def train(args):
fnames = glob.glob('../mp3/*.mp3')[:1]
traces = [util.loadf(fname) for fname in fnames]
traces = np.hstack(traces)
dirname = 'save-vrnn'
if not os.path.exists(dirname):
os.makedirs(dirname)
with open(os.path.join(dirname, 'config.pkl'), 'w') as f:
cPickle.dump(args, f)
model = VRNN(args)
# load previously trained model if applicable
ckpt = tf.train.get_checkpoint_state(dirname)
if ckpt:
model.load_model(dirname)
with tf.Session() as sess:
summary_writer = tf.train.SummaryWriter('logs/'+datetime.now().isoformat().replace(':','-'), sess.graph)
check = tf.add_check_numerics_ops()
merged = tf.merge_all_summaries()
tf.initialize_all_variables().run()
saver = tf.train.Saver(tf.all_variables())
start = time.time()
for e in xrange(args.num_epochs):
sess.run(tf.assign(model.lr, args.learning_rate * (args.decay_rate ** e)))
state = model.initial_state
for b in xrange(100):
#t0 = np.random.randn(args.batch_size,1,(args.chunk_samples))
#x = np.sin(2*np.pi*(np.arange(args.seq_length)[np.newaxis,:,np.newaxis]/30.+t0)) + np.random.randn(args.batch_size,args.seq_length,(args.chunk_samples))*0.1
#y = np.sin(2*np.pi*(np.arange(1,args.seq_length+1)[np.newaxis,:,np.newaxis]/30.+t0)) + np.random.randn(args.batch_size,args.seq_length,(args.chunk_samples))*0.1
if (e * 100 + b)%int(traces.shape[0]/(args.chunk_samples*args.batch_size)) == 0:
data, _, _ = util.load_augment_data(traces,args.chunk_samples)
print "Refreshed data"
#x,y = next_batch(data,args)
slopes = 10*np.random.random((1,1,2*args.chunk_samples))+1
x,y = (slopes*np.arange(args.seq_length)[np.newaxis,:,np.newaxis])-1,(slopes*np.arange(args.seq_length)[np.newaxis,:,np.newaxis])
y[:,:,args.chunk_samples:] = 0.
x[:,:,args.chunk_samples:] = 0.
feed = {model.input_data: x, model.target_data: y}
train_loss, _, cr, summary, sigma = sess.run([model.cost, model.train_op, check, merged, model.sigma], feed)
#train_loss, state, _ = sess.run([model.cost, model.final_state, model.train_op], feed)
summary_writer.add_summary(summary, e * 100 + b)
if (e * 100 + b) % args.save_every == 0 and ((e * 100 + b) > 0):
checkpoint_path = os.path.join('save', 'model.ckpt')
saver.save(sess, checkpoint_path, global_step = e * 100 + b)
print "model saved to {}".format(checkpoint_path)
end = time.time()
print "{}/{} (epoch {}), train_loss = {:.6f}, time/batch = {:.1f}, std = {:.3f}/{:.3f}" \
.format(e * 100 + b,
args.num_epochs * 100,
e, args.chunk_samples*train_loss, end - start, (sigma[:,200:]).mean(axis=0).mean(axis=0),(sigma[:,:200]).mean(axis=0).mean(axis=0))
start = time.time()
def main():
'''
Main function
'''
# Laod the saved arguments
with open(os.path.join('save-vrnn', 'config.pkl')) as f:
saved_args = cPickle.load(f)
# Initialize the model with the saved arguments in inference mode
model = VRNN(saved_args, True)
# Initialize the TensorFlow session
sess = tf.InteractiveSession()
# Initialize the saver
saver = tf.train.Saver(tf.all_variables())
# Get model checkpoint
ckpt = tf.train.get_checkpoint_state('save-vrnn')
print "loading model: ", ckpt.model_checkpoint_path
# Restore the model from the saved file
saver.restore(sess, ckpt.model_checkpoint_path)
# Sample the model
sample_data, mus, sigmas = model.sample(sess, saved_args)
def train(args):
dirname = 'save-vrnn'
if not os.path.exists(dirname):
os.makedirs(dirname)
with open(os.path.join(dirname, 'config.pkl'), 'w') as f:
cPickle.dump(args, f)
model = VRNN(args)
ckpt = tf.train.get_checkpoint_state(dirname)
with tf.Session() as sess:
summary_writer = tf.train.SummaryWriter('logs/' + datetime.now().isoformat().replace(':', '-'), sess.graph)
check = tf.add_check_numerics_ops()
merged = tf.merge_all_summaries()
tf.initialize_all_variables().run()
saver = tf.train.Saver(tf.all_variables())
if ckpt:
saver.restore(sess, ckpt.model_checkpoint_path)
print "Loaded model"
start = time.time()
for e in xrange(args.num_epochs):
sess.run(tf.assign(model.lr, args.learning_rate * (args.decay_rate ** e)))
state = model.initial_state
for b in xrange(100):
x, y = next_batch(args)
feed = {model.input_data: x, model.target_data: y}
train_loss, _, cr, summary, sigma, mu, input, target= sess.run(
[model.cost, model.train_op, check, merged, model.sigma, model.mu, model.flat_input, model.target],
feed)
summary_writer.add_summary(summary, e * 100 + b)
if (e * 100 + b) % args.save_every == 0 and ((e * 100 + b) > 0):
