load .pickle file:utf-8 invalid start byte

when I am running the evaluate script of my model,I came across this:

Traceback (most recent call last):  File "/mypath/evaluate.py", line 82, in <module>    main()  File "/mypath/evaluate.py", line 66, in main    preds = pickle.load(open(args.preds_file, "rb"))            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^UnicodeDecodeError: 'utf-8' codec can't decode byte 0x95 in position 152: invalid start byte

I browse the internet, try to change add an encoding parameter,but still says utf-8 invalid.

Traceback (most recent call last):  File "/home/test/test/ReProver-main/retrieval/evaluate.py", line 82, in <module>    main()  File "/home/test/test/ReProver-main/retrieval/evaluate.py", line 66, in main    preds = pickle.load(open(args.preds_file, "rb"),encoding='ascii')            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^UnicodeDecodeError: 'utf-8' codec can't decode byte 0x95 in position 152: invalid start byte

I was trying to replicate others' model,they trained the model WITH A SINGLE GPU,while I train and predict with 8 GPUS.Is that why I got wrong because I generate the wrong pickles?