search

continuedev / continue

⏩ Continue is the leading open-source AI code assistant. You can connect any models and any context to build custom autocomplete and chat experiences inside VS Code and JetBrains
https://docs.continue.dev/
Apache License 2.0
14.93k stars 1.1k forks source link

Rename `llm` abstraction `model` #30

Closed TyDunn closed 1 year ago

sestinj commented 1 year ago

@TyDunn I started using StarCoder yesterday and so wanted to group all llms/models under some namespace in the SDK. So it's now called sdk.models.starcoder/gpt35/etc.... So developers will primarily interact with the term "model." I've kept the LLM class because those are defined by capabilities, and nothing stops us from putting a different class in the models namespace. It would be misleading I think to name it anything else right now, because the capabilities are exactly that of a LLM, not more, and this is the most common naming convention (see https://github.com/jerryjliu/llama_index/tree/main, https://github.com/Significant-Gravitas/Auto-GPT, https://github.com/hwchase17/langchain, https://github.com/getcursor/cursor)

TyDunn commented 1 year ago

This is a good compromise. After your points yesterday, I started using LLMs all over the docs for now. We are far from a standardization here. OpenAI (calling GPT-4 a Large Multimodal Model) and Stanford (naming their research center CRFM), for example, will likely have more of a say in the ultimate name than many of those projects listed (though I agree let's align with them for now)