# Get all parameters as a state object
state = nnx.state(model)

# Or specifically just the trainable parameters
params = nnx.state(model, nnx.Param)

        __
        __