Rob Zinkov

Hey @rusty-robin, generally `NUTS` will be slower than `HMC` as it does strictly more work. That it's taking 1-2s per iteration is more likely due to needing a non-centred reparameterisation....

I'm happy to leave in the arviz import. Also with the contents of to_inference_data merged, we could remove it as well. On Thu, 15 Dec 2022, 20:49 Xiaoyan Wang, ***@***.***>...

Are the current tests failing suppose to fail?

Can we reuse how numpy implements tensordot? On Tue, 7 May 2024, 11:52 Ricardo Vieira, ***@***.***> wrote: > All cases except those requiring tensordot with batch dims not on the...

Sure thing! This license complaint isn't something I was able to replicate on my end. Would an empty file be sufficient here?

Oh I can do that! On Wed, 12 Oct 2022, 05:44 Xiaoyan Wang, ***@***.***> wrote: > :) By "license" the linter is referring to the few lines of comments at...

Yes, these comments and suggestions are very good! I think they should all be integrated

Just for clarity, we don't have any sections of the code that use a tensor of RNG seeds right?

Looks like some changes still needed till tests pass.

Oh ok. I know it's not the most essential for this PR, but why does the Scan require a shared variable as an argument? On Thu, 28 Mar 2024, 22:35...