Update to TF1.1 TL1.4.5

tensorlayer · May 27, 2017 · 8ebb42f · 8ebb42f
1 parent 27d83c3
commit 8ebb42f
Show file tree

Hide file tree

Showing 28 changed files with 437 additions and 332 deletions.
diff --git a/main.py b/main.py
@@ -43,7 +43,6 @@ def main(_):
     tl.files.exists_or_mkdir(FLAGS.sample_dir)
 
     z_dim = 100
-
     with tf.device("/gpu:0"):
         ##========================= DEFINE MODEL ===========================##
         z = tf.placeholder(tf.float32, [FLAGS.batch_size, z_dim], name='z_noise')
@@ -94,15 +93,14 @@ def main(_):
     net_d_name = os.path.join(save_dir, 'net_d.npz')
 
     data_files = glob(os.path.join("./data", FLAGS.dataset, "*.jpg"))
-        # sample_seed = np.random.uniform(low=-1, high=1, size=(FLAGS.sample_size, z_dim)).astype(np.float32)
-    sample_seed = np.random.normal(loc=0.0, scale=1.0, size=(FLAGS.sample_size, z_dim)).astype(np.float32)
+
+    sample_seed = np.random.normal(loc=0.0, scale=1.0, size=(FLAGS.sample_size, z_dim)).astype(np.float32)# sample_seed = np.random.uniform(low=-1, high=1, size=(FLAGS.sample_size, z_dim)).astype(np.float32)
 
     ##========================= TRAIN MODELS ================================##
     iter_counter = 0
     for epoch in range(FLAGS.epoch):
         ## shuffle data
         shuffle(data_files)
-        print("[*] Dataset shuffled!")
 
         ## update sample files based on shuffled data
         sample_files = data_files[0:FLAGS.sample_size]
@@ -119,46 +117,28 @@ def main(_):
             # more image augmentation functions in http://tensorlayer.readthedocs.io/en/latest/modules/prepro.html
             batch = [get_image(batch_file, FLAGS.image_size, is_crop=FLAGS.is_crop, resize_w=FLAGS.output_size, is_grayscale = 0) for batch_file in batch_files]
             batch_images = np.array(batch).astype(np.float32)
-                # batch_z = np.random.uniform(low=-1, high=1, size=(FLAGS.batch_size, z_dim)).astype(np.float32)
-            batch_z = np.random.normal(loc=0.0, scale=1.0, size=(FLAGS.sample_size, z_dim)).astype(np.float32)
+            batch_z = np.random.normal(loc=0.0, scale=1.0, size=(FLAGS.sample_size, z_dim)).astype(np.float32)  # batch_z = np.random.uniform(low=-1, high=1, size=(FLAGS.batch_size, z_dim)).astype(np.float32)
             start_time = time.time()
             # updates the discriminator
             errD, _ = sess.run([d_loss, d_optim], feed_dict={z: batch_z, real_images: batch_images })
             # updates the generator, run generator twice to make sure that d_loss does not go to zero (difference from paper)
             for _ in range(2):
                 errG, _ = sess.run([g_loss, g_optim], feed_dict={z: batch_z})
             print("Epoch: [%2d/%2d] [%4d/%4d] time: %4.4f, d_loss: %.8f, g_loss: %.8f" \
-                    % (epoch, FLAGS.epoch, idx, batch_idxs,
-                        time.time() - start_time, errD, errG))
-            sys.stdout.flush()
+                    % (epoch, FLAGS.epoch, idx, batch_idxs, time.time() - start_time, errD, errG))
 
             iter_counter += 1
             if np.mod(iter_counter, FLAGS.sample_step) == 0:
                 # generate and visualize generated images
                 img, errD, errG = sess.run([net_g2.outputs, d_loss, g_loss], feed_dict={z : sample_seed, real_images: sample_images})
-                save_images(img, [8, 8],
-                            './{}/train_{:02d}_{:04d}.png'.format(FLAGS.sample_dir, epoch, idx))
+                tl.visualize.save_images(img, [8, 8], './{}/train_{:02d}_{:04d}.png'.format(FLAGS.sample_dir, epoch, idx))
                 print("[Sample] d_loss: %.8f, g_loss: %.8f" % (errD, errG))
-                sys.stdout.flush()
 
             if np.mod(iter_counter, FLAGS.save_step) == 0:
                 # save current network parameters
                 print("[*] Saving checkpoints...")
-                img, errD, errG = sess.run([net_g2.outputs, d_loss, g_loss], feed_dict={z : sample_seed, real_images: sample_images})
-                model_dir = "%s_%s_%s" % (FLAGS.dataset, FLAGS.batch_size, FLAGS.output_size)
-                save_dir = os.path.join(FLAGS.checkpoint_dir, model_dir)
-                if not os.path.exists(save_dir):
-                    os.makedirs(save_dir)
-                # the latest version location
-                net_g_name = os.path.join(save_dir, 'net_g.npz')
-                net_d_name = os.path.join(save_dir, 'net_d.npz')
-                # # this version is for future re-check and visualization analysis
-                # net_g_iter_name = os.path.join(save_dir, 'net_g_%d.npz' % iter_counter)
-                # net_d_iter_name = os.path.join(save_dir, 'net_d_%d.npz' % iter_counter)
-                # tl.files.save_npz(net_g.all_params, name=net_g_name, sess=sess)
-                # tl.files.save_npz(net_d.all_params, name=net_d_name, sess=sess)
-                # tl.files.save_npz(net_g.all_params, name=net_g_iter_name, sess=sess)
-                # tl.files.save_npz(net_d.all_params, name=net_d_iter_name, sess=sess)
+                tl.files.save_npz(net_g.all_params, name=net_g_name, sess=sess)
+                tl.files.save_npz(net_d.all_params, name=net_d_name, sess=sess)
                 print("[*] Saving checkpoints SUCCESS!")
 
