无法创建cudnn句柄: CUDNN_STATUS_ALLOC_FAILED

Question

下面是我正在使用的模型：

#import tensorflow as tf
def create_model():
    return tf.keras.models.Sequential([
    #tf.keras.layers.Flatten(input_shape=(2,)),
    tf.keras.layers.Conv2D(filters=32,kernel_size=(3,3),strides=(1,1),input_shape=(156,256,3),padding='valid',data_format='channels_last',
              activation='relu',kernel_initializer=tf.keras.initializers.he_normal(seed=0),name='Conv1'),
    tf.keras.layers.MaxPool2D(pool_size=(2,2),strides=(2,2),padding='valid',data_format='channels_last',name='Pool1'), 
    tf.keras.layers.Conv2D(filters=64,kernel_size=(3,3),strides=(1,1),padding='valid',data_format='channels_last',
              activation='relu',kernel_initializer=tf.keras.initializers.he_normal(seed=3),name='Conv2'),

    tf.keras.layers.Conv2D(filters=64,kernel_size=(3,3),strides=(2,2),padding='valid',data_format='channels_last',
              activation='relu',kernel_initializer=tf.keras.initializers.he_normal(seed=5),name='Conv3'),
    tf.keras.layers.MaxPool2D(pool_size=(2,2),strides=(1,1),padding='valid',data_format='channels_last',name='Pool2'),
    tf.keras.layers.Conv2D(filters=128,kernel_size=(3,3),strides=(2,2),padding='valid',data_format='channels_last',
              activation='relu',kernel_initializer=tf.keras.initializers.he_normal(seed=9),name='Conv4'),

    tf.keras.layers.MaxPool2D(pool_size=(2,2),strides=(2,2),padding='valid',data_format='channels_last',name='Pool3'),
    tf.keras.layers.Flatten(data_format='channels_last',name='Flatten'),    
    tf.keras.layers.Dense(units=30,activation='relu',kernel_initializer=tf.keras.initializers.glorot_normal(seed=32),name='FC1'),
    tf.keras.layers.Dense(units=15,activation='relu',kernel_initializer=tf.keras.initializers.glorot_normal(seed=33),name='FC2'),
    tf.keras.layers.Dense(units=8,activation='softmax',kernel_initializer=tf.keras.initializers.glorot_normal(seed=3),name='Output'),


  ])

下面是我正在犯的错误：

    UnknownError                              Traceback (most recent call last)
<ipython-input-47-264c0fcc37e1> in <module>
      1 ##fitting generator
----> 2 model.fit_generator(ImageGenerator,steps_per_epoch=216,epochs=3)

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training.py in fit_generator(self, generator, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, validation_freq, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch)
   1295         shuffle=shuffle,
   1296         initial_epoch=initial_epoch,
-> 1297         steps_name='steps_per_epoch')
   1298 
   1299   def evaluate_generator(self,

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training_generator.py in model_iteration(model, data, steps_per_epoch, epochs, verbose, callbacks, validation_data, validation_steps, validation_freq, class_weight, max_queue_size, workers, use_multiprocessing, shuffle, initial_epoch, mode, batch_size, steps_name, **kwargs)
    263 
    264       is_deferred = not model._is_compiled
--> 265       batch_outs = batch_function(*batch_data)
    266       if not isinstance(batch_outs, list):
    267         batch_outs = [batch_outs]

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training.py in train_on_batch(self, x, y, sample_weight, class_weight, reset_metrics)
    971       outputs = training_v2_utils.train_on_batch(
    972           self, x, y=y, sample_weight=sample_weight,
--> 973           class_weight=class_weight, reset_metrics=reset_metrics)
    974       outputs = (outputs['total_loss'] + outputs['output_losses'] +
    975                  outputs['metrics'])

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training_v2_utils.py in train_on_batch(model, x, y, sample_weight, class_weight, reset_metrics)
    262       y,
    263       sample_weights=sample_weights,
--> 264       output_loss_metrics=model._output_loss_metrics)
    265 
    266   if reset_metrics:

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training_eager.py in train_on_batch(model, inputs, targets, sample_weights, output_loss_metrics)
    309           sample_weights=sample_weights,
    310           training=True,
--> 311           output_loss_metrics=output_loss_metrics))
    312   if not isinstance(outs, list):
    313     outs = [outs]

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training_eager.py in _process_single_batch(model, inputs, targets, output_loss_metrics, sample_weights, training)
    250               output_loss_metrics=output_loss_metrics,
    251               sample_weights=sample_weights,
--> 252               training=training))
    253       if total_loss is None:
    254         raise ValueError('The model cannot be run '

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\training_eager.py in _model_loss(model, inputs, targets, output_loss_metrics, sample_weights, training)
    125     inputs = nest.map_structure(ops.convert_to_tensor, inputs)
    126 
--> 127   outs = model(inputs, **kwargs)
    128   outs = nest.flatten(outs)
    129 

