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.

[Blazeolmo]

What ze hell does that mean?

[kiasar]

My question too!

[Choitsugun]

That is used to solve the problem of unaligned generation caused by padding during the batch inference of the decoder-only model.

[Vectorrent]

I think this must be a hard-coded warning, coming from the upstream Transformers package. No matter how hard I've tried, I can't seem to suppress the message.