Simeon Carstens

I assume it is still not possible to expose the log-likelihood gradient w.r.t. the parameters in, say, the `transformed parameters` block? If so, I'd love to +1 this issue -...

Oh wow, I did not know about BridgeStan! That looks super useful for my use case - I am currently using `httpstan` (and not even `pystan` :facepalm:) to access the...

Thanks a lot, @bob-carpenter! Good to have a definite answer and to know that this is not directly possible. But even if in fact a bit messy, I think your...

Same here, connecting from France. Some info on this would be highly appreciated :-) The function raising this error is `spot4e-refresh`.

We (@zz1874, I) wrote a hacky demo script that uses a CLI based on `pydantic-argparse` to run the [`hello_looper`](http://looper.databio.org/en/latest/hello-world/) example. You can find it on a [branch](https://github.com/tweag/looper/tree/tweag/pydantic-argparse-demo) in our fork...

Going forward, I (this has not been discussed with neither @nsheff nor @zz1874) propose to organize the `pydantic`-based CLI rewrite as follows: At the core, we introduce a `dataclass` called...

@donaldcampbelljr interesting issue! I investigated a bit and the problem seems to be that `pydantic-argparse` doesn't know how to distinguish between list elements (such as "failed" and "running") and the...

Another option is to monkey-patch the`pydantic2-argparse`. For example, we could require all list arguments to be given as comma-separated strings, and then add the following in `cli_pydantic.py` right after the...

Putting the majority of shared arguments after the subcommand would also be possible, of course. As you say, there's then a duplication of arguments across various subcommands. But then again,...

I'm sorry about the late realization that short arguments are not supported by `pydantic-argparse`! We probably should have pointed that out more explicitly, rather than only in a comment on...