Colin Carroll

> Did you mean add it to this? [test_model_graph.py](https://github.com/pymc-devs/pymc/blob/main/pymc/tests/test_model_graph.py) Yes! apologies. A test or two would be super helpful here. Right now formatting is causing the test suite to fail...

Test looks good -- it looks like it mostly verifies that this works on certain inputs, and I'm ok for now with that (this seems like a cool feature, and...

:( thanks for bearing with it -- it looks like it is still a little mucked up. Sorry I can't offer more guidance - I haven't been contributing a ton...

I think the right way to do this would be to build a logisitic regression model that will be very similar to the linear regression model [here](https://www.tensorflow.org/probability/examples/JointDistributionAutoBatched_A_Gentle_Tutorial) using a joint...

What version are you seeing that on? In a fresh colab (running tensorflow 2.7.0 and tensorflow_probability 0.15.0), this outputs 

Thanks for following up! It looks like this is a known reason not to subclass `Lambda` (see https://www.tensorflow.org/api_docs/python/tf/keras/layers/Lambda), but I'm not immediately sure the solution here - maybe someone else...

Thanks, Chris! I started taking a look at this, but feel free to ping in a few days if there are no comments.

Thanks for providing some code! Can you point out which parts of the pastebin you're running, what the output looked like, and what you were expecting? I'm not quite sure...

Hey Matt! I can reproduce, and add that it also happens with the nightly tensorflow/tfp, and with the release candidate and tensorflow==2.5.0, but I don't yet have good leads on...

It looks like this should be running multiple chains just fine -- is there a reason you think it is not? A few tips on performance -- 1. You might...