Hi There !
i’m sorry if this been asked many times (i’m on it since yesterday and crawled on google since then :x )
I’m trying to vaccum clean VRAM on a small cuda device within a notebook, and then re-using it (reloading a model in it right after resetting it) but unfortunately i got this error each time :
RuntimeError: CUDA error: invalid argument
to be more detailed here is what I do :
device = numba.cuda.get_current_device()
device.reset()
torch.cuda.is_available()
numba.cuda.detect()
which gives me :
Found 1 CUDA devices
id 0 'GeForce GTX 1080 Ti' [SUPPORTED]
Compute Capability: 6.1
PCI Device ID: 0
PCI Bus ID: 1
Watchdog: Enabled
FP32/FP64 Performance Ratio: 32
Summary:
1/1 devices are supported
then I do :
cuda.select_device(0)
and try to load again my model in it but got the
RuntimeError: CUDA error: invalid argument
So is it possible to reset a device and re-use within a notebook process ?