Hello, when I was doing model inference, under the nnUNetV1-based project, an error occurred when reading the model you provided, whether it was small, base, large or huge, the specific error was in 'all_params = [torch The error message displayed at .load(i, map_location=torch.device('cpu')) for i in all_best_model_files]' is: '_pickle.UnpicklingError: invalid load key, '%'.', that is, to read the provided **_ep4k.model.
I would like to ask, according to the paper mentioned that stunet is trained on a graphics card with 80G memory, my device is an RTX4090 with 24G memory, is it caused by insufficient memory (but there is no such prompt in the nnunet error message), or because of the model The data is caused by the destruction of upload and download. Thank you so much.