Get Input/Output Names

 Exec Inputs/Outputs Names : cluster, this cluster defines the tensor names used during a full execution cycle of the model, including the forward pass and loss/backward computation.

Forward Inputs Names : array, list of input tensor names required to execute the forward pass. These typically represent the data fed into the model during inference or training forward.
Forward Outputs Names : array, list of output tensor names produced by the forward pass. These are generally the raw predictions generated by the model, before any loss is computed.
Loss/Backward Inputs Names : array, list of tensor names used to compute the loss and perform the backward pass. This usually includes the ground truth labels, and may also include additional inputs from the loss subgraph such as sample weights or auxiliary values.