AnswerDotAI / cold-compress
Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of GPT-Fast, a simple, PyTorch-native generation codebase.
☆124Updated 7 months ago
Alternatives and similar repositories for cold-compress:
Users that are interested in cold-compress are comparing it to the libraries listed below
- Layer-Condensed KV cache w/ 10 times larger batch size, fewer params and less computation. Dramatic speed up with better task performance…☆148Updated 2 months ago
- ☆195Updated 4 months ago
- The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.☆59Updated 2 months ago
- Fast Matrix Multiplications for Lookup Table-Quantized LLMs☆236Updated last month
- Explorations into some recent techniques surrounding speculative decoding☆252Updated 3 months ago
- ☆122Updated last month
- The official repo for "LLoCo: Learning Long Contexts Offline"☆116Updated 9 months ago
- Triton-based implementation of Sparse Mixture of Experts.☆210Updated 4 months ago
- Boosting 4-bit inference kernels with 2:4 Sparsity☆72Updated 7 months ago
- ☆234Updated 11 months ago
- Single File, Single GPU, From Scratch, Efficient, Full Parameter Tuning library for "RL for LLMs"☆119Updated this week
- Code for studying the super weight in LLM☆94Updated 4 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆125Updated 4 months ago
- Efficient triton implementation of Native Sparse Attention.☆132Updated this week
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆158Updated 8 months ago
- ☆99Updated 10 months ago
- Official repository for LightSeq: Sequence Level Parallelism for Distributed Training of Long Context Transformers☆207Updated 7 months ago
- Load compute kernels from the Hub☆111Updated this week
- 🔥 A minimal training framework for scaling FLA models☆94Updated this week
- ☆125Updated last year
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆111Updated 4 months ago
- 🚀 Collection of components for development, training, tuning, and inference of foundation models leveraging PyTorch native components.☆190Updated this week
- Simple implementation of Speculative Sampling in NumPy for GPT-2.☆92Updated last year
- REST: Retrieval-Based Speculative Decoding, NAACL 2024☆198Updated 4 months ago
- The source code of our work "Prepacking: A Simple Method for Fast Prefilling and Increased Throughput in Large Language Models"☆59Updated 5 months ago
- Unofficial implementation for the paper "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models"☆155Updated 9 months ago
- Repo for "LoLCATs: On Low-Rank Linearizing of Large Language Models"☆229Updated 2 months ago
- Code for exploring Based models from "Simple linear attention language models balance the recall-throughput tradeoff"☆224Updated last month
- Simple and efficient pytorch-native transformer training and inference (batched)☆72Updated last year
- KV cache compression for high-throughput LLM inference☆123Updated 2 months ago