Devaansh Gupta

Not stale, still looking forward to a response!

Hey! While the output of the tokenizer is correct(both input_ids and labels in the same format), the labels are going to pass through the `shift_tokens_right` to create the `decoder_input_ids`. The...

I agree with @LoicGrobol. I also want to clarify this example from the [docs](https://huggingface.co/docs/transformers/model_doc/mbart#training-of-mbart50): ```python from transformers import MBartForConditionalGeneration, MBart50TokenizerFast article_hi = "संयुक्त राष्ट्र के प्रमुख का कहना है कि...

Got it! I guess the only change then would be in the "labels" from the tokenizer then - where there was LANGID initially, we would have a 1/-100?

Thank you for all the help!

Hi! I'm not the author but I may be able to help with some of the questions here. > Set the "model_name_or_path" parameter to "liuhaotian/llava-v1.5-13b" to initialize the model weights...

I'm not completely sure how that would work. Ideally, to load the entire HuggingFace model, you would use `load_pretrained_model` from `llava/model/builder.py`. However, that is not used during training. One hack...