer = optim.RMSprop(model.parameters(), lr=1e-3, weight_decay=1e-5)
num_epochs = 500
params = {
'model': model,
'train_loader': train_d,
'test_loader': valid_d,
'num_epochs': num_epochs,
'loss_fn': loss_fn,
'optimizer': optimizer,
}
loss_train, loss_test, model = network.train(**params)
network.write_loss(loss_train, "train.txt")
torch.save(model.state_dict(), 'model.pt')
def evaluateFitnessNN(self, nn): return(nn.train(self.train_train_data[0], self.train_train_data[1])/1797.0)
import numpy as np
import matplotlib.pyplot as plt
import utilities
import preproc
import network
import data
test_comb = [
'0b11100',
'0b11011',
'0b11111',
'0b10101',
'0b00100',
'0b10111'
]
train_data_unscaled, train_labels, test_data_unscaled, test_labels = data.data_init_comb('data/real_world_new/', train_comb=test_comb, test_comb=test_comb)
_,_, NET = network.train(train_data_unscaled, train_labels, test_data_unscaled, test_labels)
def main():
args = cfg.parse_args()
torch.cuda.manual_seed(args.random_seed)
# set visible GPU ids
if len(args.gpu_ids) > 0:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_ids
# set TensorFlow environment for evaluation (calculate IS and FID)
_init_inception()
inception_path = check_or_download_inception('./tmp/imagenet/')
create_inception_graph(inception_path)
# the first GPU in visible GPUs is dedicated for evaluation (running Inception model)
str_ids = args.gpu_ids.split(',')
args.gpu_ids = []
for id in range(len(str_ids)):
if id >= 0:
args.gpu_ids.append(id)
if len(args.gpu_ids) > 1:
args.gpu_ids = args.gpu_ids[1:]
else:
args.gpu_ids = args.gpu_ids
# genotype G
genotypes_root = os.path.join('exps', args.genotypes_exp, 'Genotypes')
genotype_G = np.load(os.path.join(genotypes_root, 'latest_G.npy'))
# import network from genotype
basemodel_gen = eval('archs.' + args.arch + '.Generator')(args, genotype_G)
gen_net = torch.nn.DataParallel(
basemodel_gen, device_ids=args.gpu_ids).cuda(args.gpu_ids[0])
basemodel_dis = eval('archs.' + args.arch + '.Discriminator')(args)
dis_net = torch.nn.DataParallel(
basemodel_dis, device_ids=args.gpu_ids).cuda(args.gpu_ids[0])
# basemodel_gen = eval('archs.' + args.arch + '.Generator')(args=args)
# gen_net = torch.nn.DataParallel(basemodel_gen, device_ids=args.gpu_ids).cuda(args.gpu_ids[0])
# basemodel_dis = eval('archs.' + args.arch + '.Discriminator')(args=args)
# dis_net = torch.nn.DataParallel(basemodel_dis, device_ids=args.gpu_ids).cuda(args.gpu_ids[0])
# weight init
def weights_init(m):
classname = m.__class__.__name__
if classname.find('Conv2d') != -1:
if args.init_type == 'normal':
nn.init.normal_(m.weight.data, 0.0, 0.02)
elif args.init_type == 'orth':
nn.init.orthogonal_(m.weight.data)
elif args.init_type == 'xavier_uniform':
nn.init.xavier_uniform(m.weight.data, 1.)
