Large Language Models
FAQ
1. How to define a custom LLM?
You can access Usage Custom to define a custom LLM.
2. How to use a different OpenAI model?
To use a different OpenAI model you can access Configure Model to set your own custom model.
4. Is it required to fine-tune my model?
No. there’s isolated modules which might provide better results, but isn’t required, you can use llamaindex without needing to fine-tune the model.
5. I want to the LLM answer in Chinese/Italian/French but only answers in English, how to proceed?
To the LLM answer in another language more accurate you can update the prompts to enforce more the output language.
response = query_engine.query("Rest of your query... \nRespond in Italian")
Alternatively:
from llama_index import LLMPredictor, ServiceContext
from llama_index.llms import OpenAI
llm_predictor = LLMPredictor(system_prompt="Always respond in Italian.")
service_context = ServiceContext.from_defaults(llm_predictor=llm_predictor)
query_engine = load_index_from_storage(
storage_context, service_context=service_context
).as_query_engine()
6. Is LlamaIndex GPU accelerated?
Yes, you can run a language model (LLM) on a GPU when running it locally. You can find an example of setting up LLMs with GPU support in the llama2 setup documentation.