completion_to_prompt and messages_to_prompt, but those using a HuggingFaceLLM() seem to use a system_prompt and query_wrapper_prompt. How do I migrate from the former to the latter correctly?device_map had weights offloaded to the disk. Please provide an offload_folder for them. Alternatively, make sure you have safetensors installed if the model you are using offers the weights in this format."from langchain.llms import GPT4All and chat with it directly, but it seems index = GPTVectorStoreIndex.from_documents(documents, service_context=service_context) is specifically for OpenAI?)llm = HuggingFaceLLM( ... tokenizer_name="meta-llama/Llama-2-7b-chat-hf", model_name="meta-llama/Llama-2-7b-chat-hf", ...)
state_dict or a save_folder containing offloaded weights.". I've tried specifying an empty save_folder right in the HuggingFaceLLM() call, but that's an unexpected keyword, and I've also tried adding it to generate_kwargs={} and tokenizer_kwargs={} without success. I suspect it's not just looking for a blank folder, either. Any ideas?llm = HuggingFaceLLM(), where do I set trust_remote_code?llm = HuggingFaceHub(...), but (a) I seem to still need a local embedding model? and (b) Even when I use a local embedding model, I get "Empty Response" in an app where using llm = GPT4All(...) works well.HuggingFacePipeline.from_model_id() alongisde HuggingFaceEmbeddings() and pass that as the ServiceContext to a GPTVectorStoreIndex.from_documents().as_query_engine(), but I'm getting a few lines of sensible responding followed by a bunch of repetition and nonsense. Not sure if I just need to tweak parameters and response length, or if I'm producing Frankenstein's Monster here.index.as_query_engine. However, it seems that each question is distinct (e.g., there's no continuity from message to message). Is there a way to begin with the prompt, and then ask follow-up questions in this context? A la ChatGPT?ImportError: cannot import name 'GPTSimpleVectorIndex'