When running python (or TACO) code, some of the suitesparse tensors appear to cause OOM's while either loading or running operations on them. These ooms result in SLURM killing the job and us losing all progress in the benchmark. Since there are 3000 of them, doing something by hand isn't going to work. We need some sort of systematic way to exclude tensors that are too large to run (and do this consistently between taco and numpy).
When running python (or TACO) code, some of the suitesparse tensors appear to cause OOM's while either loading or running operations on them. These ooms result in SLURM killing the job and us losing all progress in the benchmark. Since there are 3000 of them, doing something by hand isn't going to work. We need some sort of systematic way to exclude tensors that are too large to run (and do this consistently between taco and numpy).