The real_nvp example code seems wrong in computing log_det from the decoder function

[haowen-xu]

#7177 # System information

• What is the top-level directory of the model you are using: research/real_nvp
• Have I written custom code (as opposed to using a stock example script provided in TensorFlow): no
• OS Platform and Distribution (e.g., Linux Ubuntu 16.04): Ubuntu 18.04 (in docker)
• TensorFlow installed from (source or binary): Binary
• TensorFlow version (use command below): 1.14.0
• Bazel version (if compiling from source):
• CUDA/cuDNN version: 9.0.176
• GPU model and memory: GTX 1080 Ti
• Exact command to reproduce:

Describe the problem

Describe the problem clearly here. Be sure to convey here why it's a bug in TensorFlow or a feature request.

In real_nvp_multiscale_dataset.py, at L806 and L818, you mistakenly override log_diff_1 and log_diff. The outputs are supposed to be assigned to inc_log_diff and then add to log_diff_1 or log_diff respectively, as is done in the encoder at L744 and L756.

This problem have not affected the results of your example code, since the log_diff output from the decoder is never used. However, I think it might be better fixed, since it is an example code from the TensorFlow models repo.

Source code / logs

Include any logs or source code that would be helpful to diagnose the problem. If including tracebacks, please include the full traceback. Large logs and files should be attached. Try to provide a reproducible test case that is the bare minimum necessary to generate the problem.