Casting a model dict to a different dataype

Hello. I have an exported model in .pth format. It is loaded in an Ordered Dictionary format, from which I would like to separate the numeric tensors (not the key values i.e. strings) and cast them to a different datatype, say int. However, upon calling dict.values(), I get a huge output, an excerpt of which is given below.

[-7.6345e-03, -4.4007e-03, -7.9214e-03,  ..., -8.5340e-04,
           6.4599e-04, -5.3076e-04]], device='cuda:0')}, 128: {'momentum_buffer': tensor([ 0.0129, -0.0025, -0.0054, -0.0146, -0.0153,  0.0049,  0.0010, -0.0075,
          0.0060, -0.0038,  0.0023,  0.0005, -0.0026,  0.0036, -0.0013,  0.0051,
          0.0073,  0.0064, -0.0050,  0.0048,  0.0024, -0.0002,  0.0051,  0.0052,
         -0.0051,  0.0013, -0.0116,  0.0018,  0.0002,  0.0051, -0.0054,  0.0053,
         -0.0014,  0.0086, -0.0128,  0.0101, -0.0145,  0.0103,  0.0042, -0.0055],
        device='cuda:0')}}

I want to extract only the numeric values from this. How can I do that?