Even though I explicitly set the torch device to gpu-0
with the torch.cuda.set_device(0)
, I notice that my model loads in multiple GPUs.
ae = torch.load(MODEL_PATH).eval()
Is there a way to circumvent this issue?
Even though I explicitly set the torch device to gpu-0
with the torch.cuda.set_device(0)
, I notice that my model loads in multiple GPUs.
ae = torch.load(MODEL_PATH).eval()
Is there a way to circumvent this issue?