[FEAT]Remove generator vars from gradient of the Supervised loss in TimeGAN (as loss does not depend on it)

[itakatz]

(This is not a problem in the algorithm per-se, but more a matter of code clarity. So I do not tag it as a bug)

In the definition of the train_supervisor method:

ydata-synthetic/src/ydata_synthetic/synthesizers/timeseries/timegan/model.py

Lines 159 to 169 in 1279e5e

	def train_supervisor(self, x, opt):
	with GradientTape() as tape:
	h = self.embedder(x)
	h_hat_supervised = self.supervisor(h)
	generator_loss_supervised = self._mse(h[:, 1:, :], h_hat_supervised[:, :-1, :])
	var_list = self.supervisor.trainable_variables + self.generator.trainable_variables
	gradients = tape.gradient(generator_loss_supervised, var_list)
	apply_grads = [(grad, var) for (grad, var) in zip(gradients, var_list) if grad is not None]
	opt.apply_gradients(apply_grads)
	return generator_loss_supervised

the generator variables are added to var_list. This is not needed, as generator_loss_supervised does not depend on these variables (so gradient will be 0 and no update on these parameters will take place). This is probably inherited from the original TimeGAN implementation.

If I am correct, I suggest removing these variables from var_list, for better code clarity.