 if __name__ == '__main__':

diff --git a/model.py b/model.py
@@ -12,10 +12,8 @@ def generator_simplified_api(inputs, is_train=True, reuse=False):
     gf_dim = 64 # Dimension of gen filters in first conv layer. [64]
     c_dim = FLAGS.c_dim # n_color 3
     batch_size = FLAGS.batch_size # 64
-
     w_init = tf.random_normal_initializer(stddev=0.02)
     gamma_init = tf.random_normal_initializer(1., 0.02)
-
     with tf.variable_scope("generator", reuse=reuse):
         tl.layers.set_name_reuse(reuse)
 
@@ -47,15 +45,12 @@ def generator_simplified_api(inputs, is_train=True, reuse=False):
         net_h4.outputs = tf.nn.tanh(net_h4.outputs)
     return net_h4, logits
 
-
 def discriminator_simplified_api(inputs, is_train=True, reuse=False):
     df_dim = 64 # Dimension of discrim filters in first conv layer. [64]
     c_dim = FLAGS.c_dim # n_color 3
     batch_size = FLAGS.batch_size # 64
-
     w_init = tf.random_normal_initializer(stddev=0.02)
     gamma_init = tf.random_normal_initializer(1., 0.02)
-
     with tf.variable_scope("discriminator", reuse=reuse):
         tl.layers.set_name_reuse(reuse)
 

diff --git a/tensorlayer/__init__.py b/tensorlayer/__init__.py
@@ -25,7 +25,7 @@
 from . import rein
 
 
-__version__ = "1.4.2"
+__version__ = "1.4.5"
 
 global_flag = {}
 global_dict = {}
diff --git a/tensorlayer/__pycache__/__init__.cpython-35.pyc b/tensorlayer/__pycache__/__init__.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/activation.cpython-35.pyc b/tensorlayer/__pycache__/activation.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/cost.cpython-35.pyc b/tensorlayer/__pycache__/cost.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/files.cpython-35.pyc b/tensorlayer/__pycache__/files.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/iterate.cpython-35.pyc b/tensorlayer/__pycache__/iterate.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/layers.cpython-35.pyc b/tensorlayer/__pycache__/layers.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/nlp.cpython-35.pyc b/tensorlayer/__pycache__/nlp.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/ops.cpython-35.pyc b/tensorlayer/__pycache__/ops.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/prepro.cpython-35.pyc b/tensorlayer/__pycache__/prepro.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/rein.cpython-35.pyc b/tensorlayer/__pycache__/rein.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/utils.cpython-35.pyc b/tensorlayer/__pycache__/utils.cpython-35.pyc
diff --git a/tensorlayer/__pycache__/visualize.cpython-35.pyc b/tensorlayer/__pycache__/visualize.cpython-35.pyc
diff --git a/tensorlayer/activation.py b/tensorlayer/activation.py
diff --git a/tensorlayer/cost.py b/tensorlayer/cost.py
@@ -32,11 +32,11 @@ def cross_entropy(output, target, name=None):
     - About cross-entropy: `wiki <https://en.wikipedia.org/wiki/Cross_entropy>`_.\n
     - The code is borrowed from: `here <https://en.wikipedia.org/wiki/Cross_entropy>`_.
     """
-    # try: # old
-    #     return tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits=output, targets=target))
-    # except: # TF 1.0
-    #     assert name is not None, "Please give a unique name to tl.cost.cross_entropy for TF1.0+"
-    return tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=target, logits=output, name=name))
+    try: # old
+        return tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(logits=output, targets=target))
+    except: # TF 1.0
+        assert name is not None, "Please give a unique name to tl.cost.cross_entropy for TF1.0+"
+        return tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=target, logits=output, name=name))
 
 def sigmoid_cross_entropy(output, target, name=None):
     """It is a sigmoid cross-entropy operation, see ``tf.nn.sigmoid_cross_entropy_with_logits``.
@@ -72,8 +72,8 @@ def binary_cross_entropy(output, target, epsilon=1e-8, name='bce_loss'):
 #         output = ops.convert_to_tensor(output, name="preds")
 #         target = ops.convert_to_tensor(targets, name="target")
     with tf.name_scope(name):
-        return tf.reduce_mean(-(target * tf.log(output + epsilon) +
-                              (1. - target) * tf.log(1. - output + epsilon)))
+        return tf.reduce_mean(tf.reduce_sum(-(target * tf.log(output + epsilon) +
+                              (1. - target) * tf.log(1. - output + epsilon)), axis=1))
 
 
 def mean_squared_error(output, target, is_mean=False):
@@ -223,7 +223,7 @@ def cross_entropy_seq(logits, target_seqs, batch_size=None):#, batch_size=1, num
     >>> see PTB tutorial for more details
     >>> input_data = tf.placeholder(tf.int32, [batch_size, num_steps])
     >>> targets = tf.placeholder(tf.int32, [batch_size, num_steps])
-    >>> cost = tf.cost.cross_entropy_seq(network.outputs, targets)
+    >>> cost = tl.cost.cross_entropy_seq(network.outputs, targets)
     """
     try: # TF 1.0
         sequence_loss_by_example_fn = tf.contrib.legacy_seq2seq.sequence_loss_by_example