else:
raise NotImplementedError('{} unknown inital type'.format(
args.init_type))
elif classname.find('BatchNorm2d') != -1:
nn.init.normal_(m.weight.data, 1.0, 0.02)
nn.init.constant_(m.bias.data, 0.0)
gen_net.apply(weights_init)
dis_net.apply(weights_init)
# set up data_loader
dataset = datasets.ImageDataset(args)
train_loader = dataset.train
# epoch number for dis_net
args.max_epoch_D = args.max_epoch_G * args.n_critic
if args.max_iter_G:
args.max_epoch_D = np.ceil(args.max_iter_G * args.n_critic /
len(train_loader))
max_iter_D = args.max_epoch_D * len(train_loader)
# set optimizer
gen_optimizer = torch.optim.Adam(
filter(lambda p: p.requires_grad, gen_net.parameters()), args.g_lr,
(args.beta1, args.beta2))
dis_optimizer = torch.optim.Adam(
filter(lambda p: p.requires_grad, dis_net.parameters()), args.d_lr,
(args.beta1, args.beta2))
gen_scheduler = LinearLrDecay(gen_optimizer, args.g_lr, 0.0, 0, max_iter_D)
dis_scheduler = LinearLrDecay(dis_optimizer, args.d_lr, 0.0, 0, max_iter_D)
# fid stat
if args.dataset.lower() == 'cifar10':
fid_stat = 'fid_stat/fid_stats_cifar10_train.npz'
elif args.dataset.lower() == 'stl10':
fid_stat = 'fid_stat/stl10_train_unlabeled_fid_stats_48.npz'
else:
raise NotImplementedError(f'no fid stat for {args.dataset.lower()}')
assert os.path.exists(fid_stat)
# initial
gen_avg_param = copy_params(gen_net)
start_epoch = 0
best_fid = 1e4
# set writer
if args.checkpoint:
# resuming
print(f'=> resuming from {args.checkpoint}')
assert os.path.exists(os.path.join('exps', args.checkpoint))
checkpoint_file = os.path.join('exps', args.checkpoint, 'Model',
'checkpoint_best.pth')
assert os.path.exists(checkpoint_file)
checkpoint = torch.load(checkpoint_file)
start_epoch = checkpoint['epoch']
best_fid = checkpoint['best_fid']
gen_net.load_state_dict(checkpoint['gen_state_dict'])
dis_net.load_state_dict(checkpoint['dis_state_dict'])
gen_optimizer.load_state_dict(checkpoint['gen_optimizer'])
dis_optimizer.load_state_dict(checkpoint['dis_optimizer'])
avg_gen_net = deepcopy(gen_net)
avg_gen_net.load_state_dict(checkpoint['avg_gen_state_dict'])
gen_avg_param = copy_params(avg_gen_net)
del avg_gen_net
args.path_helper = checkpoint['path_helper']
logger = create_logger(args.path_helper['log_path'])
logger.info(
f'=> loaded checkpoint {checkpoint_file} (epoch {start_epoch})')
else:
# create new log dir
assert args.exp_name
args.path_helper = set_log_dir('exps', args.exp_name)
logger = create_logger(args.path_helper['log_path'])
logger.info(args)
writer_dict = {
'writer': SummaryWriter(args.path_helper['log_path']),
'train_global_steps': start_epoch * len(train_loader),
'valid_global_steps': start_epoch // args.val_freq,
}
# model size
logger.info('Param size of G = %fMB', count_parameters_in_MB(gen_net))
logger.info('Param size of D = %fMB', count_parameters_in_MB(dis_net))
print_FLOPs(basemodel_gen, (1, args.latent_dim), logger)
print_FLOPs(basemodel_dis, (1, 3, args.img_size, args.img_size), logger)
# for visualization
if args.draw_arch:
from utils.genotype import draw_graph_G
draw_graph_G(genotype_G,
save=True,
file_path=os.path.join(args.path_helper['graph_vis_path'],
'latest_G'))
fixed_z = torch.cuda.FloatTensor(
np.random.normal(0, 1, (100, args.latent_dim)))
# train loop
for epoch in tqdm(range(int(start_epoch), int(args.max_epoch_D)),
desc='total progress'):
lr_schedulers = (gen_scheduler,
dis_scheduler) if args.lr_decay else None
train(args, gen_net, dis_net, gen_optimizer, dis_optimizer,
gen_avg_param, train_loader, epoch, writer_dict, lr_schedulers)
if epoch % args.val_freq == 0 or epoch == int(args.max_epoch_D) - 1:
backup_param = copy_params(gen_net)
load_params(gen_net, gen_avg_param)
inception_score, std, fid_score = validate(args, fixed_z, fid_stat,
gen_net, writer_dict)
logger.info(
f'Inception score mean: {inception_score}, Inception score std: {std}, '
f'FID score: {fid_score} || @ epoch {epoch}.')
