Error running adanet_objective on Windows 7 (Japanese) with Python 3 about adanet HOT 9 CLOSED

@GitHub30
Even if I change to a python2 environment, I will still get an error:

TypeError                                 Traceback (most recent call last)
<ipython-input-7-7612596cd623> in <module>()
     63 
     64 
---> 65 results, _ = train_and_evaluate()
     66 print("Loss:", results["average_loss"])
     67 print("Architecture:", ensemble_architecture(results))

TypeError: 'NoneType' object is not iterable

from adanet.

This colab completely work with Python 3.

https://colab.research.google.com/github/GitHub30/adanet/blob/google-colaboratory-python3/adanet/examples/tutorials/adanet_objective.colab.python3.ipynb

From the top menu bar, choose Runtime - Change runtime type.
Please select Python 3 from the runtime type.

from adanet.

I also got same issue on ubuntu 16.04
ValueError: b"\n\x83\x01\n;adanet/iteration_2/ensemble_3_layer_dnn/architecture/adanetB:\x08\x07\x12\x00B4| b'1_layer_dnn' | b'2_layer_dnn' | b'3_layer_dnn' |J\x08\n\x06\n\x04text" has type str, but isn't valid UTF-8 encoding. Non-UTF-8 strings must be converted to unicode objects before being added.

from adanet.

Please try with Python 2

from adanet.

The model started with no errors on python2

from adanet.

@GitHub30
How can I use python3 to make the code work? Because python2 is no longer applicable.

from adanet.

Working on Ubuntu 16.02 with python 2, NOT working on Windows 10 with Python 3. I get the same error as reported.
@GitHub30 Is there Python 3 compatibility in the roadmap? As @zhouhao-learning also said, we should stop using Python 2.

from adanet.

@Efaq: We do test this package on Python 3, so I'm surprised you are running into such an error. In fact, normally six.u(string) is supposed to convert a string to unicode in both Python 2 and 3, so it's not clear to me how that line is raising an exception:

field[key].string_value = six.u(value)

from adanet.

@GitHub30 Does it mean you cannot reproduce the error, or is this a suggested workaround to play with the code? Is the pull request mentioned aiming to fix it?
Here is a more detailed specs from my environment, where I still can reproduce the error:
Windows 10, Python 3.6, virtualenv with Tensorflow 1.12.0rc2 and adanet 0.1.0.dev2. Some checkpoints saved, and then the error:
ValueError: b"\n\x83\x01\n;adanet/iteration_2/ensemble_2_layer_dnn/architecture/adanetB:\x08\x07\x12\x00B4| b'1_layer_dnn' | b'2_layer_dnn' | b'2_layer_dnn' |J\x08\n\x06\n\x04text" has type str, but isn't valid UTF-8 encoding. Non-UTF-8 strings must be converted to unicode objects before being added.

from adanet.