is there an argument that allows the user to specify that they want to target all available CPUs plus the GPU? is there also an argument to rank by GPU speed?
Doesn’t the current code target everything a worker node has to offer already? Also, since Xbox clusters are user-designed and homogeneous, does this really apply?
is there an argument that allows the user to specify that they want to target all available CPUs plus the GPU? is there also an argument to rank by GPU speed?