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py in __call__(self, inputs, *args, **kwargs)
    889           with base_layer_utils.autocast_context_manager(
    890               self._compute_dtype):
--> 891             outputs = self.call(cast_inputs, *args, **kwargs)
    892           self._handle_activity_regularization(inputs, outputs)
    893           self._set_mask_metadata(inputs, outputs, input_masks)

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\sequential.py in call(self, inputs, training, mask)
    254       if not self.built:
    255         self._init_graph_network(self.inputs, self.outputs, name=self.name)
--> 256       return super(Sequential, self).call(inputs, training=training, mask=mask)
    257 
    258     outputs = inputs  # handle the corner case where self.layers is empty

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\network.py in call(self, inputs, training, mask)
    706     return self._run_internal_graph(
    707         inputs, training=training, mask=mask,
--> 708         convert_kwargs_to_constants=base_layer_utils.call_context().saving)
    709 
    710   def compute_output_shape(self, input_shape):

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\network.py in _run_internal_graph(self, inputs, training, mask, convert_kwargs_to_constants)
    858 
    859           # Compute outputs.
--> 860           output_tensors = layer(computed_tensors, **kwargs)
    861 
    862           # Update tensor_dict.

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\engine\base_layer.py in __call__(self, inputs, *args, **kwargs)
    889           with base_layer_utils.autocast_context_manager(
    890               self._compute_dtype):
--> 891             outputs = self.call(cast_inputs, *args, **kwargs)
    892           self._handle_activity_regularization(inputs, outputs)
    893           self._set_mask_metadata(inputs, outputs, input_masks)

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\keras\layers\convolutional.py in call(self, inputs)
    195 
    196   def call(self, inputs):
--> 197     outputs = self._convolution_op(inputs, self.kernel)
    198 
    199     if self.use_bias:

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\nn_ops.py in __call__(self, inp, filter)
   1132           call_from_convolution=False)
   1133     else:
-> 1134       return self.conv_op(inp, filter)
   1135     # copybara:strip_end
   1136     # copybara:insert return self.conv_op(inp, filter)

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\nn_ops.py in __call__(self, inp, filter)
    637 
    638   def __call__(self, inp, filter):  # pylint: disable=redefined-builtin
--> 639     return self.call(inp, filter)
    640 
    641 

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\nn_ops.py in __call__(self, inp, filter)
    236         padding=self.padding,
    237         data_format=self.data_format,
--> 238         name=self.name)
    239 
    240 

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\nn_ops.py in conv2d(input, filter, strides, padding, use_cudnn_on_gpu, data_format, dilations, name, filters)
   2008                            data_format=data_format,
   2009                            dilations=dilations,
-> 2010                            name=name)
   2011 
   2012 

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\gen_nn_ops.py in conv2d(input, filter, strides, padding, use_cudnn_on_gpu, explicit_paddings, data_format, dilations, name)
   1029             input, filter, strides=strides, use_cudnn_on_gpu=use_cudnn_on_gpu,
   1030             padding=padding, explicit_paddings=explicit_paddings,
-> 1031             data_format=data_format, dilations=dilations, name=name, ctx=_ctx)
   1032       except _core._SymbolicException:
   1033         pass  # Add nodes to the TensorFlow graph.

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\ops\gen_nn_ops.py in conv2d_eager_fallback(input, filter, strides, padding, use_cudnn_on_gpu, explicit_paddings, data_format, dilations, name, ctx)
   1128   explicit_paddings, "data_format", data_format, "dilations", dilations)
   1129   _result = _execute.execute(b"Conv2D", 1, inputs=_inputs_flat, attrs=_attrs,
-> 1130                              ctx=_ctx, name=name)
   1131   _execute.record_gradient(
   1132       "Conv2D", _inputs_flat, _attrs, _result, name)

D:\anaconda\envs\tf_gpu\lib\site-packages\tensorflow_core\python\eager\execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     65     else:
     66       message = e.message
---> 67     six.raise_from(core._status_to_exception(e.code, message), None)
     68   except TypeError as e:
     69     keras_symbolic_tensors = [

D:\anaconda\envs\tf_gpu\lib\site-packages\six.py in raise_from(value, from_value)

UnknownError: Failed to get convolution algorithm. This is probably because cuDNN failed to initialize, so try looking to see if a warning log message was printed above. [Op:Conv2D]

我正在使用tensorflow 2.0，安装在anaconda版本10.2中。当我不使用cnn的时候，谁能帮我做同样的安装吗？。

是因为我在使用CONV2d，还是因为我使用了生成器？我在一个windows 10机器上，16 gb内存和4gb nvidia 1650显卡。

Answer 1

得到相同的错误，并通过以下方法解决：

gpus = tf.config.experimental.list_physical_devices('GPU')
tf.config.experimental.set_virtual_device_configuration(gpus[0], 
           [tf.config.experimental.VirtualDeviceConfiguration(memory_limit=4000)])

(使用GTX 1660，6G内存)