I find CTranslate2 has the target_prefiix support, which is used for autocompletion.
But how can I use tgt_prefix by opennmt_py, for example, in .yml format?
I can’t make sense to use it, like in add target_prefix decoding feature by Zenglinxiao · Pull Request #1800 · OpenNMT/OpenNMT-py · GitHub
model: xxxx.pt
src: xxxxx.txt
output: xxxx.txt
gpu: 0
beam_size: 10
n_best: 10
batch_size: 16384
batch_type: tokens
max_length: 500
seed: 0