FEATURE: load preprocessed file

[StempelLab]

Feature you'd like to see:

Dear Devs,
in the GUI version, when fine-tuning sorting parameters for very long recordings (or any for that matter), it would be useful if the data reading and preprocessed data writing could be skipped, so the program runs on the temp.dat file directly. skipping preprocessing steps.

Additional Context

No response

[jacobpennington]

I'll add this to the todo list, but it may take some time to get to.

[jacobpennington]

Just to clarify in the meantime: where are you hoping to save on runtime with this? Unlike previous version of Kilosort, Kilosort4 does not rely on temp_wh.dat for later steps of the pipeline. That file is only provided as an optional feature for visualization in Phy. All preprocessing steps are computed on the fly every time, so re-using temp_wh.dat would only skip the drift correction step which is typically around 1/4 of the total runtime.