get_input_tensors - 2.0 English

Each element of the list returned by get_input_tensors() corresponds to a DPU runner input. Each list element has a number of class attributes which can be displayed like this:

inputTensors = dpu_runner.get_input_tensors()
print(dir(inputTensors[0])

The most useful of these attributes are name, dims and dtype:

for inputTensor in inputTensors:
print(inputTensor.name)
print(inputTensor.dims)
print(inputTensor.dtype)

Note that the dimensions (.dim) of an input tensor are in the form NHWC (batchsize, height,width,channels).