Recurrent Drafter for Fast Speculative Decoding in Large Language Models
Reviewed by Pith T0 review T1 audit T2 compute T3 formal T4 kernel pith:UYTKMATYrecord.jsonopen to challenge →
read the original abstract
We present Recurrent Drafter (ReDrafter), an advanced speculative decoding approach that achieves state-of-the-art speedup for large language models (LLMs) inference. The performance gains are driven by three key aspects: (1) leveraging a recurrent neural network (RNN) as the draft model conditioning on LLM's hidden states, (2) applying a dynamic tree attention algorithm over beam search results to eliminate duplicated prefixes in candidate sequences, and (3) training 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. To demonstrate its practicality in real environments, we also validated its effectiveness for on-device applications by implementing the approach in MLX and benchmarking performance on Metal GPUs in Apple Silicon chips, achieving up to 2.3x speedup.
This paper has not been read by Pith yet.
Forward citations
Cited by 6 Pith papers
-
TAPS: Target-Aware Prefix Tree Selection for Diffusion-Drafted Speculative Decoding
TAPS converts diffusion marginal probabilities into path-conditioned acceptance estimates to select prefix-closed subtrees under a fixed verification budget, achieving up to 7.9x end-to-end speedup over autoregressive...
-
SpecBlock: Block-Iterative Speculative Decoding with Dynamic Tree Drafting
SpecBlock achieves 8-13% higher mean speedup than EAGLE-3 at 44-52% drafting cost via block-iterative drafting with hidden-state inheritance, dynamic rank-head branching, valid-prefix masking, and optional cost-aware ...
-
SpecBlock: Block-Iterative Speculative Decoding with Dynamic Tree Drafting
SpecBlock achieves 8-19% higher speedup than EAGLE-3 in LLM speculative decoding by using repeated block expansions with hidden-state inheritance, a dynamic rank head, and a valid-prefix training mask.
-
SnapKV: LLM Knows What You are Looking for Before Generation
SnapKV selects clustered important KV positions per attention head from an observation window at the prompt end, yielding 3.6x faster generation and 8.2x better memory efficiency on 16K-token inputs with comparable pe...
-
D-PACE: Dynamic Position-Aware Cross-Entropy for Parallel Speculative Drafting
D-PACE derives per-position weights from a surrogate of expected accepted draft length to shift training focus toward currently limiting positions, yielding measured gains in wall-clock speedup and emitted length acro...
-
Token-Operations-Oriented Inference Optimization Techniques for Large Models
The paper introduces a four-layer technical architecture for token-operations-oriented inference optimization in large models and reviews key technologies and industry status at each layer.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.