Infini-AI-Lab / Sirius
Sirius, an efficient correction mechanism, which significantly boosts Contextual Sparsity models on reasoning tasks while maintaining its efficiency gain.
☆19Updated 2 months ago
Related projects ⓘ
Alternatives and complementary repositories for Sirius
- 16-fold memory access reduction with nearly no loss☆57Updated this week
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆146Updated 4 months ago
- Implementation of Kangaroo: Lossless Self-Speculative Decoding via Double Early Exiting☆41Updated 4 months ago
- [NeurIPS'23] Speculative Decoding with Big Little Decoder☆85Updated 9 months ago
- Triton-based implementation of Sparse Mixture of Experts.☆184Updated last month
- This repo contains the source code for: Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs☆32Updated 2 months ago
- Ouroboros: Speculative Decoding with Large Model Enhanced Drafting (EMNLP 2024 main)☆76Updated 3 weeks ago
- Code for Palu: Compressing KV-Cache with Low-Rank Projection☆56Updated this week
- Boosting 4-bit inference kernels with 2:4 Sparsity☆51Updated 2 months ago
- MagicPIG: LSH Sampling for Efficient LLM Generation☆45Updated 2 weeks ago
- Fast and Robust Early-Exiting Framework for Autoregressive Language Models with Synchronized Parallel Decoding (EMNLP 2023 Long)☆52Updated last month
- A sparse attention kernel supporting mix sparse patterns☆53Updated 3 weeks ago
- ☆95Updated last month
- Triton implementation of FlashAttention2 that adds Custom Masks.☆74Updated 2 months ago
- TidalDecode: A Fast and Accurate LLM Decoding with Position Persistent Sparse Attention☆21Updated last month
- ☆31Updated 2 months ago
- The official implementation of paper: SimLayerKV: A Simple Framework for Layer-Level KV Cache Reduction.☆36Updated 3 weeks ago
- ☆42Updated 6 months ago
- Official repository for LightSeq: Sequence Level Parallelism for Distributed Training of Long Context Transformers☆196Updated 2 months ago
- ☆31Updated 2 months ago
- 📰 Must-read papers on KV Cache Compression (constantly updating 🤗).☆123Updated this week
- [ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference☆196Updated last week
- Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆70Updated last week
- ☆88Updated 2 months ago
- ☆183Updated 6 months ago
- ☆20Updated this week
- Repository of the paper "Accelerating Transformer Inference for Translation via Parallel Decoding"☆108Updated 7 months ago
- PyTorch implementation of paper "Response Length Perception and Sequence Scheduling: An LLM-Empowered LLM Inference Pipeline".☆74Updated last year
- Activation-aware Singular Value Decomposition for Compressing Large Language Models☆49Updated 3 weeks ago
- My own implementation of "Fast Inference from Transformers via Speculative Decoding"☆11Updated 11 months ago