snu-mllab / KVzipLinks
Query-agnostic KV cache eviction: 3–4× reduction in memory and 2× decrease in latency (Qwen3/2.5, Gemma3, LLaMA3)
☆93Updated this week
Alternatives and similar repositories for KVzip
Users that are interested in KVzip are comparing it to the libraries listed below
Sorting:
- RWKV-7: Surpassing GPT☆94Updated 8 months ago
- Lightweight toolkit package to train and fine-tune 1.58bit Language models☆81Updated 2 months ago
- ☆53Updated 2 months ago
- ☆82Updated 6 months ago
- Training-free Post-training Efficient Sub-quadratic Complexity Attention. Implemented with OpenAI Triton.☆141Updated this week
- ☆60Updated 4 months ago
- Easy to use, High Performant Knowledge Distillation for LLMs☆88Updated 2 months ago
- KV cache compression for high-throughput LLM inference☆134Updated 5 months ago
- Repo hosting codes and materials related to speeding LLMs' inference using token merging.☆36Updated 2 weeks ago
- This repository contains the code for the paper: SirLLM: Streaming Infinite Retentive LLM☆59Updated last year
- Work in progress.☆70Updated last month
- Simple extension on vLLM to help you speed up reasoning model without training.☆169Updated 2 months ago
- ☆79Updated 8 months ago
- ☆51Updated last year
- Official repository for the paper "NeuZip: Memory-Efficient Training and Inference with Dynamic Compression of Neural Networks". This rep…☆59Updated 9 months ago
- ☆143Updated last month
- Repository for the Q-Filters method (https://arxiv.org/pdf/2503.02812)☆34Updated 4 months ago
- ☆42Updated last month
- ☆87Updated 8 months ago
- QuIP quantization☆54Updated last year
- ☆77Updated 4 months ago
- From GaLore to WeLore: How Low-Rank Weights Non-uniformly Emerge from Low-Rank Gradients. Ajay Jaiswal, Lu Yin, Zhenyu Zhang, Shiwei Liu,…☆47Updated 3 months ago
- Block Transformer: Global-to-Local Language Modeling for Fast Inference (NeurIPS 2024)☆160Updated 3 months ago
- The official implementation for paper "Agentic-R1: Distilled Dual-Strategy Reasoning"☆84Updated 2 weeks ago
- Chain of Experts (CoE) enables communication between experts within Mixture-of-Experts (MoE) models☆219Updated last month
- A pipeline for LLM knowledge distillation☆106Updated 4 months ago
- Fira: Can We Achieve Full-rank Training of LLMs Under Low-rank Constraint?☆112Updated 9 months ago
- Fused Qwen3 MoE layer for faster training, compatible with HF Transformers, LoRA, 4-bit quant, Unsloth☆137Updated last week
- A collection of tricks and tools to speed up transformer models☆169Updated 2 months ago
- Fine-tunes a student LLM using teacher feedback for improved reasoning and answer quality. Implements GRPO with teacher-provided evaluati…☆46Updated 2 months ago