What’s the difference between torch.nn.quantized.FloatFunctional.add and torch.nn.quantized.FloatFunctional.add_scalar? Is the former one used for tensor + tensor and the latter one used for tensor + scalar ? Or anything other?
Yes, this is correct. You can check this out in the source code, here: https://github.com/pytorch/pytorch/blob/b89827b73f7881a8108518452768654894862e5d/torch/nn/quantized/modules/functional_modules.py#L42