この例では、dpu_subgraph という名前の DPU サブグラフから DPU ランナーを作成します。
# create DPU runner
dpu_runner = vart.Runner.create_runner(dpu_subgraph, "run")
# get a list of runner inputs
inputTensors = dpu.get_input_tensors()
# optional – print names and shapes of each input tensor
for inputTensor in inputTensors:
print('Input tensor :',inputTensor.name, inputTensors.dims)
# create input buffer
# Important: Order of values passed to DPU thru’ input data buffer must match the order of tensor objects returned by get_input_tensor()
inputData = []
for inputTensor in inputTensors:
inputData.append(some_input_data.reshape(inputTensor.dims))
# pass input buffer to DPU runner, launch and wait for completion
job_id = dpu_runner.execute_async(inputData,outputData)
dpu_runner.wait(job_id)