Janos Gabler

Thanks for the detailed issue and debugging! I think this is a general problem that occurs if a query or selector uses parameter values. The reason why we replace params...

We use the "value" field because it avoids rewriting tree flattening rules. But we might be able to avoid it. The whole problem would be trivial to fix if we...

Yes, I would run this on your start params before you pass the constraints to estimagic.

I am very surprised. We define helper as: ```python n_params = len(tree_converter.params_flatten(params)) helper = tree_converter.params_unflatten(np.arange(n_params)) ``` This is before any constraints handling is done, i.e. at this point we do...

To explain a bit more: We always treat parameters as a general pytree that can be flattened and unflattened. Even if params is just one DataFrame. A solution like: ```python...

As you say, this is sufficiently rare that we don't need to support it. Moreover, it's made partially obsolete with #495 where we deprecate queries as a means to select...

Hi @Kaniyaki, looks great. I would replace "Invalid model specification" by "Violated constraint at start params". Estimagic is a very general library and not everything you optimize could be called...

This is a very important usecase for us and we should offer a batch evaluator that support jax functions. It's not just for multistart but also for bootstrap or parallelizing...

> The link to the logo in the navigation bar is not being found. Using the [pandas dark mode logo](https://github.com/pandas-dev/pandas/blob/main/doc/source/conf.py#L250) and our [standard estimagic logo](https://github.com/OpenSourceEconomics/estimagic/blob/main/docs/source/_static/images/estimagic_logo.svg) as dummies worked. Maybe, once...

Hi Alan, thanks for the PR. I think this would be nice to have in estimagic. As Tim said, we are planning major refactorings and would like to have few...