Skip to content

GitLab

Commit f305d2df authored by Dougal J. Sutherland's avatar Dougal J. Sutherland Committed by Sergio Guadarrama
Browse files

change no-longer-existing concat_v2 to concat (#1701)

parent 057203e7
Hide whitespace changes
Inline Side-by-side
Showing with 25 additions and 25 deletions
cognitive_mapping_and_planning/tfcode/vision_baseline_lstm.py
View file @ f305d2df
...@@ -145,7 +145,7 @@ def visit_count_fc(visit_count, last_visit, embed_neurons, wt_decay, fc_dropout) ...@@ -145,7 +145,7 @@ def visit_count_fc(visit_count, last_visit, embed_neurons, wt_decay, fc_dropout)
on_value=10., off_value=0.) on_value=10., off_value=0.)
last_visit = tf.one_hot(last_visit, depth=16, axis=1, dtype=tf.float32, last_visit = tf.one_hot(last_visit, depth=16, axis=1, dtype=tf.float32,
on_value=10., off_value=0.) on_value=10., off_value=0.)
f = tf.concat_v2([visit_count, last_visit], 1) f = tf.concat([visit_count, last_visit], 1)
x, _ = tf_utils.fc_network( x, _ = tf_utils.fc_network(
f, neurons=embed_neurons, wt_decay=wt_decay, name='visit_count_embed', f, neurons=embed_neurons, wt_decay=wt_decay, name='visit_count_embed',
offset=0, batch_norm_param=None, dropout_ratio=fc_dropout, offset=0, batch_norm_param=None, dropout_ratio=fc_dropout,
...@@ -201,7 +201,7 @@ def combine_setup(name, combine_type, embed_img, embed_goal, num_img_neuorons=No ...@@ -201,7 +201,7 @@ def combine_setup(name, combine_type, embed_img, embed_goal, num_img_neuorons=No
def preprocess_egomotion(locs, thetas): def preprocess_egomotion(locs, thetas):
with tf.name_scope('pre_ego'): with tf.name_scope('pre_ego'):
pre_ego = tf.concat_v2([locs, tf.sin(thetas), tf.cos(thetas)], 2) pre_ego = tf.concat([locs, tf.sin(thetas), tf.cos(thetas)], 2)
sh = pre_ego.get_shape().as_list() sh = pre_ego.get_shape().as_list()
pre_ego = tf.reshape(pre_ego, [-1, sh[-1]]) pre_ego = tf.reshape(pre_ego, [-1, sh[-1]])
return pre_ego return pre_ego
......
real_nvp/real_nvp_multiscale_dataset.py
View file @ f305d2df
...@@ -321,8 +321,8 @@ def masked_conv_aff_coupling(input_, mask_in, dim, name, ...@@ -321,8 +321,8 @@ def masked_conv_aff_coupling(input_, mask_in, dim, name,
input_=res, dim=channels, name="bn_in", scale=False, input_=res, dim=channels, name="bn_in", scale=False,
train=train, epsilon=1e-4, axes=[0, 1, 2]) train=train, epsilon=1e-4, axes=[0, 1, 2])
res *= 2. res *= 2.
res = tf.concat_v2([res, -res], 3) res = tf.concat([res, -res], 3)
res = tf.concat_v2([res, mask], 3) res = tf.concat([res, mask], 3)
dim_in = 2. * channels + 1 dim_in = 2. * channels + 1
res = tf.nn.relu(res) res = tf.nn.relu(res)
res = resnet(input_=res, dim_in=dim_in, dim=dim, res = resnet(input_=res, dim_in=dim_in, dim=dim,
...@@ -411,8 +411,8 @@ def masked_conv_add_coupling(input_, mask_in, dim, name, ...@@ -411,8 +411,8 @@ def masked_conv_add_coupling(input_, mask_in, dim, name,
input_=res, dim=channels, name="bn_in", scale=False, input_=res, dim=channels, name="bn_in", scale=False,
train=train, epsilon=1e-4, axes=[0, 1, 2]) train=train, epsilon=1e-4, axes=[0, 1, 2])
res *= 2. res *= 2.
