Jan Kaiser
Jan Kaiser
We are getting to the point where `accelerator.py` is getting quite large and it might soon make sense to split it into separate files. This is the place to discuss...
It came up that adding a Cheetah-native interface directly into [distgen](https://github.com/ColwynGulliford/distgen) would make it very easy to generate all kinds of beam distributions. **This issue is to discuss the design...
Elements have a `is_skippable` property, which Cheetah uses to determine whether its linear transfer map can be merged with other transfer maps adjacent to the element. This name may be...
The image link on PyPI is broken and does not show the Cheetah logo. It would be nice to have that fixed. [Stable Baselines3](https://pypi.org/project/stable-baselines3/) seems to fix the same problem...
In #116, we added a vectorised way of using Cheetah. This means that instead of either one of these ```python elem1 = Drift(length=0.3142) elem2 = Drift(length=torch.tensor(0.3142)) ``` you have to...
Currently, Cheetah only considers `"cpu"` and `"cuda"` as possible devices. It would be nice to also make use of `"mps"` on Apple Silicon Macs. First, we need to test that...
Add this link somewhere in the README: https://github.com/desy-ml/cheetah-demos
It should be possible, like with any other PyTorch Module, to move Elements and Beams by calling ```python element = element.to(device) ``` or ```python element = element.cpu() ``` or ```python...
Adding type checking would improve the code quality even more. It would probably take a moment to make sure the type checking passes though.