Summary of Recurrent Drafter For Fast Speculative Decoding in Large Language Models, by Yunfei Cheng and Aonan Zhang and Xuanyu Zhang and Chong Wang and Yi Wang
Recurrent Drafter for Fast Speculative Decoding in Large Language Models
by Yunfei Cheng, Aonan Zhang, Xuanyu Zhang, Chong Wang, Yi Wang
First submitted to arxiv on: 14 Mar 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Recurrent Drafter (ReDrafter) is an advanced decoding approach that achieves state-of-the-art speedup for large language models (LLMs) inference. The approach uses a recurrent neural network (RNN) as the draft model, conditioning on LLM’s hidden states, and applies dynamic tree attention over beam search results to eliminate duplicated prefixes in candidate sequences. This is trained through knowledge distillation from the LLM. ReDrafter accelerates Vicuna inference in MT-Bench by up to 2.8x with a PyTorch implementation on Nvidia H100 GPUs. It was also validated for on-device applications, achieving up to 2.3x speedup on Metal GPUs in Apple Silicon chips. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Recurrent Drafter is a new way to make computers understand language faster and better. It uses special math to help big computer models work more efficiently. This approach makes large language models run 2.8 times faster than before, which is very fast! To show how it works in real life, the researchers tested it on Apple’s devices and made it run even faster, up to 2.3 times! |
Keywords
* Artificial intelligence * Attention * Inference * Knowledge distillation * Neural network * Rnn