res = tf.concat_v2([res, -res], 3) res = tf.concat([res, -res], 3)
res = tf.concat_v2([res, mask], 3) res = tf.concat([res, mask], 3)
dim_in = 2. * channels + 1 dim_in = 2. * channels + 1
res = tf.nn.relu(res) res = tf.nn.relu(res)
shift = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=channels, shift = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=channels,
...@@ -501,7 +501,7 @@ def conv_ch_aff_coupling(input_, dim, name, ...@@ -501,7 +501,7 @@ def conv_ch_aff_coupling(input_, dim, name,
res = batch_norm( res = batch_norm(
input_=res, dim=channels, name="bn_in", scale=False, input_=res, dim=channels, name="bn_in", scale=False,
train=train, epsilon=1e-4, axes=[0, 1, 2]) train=train, epsilon=1e-4, axes=[0, 1, 2])
res = tf.concat_v2([res, -res], 3) res = tf.concat([res, -res], 3)
dim_in = 2. * channels dim_in = 2. * channels
res = tf.nn.relu(res) res = tf.nn.relu(res)
res = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=2 * channels, res = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=2 * channels,
...@@ -551,11 +551,11 @@ def conv_ch_aff_coupling(input_, dim, name, ...@@ -551,11 +551,11 @@ def conv_ch_aff_coupling(input_, dim, name,
res *= tf.exp(-.5 * log_var) res *= tf.exp(-.5 * log_var)
log_diff -= .5 * log_var log_diff -= .5 * log_var
if change_bottom: if change_bottom:
res = tf.concat_v2([input_, res], 3) res = tf.concat([input_, res], 3)
log_diff = tf.concat_v2([tf.zeros_like(log_diff), log_diff], 3) log_diff = tf.concat([tf.zeros_like(log_diff), log_diff], 3)
else: else:
res = tf.concat_v2([res, input_], 3) res = tf.concat([res, input_], 3)
log_diff = tf.concat_v2([log_diff, tf.zeros_like(log_diff)], 3) log_diff = tf.concat([log_diff, tf.zeros_like(log_diff)], 3)
return res, log_diff return res, log_diff
...@@ -582,7 +582,7 @@ def conv_ch_add_coupling(input_, dim, name, ...@@ -582,7 +582,7 @@ def conv_ch_add_coupling(input_, dim, name,
res = batch_norm( res = batch_norm(
input_=res, dim=channels, name="bn_in", scale=False, input_=res, dim=channels, name="bn_in", scale=False,
train=train, epsilon=1e-4, axes=[0, 1, 2]) train=train, epsilon=1e-4, axes=[0, 1, 2])
res = tf.concat_v2([res, -res], 3) res = tf.concat([res, -res], 3)
dim_in = 2. * channels dim_in = 2. * channels
res = tf.nn.relu(res) res = tf.nn.relu(res)
shift = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=channels, shift = resnet(input_=res, dim_in=dim_in, dim=dim, dim_out=channels,
...@@ -616,11 +616,11 @@ def conv_ch_add_coupling(input_, dim, name, ...@@ -616,11 +616,11 @@ def conv_ch_add_coupling(input_, dim, name,
res *= tf.exp(-.5 * log_var) res *= tf.exp(-.5 * log_var)
log_diff -= .5 * log_var log_diff -= .5 * log_var
if change_bottom: if change_bottom:
res = tf.concat_v2([input_, res], 3) res = tf.concat([input_, res], 3)
log_diff = tf.concat_v2([tf.zeros_like(log_diff), log_diff], 3) log_diff = tf.concat([tf.zeros_like(log_diff), log_diff], 3)
else: else:
res = tf.concat_v2([res, input_], 3) res = tf.concat([res, input_], 3)
log_diff = tf.concat_v2([log_diff, tf.zeros_like(log_diff)], 3) log_diff = tf.concat([log_diff, tf.zeros_like(log_diff)], 3)
return res, log_diff return res, log_diff
...@@ -742,9 +742,9 @@ def rec_masked_conv_coupling(input_, hps, scale_idx, n_scale, ...@@ -742,9 +742,9 @@ def rec_masked_conv_coupling(input_, hps, scale_idx, n_scale,
input_=res_1, hps=hps, scale_idx=scale_idx + 1, n_scale=n_scale, input_=res_1, hps=hps, scale_idx=scale_idx + 1, n_scale=n_scale,
