AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |
Back to Blog
![]() If you are just looking to quickly play with the models they are your best shot. ![]() These experimental variants have been finetuned on instructions and conversational data they thus lend better to popular assistant-style tasks. TII has also made available instruct versions of the models, Falcon-7B-Instruct and Falcon-40B-Instruct. (Later in this blog, we will discuss how we can leverage quantization to make Falcon-40B accessible even on cheaper GPUs!) On the other hand, Falcon-7B only needs ~15GB, making inference and finetuning accessible even on consumer hardware. The 40B parameter model currently tops the charts of the Open LLM Leaderboard, while the 7B model is the best in its weight class.įalcon-40B requires ~90GB of GPU memory - that’s a lot, but still less than LLaMA-65B, which Falcon outperforms. The Falcon family is composed of two base models: Falcon-40B and its little brother Falcon-7B. ![]() ![]() In this blog, we will be taking a deep dive into the Falcon models: first discussing what makes them unique and then showcasing how easy it is to build on top of them (inference, quantization, finetuning, and more) with tools from the Hugging Face ecosystem. September 2023 Update: Falcon 180B has just been released! It's currently the largest openly available model, and rivals proprietary models like PaLM-2. ![]()
0 Comments
Read More
Leave a Reply. |