Closed pietro14 closed 1 year ago
Thanks Pietro. In the meanwhile the uproot developers discovered the issue, it is going to be debugged. But if the h5 format is small enough, I think we can use it.
Yes, the h5 format takes around 5 Mb per image, while ROOT 4 Mb. At least for the moment, I think it can be a temporary solution.
Any news from the uproot developers? @emanueledimarco if you opened an issue on the uproot GitHub page, could you share it here?
To solve uproot memory leak, at least for the moment, I added the option to set 'rawdata_tier' : 'h5', for simulated images (the simulation now produces h5 files). This way the leak is partially solved. The memory still increases a bit, probably due to the use of uproot when subtracting pedestals.
The code runs smoothly, the images seem to be okay: they appear in the right position and with the right shape.
I also updated the instructions to run the reconstruction on simulated images.