hao-ai-lab / DynasorLinks
[NeurIPS 2025] Simple extension on vLLM to help you speed up reasoning model without training.
☆197Updated 4 months ago
Alternatives and similar repositories for Dynasor
Users that are interested in Dynasor are comparing it to the libraries listed below
Sorting:
- ☆284Updated 3 months ago
- ☆85Updated 9 months ago
- Unofficial implementation for the paper "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models"☆174Updated last year
- Ouroboros: Speculative Decoding with Large Model Enhanced Drafting (EMNLP 2024 main)☆110Updated 7 months ago
- [ICLR2025 Spotlight] MagicPIG: LSH Sampling for Efficient LLM Generation☆238Updated 10 months ago
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆130Updated 10 months ago
- [ICLR 2025] DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads☆494Updated 8 months ago
- KV cache compression for high-throughput LLM inference☆142Updated 8 months ago
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆169Updated last year
- Implementation for FP8/INT8 Rollout for RL training without performence drop.☆261Updated 3 weeks ago
- Official Implementation of SAM-Decoding: Speculative Decoding via Suffix Automaton☆33Updated 8 months ago
- ☆119Updated 4 months ago
- ☆145Updated 8 months ago
- The official repo for "LLoCo: Learning Long Contexts Offline"☆117Updated last year
- SeerAttention: Learning Intrinsic Sparse Attention in Your LLMs☆158Updated last month
- REST: Retrieval-Based Speculative Decoding, NAACL 2024☆210Updated last month
- [ACL 2025 main] FR-Spec: Frequency-Ranked Speculative Sampling☆45Updated 3 months ago
- Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs☆191Updated 3 weeks ago
- Code associated with the paper **Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding**☆205Updated 8 months ago
- Async pipelined version of Verl☆121Updated 6 months ago
- [ICML 2025] Reward-guided Speculative Decoding (RSD) for efficiency and effectiveness.☆49Updated 5 months ago
- The official implementation of the paper "Towards Efficient Mixture of Experts: A Holistic Study of Compression Techniques (TMLR)".☆79Updated 7 months ago
- EE-LLM is a framework for large-scale training and inference of early-exit (EE) large language models (LLMs).☆69Updated last year
- Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference☆147Updated last week
- [CoLM'25] The official implementation of the paper <MoA: Mixture of Sparse Attention for Automatic Large Language Model Compression>☆147Updated 3 months ago
- [NeurIPS 2024] The official implementation of "Kangaroo: Lossless Self-Speculative Decoding for Accelerating LLMs via Double Early Exitin…☆60Updated last year
- Layer-Condensed KV cache w/ 10 times larger batch size, fewer params and less computation. Dramatic speed up with better task performance…☆155Updated 6 months ago
- Parallel Scaling Law for Language Model — Beyond Parameter and Inference Time Scaling☆448Updated 5 months ago
- The official implementation of paper: SimLayerKV: A Simple Framework for Layer-Level KV Cache Reduction.☆49Updated last year
- [ICLR‘24 Spotlight] Code for the paper "Merge, Then Compress: Demystify Efficient SMoE with Hints from Its Routing Policy"☆96Updated 4 months ago