uppate with bugs but can run

Author: Smellly

show-adapt-tell/highway.py

@@ -0,0 +1,19 @@
+import tensorflow as tf
+
+# highway layer that borrowed from https://github.com/carpedm20/lstm-char-cnn-tensorflow
+def highway(input_, size, layer_size=1, bias=-2, f=tf.nn.relu):
+    """Highway Network (cf. http://arxiv.org/abs/1505.00387).
+
+    t = sigmoid(Wy + b)
+    z = t * g(Wy + b) + (1 - t) * y
+    where g is nonlinearity, t is transform gate, and (1 - t) is carry gate.
+    """
+    output = input_
+    for idx in xrange(layer_size):
+        output = f(tf.nn.rnn_cell._linear(output, size, 0, scope='output_lin_%d' % idx))
+        transform_gate = tf.sigmoid(
+            tf.nn.rnn_cell._linear(input_, size, 0, scope='transform_lin_%d' % idx) + bias)
+        carry_gate = 1. - transform_gate
+        output = transform_gate * output + carry_gate * input_
+    return output
+

show-adapt-tell/highway_tf1.4.py

@@ -0,0 +1,31 @@
+import tensorflow as tf
+
+import sys
+sys.path.append('/home/smelly/miniconda2/envs/tensorflow/lib/python2.7/site-packages/tensorflow/python/ops')
+import rnn_cell_impl
+
+# highway layer that borrowed from https://github.com/carpedm20/lstm-char-cnn-tensorflow
+def highway(input_, size, layer_size=1, bias=-2, f=tf.nn.relu):
+    """Highway Network (cf. http://arxiv.org/abs/1505.00387).
+
+    t = sigmoid(Wy + b)
+    z = t * g(Wy + b) + (1 - t) * y
+    where g is nonlinearity, t is transform gate, and (1 - t) is carry gate.
+    """
+    output = input_
+    for idx in xrange(layer_size):
+        # output = f(tf.nn.rnn_cell._linear(output, size, 0, scope='output_lin_%d' % idx))
+        with tf.variable_scope('output_lin_%d' % idx):
+            output = f(
+                    rnn_cell_impl._linear(
+                        output, size, 0))
+            # for tf.1.2.1 above
+            # tf.nn.rnn_cell._linear(input_, size, 0, scope='transform_lin_%d' % idx) + bias)
+        with tf.variable_scope('transform_lin_%d' % idx):
+            transform_gate = tf.sigmoid(
+                rnn_cell_impl._linear(
+                    input_, size, 0) + bias)
+        carry_gate = 1. - transform_gate
+        output = transform_gate * output + carry_gate * input_
+    return output
+

show-adapt-tell/main.py

@@ -38,6 +38,7 @@
 
 FLAGS = flags.FLAGS
 pp = pprint.PrettyPrinter()
+
 def main(_):
     pp.pprint(flags.FLAGS.__flags)
 
@@ -60,7 +61,7 @@ def main(_):
                 print('G_pretrained_model train')
 		G_pretrained_model.train()
             print('G_pretrained_model evaluate')
-	    G_pretrained_model.evaluate('test', 0, )
+	    G_pretrained_model.evaluate('test', 0)
 	if FLAGS.D_is_pretrain:
 	    negative_dataset = mscoco_negative(dataset, FLAGS)
 	    D_pretrained_model = D_pretrained(sess, dataset, negative_dataset, info, conf=FLAGS)

show-adapt-tell/model.py

@@ -1,3 +1,12 @@
+# -*- coding: utf-8 -*-
+'''
+// Author: Jay Smelly.
+// Last modify: 2018-04-20 10:59:19.
+// File name: pretrain_G.py
+//
+// Description: tensorflow 0.12.1
+'''
+
 from __future__ import division
 import os
 import time
@@ -293,9 +302,11 @@ def evaluate(self, split, count, eval_algo='max'):
                     break
                 meteor_pd[str(int(image_id[j]))] = [{'image_id':str(int(image_id[j])), 'caption':samples[j][0]}]
                 meteor_id.append(str(int(image_id[j])))
-            #np.savez("result_%s"%str(count), meteor_pd=meteor_pd, meteor_id=meteor_id)
+            # np.savez("result_%s"%str(count), meteor_pd=meteor_pd, meteor_id=meteor_id)
+            # test_annotation is a json file
             scorer = COCOEvalCap(test_annotation, meteor_pd, meteor_id)
-            scorer.evaluate(verbose=True)		
+            # scorer.evaluate(verbose=True)		
+            scorer.evaluate()		
 
     def save(self, checkpoint_dir, step):
         model_name = "G_pretrained"
@@ -320,3 +331,4 @@ def load(self, checkpoint_dir):
         else:
             return False
 
+