checkpoint_path = os.path.join(dirname, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=e * 100 + b)
print "model saved to {}".format(checkpoint_path)
end = time.time()
print "{}/{} (epoch {}), train_loss = {:.6f}, time/batch = {:.1f}, std = {:.3f}" \
.format(e * 100 + b,
args.num_epochs * 100,
e, args.chunk_samples * train_loss, end - start, sigma.mean(axis=0).mean(axis=0))
start = time.time()
help='RNN sequence length')
parser.add_argument('--num_epochs',
type=int,
default=100,
help='number of epochs')
parser.add_argument('--save_every',
type=int,
default=500,
help='save frequency')
parser.add_argument('--grad_clip',
type=float,
default=10.,
help='clip gradients at this value')
parser.add_argument('--learning_rate',
type=float,
default=0.0005,
help='learning rate')
parser.add_argument('--decay_rate',
type=float,
default=1.,
help='decay of learning rate')
parser.add_argument('--chunk_samples',
type=int,
default=1,
help='number of samples per mdct chunk')
args = parser.parse_args()
model = VRNN(args)
train(args, model)
import tensorflow as tf
import os
import cPickle
from model_vrnn import VRNN
import numpy as np
from train_vrnn import next_batch
with open(os.path.join('save-vrnn', 'config.pkl')) as f:
saved_args = cPickle.load(f)
model = VRNN(saved_args, True)
sess = tf.InteractiveSession()
saver = tf.train.Saver(tf.all_variables())
ckpt = tf.train.get_checkpoint_state('save-vrnn')
print "loading model: ", ckpt.model_checkpoint_path
saver.restore(sess, ckpt.model_checkpoint_path)
sample_data, mus, sigmas = model.sample(sess, saved_args)
import tensorflow as tf
import os
import cPickle
from model_vrnn import VRNN
import numpy as np
from train_vrnn import next_batch
with open(os.path.join('save-vrnn', 'config.pkl')) as f:
saved_args = cPickle.load(f)
model = VRNN(saved_args, True)
sess = tf.InteractiveSession()
saver = tf.train.Saver(tf.all_variables())
ckpt = tf.train.get_checkpoint_state('save-vrnn')
print "loading model: ",ckpt.model_checkpoint_path
saver.restore(sess, ckpt.model_checkpoint_path)
sample_data,mus,sigmas = model.sample(sess,saved_args)
def train(args):
'''
The train function
Params:
args : Input arguments
'''
# Initialize the model
model = VRNN(args)
# Initialize the data loader object
dataloader = DataLoader(args)
pdb.set_trace()
# Directory to save the trained model
dirname = 'save-vrnn'
if not os.path.exists(dirname):
os.makedirs(dirname)
# Dump the input arguments into the file
with open(os.path.join(dirname, 'config.pkl'), 'w') as f:
cPickle.dump(args, f)
# get checkpoint
ckpt = tf.train.get_checkpoint_state(dirname)
# Initialize TensorFlow session
with tf.Session() as sess:
# Initialize summary writer
summary_writer = tf.train.SummaryWriter(
'logs/' + datetime.now().isoformat().replace(':', '-'), sess.graph)
check = tf.add_check_numerics_ops()
# Write all summaries
merged = tf.merge_all_summaries()
# Initialize all variables in the graph
tf.initialize_all_variables().run()
# Initialize a saver for all variables
saver = tf.train.Saver(tf.all_variables())
# Load already saved model
if ckpt:
saver.restore(sess, ckpt.model_checkpoint_path)
print "Loaded model"
# Initialize timer
start = time.time()
# For each epoch
for e in xrange(args.num_epochs):
# Set the learning rate for this epoch
sess.run(
tf.assign(model.lr, args.learning_rate * (args.decay_rate**e)))
# state = model.initial_state_c, model.initial_state_h
# For each minibatch
for b in xrange(100):
# Get the input and target data of the next minibatch
x, y = dataloader.next_batch()
# Create the feed dict
feed = {model.input_data: x, model.target_data: y}
# Run the session and get loss
train_loss, _, cr, summary, sigma, mu, input, target = sess.run(
[
model.cost, model.train_op, check, merged, model.sigma,
model.mu, model.flat_input, model.target
], feed)
# Write summary
summary_writer.add_summary(summary, e * 100 + b)
# Save model
if (e * 100 + b) % args.save_every == 0 and (
(e * 100 + b) > 0):
checkpoint_path = os.path.join(dirname, 'model.ckpt')
saver.save(sess, checkpoint_path, global_step=e * 100 + b)
print "model saved to {}".format(checkpoint_path)
# End timer
end = time.time()
# Print info
print "{}/{} (epoch {}), train_loss = {:.6f}, time/batch = {:.1f}, std = {:.3f}" \
.format(e * 100 + b,
args.num_epochs * 100,
e, args.chunk_samples * train_loss, end - start, sigma.mean(axis=0).mean(axis=0))
start = time.time()