Fix autocast in 2d_ldm tutorial

[Warvito]

In the 2d LDM tutorial, the scalers for autocast are specified in

GenerativeModels/tutorials/generative/2d_ldm/2d_ldm_tutorial.py

Line 181 in 9cd35d0

scaler_g = torch.cuda.amp.GradScaler()

but they were not used.
Consider add the autocast functionality or remove the scalers.

[Warvito]

besides that, there are these lines that are not necessary for the validation loop

GenerativeModels/tutorials/generative/2d_ldm/2d_ldm_tutorial.py

Line 265 in 9cd35d0

optimizer_g.zero_grad(set_to_none=True)

GenerativeModels/tutorials/generative/2d_ldm/2d_ldm_tutorial.py

Line 344 in 9cd35d0

optimizer.zero_grad(set_to_none=True)