tsinghua-ideal / TwilightLinks
Adaptive Attention Sparsity with Hierarchical Top-p Pruning
☆19Updated 7 months ago
Alternatives and similar repositories for Twilight
Users that are interested in Twilight are comparing it to the libraries listed below
Sorting:
- [ICLR 2025] TidalDecode: A Fast and Accurate LLM Decoding with Position Persistent Sparse Attention☆47Updated last month
- The Official Implementation of Ada-KV: Optimizing KV Cache Eviction by Adaptive Budget Allocation for Efficient LLM Inference☆92Updated 2 months ago
- ArkVale: Efficient Generative LLM Inference with Recallable Key-Value Eviction (NIPS'24)☆43Updated 9 months ago
- ☆55Updated last year
- [COLM 2024] SKVQ: Sliding-window Key and Value Cache Quantization for Large Language Models☆24Updated 11 months ago
- 16-fold memory access reduction with nearly no loss☆105Updated 5 months ago
- ☆15Updated last year
- Pytorch implementation of our paper accepted by ICML 2024 -- CaM: Cache Merging for Memory-efficient LLMs Inference☆45Updated last year
- SQUEEZED ATTENTION: Accelerating Long Prompt LLM Inference☆52Updated 10 months ago
- [ICLR 2025] PEARL: Parallel Speculative Decoding with Adaptive Draft Length☆111Updated 5 months ago
- SpInfer: Leveraging Low-Level Sparsity for Efficient Large Language Model Inference on GPUs☆57Updated 5 months ago
- ☆57Updated 9 months ago
- ☆15Updated 6 months ago
- ☆78Updated 5 months ago
- Implement some method of LLM KV Cache Sparsity☆38Updated last year
- [NeurIPS 2024] Efficient LLM Scheduling by Learning to Rank☆59Updated 10 months ago
- PyTorch implementation of paper "Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline".☆90Updated 2 years ago
- Multi-Candidate Speculative Decoding☆36Updated last year
- [ICLR 2025] Palu: Compressing KV-Cache with Low-Rank Projection☆139Updated 7 months ago
- PoC for "SpecReason: Fast and Accurate Inference-Time Compute via Speculative Reasoning" [arXiv '25]☆49Updated 2 months ago
- Scalable long-context LLM decoding that leverages sparsity—by treating the KV cache as a vector storage system.☆80Updated last month
- ☆52Updated last year
- ☆45Updated 9 months ago
- ☆43Updated 4 months ago
- Official implementation for Yuan & Liu & Zhong et al., KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark o…☆84Updated 6 months ago
- ☆71Updated last year
- [NeurIPS 2024] The official implementation of "Kangaroo: Lossless Self-Speculative Decoding for Accelerating LLMs via Double Early Exitin…☆60Updated last year
- ☆14Updated last year
- official code for GliDe with a CaPE☆18Updated last year
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆168Updated last year