Teams at the MistralAI hackathon in SF successfully trained the Mistral 7B model to play DOOM using ASCII representations of the game frames. The winning project, developed by a team including Paul Chu, won first place in the finetuning track.
Our project to run DOOM on LLMs won π₯ place in the finetuning track the @MistralAI hackathon. Yes, there are better approaches you could use to train a game playing agent, but we just HAD to find out if Mistral-7B could run DOOM. More details on what we did..
The first @MistralAI hackathon was a success, thank you to all participants! Here are the winners (links in thread): - Fine-tuned Mistral 7B playing DOOM - Optimize prompts through tests - Mistral-powered search engine for ML research - Mistral playing Pac-Man https://t.co/ldtn2ewrir
This is wild! A team trained Mistral 7B playing DOOM π€― Trained on ASCII representations - love the ingenuity and sheer creativity! Open/ acc. https://t.co/qYFgAWymWG
We built an llm that can run doom at the @MistralAI hackathon. Built with @hopingtocode, Paul chu and finetuned on @FireworksAI_HQ https://t.co/AramfzztwX
π« Badass! A team at the @MistralAI hackathon in SF trained the 7B open-source model to play DOOM, based on an ASCII representation of the current frame in the game. π€― @ID_AA_Carmack https://t.co/B2Tn8WPX1x