I get the following error when fitting the model:
TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [None, 16, 16384, 128]. Consider casting elements to a supported type.
This is where I define the input shape in the model:
from tensorflow.keras.layers import Input
def model(input_shape=(4, 128, 128, 128),n_base_filters=16, depth=5, dropout_rate=0.3,
n_segmentation_levels=3, n_labels=4, optimizer=Adam, initial_learning_rate=5e-4,
loss_function=bin_crossentropy, activation_name="sigmoid",metrics=dice_coefficient):
inputs = Input(input_shape)
......
Traceback:
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/tensor_util.py in make_tensor_proto(values, dtype, shape, verify_shape, allow_broadcast)
548 try:
--> 549 str_values = [compat.as_bytes(x) for x in proto_values]
550 except TypeError:
24 frames
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/tensor_util.py in <listcomp>(.0)
548 try:
--> 549 str_values = [compat.as_bytes(x) for x in proto_values]
550 except TypeError:
/usr/local/lib/python3.7/dist-packages/tensorflow/python/util/compat.py in as_bytes(bytes_or_text, encoding)
86 raise TypeError('Expected binary or unicode string, got %r' %
---> 87 (bytes_or_text,))
88
TypeError: Expected binary or unicode string, got None
During handling of the above exception, another exception occurred:
TypeError Traceback (most recent call last)
<ipython-input-56-c79d6965b744> in <module>()
1 model=model(input_shape=(4, 128, 128, 128),n_base_filters=16, depth=5, dropout_rate=0.3,
2 n_segmentation_levels=3, n_labels=4, optimizer=Adam, initial_learning_rate=5e-4,
----> 3 loss_function=weighted_dice_coefficient_loss, activation_name="sigmoid",metrics=dice_coefficient)
4
5 print(len(model.layers))
<ipython-input-55-051eee07240c> in model(input_shape, n_base_filters, depth, dropout_rate, n_segmentation_levels, n_labels, optimizer, initial_learning_rate, loss_function, activation_name, metrics)
48
49 if current_layer is inputs:
---> 50 in_conv = create_convolution_block_flip(current_layer, n_level_filters)
51 else:
52 in_conv = create_convolution_block_flip(current_layer,n_level_filters, strides=(2, 2, 2))
<ipython-input-42-e2034c0ac3eb> in create_convolution_block_flip(input_layer, n_filters, num_train_examples, batch_normalization, kernel, activation, padding, strides, instance_normalization)
7 tf.cast(num_train_examples, dtype=tf.float32))
8
----> 9 layer = tfp.layers.Convolution3DFlipout(n_filters, kernel, padding=padding, strides=strides, data_format="channels_first",kernel_divergence_fn=kl_diverge_func)(input_layer)
10
11
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py in __call__(self, *args, **kwargs)
968 if _in_functional_construction_mode(self, inputs, args, kwargs, input_list):
969 return self._functional_construction_call(inputs, args, kwargs,
--> 970 input_list)
971
972 # Maintains info about the `Layer.call` stack.
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py in _functional_construction_call(self, inputs, args, kwargs, input_list)
1106 # Check input assumptions set after layer building, e.g. input shape.
1107 outputs = self._keras_tensor_symbolic_call(
-> 1108 inputs, input_masks, args, kwargs)
1109
1110 if outputs is None:
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py in _keras_tensor_symbolic_call(self, inputs, input_masks, args, kwargs)
838 return nest.map_structure(keras_tensor.KerasTensor, output_signature)
839 else:
--> 840 return self._infer_output_signature(inputs, args, kwargs, input_masks)
841
842 def _infer_output_signature(self, inputs, args, kwargs, input_masks):
/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/engine/base_layer.py in _infer_output_signature(self, inputs, args, kwargs, input_masks)
878 self._maybe_build(inputs)
879 inputs = self._maybe_cast_inputs(inputs)
--> 880 outputs = call_fn(inputs, *args, **kwargs)
881
882 self._handle_activity_regularization(inputs, outputs)
/usr/local/lib/python3.7/dist-packages/tensorflow/python/autograph/impl/api.py in wrapper(*args, **kwargs)
690 try:
691 with conversion_ctx:
--> 692 return converted_call(f, args, kwargs, options=options)
693 except Exception as e: # pylint:disable=broad-except
694 if hasattr(e, 'ag_error_metadata'):
/usr/local/lib/python3.7/dist-packages/tensorflow/python/autograph/impl/api.py in converted_call(f, args, kwargs, caller_fn_scope, options)
334 if conversion.is_in_allowlist_cache(f, options):
335 logging.log(2, 'Allowlisted %s: from cache', f)
--> 336 return _call_unconverted(f, args, kwargs, options, False)
337
338 if ag_ctx.control_status_ctx().status == ag_ctx.Status.DISABLED:
/usr/local/lib/python3.7/dist-packages/tensorflow/python/autograph/impl/api.py in _call_unconverted(f, args, kwargs, options, update_cache)
461
462 if kwargs is not None:
--> 463 return f(*args, **kwargs)
464 return f(*args)
465
/usr/local/lib/python3.7/dist-packages/tensorflow_probability/python/layers/conv_variational.py in call(self, inputs)
230
231 outputs = self._apply_variational_kernel(inputs)
--> 232 outputs = self._apply_variational_bias(outputs)
233 if self.activation is not None:
234 outputs = self.activation(outputs)
/usr/local/lib/python3.7/dist-packages/tensorflow_probability/python/layers/conv_variational.py in _apply_variational_bias(self, inputs)
386 [outputs_shape[0], outputs_shape[1],
387 outputs_shape[2] * outputs_shape[3],
--> 388 outputs_shape[4]])
389 outputs_4d = tf.nn.bias_add(outputs_4d,
390 self.bias_posterior_tensor,
/usr/local/lib/python3.7/dist-packages/tensorflow/python/util/dispatch.py in wrapper(*args, **kwargs)
204 """Call target, and fall back on dispatchers if there is a TypeError."""
