Show HN: LlamaGym – fine-tune LLM agents with online reinforcement learning

I want to make a Discord bot that impersonates all my friends and continues to refine the model as the conversations continue. Basically this [1] post, but with a more modern model and, ideally, reinforcement learning. Seems like this would fit the bill.... Is there anything else that would make this easier?<p>[1] <a href="https:&#x2F;&#x2F;www.izzy.co&#x2F;blogs&#x2F;robo-boys.html" rel="nofollow">https:&#x2F;&#x2F;www.izzy.co&#x2F;blogs&#x2F;robo-boys.html</a>

From the title I misunderstood what it does. However, now I&#x27;m wondering if what I thought is was (don&#x27;t ask my why I thought it) is possible:<p>I have a PC that is able to run e.g. Mistral Instruct 7B Q4 inference with around 30 token&#x2F;s.<p>How (computation and memory) expensive would it be to also run backpropagation in addition to inference?<p>I&#x27;m aware that the models are typically fed with much more and better data than what is typically provided during normal conversations but on the other hand if I could finetune my local model a teeny tiny bit during during &#x2F; after each conversation I have with it anyways, it would after a while be perfectly customize for me.<p>I&#x27;m also aware that this could be problematic for models that are used by multiple users but my intended use case would be personal use by a single user.

Thank you for making this. Simplifying any aspect of RL is always welcome.

Could someone help me understand the kinds of things you can build with this? Is this like RLHF?

Can this be used outside of OpenAI environments? If yes I think an example would be great!

Twitter thread: <a href="https:&#x2F;&#x2F;x.com&#x2F;khoomeik&#x2F;status&#x2F;1766805213644800011?s=46" rel="nofollow">https:&#x2F;&#x2F;x.com&#x2F;khoomeik&#x2F;status&#x2F;1766805213644800011?s=46</a>

Thanks for making this! Helps simplify it nicely

When 150 lines of boilerplate can land you the first page on HN, maybe it is, in fact, the end of programming?

Interesting project, basically a wrapper too around openai gym-like functionality that can handle open llms.

Thanks for creating this!

llamagym.com for sale

Very interesting!

Simplified the concept. Nicely done!