LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding
Paper • 2410.03355 • Published
How to use jadohu/llamagen2_drafter with Transformers:
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("jadohu/llamagen2_drafter")
model = AutoModelForCausalLM.from_pretrained("jadohu/llamagen2_drafter")This repository contains the official PyTorch implementation of the paper LANTERN: Accelerating Visual Autoregressive Models via Relaxed Speculative Decoding.