205 try:
--> 206 return target(*args, **kwargs)
207 except (TypeError, ValueError):
208 # Note: convert_to_eager_tensor currently raises a ValueError, not a
/usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/array_ops.py in reshape(tensor, shape, name)
193 A `Tensor`. Has the same type as `tensor`.
194 """
--> 195 result = gen_array_ops.reshape(tensor, shape, name)
196 tensor_util.maybe_set_static_shape(result, shape)
197 return result
/usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/gen_array_ops.py in reshape(tensor, shape, name)
8396 # Add nodes to the TensorFlow graph.
8397 _, _, _op, _outputs = _op_def_library._apply_op_helper(
-> 8398 "Reshape", tensor=tensor, shape=shape, name=name)
8399 _result = _outputs[:]
8400 if _execute.must_record_gradient():
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/op_def_library.py in _apply_op_helper(op_type_name, name, **keywords)
523 except TypeError as err:
524 if dtype is None:
--> 525 raise err
526 else:
527 raise TypeError(
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/op_def_library.py in _apply_op_helper(op_type_name, name, **keywords)
513 name=input_arg.name,
514 as_ref=input_arg.is_ref,
--> 515 preferred_dtype=default_dtype)
516 else:
517 values = ops.convert_to_tensor(
/usr/local/lib/python3.7/dist-packages/tensorflow/python/profiler/trace.py in wrapped(*args, **kwargs)
161 with Trace(trace_name, **trace_kwargs):
162 return func(*args, **kwargs)
--> 163 return func(*args, **kwargs)
164
165 return wrapped
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/ops.py in convert_to_tensor(value, dtype, name, as_ref, preferred_dtype, dtype_hint, ctx, accepted_result_types)
1564
1565 if ret is None:
-> 1566 ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
1567
1568 if ret is NotImplemented:
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/constant_op.py in _constant_tensor_conversion_function(v, dtype, name, as_ref)
337 as_ref=False):
338 _ = as_ref
--> 339 return constant(v, dtype=dtype, name=name)
340
341
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/constant_op.py in constant(value, dtype, shape, name)
263 """
264 return _constant_impl(value, dtype, shape, name, verify_shape=False,
--> 265 allow_broadcast=True)
266
267
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/constant_op.py in _constant_impl(value, dtype, shape, name, verify_shape, allow_broadcast)
281 tensor_util.make_tensor_proto(
282 value, dtype=dtype, shape=shape, verify_shape=verify_shape,
--> 283 allow_broadcast=allow_broadcast))
284 dtype_value = attr_value_pb2.AttrValue(type=tensor_value.tensor.dtype)
285 attrs = {"value": tensor_value, "dtype": dtype_value}
/usr/local/lib/python3.7/dist-packages/tensorflow/python/framework/tensor_util.py in make_tensor_proto(values, dtype, shape, verify_shape, allow_broadcast)
551 raise TypeError("Failed to convert object of type %s to Tensor. "
552 "Contents: %s. Consider casting elements to a "
--> 553 "supported type." % (type(values), values))
554 tensor_proto.string_val.extend(str_values)
555 return tensor_proto
TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [None, 16, 16384, 128]. Consider casting elements to a supported type.
I know the error is arising from layer = tfp.layers.Convolution3DFlipout(n_filters, kernel, padding=padding, strides=strides,data_format='channels_first')(input_layer). Default data_format in Convolution3DFlipout is 'channels_last'. When I change it to data_format='channels_first', it started to throw this.
What am I doing wrong here?
If you inspect the following line from the source code:
outputs_shape = outputs.shape.as_list()
outputs_4d = tf.reshape(outputs,
[outputs_shape[0], outputs_shape[1],
outputs_shape[2] * outputs_shape[3],
outputs_shape[4]])
So your Convolution has 16 filters, and according to the reshaping 128*128 = 16384 which explains the shape in the error above.
The workaround is, you can edit line:
outputs_shape = outputs.shape.as_list()
To:
outputs_shape = tf.shape(outputs)
Here outputs.shape was static, with tf.shape it is dynamic now. I've run some unit tests with the following change, everything seemed fine so far.
Reproducible example, before changing:
import tensorflow as tf
import tensorflow_probability as tfp
print(tf.__version__) # 2.5.0
print(tfp.__version__) # 0.13.0
inputs = (3,32,32,32)
model2 = tf.keras.Sequential([
tf.keras.layers.Input(shape = inputs),
tfp.layers.Convolution3DFlipout(
64, kernel_size=5, padding='SAME', activation=tf.nn.relu,
data_format = 'channels_first'),
tf.keras.layers.MaxPooling3D(pool_size=(2, 2, 2),
strides=(2, 2, 2),
padding='SAME'),
tf.keras.layers.Flatten(),
tfp.layers.DenseFlipout(10),
])
TypeError: Failed to convert object of type <class 'list'> to Tensor. Contents: [None, 64, 1024, 32]. Consider casting elements to a supported type.
After changing to following line above in the source code model.summary() shows:
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
conv3d_flipout (Conv3DFlipou (None, 64, 32, 32, 32) 48064
_________________________________________________________________
max_pooling3d (MaxPooling3D) (None, 32, 16, 16, 32) 0
_________________________________________________________________
flatten (Flatten) (None, 262144) 0
_________________________________________________________________
dense_flipout (DenseFlipout) (None, 10) 5242890
=================================================================
Total params: 5,290,954
Trainable params: 5,290,954
Non-trainable params: 0