Ola Rønning

Support inference with RenyiELBO for local latent variable models

5

Addresses #3104 ### TODO - [x] Add Local latent variable model test case - [x] Correct RenyiELBO `logsumexp` reduction dim for LLV model

Variance reduction for (Renyi)ELBO

6

From the forum [thread](https://forum.pyro.ai/t/iwae-elbo-in-pyro-numpyro/3604/2). [Tighter Variational Bounds are Not Necessarily Better](https://arxiv.org/pdf/1802.04537.pdf) demonstrates that tighter VBs can lead to deteriorating gradient estimates for VAEs trained with [IWAE](https://arxiv.org/pdf/1509.00519.pdf) due to the variance...

Vectorized operations in `gmm_objective.py`

1

Hi, Vectorized the GMM objective function. Should be around 10x faster now.

[FR] Predictive with deterministic site in the guide

6

Hi, I'm working on a project where we would like to access the output of an NN in the guide when using `Predictive`. We've implemented it using a deterministic site...

Related to #3358.

FunctionalFlowStep

2

Why not use **kwargs here? https://github.com/stoyan-stoyanov/llmflows/blob/3c4ceb7964dd85d0ab59f7865aa37337201454fa/llmflows/flows/functional_flowstep.py#L42 Then you can name keyword args to the function (i.e., `generate`), and the remaining will be the arguments for `flowstep_fn`, so this will go...