microsoft/DialoGPT-medium

Hello, I am experiencing a similar issue. The following message shows up in between each response:

"A decoder-only architecture is being used, but right-padding was detected! For correct generation results, please set padding_side='left' when initializing the tokenizer."

Technically, I am using a fine-tuned version that I trained using the DialoGPT-large checkpoint, but when I saw this post, I felt it was best to post this here. I have tried adding the padding_side during the fine-tuning training, in the from_pretrained instantiation of the tokenizer in the chatbot, as well as adding it as an argument in the tokenizer.encode method.

Any assistance would be greatly appreciated!