Hello, so there is this transformer.Wave2Vec, “facebook/wav2vec2-large-xlsr-53” with “transformers_version”: “4.6.0.dev0”, which I used to load in my project and used it for inference. The model works pretty fine, if needed I will share the config.json file for model.bin. When I try to load_pretrain this model on another device with " “transformers_version”: “4.32.1”( again if needed will upload the config file) the model performance drops drastically. I have tried using the old config,Jason file but couldn’t achieve any success( the new device doesn’t support GPU and therefore CUDA). Don’t know what I am doing wrong here. what factor should I take into consideration, The config files are different some have attributes, the other doesn’t and the value of some attributes difffer too. Is the problem eveb rise from confiig files , if so why then when I used the old config file,even then it didn’t work.