how to use pretrained llm from hugging face

If my finetuned model involves an extra step say mean pooling of the model output.It is saved into hugging face repositry, how can I use this model in say Sub Question Query Engine

For the purpose of fine tuning embedding, my textnode has both metadata and text. So, when preparing training and validation dataset for fine tuning metadata+text will pair with questions to form a datapoint or will it only be text?

For fine tuning my embedding model I need to send custom prompts regarding the data contained in nodes and what I kind of questions I want to be generated. How can I do this using generate_qa_embedding_pairs?

I have a different finetuned embedding model than my fine tuned llm. How can I connect these two in SubQuestionQueryEngine such that my index retrieves relevant documents rather than embeddings and send the relevant documents to llm?

@kapa.ai You need to use a chat model in order the use role blocks like with user():! Perhaps you meant to use the TransformersChat class? How can I deal with this error? I am using the llama-index-question-gen-guidance library?

I am installing llama-index-question-gen-guidance in order to use guidance with an open-source LLM (https://docs.llamaindex.ai/en/stable/examples/output_parsing/guidance_sub_question.html) but upon execution of from llama_index.question_gen.guidance import GuidanceQuestionGenerator I get the error ImportError: cannot import name 'LLM' from 'llama_index.core.llms. Installing llama-index-question-gen-guidance also breaks my llama_index installation. Any advise?

@kapa.ai I am getting a "No valid JSON found in output" ValueError in subquestionoutputparser.parse(self, output). What might be the reason?

@kapa.ai how can I run hugging face embedding model over multiple gpus

@kapa.ai @kapa.ai Where to find VectorStore module source code. It is being imported from llama_index.core? I need to adjust context_window?

@kapa.ai I am getting the following error when extracting top 5 indices from an index query engine wrt a query
ValueError: Calculated available context size -42 was not non-negative.