repoGitHubTrust 82 · PrimaryPublished 1mo agoLive · 1mo ago
lightseekorg/TorchSpec
A PyTorch native library for training speculative decoding models
Lineage graph
Paper → model → repo connections mined from source citations (Tier-1 exact match).
Why these links exist
Every edge carries a method, confidence, and the source snippet that justified it — so bad links are debuggable.
- PossiblePossibly related (embedding) · 66%Speculative decoding with draft models →
- PossiblePossibly related (embedding) · 61%BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding →
- PossiblePossibly related (embedding) · 59%DSpark: Speculative decoding accelerates LLM inference [pdf] →
- PossiblePossibly related (embedding) · 59%[Research] JetSpec: Speculative Decoding with Parallel Tree Drafting Enables up to 9.64x Lossless LLM Inference Speedup with more than 1000TPS →
- PossiblePossibly related (embedding) · 53%Spec-AUF: Accept-Until-Fail Training under Train-Inference Misalignment for Masked Block Drafters →
- PossiblePossibly related (embedding) · 59%DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation →
- PossiblePossibly related (embedding) · 57%DominoTree: Conditional Tree-Structured Drafting with Domino for Speculative Decoding →
- PossiblePossibly related (embedding) · 67%A Practical Investigation of Training-free Relaxed Speculative Decoding →
Implements
Covers
Implements (incoming)
paperDSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive GenerationpaperDominoTree: Conditional Tree-Structured Drafting with Domino for Speculative DecodingpaperA Practical Investigation of Training-free Relaxed Speculative DecodingpaperLess Experts, Faster Decoding: Cost-Aware Speculative Decoding for Mixture-of-Experts
Covers (incoming)
Related across the graph
newsmodel: add Hy3 (hy_v3) support with MTP speculative decoding by satindergrewal · Pull Request #25395 · ggml-org/llama.cpppaperBlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative DecodingpaperDominoTree: Conditional Tree-Structured Drafting with Domino for Speculative DecodingnewsGPT-2 Fully Decoded Internally Black Box Fully Open With Demonews[Research] JetSpec: Speculative Decoding with Parallel Tree Drafting Enables up to 9.64x Lossless LLM Inference Speedup with more than 1000TPSpaperDSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive GenerationnewsDSpark: Speculative decoding accelerates LLM inference [pdf]newsFastest speculative decoding for qwenpaperSpec-AUF: Accept-Until-Fail Training under Train-Inference Misalignment for Masked Block DrafterspaperA Practical Investigation of Training-free Relaxed Speculative DecodingpaperSpeculative decoding with draft modelspaperLess Experts, Faster Decoding: Cost-Aware Speculative Decoding for Mixture-of-Experts