To load the MNTP trained model using LLM2Vec.from_pretrained in SimCSE face, it errors saying that no config.json file is present in the MNTP training output.
I see that this file was manually created here: https://huggingface.co/McGill-NLP/LLM2Vec-Meta-Llama-3-8B-Instruct-mntp/blob/main/config.json.
The save function run_mntp does not have the code to save the config: def _save(self, output_dir: Optional[str] = None, state_dict=None):
Or is the expectation that we copy/paste the base model's config.json as is post this training?