Lesson 13 in-class

Part 2 (2017)
41

A question about the spelling_bee_RNN.ipynb notebook with TENSORFLOW :

The seq2seq.simple_decoder_fn_inference(
encode_input, decode_input, stacked_lstm, input_vocab_size, output_vocab_size, 64)

outputs a TypeError: simple_decoder_fn_inference() missing 1 required positional argument: ‘num_decoder_symbols’

And when I use the simple_decoder_fn_inference function documentation on tensorflow to fix the num_decoder_symbols to a value I still get the error ‘function’ object is not iterable :
The function definition reference link on the documentation page also doesn’t work no more.

Did someone get the same error please? How did he solve it ?
I run my code on a CPU system.

42

@Kjeanclaude note that this thread is for in class discussion, so using it after class may cause us to miss your question.

The parts of the notebook that I didn’t present in class (which includes all the tensorflow stuff) aren’t working. Tensorflow has changed their seq2seq API.

1 Like
43

OK, thank you Jeremy for your response.
I thought that I didn’t get something when this part didn’t work during my tests.
So I will come back on when the new version of the API will appear.

And thank you again to you and Rachel also, I have learned a lot of things during these sessions (research paper understanding, a deep understanding of GANs and of LSTMs with the concept of attention, etc.)