⚡ Build your chatbot within minutes on your favorite device; offer SOTA compression techniques for LLMs; run LLMs efficiently on Intel Platforms⚡

Large-scale LLM inference engine

Official Implementation of EAGLE-1 (ICML'24) and EAGLE-2 (EMNLP'24)

scalable and robust tree-based speculative decoding algorithm

Code for "LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding", ACL 2024

[COLM 2024] TriForce: Lossless Acceleration of Long Sequence Generation with Hierarchical Speculative Decoding

REST: Retrieval-Based Speculative Decoding, NAACL 2024

[NeurIPS'23] Speculative Decoding with Big Little Decoder

Codes for our paper "Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generation" (EMNLP 2023 Findings)

Implementation of the paper Fast Inference from Transformers via Speculative Decoding, Leviathan et al. 2023.

SWIFT: On-the-Fly Self-Speculative Decoding for LLM Inference Acceleration

PipeInfer: Accelerating LLM Inference using Asynchronous Pipelined Speculation

minimal C implementation of speculative decoding based on llama2.c

Accelerating LLM inference with techniques like speculative decoding, quantization, and kernel fusion, focusing on implementing state-of-the-art research papers.

Implementation of Speculative Sampling in "Accelerating Large Language Model Decoding with Speculative Sampling"

Verification of the effect of speculative decoding in Japanese.

Dynasurge: Dynamic Tree Speculation for Prompt-Specific Decoding

一个简单的投机推理实现

Unofficial implementation of Token Recycling self-speculative decoding method.

Reproducibility Project for [NeurIPS'23] Speculative Decoding with Big Little Decoder