Jan
Jan
if `train` is called twice on the same instance, this error will be raised unless you provide `force_first_round_loss=True` **or** `resume_training=True`. Importantly, If `resume_training=True` is passed, training will resume only on...
talk to @manuelgloeckler for more details.
related to #932
another one: https://github.com/blackjax-devs/blackjax maybe we can write a wrapper to `jax` or use this workaround: https://blackjax-devs.github.io/blackjax/examples/howto_other_frameworks.html
Probably an advanced `good first issue` but it does not require knowing `sbi` really. The main task is to find an interface between our MCMC API and that of the...
solved for pymc via #1053 numpyro and blackjax not planned at the moment because we are based on `torch`
Absolutely! Please feel free to create a PR. If you haven't already, check out the contribution guidelines 😊 Thanks a lot for tackling this! 🙏
This will become even more relevant when we have a common dataloader interface and agnostic `loss` functions for all SBI methods. But I am removing the hackathon label for now...
solution for now as proposed in #848 : pin `pyright` to last passing version and fix type issues before the next major release, or when changing to `nflows`.
idea: change to `ruff` to handle type checking more flexibly: https://docs.astral.sh/ruff/ related to #928