def plot_Salah_DAE_samples(samples, image_output_file):
import yann_dauphin_utils
import PIL
#image_output_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/digits.png"
#samples_pkl_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/samples.pkl"
#samples = cPickle.load(open(samples_pkl_file,"r"))
assert len(samples.shape) == 2
N = samples.shape[0]
n_inputs = samples.shape[1]
tile_j = int(np.ceil(np.sqrt(N)))
tile_i = int(np.ceil(float(N) / tile_j))
img_j = int(np.ceil(np.sqrt(n_inputs)))
img_i = int(np.ceil(float(n_inputs) / img_j))
from PIL import Image
image = Image.fromarray(
yann_dauphin_utils.tile_raster_images(X=samples,
img_shape=(img_j, img_i),
tile_shape=(tile_j, tile_i),
tile_spacing=(1, 1)))
image.save(image_output_file)
def plot_Salah_DAE_samples(samples, image_output_file):
import yann_dauphin_utils
import PIL
# image_output_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/digits.png"
# samples_pkl_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/samples.pkl"
# samples = cPickle.load(open(samples_pkl_file,"r"))
assert len(samples.shape) == 2
N = samples.shape[0]
n_inputs = samples.shape[1]
tile_j = int(np.ceil(np.sqrt(N)))
tile_i = int(np.ceil(float(N) / tile_j))
img_j = int(np.ceil(np.sqrt(n_inputs)))
img_i = int(np.ceil(float(n_inputs) / img_j))
from PIL import Image
image = Image.fromarray(
yann_dauphin_utils.tile_raster_images(
X=samples, img_shape=(img_j, img_i), tile_shape=(tile_j, tile_i), tile_spacing=(1, 1)
)
)
image.save(image_output_file)
import numpy as np
import cPickle
import yann_dauphin_utils
import PIL
image_output_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/digits.png"
samples_pkl_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/samples.pkl"
samples = cPickle.load(open(samples_pkl_file, "r"))
assert len(samples.shape) == 2
N = samples.shape[0]
n_inputs = samples.shape[1]
tile_j = int(np.ceil(np.sqrt(N)))
tile_i = int(np.ceil(float(N) / tile_j))
img_j = int(np.ceil(np.sqrt(n_inputs)))
img_i = int(np.ceil(float(n_inputs) / img_j))
from PIL import Image
image = Image.fromarray(
yann_dauphin_utils.tile_raster_images(X=samples,
img_shape=(img_j, img_i),
tile_shape=(tile_j, tile_i),
tile_spacing=(1, 1)))
image.save(image_output_file)
import numpy as np
import cPickle
import yann_dauphin_utils
import PIL
image_output_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/digits.png"
samples_pkl_file = "/u/alaingui/Documents/tmp/Salah_DAE_2013_02_06/metropolis_hastings_langevin_grad_E/1360188911/samples.pkl"
samples = cPickle.load(open(samples_pkl_file,"r"))
assert len(samples.shape) == 2
N = samples.shape[0]
n_inputs = samples.shape[1]
tile_j = int(np.ceil(np.sqrt(N)))
tile_i = int(np.ceil(float(N) / tile_j))
img_j = int(np.ceil(np.sqrt(n_inputs)))
img_i = int(np.ceil(float(n_inputs) / img_j))
from PIL import Image
image = Image.fromarray(yann_dauphin_utils.tile_raster_images(
X = samples,
img_shape = (img_j,img_i), tile_shape = (tile_j, tile_i),
tile_spacing=(1,1)))
image.save(image_output_file)