use_batch_norm=use_batch_norm, weight_norm=weight_norm, use_batch_norm=use_batch_norm, weight_norm=weight_norm,
train=train) train=train)
res = tf.concat_v2([res_1, res_2], 3) res = tf.concat([res_1, res_2], 3)
log_diff_1 += inc_log_diff log_diff_1 += inc_log_diff
log_diff = tf.concat_v2([log_diff_1, log_diff_2], 3) log_diff = tf.concat([log_diff_1, log_diff_2], 3)
res = squeeze_2x2_ordered(res, reverse=True) res = squeeze_2x2_ordered(res, reverse=True)
log_diff = squeeze_2x2_ordered(log_diff, reverse=True) log_diff = squeeze_2x2_ordered(log_diff, reverse=True)
else: else:
...@@ -805,8 +805,8 @@ def rec_masked_deconv_coupling(input_, hps, scale_idx, n_scale, ...@@ -805,8 +805,8 @@ def rec_masked_deconv_coupling(input_, hps, scale_idx, n_scale,
scale_idx=scale_idx + 1, n_scale=n_scale, scale_idx=scale_idx + 1, n_scale=n_scale,
use_batch_norm=use_batch_norm, weight_norm=weight_norm, use_batch_norm=use_batch_norm, weight_norm=weight_norm,
train=train) train=train)
res = tf.concat_v2([res_1, res_2], 3) res = tf.concat([res_1, res_2], 3)
log_diff = tf.concat_v2([log_diff_1, log_diff_2], 3) log_diff = tf.concat([log_diff_1, log_diff_2], 3)
res = squeeze_2x2_ordered(res, reverse=True) res = squeeze_2x2_ordered(res, reverse=True)
log_diff = squeeze_2x2_ordered(log_diff, reverse=True) log_diff = squeeze_2x2_ordered(log_diff, reverse=True)
else: else:
...@@ -1018,7 +1018,7 @@ class RealNVP(object): ...@@ -1018,7 +1018,7 @@ class RealNVP(object):
width = tf.cast(width, tf.int32) width = tf.cast(width, tf.int32)
depth = tf.reshape((features["depth"], tf.int64)[0], [1]) depth = tf.reshape((features["depth"], tf.int64)[0], [1])
depth = tf.cast(depth, tf.int32) depth = tf.cast(depth, tf.int32)
image = tf.reshape(image, tf.concat_v2([height, width, depth], 0)) image = tf.reshape(image, tf.concat([height, width, depth], 0))
image = tf.random_crop(image, [64, 64, 3]) image = tf.random_crop(image, [64, 64, 3])
if FLAGS.mode == "train": if FLAGS.mode == "train":
image = tf.image.random_flip_left_right(image) image = tf.image.random_flip_left_right(image)
...@@ -1309,19 +1309,19 @@ class RealNVP(object): ...@@ -1309,19 +1309,19 @@ class RealNVP(object):
z_compressed = z_lost z_compressed = z_lost
z_noisy = z_lost z_noisy = z_lost
for _ in xrange(scale_idx + 1): for _ in xrange(scale_idx + 1):
z_compressed = tf.concat_v2( z_compressed = tf.concat(
[z_compressed, tf.zeros_like(z_compressed)], 3) [z_compressed, tf.zeros_like(z_compressed)], 3)
z_compressed = squeeze_2x2_ordered( z_compressed = squeeze_2x2_ordered(
z_compressed, reverse=True) z_compressed, reverse=True)
z_noisy = tf.concat_v2( z_noisy = tf.concat(
[z_noisy, tf.random_normal( [z_noisy, tf.random_normal(
z_noisy.get_shape().as_list())], 3) z_noisy.get_shape().as_list())], 3)
z_noisy = squeeze_2x2_ordered(z_noisy, reverse=True) z_noisy = squeeze_2x2_ordered(z_noisy, reverse=True)
z_compressed_list.append(z_compressed) z_compressed_list.append(z_compressed)
z_noisy_list.append(z_noisy) z_noisy_list.append(z_noisy)
self.z_reduced = z_lost self.z_reduced = z_lost
z_compressed = tf.concat_v2(z_compressed_list, 0) z_compressed = tf.concat(z_compressed_list, 0)
z_noisy = tf.concat_v2(z_noisy_list, 0) z_noisy = tf.concat(z_noisy_list, 0)
noisy_images, _ = decoder( noisy_images, _ = decoder(
input_=z_noisy, hps=hps, n_scale=hps.n_scale, input_=z_noisy, hps=hps, n_scale=hps.n_scale,
use_batch_norm=hps.use_batch_norm, weight_norm=True, use_batch_norm=hps.use_batch_norm, weight_norm=True,
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment