import tensorflow as tf
from forkan.models import VAE
from forkan.datasets import load_atari_normalized
learning_rate = 1e-4
beta = 5.5
latents = 20
for name in ['pong', 'breakout', 'boxing', 'gopher', 'upndown']:
data = load_atari_normalized(name)
v = VAE(data.shape[1:], name=name, lr=learning_rate, beta=beta, latent_dim=latents)
v.train(data, num_episodes=50, print_freq=-1)
tf.reset_default_graph()
del data
del v
import logging
import numpy as np
import scipy
from forkan import dataset_path
from forkan.common.utils import create_dir
from forkan.datasets import load_atari_normalized
logger = logging.getLogger(__name__)
logger.info('loading dataset ...')
data = load_atari_normalized('breakout-small')
logger.info('done loading')
np.random.seed(0)
idxs = [5, 6, 7, 305711, 244444]
rand_frames = data[idxs]
print('dumping file')
np.savez_compressed('{}/breakout-eval.npz'.format(dataset_path),
data=rand_frames)
print('storing some pngs')
create_dir('{}/breakout-eval/'.format(dataset_path))
for n, f in enumerate(rand_frames[:, ...]):
scipy.misc.imsave('{}/breakout-eval/frame{}.png'.format(dataset_path, n),
np.squeeze(f))
print('done')
import tensorflow as tf
from forkan.datasets import load_atari_normalized
from forkan.models import VAE
betas = [1.28, 2.0, 3.5, 5.0]
for beta in betas:
data = load_atari_normalized('breakout')
v = VAE(data.shape[1:],
network='atari',
name='breakout',
beta=beta,
lr=1e-4,
latent_dim=20)
v.train(data, num_episodes=100, print_freq=200, batch_size=128)
v.s.close()
tf.reset_default_graph()
del data
del v