Questions about params

Hi, thanks for your great work! We’re trying to reproduce the model parameter count reported in your paper. We're using the standard PyTorch approach: total_params = sum(p.numel() for p in model.parameters()) / 1e6 # in millions However, the result differs significantly from the numbers reported in your paper. Could you please clarify how you computed the parameter count? If possible, would you mind sharing the code or method you used for parameter calculation? Thanks in advance!