Parameter

A parameter contains a value which can be optimized by an optimizer when training. The node will be filled with some data, for example randomly generated or set to a constant. Then, when training, an optimizer can modify the stored value, similar to how it can optimize a linear or Conv1D node.

You can change the data type and shape of the stored value by clicking the node and configuring it on the right panel. There, you can also configure what value should be filled initially.