Abhinav Gupta

The option_term_prob gives the option termination probability for the current option and done indicates a transition from current state to the next state. In that case, we need an advantage...

I login to this compute node via a login node. The `tmux` session was opened on the login node but I guess it is supposed to update after the update-interval...

I think adding just a stop button won't overwhelm the UI since it makes sense to put it next to those buttons

Same for VQA1 too?

And you used the whole training set for training 100 epochs?

But you only train with the examples whose multiple choice answer is in the top n% (n determined by nans) and discard other examples. Is that correct?

Oh I thought you used nn.NLLLoss but I see you use nn.CrossEntropy. Thats fine. So will you merging that private repo? Yeah I am saying that the MLB authors choose...

I am working on this bug. What I understand is that the TER record should be written (before or after every ENDML, doesn't matter, right?)

Since the TER issue is now fixed, is the bug solved?