Dieter Weber

@sivborg and @magnunor are using Zarr for their 4D STEM work with HyperSpy/pyxem, so this could perhaps be moved up in priority? :-) Refs #1277

JSON export of https://zenodo.org/records/10091431 where I fixed the license manually [10091431.json](https://github.com/LiberTEM/LiberTEM/files/13309464/10091431.json)

Update: See #1624, correct is GPL 3 only

The installation with current torch version runs with Win 11 and latest LiberTEM on Python 3.11. However, I did not confirm that it is actually using torch. I ran libertem-server...

Maybe include checks if a block is in the file system cache? http://man7.org/linux/man-pages/man1/fincore.1.html For example, have dedicated workers for cached files that will not load IO, but heavily load memory...

@sk1p 👍 for testing! Actually, the point here is that the IO-heavy and CPU-heavy tasks can run efficiently in _parallel_, using hyperthreading and the scheduling of the operating system. What...

Another point is allowing file browsing while a computation is running. Currently, the file browser lags/freezes while a computation is running because it needs to run code on the server...

As a comment, we have service workers in a Dask cluster now that improve the file browsing responsiveness. We support threaded workers, including a mixture of threads and processes. Also,...

This hasn't been a big issue since we process data from only one source at a time in real-world use cases. Closing until this comes up as a real-world issue.

About the older Python versions: Are they still relevant?