Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

Save SavedModel fails for SetInputFeature

See original GitHub issue

Describe the bug Error when attempting to call save_savedmodel for a model that containers a set property. See error:

2020-10-19 03:06:47.001119: I tensorflow/compiler/xla/service/service.cc:176]   StreamExecutor device (0): Host, Default Version
Traceback (most recent call last):
  File "train.py", line 78, in <module>
    train(args)
  File "train.py", line 41, in train
    ludwig_model.save_savedmodel(args.model_dir)
  File "/usr/local/lib/python3.6/dist-packages/ludwig/api.py", line 1445, in save_savedmodel
    self.model.save_savedmodel(save_path)
  File "/usr/local/lib/python3.6/dist-packages/ludwig/models/ecd.py", line 82, in save_savedmodel
    keras_model = self.get_connected_model(training=False)
  File "/usr/local/lib/python3.6/dist-packages/ludwig/models/ecd.py", line 77, in get_connected_model
    inputs = inputs or self.get_model_inputs(training)
  File "/usr/local/lib/python3.6/dist-packages/ludwig/models/ecd.py", line 63, in get_model_inputs
    self.input_features.items()
  File "/usr/local/lib/python3.6/dist-packages/ludwig/models/ecd.py", line 62, in <dictcomp>
    for input_feature_name, input_feature in
  File "/usr/local/lib/python3.6/dist-packages/ludwig/features/base_feature.py", line 70, in create_input
    return tf.keras.Input(shape=self.get_input_shape(),
  File "/usr/local/lib/python3.6/dist-packages/ludwig/features/set_feature.py", line 127, in get_input_shape
    return len(self.vocab),
AttributeError: 'SetInputFeature' object has no attribute 'vocab'

https://github.com/uber/ludwig/blob/master/ludwig/features/set_feature.py#L127

To Reproduce

Create a model definition with a set property - in my case I had two set properties for a regression output.

preprocessing:
    text:
        most_common_word: 10000
input_features:
# set features
    -   name: base_colour
        type: set
    -   name: manufacturer_colour
        type: set
# more features...
# regression output
output_features:
    -   name: days_online
        type: numerical

Sample dataset for the three fields above:

days_online,base_colour,manufacturer_colour
14,white,white
8,meteor grey,
4,brown,
11,white,
44,white,fuji white

Expected behavior Expected to be able to create the tensorflow model.

Environment (please complete the following information):

OS: Ubuntu
Version: 18.04.3 LTS
Python: 3.6.9
Ludwig version: master (commit hash ab9bb7cac6287578e73b46579c18d518e2ddb5a6) ~0.3
TF version: tensorflow:2.3.0-gpu

Additional context I was testing this so that I could produce the the TF pb file for visualizing the TF graph in tensorboard as I wasn’t able to do this with just the checkpoint and logs.