Type demotion in pytorch

torch.compile FX (Functional Transformations)
1

I want to convert all tensors and operations in the model from int64 to int32. Is there any existing pass with which I can achieve this?