ValueError: Trying to share variable rnn/attention_cell_wrapper/multi_rnn_cell/cell_0/gru_cell/gates/kernel, but specified shape (200, 200) and found shape (2428, 200).

[zhivko]

Training the RNN¶

with tf.Graph().as_default():

model = RNNModel()

input_ = train[0]

target = train[1]

with tf.Session() as sess:

    init = tf.initialize_all_variables()

    sess.run([init])

    loss = 2000

    

    for e in range(NUM_EPOCHS):

        state = sess.run(model.zero_state)

        epoch_loss =0

        for batch in range(0,NUM_TRAIN_BATCHES):

            start = batch*BATCH_SIZE

            end = start + BATCH_SIZE 

            feed = {

                model.input_data:input_[start:end],

                model.target_data:target[start:end],

                model.dropout_prob:0.5,

                model.start_state:state

                        }

            _,loss,acc,state = sess.run(

                [

                    model.train_op,

                    model.loss,

                    model.accuracy,

                    model.end_state

                ]

                ,feed_dict=feed

            )

            epoch_loss+=loss

            

        print('step - {0} loss - {1} acc - {2}'.format((e),epoch_loss,acc))

    final_preds =np.array([])

    for batch in range(0,NUM_VAL_BATCHES):

            start = batch*BATCH_SIZE

            end = start + BATCH_SIZE 

            feed = {

                model.input_data:val[0][start:end],

                model.target_data:val[1][start:end],

                model.dropout_prob:1,

                model.start_state:state

                        }

            acc,preds,state = sess.run(

                [

                    model.accuracy,

                    model.predictions,

                    model.end_state

                ]

                ,feed_dict=feed

            )

            print(acc)

            assert len(preds) == BATCH_SIZE

            final_preds = np.concatenate((final_preds,preds),axis=0)

WARNING:tensorflow:<tensorflow.contrib.rnn.python.ops.rnn_cell.AttentionCellWrapper object at 0x0000021E0CFB02B0>: Using a concatenated state is slower and will soon be deprecated. Use state_is_tuple=True.

---

ValueError Traceback (most recent call last) in () 1 with tf.Graph().as_default(): ----> 2 model = RNNModel() 3 input_ = train[0] 4 target = train[1] 5 with tf.Session() as sess:

in init(self) 44 scope.reuse_variables() 45 ---> 46 output, state = self.gru_cell(inp, state) 47 states.append(state) 48 outputs.append(output)

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in call(self, inputs, state, scope) 178 with vs.variable_scope(vs.get_variable_scope(), 179 custom_getter=self._rnn_get_variable): --> 180 return super(RNNCell, self).call(inputs, state) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs):

c:\python35\lib\site-packages\tensorflow\python\layers\base.py in call(self, inputs, *args, **kwargs) 439 # Check input assumptions set after layer building, e.g. input shape. 440 self._assert_input_compatibility(inputs) --> 441 outputs = self.call(inputs, *args, **kwargs) 442 443 # Apply activity regularization.

c:\python35\lib\site-packages\tensorflow\contrib\rnn\python\ops\rnn_cell.py in call(self, inputs, state) 1111 input_size = inputs.get_shape().as_list()[1] 1112 inputs = _linear([inputs, attns], input_size, True) -> 1113 lstm_output, new_state = self._cell(inputs, state) 1114 if self._state_is_tuple: 1115 new_state_cat = array_ops.concat(nest.flatten(new_state), 1)

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in call(self, inputs, state, scope) 178 with vs.variable_scope(vs.get_variable_scope(), 179 custom_getter=self._rnn_get_variable): --> 180 return super(RNNCell, self).call(inputs, state) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs):

c:\python35\lib\site-packages\tensorflow\python\layers\base.py in call(self, inputs, *args, **kwargs) 439 # Check input assumptions set after layer building, e.g. input shape. 440 self._assert_input_compatibility(inputs) --> 441 outputs = self.call(inputs, *args, **kwargs) 442 443 # Apply activity regularization.

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in call(self, inputs, state) 914 [-1, cell.state_size]) 915 cur_state_pos += cell.state_size --> 916 cur_inp, new_state = cell(cur_inp, cur_state) 917 new_states.append(new_state) 918

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in call(self, inputs, state, scope) 178 with vs.variable_scope(vs.get_variable_scope(), 179 custom_getter=self._rnn_get_variable): --> 180 return super(RNNCell, self).call(inputs, state) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs):

c:\python35\lib\site-packages\tensorflow\python\layers\base.py in call(self, inputs, *args, **kwargs) 439 # Check input assumptions set after layer building, e.g. input shape. 440 self._assert_input_compatibility(inputs) --> 441 outputs = self.call(inputs, *args, **kwargs) 442 443 # Apply activity regularization.

