Using a `tf.Tensor` as a Python `bool` is not allowed

[VibhuSrivastava]

When training the model by running the train.py command, I encounter the following issue:

python train.py
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcuda.so locally
I tensorflow/stream_executor/dso_loader.cc:108] successfully opened CUDA library libcurand.so locally
You are using the decoding_enhanced library!
Initializer Library -- Will add to this if necessary.
Unitary Linear has been imported
Preparing dialog data in /var/lib/tf_seq2seq_chatbot/data
E tensorflow/stream_executor/cuda/cuda_driver.cc:491] failed call to cuInit: CUDA_ERROR_NO_DEVICE
I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:140] kernel driver does not appear to be running on this host (instance-1): /proc/driver/nvidia/version does not exist
I tensorflow/core/common_runtime/gpu/gpu_init.cc:81] No GPU devices available on machine.
Creating 4 layers of 2048 units.
Traceback (most recent call last):
host (instance-1): /proc/driver/nvidia/version does not exist
File "train.py", line 15, in
tf.app.run()
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/platform/app.py", line 30, in run
sys.exit(main(sys.argv))
File "train.py", line 12, in main
train()
File "/home/vibhu_vandita/tf_seq2seq_chatbot/tf_seq2seq_chatbot/lib/train.py", line 28, in train
model = create_model(sess, forward_only=False)
File "/home/vibhu_vandita/tf_seq2seq_chatbot/tf_seq2seq_chatbot/lib/seq2seq_model_utils.py", line 27, in cre
ate_model
forward_only=forward_only)
File "/home/vibhu_vandita/tf_seq2seq_chatbot/tf_seq2seq_chatbot/lib/seq2seq_model.py", line 149, in init
softmax_loss_function=softmax_loss_function)
File "build/bdist.linux-x86_64/egg/rnn_enhancement/seq2seq_enhanced.py", line 629, in model_with_buckets
File "/home/vibhu_vandita/tf_seq2seq_chatbot/tf_seq2seq_chatbot/lib/seq2seq_model.py", line 148, in
lambda x, y: seq2seq_f(x, y, False),
File "/home/vibhu_vandita/tf_seq2seq_chatbot/tf_seq2seq_chatbot/lib/seq2seq_model.py", line 112, in seq2seq_
f
feed_previous=do_decode)
File "build/bdist.linux-x86_64/egg/rnn_enhancement/seq2seq_enhanced.py", line 341, in embedding_attention_se
q2seq
File "build/bdist.linux-x86_64/egg/rnn_enhancement/seq2seq_enhanced.py", line 259, in embedding_attention_de
coder
File "build/bdist.linux-x86_64/egg/rnn_enhancement/seq2seq_enhanced.py", line 165, in attention_decoder
File "build/bdist.linux-x86_64/egg/rnn_enhancement/seq2seq_enhanced.py", line 112, in attention
File "build/bdist.linux-x86_64/egg/rnn_enhancement/linear_enhanced.py", line 42, in linear
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/ops.py", line 528, in nonzero
raise TypeError("Using a tf.Tensor as a Python bool is not allowed. "
TypeError: Using a tf.Tensor as a Python bool is not allowed. Use if t is not None: instead of if t: to test if a tensor is defined, and use the logical TensorFlow ops to test the value of a tensor.

What can be done in order to resolve this issue?

[maeda]

I don't know exactly why, but when you comment or remove that validation (https://github.com/nicolas-ivanov/Seq2Seq_Upgrade_TensorFlow/blob/master/rnn_enhancement/linear_enhanced.py#L44), this error not happen anymore.

I'm just testing this code, so I won't invest more time to investigate the reason. Anyway, I'm sharing to unblock you.