question regarding _get_parallel_step_context

[jingweiz]

Hi, thanks for making the code public!
I have a question regarding the function _get_parallel_step_context:
Here,

attention-learn-to-route/nets/attention_model.py

Line 378 in c66da2c

batch_size, num_steps = current_node.size()

, num_steps would always be 1 as the current_node reads the prev_a of the tsp state, so then this means that

attention-learn-to-route/nets/attention_model.py

Line 427 in c66da2c

if num_steps == 1: # We need to special case if we have only 1 step, may be the first or not

will always be hit, and the lines from 436 to 449 will never be used, is this correct or am I missing sth here?
Thanks in advance and looking forward for your reply!
Jingwei

[wouterkool]

Hi!

This is correct. The reason that there is a steps dimension is that this can be used to evaluate the model on a given tour in a single forward pass, which is much more efficient than one step at the time. This could be useful for, e.g., supervised training (teacher forcing) or things like experience replay. This code is a leftover of some early experiments in that direction which I thought may still be useful to somebody.

Wouter