load_params(gen_net, backup_param)
if fid_score < best_fid:
best_fid = fid_score
is_best = True
else:
is_best = False
else:
is_best = False
# save model
avg_gen_net = deepcopy(gen_net)
load_params(avg_gen_net, gen_avg_param)
save_checkpoint(
{
'epoch': epoch + 1,
'model': args.arch,
'gen_state_dict': gen_net.state_dict(),
'dis_state_dict': dis_net.state_dict(),
'avg_gen_state_dict': avg_gen_net.state_dict(),
'gen_optimizer': gen_optimizer.state_dict(),
'dis_optimizer': dis_optimizer.state_dict(),
'best_fid': best_fid,
'path_helper': args.path_helper
}, is_best, args.path_helper['ckpt_path'])
del avg_gen_net
import network
import tool
if __name__ == '__main__':
network.train()
'''times = tool.loadList('./times.txt')
accus = tool.loadList('./accus.txt')
ttimes = []
aaccus = []
for i in xrange(len(times)):
if i % 3 == 0:
ttimes.append(times[i] / 3)
aaccus.append(accus[i])
tool.showXYData(ttimes, aaccus, './times_accus.png')'''
l.FullyConnectedLayer(height=10, init_func=f.glorot_uniform, act_func=f.softmax)
], f.categorical_crossentropy)
optimizer = o.SGD(0.1)
num_epochs = 2
batch_size = 8
return net, optimizer, num_epochs, batch_size
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("data", help="the path to the MNIST data set in .npz format (generated using utils.py)")
parser.add_argument("func", help="the function name of the example to be run")
args = parser.parse_args()
np.random.seed(314)
u.print("Loading '%s'..." % args.data, bcolor=u.bcolors.BOLD)
trn_set, tst_set = u.load_mnist_npz(args.data)
trn_set, vld_set = (trn_set[0][:50000], trn_set[1][:50000]), (trn_set[0][50000:], trn_set[1][50000:])
u.print("Loading '%s'..." % args.func, bcolor=u.bcolors.BOLD)
net, optimizer, num_epochs, batch_size = locals()[args.func]()
u.print(inspect.getsource(locals()[args.func]).strip())
u.print("Training network...", bcolor=u.bcolors.BOLD)
n.train(net, optimizer, num_epochs, batch_size, trn_set, vld_set)
u.print("Testing network...", bcolor=u.bcolors.BOLD)
accuracy = n.test(net, tst_set)
u.print("Test accuracy: %0.2f%%" % (accuracy*100))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument(
"data",
help=
"the path to the MNIST data set in .npz format (generated using utils.py)"
)
parser.add_argument("func",
help="the function name of the example to be run")
args = parser.parse_args()
np.random.seed(314)
u.print("Loading '%s'..." % args.data, bcolor=u.bcolors.BOLD)
trn_set, tst_set = u.load_mnist_npz(args.data)
trn_set, vld_set = (trn_set[0][:50000],
trn_set[1][:50000]), (trn_set[0][50000:],
trn_set[1][50000:])
u.print("Loading '%s'..." % args.func, bcolor=u.bcolors.BOLD)
net, optimizer, num_epochs, batch_size = locals()[args.func]()
u.print(inspect.getsource(locals()[args.func]).strip())
u.print("Training network...", bcolor=u.bcolors.BOLD)
n.train(net, optimizer, num_epochs, batch_size, trn_set, vld_set)
u.print("Testing network...", bcolor=u.bcolors.BOLD)
accuracy = n.test(net, tst_set)
u.print("Test accuracy: %0.2f%%" % (accuracy * 100))
]
#Settings
# number of samples in the data set
N_SAMPLES = 1000
# ratio between training and test sets
TEST_SIZE = 0.1
#Data set
X, y = make_moons(n_samples=N_SAMPLES, noise=0.2, random_state=100)
X_train, X_test, y_train, y_test = train_test_split(X,
y,
test_size=TEST_SIZE,
random_state=42)
plt.make_plot(X, y, "Dataset")
# Training
params_values, cost = net.train(
np.transpose(X_train), np.transpose(y_train.reshape(
(y_train.shape[0], 1))), NN_ARCHITECTURE, 10000, 0.01)
# Prediction
Y_test_hat, _ = net.full_forward_propagation(np.transpose(X_test),
params_values, NN_ARCHITECTURE)
# Accuracy achieved on the test set
acc_test = co.get_accuracy_value(
Y_test_hat, np.transpose(y_test.reshape((y_test.shape[0], 1))))
print("Test set accuracy: {:.2f} - David".format(acc_test))
import network as nn import data as dt train_inputs, train_labels, test_inputs, test_labels = dt.prepare_data() nn.train(train_inputs, train_labels, test_inputs, test_labels) # Rodar o tensorboard com o comando: # tensorboard --logdir logs/fit # no prompt
weight0 = 2 * np.random.random((784, 50)) - 1
weight1 = 2 * np.random.random((50, 10)) - 1
"""Loading the MNIST data set into three lists two containg the training data
and the third one containing the test data"""
tr_data, val_data, test_data = ml.load_data()
"""Fitting the 28*28 input image into a numpy array of 784*1 dimension"""
tr_inputs = [np.reshape(a, (784, 1)) for a in tr_data[0]]
"""Converting the single output into a numpy array of 10 dimensions with 1 at
the index of the output an 0 elsewhere"""
tr_output = [ml.vectr_result(a) for a in tr_data[1]]
"""Loop to train the data taking an input of 10,000 images"""
for i in range(50000):
weight0, weight1 = net.train(tr_inputs[i], tr_output[i], weight0, weight1)
if (i % 500) == 0:
br.progress(i, 50000)
br.progress(50000, 50000, cond=True)
print("\n")
print("Network Trained and ready to be operated")
te_inputs = [np.reshape(a, (784, 1)) for a in test_data[0]]
te_output = test_data[1]
"""Function to check the accuracy of our trained network by testing it on
unchecked data of 10,000 images"""
tt.check(te_inputs, te_output, weight0, weight1)
print("x_train.shape = ", x_train.shape)
print("y_train.shape = ", y_train.shape)
print("x_val.shape = ", x_val.shape)
print("y_val.shape = ", y_val.shape)
TRAIN_EPOCHS = args.TRAIN_EPOCHS
BATCH_SIZES = args.BATCH_SIZES
LEARNING_RATE = args.LEARNING_RATE
TOTAL_BATCHES = int(n_examples / BATCH_SIZES + 0.5)
DISPLAY_EPOCH = args.DISPLAY_EPOCH
X = network.get_placeholder_X(img_h, img_w, n_features)
Y = network.get_placeholder_Y(n_labels)
outputs, predictions = network.forward(X, is_training=True)
loss, train_op = network.train(LEARNING_RATE, outputs, Y)
correct_prediction, acc = network.accuracy(predictions, Y)
saver = network.save()
logs_path = args.LOG_DIR
tf.summary.scalar("loss", loss)
tf.summary.scalar("accuracy", acc)
merged_summary_op = tf.summary.merge_all()
print("\nStart Training!!!\n")
with tf.Session() as sess:
sess.run(
[tf.global_variables_initializer(),
tf.local_variables_initializer()])
summary_writer = tf.summary.FileWriter(logs_path,
theta0=test_set[:, 0] + test_set[:, 2],
theta1=test_set[:, 1] + test_set[:, 3])
