support with python3.6 and tensorflow1.10 which depends on protobuf3.6

.gitignore

@@ -8,3 +8,7 @@
 # Python cache
 *.pyc
 
+.*
+
+MNIST_data/
+bvlc_alexnet

convert.py

@@ -34,7 +34,7 @@ def convert(def_path, caffemodel_path, data_output_path, code_output_path, phase
         if code_output_path:
             print_stderr('Saving source...')
             with open(code_output_path, 'wb') as src_out:
-                src_out.write(transformer.transform_source())
+                src_out.write(transformer.transform_source().encode('utf-8'))
         print_stderr('Done.')
     except KaffeError as err:
         fatal_error('Error encountered: {}'.format(err))

examples/mnist/finetune_mnist.py

@@ -9,15 +9,17 @@
 mnist = input_data.read_data_sets('MNIST_data', one_hot=True)
 batch_size = 32
 
+
 def gen_data(source):
     while True:
-        indices = range(len(source.images))
+        indices = list(range(len(source.images)))
         random.shuffle(indices)
         for i in indices:
             image = np.reshape(source.images[i], (28, 28, 1))
             label = source.labels[i]
             yield image, label
 
+
 def gen_data_batch(source):
     data_gen = gen_data(source)
     while True:
@@ -37,13 +39,13 @@ def gen_data_batch(source):
 ip2 = net.layers['ip2']
 pred = tf.nn.softmax(ip2)
 
-loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(ip2, labels), 0)
+loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=labels, logits=ip2), 0)
 opt = tf.train.RMSPropOptimizer(0.001)
 train_op = opt.minimize(loss)
 
 with tf.Session() as sess:
     # Load the data
-    sess.run(tf.initialize_all_variables())
+    sess.run(tf.global_variables_initializer())
     net.load('mynet.npy', sess)
 
     data_gen = gen_data_batch(mnist.train)

kaffe/caffe/resolver.py

@@ -14,8 +14,8 @@ def import_caffe(self):
             self.caffe = caffe
         except ImportError:
             # Fall back to the protobuf implementation
-            from . import caffepb
-            self.caffepb = caffepb
+            from . import caffe_pb2
+            self.caffepb = caffe_pb2
             show_fallback_warning()
         if self.caffe:
             # Use the protobuf code from the imported distribution.

kaffe/graph.py

@@ -121,8 +121,8 @@ def __str__(self):
             # In case of convolutions, this corresponds to the weights.
             data_shape = node.data[0].shape if node.data else '--'
             out_shape = node.output_shape or '--'
-            s.append('{:<20} {:<30} {:>20} {:>20}'.format(node.kind, node.name, data_shape,
-                                                          tuple(out_shape)))
+            s.append('{:<20} {:<30} {:>20} {:>20}'.format(node.kind, node.name, str(data_shape),
+                                                          str(out_shape)))
         return '\n'.join(s)
 
 

kaffe/tensorflow/network.py

@@ -30,7 +30,6 @@ def layer_decorated(self, *args, **kwargs):
 
 
 class Network(object):
-
     def __init__(self, inputs, trainable=True):
         # The input nodes for this network
         self.inputs = inputs
@@ -59,7 +58,7 @@ def load(self, data_path, session, ignore_missing=False):
         data_dict = np.load(data_path).item()
         for op_name in data_dict:
             with tf.variable_scope(op_name, reuse=True):
-                for param_name, data in data_dict[op_name].iteritems():
+                for param_name, data in data_dict[op_name].items():
                     try:
                         var = tf.get_variable(param_name)
                         session.run(var.assign(data))
@@ -74,7 +73,7 @@ def feed(self, *args):
         assert len(args) != 0
         self.terminals = []
         for fed_layer in args:
-            if isinstance(fed_layer, basestring):
+            if isinstance(fed_layer, str):
                 try:
                     fed_layer = self.layers[fed_layer]
                 except KeyError:
@@ -124,7 +123,7 @@ def conv(self,
         # Convolution for a given input and kernel
         convolve = lambda i, k: tf.nn.conv2d(i, k, [1, s_h, s_w, 1], padding=padding)
         with tf.variable_scope(name) as scope:
-            kernel = self.make_var('weights', shape=[k_h, k_w, c_i / group, c_o])
+            kernel = self.make_var('weights', shape=[k_h, k_w, c_i.value / group, c_o])
             if group == 1:
                 # This is the common-case. Convolve the input without any further complications.
                 output = convolve(input, kernel)
@@ -177,7 +176,7 @@ def lrn(self, input, radius, alpha, beta, name, bias=1.0):
 
     @layer
     def concat(self, inputs, axis, name):
-        return tf.concat(concat_dim=axis, values=inputs, name=name)
+        return tf.concat(inputs, axis=axis, name=name)
 
     @layer
     def add(self, inputs, name):
@@ -203,7 +202,7 @@ def fc(self, input, num_out, name, relu=True):
 
     @layer
     def softmax(self, input, name):
-        input_shape = map(lambda v: v.value, input.get_shape())
+        input_shape = [v.value for v in input.get_shape()]
         if len(input_shape) > 2:
             # For certain models (like NiN), the singleton spatial dimensions
             # need to be explicitly squeezed, since they're not broadcast-able

kaffe/tensorflow/transformer.py

@@ -44,7 +44,7 @@ def __init__(self, op, *args, **kwargs):
 
     def format(self, arg):
         '''Returns a string representation for the given value.'''
-        return "'%s'" % arg if isinstance(arg, basestring) else str(arg)
+        return "'%s'" % arg if isinstance(arg, str) else str(arg)
 
     def pair(self, key, value):
         '''Returns key=formatted(value).'''
@@ -53,7 +53,7 @@ def pair(self, key, value):
     def emit(self):
         '''Emits the Python source for this node.'''
         # Format positional arguments
-        args = map(self.format, self.args)
+        args = [self.format(arg) for arg in self.args]
         # Format any keyword arguments
         if self.kwargs:
             args += [self.pair(k, v) for k, v in self.kwargs]