-
Actually, I have already trained my emergent game now I want to generate the interactions that are learned. Specifically, I want to pass my training dataset through my trained 'game' to store the mess…
-
in the train_vae script the kl_loss is set to zero via the weight parameter and also in my elaborate runs of experiments, I found that including the KL term does more harm than it helps. @karpathy als…
-
## 🚀 Feature
Add PlackettLuce and RelaxedPlackettLuce distributions. It is a simple distribution over permutations.
## Motivation
For optimization over categorical/binary variables (i.e. variat…
-
Hello~ I recently read your brilliant paper, but confused anout BP problem mentioned in the introduction:
`Moreover, this would also hinder the back-propagation for the prediction module, which need…
-
# 🚀 Feature Request: Batch method for non-analytic acquisition functions when using fully-Bayesian GPRs
A batch method for non-analytic acquisition functions when using fully-Bayesian treated GPRs.…
-
Hello and thank you for this repo.
I was wondering, if there is a reason to use a 3-dim embedding instead of a 2-dim codebook.
Is the idea to achieve some from of multi head gumbel sampling? T…
-
```bash
tensorflow==2.7.0
tensorflow-probability==0.14.1
```
## TLDR
To perform VI on discrete RVs, should I use:
- A- the REINFORCE gradient estimator
- B- the Gumbel-Softmax reparametrizati…
-
Hello again. Thank you for sharing your work!
I have carefully read your paper (and looked through your code), but I fail to understand how LM priors are actually calculated. (Going from lstm logit…
-
After reading the unsupervised part of the paper, I can not figure out what exactly the structures of unsupervised part is.
By the way, thanks very much for your replementation of FCSN ans VSULD, I…
-
I've noticed that pyro.distributions.RelaxedOneHotCategorical tends to underflow pretty dramatically if you decrease the temperature below 0.3 or so with many categories. I've been adding a slight mod…