The parameter sweep function experiences several errors when attempting to use the n_frames keyword argument. When attempting to split a large data set into 100 frame batches, the function will load the entire data set, creating a memory overflow. I have been able to pass 100 frames to the parameter sweep function without any issues on my laptop with 32GB of RAM. However, attempting to batch large data sets (1000+ frames) into 100 frame chucks causes memory to exceed 75GBs on a high performance cluster.
By reducing the size of the original dataset, the function can run within memory limits; however, it displays a quirk in qnr the calculation returning the impossibly low value of 0. The resulting qnr chart is depicted below, despite moderate similarities in the reference image and the eSRRF image:
Please ignore the "1000 frame average" title the variable used in the f-string was outdated. Both the references and eSRRF image are from 10 frames.
The parameter sweep function experiences several errors when attempting to use the n_frames keyword argument. When attempting to split a large data set into 100 frame batches, the function will load the entire data set, creating a memory overflow. I have been able to pass 100 frames to the parameter sweep function without any issues on my laptop with 32GB of RAM. However, attempting to batch large data sets (1000+ frames) into 100 frame chucks causes memory to exceed 75GBs on a high performance cluster.
By reducing the size of the original dataset, the function can run within memory limits; however, it displays a quirk in qnr the calculation returning the impossibly low value of 0. The resulting qnr chart is depicted below, despite moderate similarities in the reference image and the eSRRF image:
Please ignore the "1000 frame average" title the variable used in the f-string was outdated. Both the references and eSRRF image are from 10 frames.