neural-combinatorial-rl-pytorch
neural-combinatorial-rl-pytorch copied to clipboard
beam search
The To Do mentions "Finish implementing beam search decoding to support > 1 beam".
It looks like you have various beam width variables in different places of the code. Was your note because:
- you implemented beam search but it didn't seem to work as well when width>1 and you wanted to investigate further, or
- there are still some places in the code that would need to be modified to before using beam search?
If 2), then which places would need to be modified?
Thanks!