Train on cityscapes. A `Concatenate` layer requires inputs with matching shapes except for the concat axis
mrengineer opened this issue · 4 comments
I am trying to train cityscapes. But I have messages hereunder. Pls help
Traceback (most recent call last):
File "train.py", line 115, in
model = DeepLabV3Plus(H, W, num_classes)
File "/home/bulkin/tensorflow/DeepLabV3_Plus-Tensorflow2.0/deeplab.py", line 70, in DeepLabV3Plus
x_a = ASPP(image_features)
File "/home/bulkin/tensorflow/DeepLabV3_Plus-Tensorflow2.0/deeplab.py", line 54, in ASPP
y = concatenate([y_pool, y_1, y_6, y_12, y_18], name='ASPP_concat')
File "/home/bulkin/.local/lib/python3.6/site-packages/tensorflow/python/keras/layers/merge.py", line 687, in concatenate
return Concatenate(axis=axis, **kwargs)(inputs)
File "/home/bulkin/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/base_layer.py", line 616, in call
self._maybe_build(inputs)
File "/home/bulkin/.local/lib/python3.6/site-packages/tensorflow/python/keras/engine/base_layer.py", line 1966, in _maybe_build
self.build(input_shapes)
File "/home/bulkin/.local/lib/python3.6/site-packages/tensorflow/python/keras/utils/tf_utils.py", line 296, in wrapper
output_shape = fn(instance, input_shape)
File "/home/bulkin/.local/lib/python3.6/site-packages/tensorflow/python/keras/layers/merge.py", line 391, in build
'Got inputs shapes: %s' % (input_shape))
ValueError: A Concatenate layer requires inputs with matching shapes except for the concat axis. Got inputs shapes: [(None, 32, 32, 256), (None, 32, 32, 256), (None, None, None, 256), (None, None, None, 256), (None, None, None, 256)]
Can you post the tensorflow version you are currently using?
Tensorflow 2.0.0-beta1
python 3.6
BTW, why dont you make docker image?
UPD: downgraded to TensorFlow 2.0.0-alpha0 and train is started.
UPD2: cuddn 7.6.3, cuda 10, TF 2.0.0 alpha0. Nvidia 1080TI. Train start, but not enough memory
Train is started with bath = 6
@mrengineer Tensorflow 2.0.0-rc0 will also work