How to run tensor flow seq2seq demo

Question

I tensor flow installed and successfully went through the MNIST demo. Now, I am trying to run the seq2seq demo, but this is not working for me.

I cloned a version of their github repo and attempted to run some of the listed commands from the repo root.

$ bazel run -c opt ./tensorflow/models/rnn/translate/translate.py 
    ERROR: Bad target pattern './tensorflow/models/rnn/translate/translate.py': package names may contain only A-Z, a-z, 0-9, '/', '-' and '_'.
    INFO: Elapsed time: 0.115s
    ERROR: Build failed. Not running target.

No surprise here, as it doesn't really make sense to have bazel execute a python script.

Later in the tutorial,

$ bazel run -c opt //tensorflow/models/rnn/translate:translate \
  --data_dir ./data_dir --train_dir ./checkpoints_directory \
  --en_vocab_size=40000 --fr_vocab_size=40000

Unrecognized option: --data_dir

If I remove the parameters from the invocation above, it will attempt (and fail) to build the entire tensor flow project before it executes translate. This is not what I want as I have already successfully installed tensor flow with pip.

The last thing I tried running,

$ python ./tensorflow/models/rnn/translate/translate.py 
Traceback (most recent call last):
  File "./tensorflow/models/rnn/translate/translate.py", line 28, in <module>
    from tensorflow.models.rnn.translate import data_utils
ImportError: No module named translate

Environment info: OS X 10.11.1, Python 2.7.10 (anaconda)

Answer 1

There are two ways to run the script:

1) separate the script arguments with -- as part of bazel run

bazel run -c opt //tensorflow/models/rnn/translate:translate -- \
--data_dir ./data_dir --train_dir ./checkpoints_directory \
--en_vocab_size=40000 --fr_vocab_size=40000

2) build and then run from ./bazel-bin/:

bazel build -c opt //tensorflow/models/rnn/translate:translate

./bazel-bin/tensorflow/models/rnn/translate/translate \
--data_dir ./data_dir --train_dir ./checkpoints_directory \
--en_vocab_size=40000 --fr_vocab_size=40000