Setting NPU as Default Device
The TF Adapter provides an API for registering an NPU as a valid TensroFlow device. You can add the following code to the beginning of the script file that is decorated with @tf.function and can work properly under the CPU or GPU to set an NPU as the default device:
1 2 3 |
import npu_device as npu # By default, device 0 is a compute device. npu.open().as_default() |
Perform this step before importing other Python modules. Otherwise, operators might be scheduled to devices other than your NPU.
Parent topic: Manual Porting