Ignorant here. How do you take an already trained model and execute it elsewhere? Isn’t the training phase part of the whole (ongoing) simulation?
Once that is finished, you have a trained model which you can use by providing input and getting an output (with the weights frozen).
The training phase is very expensive computationally because you have to calculate the gradient of your loss function on potentially huge tensors.
The execution phase is not that expensive and commercial laptops will be likely able to run the model without any problems.