search

weizhepei / InstructRAG

InstructRAG: Instructing Retrieval-Augmented Generation via Self-Synthesized Rationales
https://weizhepei.com/instruct-rag-page
MIT License
54 stars 4 forks source link

testing #6

Open parthmahajan008 opened 4 days ago

parthmahajan008 commented 4 days ago

I want to test out the results by myself and wanted to set this up, what sort of config should i look for , for my VM. Would 2H100 in a cluster be enough ? if yes, what would be good testes service to rent these from ?

weizhepei commented 3 days ago

Yes, I think 2x H100 GPUs should be sufficient to produce the evaluation results for both the 8B and 70B models in our work. Regarding rental platforms, I don’t have personal experience with them, so I can’t make specific recommendations. After a quick online search, I found platforms like HPC-AI and Lambda seem like viable options to try out.