# Scale inputs to range [-1. 1.]
training_input[:, :2] /= np.sum(arm_length)
test_input[:, :2] /= np.sum(arm_length)
training_input[:, 2:] /= max_delta
test_input[:, 2:] /= max_delta
# Baseline MF input to PN: required to map network response to target range
pn_baselines = np.random.normal(1.2, 0.1, number_pn)
net = create_network(pn_baselines)
# Train with algorithm by Bouvier et al. 2018
print("Train network...")
training_errors, error_estimates, training_responses = train(
net, training_input, training_targets, pn_baselines, arm_length)
# Test network performance
print("Test network performance...")
test_errors, test_responses, pn_activity = test(net, test_input, test_targets,
pn_baselines, arm_length)
filename = time.strftime("%Y%m%d%H%M")
filename = filename + "_results.npz"
filename = os.path.join(sys.argv[2], filename)
np.savez_compressed(filename,
training_errors=training_errors,
error_estimates=error_estimates,
training_angles=training_set,
training_targets=training_targets,
hdf5 = "/media/zac/easystore/dataset.hdf5"
# batches = readFromHDF5(hdf5, 100, 10)
# print(batches)
# file = tables.open_file(hdf5, mode='r')
network = CNN(.0000001, (224, 224), 140, 128)
# totalBatch = file.root.trainImg.shape[0] // network.batchSize
# testBatch = file.root.testImg.shape[0] // network.batchSize
# file.close()
# # # print(getClassfierCount("alligned_db/aligned_images_DB"))
gpu_options = tf.GPUOptions(allow_growth=True)
with tf.Session(config=tf.ConfigProto(gpu_options=gpu_options)) as sess:
#178
network.train(sess, 33, 33, 100)
# print(batches
# 10 faces
# 0.14424242424242426 0
# 0.14424242424242426 10
# 0.14454545454545453 20
# 0.14484848484848484 30
# 0.14484848484848484 40
# 0.14515151515151514 50
# 0.14515151515151514 60
# 0.14606060606060606 70
# 0.1478787878787879 90
# 0.14939393939393938 110
# 0.1481818181818182 130
file_source_lang='small_vocab_en',
file_target_lang='small_vocab_fr',
load_method=load_from_file,
)
# create mappings
x_sent_array, x_vocab_len, x_word_to_idx, x_idx_to_word = preprocess_data(
x, conf['MAX_LEN'], conf['VOCAB_SIZE'], conf['NUM_SENT'])
y_sent_array, y_vocab_len, y_word_to_idx, y_idx_to_word = preprocess_data(
y, conf['MAX_LEN'], conf['VOCAB_SIZE'], conf['NUM_SENT'])
# Find the length of the longest sequence
x_max_len = max([len(sentence) for sentence in x_sent_array])
y_max_len = max([len(sentence) for sentence in y_sent_array])
# Padding zeros to make all sequences have a same length with the longest one
print('Zero padding...')
X = pad_sequences(x_sent_array, maxlen=x_max_len, dtype='int32')
y = pad_sequences(y_sent_array, maxlen=y_max_len, dtype='int32')
# Creating the network model
print('Compiling model...')
model = create_model(x_vocab_len, x_max_len, y_vocab_len, y_max_len,
conf['HIDDEN_DIM'], conf['LAYER_NUM'])
# Finding trained weights of previous epoch if any
saved_weights = find_checkpoint_file('.')
saved_weights = []
train(X, y, y_word_to_idx, y_max_len, saved_weights, model, conf)