Converter fails with tf.keras.applications.EfficientNet

[xchoo]

Trying to use the NengoDL converter with the built-in EfficientNet TensorFlow networks will cause a failure with this error:

:\xchoo\git\nengo-dl\nengo_dl\converter.py:324: UserWarning: Layer type Rescaling does not have a registered converter. Falling back to TensorNode.
  warnings.warn(
d:\xchoo\git\nengo-dl\nengo_dl\converter.py:324: UserWarning: Layer type Normalization does not have a registered converter. Falling back to TensorNode.
  warnings.warn(
d:\xchoo\git\nengo-dl\nengo_dl\converter.py:586: UserWarning: Activation type <function swish at 0x0000017DD4D479D0> does not have a native Nengo equivalent; falling back to a TensorNode
  warnings.warn(
d:\xchoo\git\nengo-dl\nengo_dl\converter.py:324: UserWarning: Layer type DepthwiseConv2D does not have a registered converter. Falling back to TensorNode.
  warnings.warn(
d:\xchoo\git\nengo-dl\nengo_dl\converter.py:324: UserWarning: Layer type Multiply does not have a registered converter. Falling back to TensorNode.
  warnings.warn(
Traceback (most recent call last):
  File ".\test_efficientnet.py", line 59, in <module>
    conv = nengo_dl.Converter(
  File "d:\xchoo\git\nengo-dl\nengo_dl\converter.py", line 133, in __init__
    self.net = self.get_converter(model).convert(None)
  File "d:\xchoo\git\nengo-dl\nengo_dl\converter.py", line 870, in convert
    nengo_layer = layer_converter.convert(layer_node_id)
  File "d:\xchoo\git\nengo-dl\nengo_dl\converter.py", line 962, in convert
    output = self.tensor_layer(
  File "d:\xchoo\git\nengo-dl\nengo_dl\tensor_node.py", line 458, in __call__
    obj = TensorNode(
  File "D:\xchoo\miniconda3\envs\nengo-dl\lib\site-packages\nengo\base.py", line 34, in __call__
    inst.__init__(*args, **kwargs)
  File "d:\xchoo\git\nengo-dl\nengo_dl\tensor_node.py", line 198, in __init__
    self.shape_in = shape_in
  File "D:\xchoo\miniconda3\envs\nengo-dl\lib\site-packages\nengo\base.py", line 108, in __setattr__
    super().__setattr__(name, val)
  File "D:\xchoo\miniconda3\envs\nengo-dl\lib\site-packages\nengo\config.py", line 484, in __setattr__
    raise exc_info[1].with_traceback(None) from e
nengo.exceptions.ValidationError: TensorNode.shape_in: Element 0 must be an int (got type 'tuple')

Minimal code to reproduce error:

import nengo_dl
import tensorflow as tf

model = tf.keras.applications.EfficientNetB2(
    include_top=True,
    weights="imagenet",
    input_tensor=None,
    input_shape=None,
    pooling=None,
    classes=1000,
    classifier_activation="softmax",
)

conv = nengo_dl.Converter(
    model, allow_fallback=True, max_to_avg_pool=True, inference_only=True
)

I tested some of the other tf.keras.applications networks, and the only networks that seem to be affected are the EfficientNet ones.

First reported on the Nengo forums here.