position-specialist-speculative-decoding/Speed-E3-Llama3.1-8B-Instruct

SPEED: Specialized Position Experts for Efficient Speculative Decoding

This repository provides the model checkpoint for SPEED, a speculative decoding method proposed in the anonymous paper:

SPEED: Specialized Position Experts for Efficient Speculative Decoding

📦 Files

• —pytorch_model.bin — model weights
• —config.json — model configuration