[HN Gopher] Show HN: LlamaGym - fine-tune LLM agents with online...
       ___________________________________________________________________
        
       Show HN: LlamaGym - fine-tune LLM agents with online reinforcement
       learning
        
       Author : KhoomeiK
       Score  : 227 points
       Date   : 2024-03-10 12:40 UTC (1 days ago)
        
 (HTM) web link (github.com)
 (TXT) w3m dump (github.com)
        
       | 3abiton wrote:
       | Interesting project, basically a wrapper too around openai gym-
       | like functionality that can handle open llms.
        
         | KhoomeiK wrote:
         | Yup, it does simplify LLM agent inference on Gym environments
         | but the main technical contribution is reducing your would-be
         | code overhead for online RL
        
       | raidicy wrote:
       | Thanks for creating this!
        
       | KhoomeiK wrote:
       | Twitter thread:
       | https://x.com/khoomeik/status/1766805213644800011?s=46
        
       | internet101010 wrote:
       | Thank you for making this. Simplifying any aspect of RL is always
       | welcome.
        
         | KhoomeiK wrote:
         | Thanks! Yeah RL for LLMs is pretty underexplored I think beyond
         | the RLHF stuff. Pretty tough to get working tho.
        
           | dartos wrote:
           | Didn't DPO supplant rlhf?
        
       | dennisy wrote:
       | Can this be used outside of OpenAI environments? If yes I think
       | an example would be great!
        
         | KhoomeiK wrote:
         | Gymnasium is now maintained by the Farama Fpundation, an open-
         | source consortium, not OpenAI. But most RL environment work for
         | the past 5+ years has been Gym-compliant. The TextWord example
         | in the repo, for example, instantiates a Gym-style environment
         | but it doesn't import from Gymnasium (uses textworld.gym
         | instead).
        
       | neodypsis wrote:
       | Very interesting!
        
       | adawg4 wrote:
       | Thanks for making this! Helps simplify it nicely
        
       | zeroq wrote:
       | When 150 lines of boilerplate can land you the first page on HN,
       | maybe it is, in fact, the end of programming?
        
         | yen223 wrote:
         | Let's not be one of those people who measure developer
         | productivity by number of lines
        
         | KhoomeiK wrote:
         | Karpathy's micrograd [1] is literally 154 lines. Guess
         | programming ended 4 years ago.
         | 
         | [1] https://github.com/karpathy/micrograd
        
           | DSingularity wrote:
           | You think autograd is boilerplate?
        
         | klysm wrote:
         | I'm not really sure what your point is. Is it not remarkable
         | that valuable things can be done in 150 lines?
        
           | xpe wrote:
           | I agree with you. / Above, I wouldn't assume a single nor
           | clearly intended "point". Reading it I got an impression more
           | of concern, even fear. I'm guessing one underlying driver may
           | be a concern that AI is creeping into more and more
           | programming. Which is true.
        
         | yowlingcat wrote:
         | Carmack's infamous fast inverse square root was only 13 lines.
         | Measuring code by line metrics rather than its contents
         | reflects shallow, questionable comprehension.
        
           | sevagh wrote:
           | If the first line of Carmack's infamous code was "import
           | fast_inverse_square_root" from Pypi.org, it wouldn't be as
           | impressive.
        
       | kayson wrote:
       | 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?
       | 
       | [1] https://www.izzy.co/blogs/robo-boys.html
        
         | mzl wrote:
         | You could perhaps adapt the Doppel Bot slack bot from Modal
         | Labs: https://github.com/modal-labs/doppel-bot
        
       | potatoman22 wrote:
       | Could someone help me understand the kinds of things you can
       | build with this? Is this like RLHF?
        
       | katzenversteher wrote:
       | From the title I misunderstood what it does. However, now I'm
       | wondering if what I thought is was (don't ask my why I thought
       | it) is possible:
       | 
       | I have a PC that is able to run e.g. Mistral Instruct 7B Q4
       | inference with around 30 token/s.
       | 
       | How (computation and memory) expensive would it be to also run
       | backpropagation in addition to inference?
       | 
       | I'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 / after each conversation I
       | have with it anyways, it would after a while be perfectly
       | customize for me.
       | 
       | I'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.
        
         | dartos wrote:
         | Very expensive.
         | 
         | AFAIK the model can't be quantized during backprop, so right
         | there you'd need a ton of RAM.
         | 
         | Backprop is faster bc it can be parallelized, but IIRC you need
         | to hold an entire copy of the model for each backprop process.
        
           | scribu wrote:
           | Actually, there have been attempts to do quantized backprop,
           | but not sure how successfully.
        
       | SuhanaJabin wrote:
       | Simplified the concept. Nicely done!
        
       ___________________________________________________________________
       (page generated 2024-03-11 23:01 UTC)