--reinforce_learn doesn't work!

[SeekPoint]

No description provided.

[SeekPoint]

On CPU setting:
after the following changes:
@@ -331,7 +332,9 @@ class Seq2SeqModel(object):
while True:
#----[Step]----------------------------------------
encoder_state, step_loss, output_logits = self.step(session, encoder_inputs, decoder_inputs, target_weights,

•                      bucket_id, training=False, force_dec_input=False)
  

•                      bucket_id, forward_only=False, force_dec_input=False)
  

and

@@ -395,7 +399,7 @@ class Seq2SeqModel(object):

 # step
 _, _, output_logits = self.step(session, encoder_inputs, decoder_inputs, target_weights,

•                    bucket_id, training=False, force_dec_input=True)
  

•                    bucket_id, forward_only=False, force_dec_input=True)
  

I still got at function logits2tokens:

Creating 4 layers of 256 units.
Created model with fresh parameters.
Reading development and training data (limit: 0).
reading data line 100000
reading data line 200000
[INPUT]: [b'\xe8\xae\x93', b'\xe4\xbd\xa0', b'\xe8\xbd\x89\xe9\x81\x8e', b'\xe8\xba\xab\xe8\x83\x8c', b'\xe5\xb0\x8d', b'\xe8\x91\x97', b'\xe6\x88\x91', b'\xe7\x82\xba', b'\xe4\xbd\xa0', b'\xe7\xb9\xab', b'\xe4\xb8\x8a', b'\xe6\x88\x91', b'\xe7\x9a\x84', b'\xe6\x89\xbf', b'\xe8\xab\xbe', b'_PAD', b'_PAD', b'_PAD', b'_PAD', b'_PAD']
output_logits is: None
Traceback (most recent call last):
File "main.py", line 28, in
tf.app.run()
File "/ve_tf1.0_py3/venv/lib/python3.6/site-packages/tensorflow/python/platform/app.py", line 44, in run
_sys.exit(main(_sys.argv[:1] + flags_passthrough))
File "main.py", line 18, in main
train(args)
File "/ve_tf1.0_py3/tf_chatbot_seq2seq_antilm---ori/lib/train.py", line 83, in train
target_weights, bucket_id, rev_vocab=rev_vocab)
File "/ve_tf1.0_py3/tf_chatbot_seq2seq_antilm---ori/lib/seq2seq_model.py", line 347, in step_rf
resp_tokens, resp_txt = self.logits2tokens(output_logits, rev_vocab, sent_max_length)
File "/ve_tf1.0_py3/tf_chatbot_seq2seq_antilm---ori/lib/seq2seq_model.py", line 414, in logits2tokens
tokens = [int(np.argmax(t, axis=1)) for t in logits]
TypeError: 'NoneType' object is not iterable

[yogesh-0586]

did you get something about --reinforce_learn not work ?

[SeekPoint]

@yogesh-0586
still struggling

[dollyk]

For first "step()"
(original)one of parameter : training =False, which means do the predict , that is , forward_only=True
however, I don't know whether the "force_dec_input" should be set True or False.
i am working on the reinforce_learn too.