a problem during training decoder in train_ Decoder.py

[aqinofear]

Dear Professor:
Sorry to be a bother. I am a student of XiDian University. I read your paper titled as "Zero-Shot Out-of-Distribution Detection Based on the Pre-trained Model CLIP" and met a problem.
After I download your code and configure the environment, CLIP model, and dataset, I run train_ Decoder.py. when the model is trained for the second epoch, the calculation data from the previous training is not cleared and continues to occupy the CUDA memory, resulting in an error: CUDA out of memory.
So I wonder if there is a mistake I need to modify in the program, and I want to ask for the required graphics card capacity to successfully run the model training and testing programs.
I will be appreciated for your reply, thank you very much!