Hello, I’m trying to run a yolov8 custom trained model with sentis for an android app,and i managed to run it on cpu, but when i try to run it on gpu it throw everytime the error
Tensor data cannot be read from, use .CompleteOperationsAndDownload() to allow reading from tensor.
But I think I’m doing everything, ReadbackRequests, calling CompleteOperationsAndDownloadAsync, and it tried with both async and sync versions. What can be the issue here? If code is needed, im glad to share it
ok maybe i found a solution. I was trying to access an empty tensor at a given index, on cpu it doesn’t complain, but on gpu it throws that error. I leave it unresolved for a little more just in case i find out it’s not really fixed
Did you use CompleteOperationsAndDownload to mutate your gpu tensor back on the cpu?
Yess, i tried every possible download function, CompleteOperationsAndDownload/Async, complete all pendings etc… But yeah, now if i run it in a single frame it work, but still at a kinda low framerate, so im trying to use the multi layer with some issue too… But for that i think im gonna open a new thread, tensors are empty T_T Im setting this as solved, the main problem was that i was trying to access a null element of the tensor
Do you want to make a simple repro case so we can debug it?
You must be overriding the tensor somehow. I doubt that CompleteOperationsAndDownload doesn’t work