Tensorflow seq2seq 1.0 API: Why is embeddings a required argument to simple_decoder_fn_inference?

Vote count: 0

See tensorflow/contrib/seq2seq/python/ops/decoder_fn.py:106

I'm trying to build a sequence-to-sequence model which doesn't need embedding. How do I go about this with the new API?

I was trying to adapt the example from here.

asked 1 min ago

John

Let's block ads! (Why?)

Tensorflow seq2seq 1.0 API: Why is embeddings a required argument to simple_decoder_fn_inference?