tf-seq2seq
tf-seq2seq copied to clipboard
attn_input_feeding
In the documentation, it is suggested to make attn_input_feeding =True during decoding. But in the code, I don't see any place where it is set to True during decoding.
The configuration is all read from the dump formed during training and since it was set False during training, the attn_input_feeding remains False even during decoding.
Am i missing something?