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in call(self, inputs, state) 293 value = math_ops.sigmoid( 294 _linear([inputs, state], 2 * self._num_units, True, bias_ones, --> 295 self._kernel_initializer)) 296 r, u = array_ops.split(value=value, num_or_size_splits=2, axis=1) 297 with vs.variable_scope("candidate"):

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in _linear(args, output_size, bias, bias_initializer, kernel_initializer) 1015 _WEIGHTS_VARIABLE_NAME, [total_arg_size, output_size], 1016 dtype=dtype, -> 1017 initializer=kernel_initializer) 1018 if len(args) == 1: 1019 res = math_ops.matmul(args[0], weights)

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in get_variable(name, shape, dtype, initializer, regularizer, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter) 1063 collections=collections, caching_device=caching_device, 1064 partitioner=partitioner, validate_shape=validate_shape, -> 1065 use_resource=use_resource, custom_getter=custom_getter) 1066 get_variable_or_local_docstring = ( 1067 """%s

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in get_variable(self, var_store, name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter) 960 collections=collections, caching_device=caching_device, 961 partitioner=partitioner, validate_shape=validate_shape, --> 962 use_resource=use_resource, custom_getter=custom_getter) 963 964 def _get_partitioned_variable(self,

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in get_variable(self, name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource, custom_getter) 358 reuse=reuse, trainable=trainable, collections=collections, 359 caching_device=caching_device, partitioner=partitioner, --> 360 validate_shape=validate_shape, use_resource=use_resource) 361 else: 362 return _true_getter(

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in wrapped_custom_getter(getter, *args, **kwargs) 1403 return custom_getter( 1404 functools.partial(old_getter, getter), -> 1405 *args, **kwargs) 1406 return wrapped_custom_getter 1407

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in _rnn_get_variable(self, getter, *args, **kwargs) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs): --> 183 variable = getter(*args, **kwargs) 184 trainable = (variable in tf_variables.trainable_variables() or 185 (isinstance(variable, tf_variables.PartitionedVariable) and

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in wrapped_custom_getter(getter, *args, **kwargs) 1403 return custom_getter( 1404 functools.partial(old_getter, getter), -> 1405 *args, **kwargs) 1406 return wrapped_custom_getter 1407

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in _rnn_get_variable(self, getter, *args, **kwargs) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs): --> 183 variable = getter(*args, **kwargs) 184 trainable = (variable in tf_variables.trainable_variables() or 185 (isinstance(variable, tf_variables.PartitionedVariable) and

c:\python35\lib\site-packages\tensorflow\python\ops\rnn_cell_impl.py in _rnn_get_variable(self, getter, *args, **kwargs) 181 182 def _rnn_get_variable(self, getter, *args, **kwargs): --> 183 variable = getter(*args, **kwargs) 184 trainable = (variable in tf_variables.trainable_variables() or 185 (isinstance(variable, tf_variables.PartitionedVariable) and

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in _true_getter(name, shape, dtype, initializer, regularizer, reuse, trainable, collections, caching_device, partitioner, validate_shape, use_resource) 350 trainable=trainable, collections=collections, 351 caching_device=caching_device, validate_shape=validate_shape, --> 352 use_resource=use_resource) 353 354 if custom_getter is not None:

c:\python35\lib\site-packages\tensorflow\python\ops\variable_scope.py in _get_single_variable(self, name, shape, dtype, initializer, regularizer, partition_info, reuse, trainable, collections, caching_device, validate_shape, use_resource) 667 raise ValueError("Trying to share variable %s, but specified shape %s" 668 " and found shape %s." % (name, shape, --> 669 found_var.get_shape())) 670 if not dtype.is_compatible_with(found_var.dtype): 671 dtype_str = dtype.name

ValueError: Trying to share variable rnn/attention_cell_wrapper/multi_rnn_cell/cell_0/gru_cell/gates/kernel, but specified shape (200, 200) and found shape (2428, 200).

[zhivko]

I am using tensorflow 1.2. It seems this error was not there if using tensorflow 1.0