synchronize

paddle.device. synchronize ( device: PlaceLike | None = None ) → None [source]

Wait for the compute on the given device to finish.

Parameters: device (str|paddle.CUDAPlace(n)|paddle.XPUPlace(n)|paddle.CustomPlace(n)) – The device which want to wait for. If device is None, the device is the current device. Default: None. It can be gpu, gpu:x, xpu, xpu:x, custom_device, custom_device:x, where custom_device is the name of CustomDevice, where x is the index of the GPUs, XPUs. And it can be paddle.CUDAPlace(n) or paddle.XPUPlace(n) or paddle.CustomPlace(n).

Examples

           >>> 
>>> import paddle

>>> paddle.set_device('custom_cpu')
>>> paddle.device.synchronize()
>>> paddle.device.synchronize("custom_cpu:0")
>>> place = paddle.CustomPlace('custom_cpu', 0)
>>> paddle.device.synchronize(place)

          