amazon-science / mxfp4-llmLinks
Official implementation for Training LLMs with MXFP4
☆118Updated 9 months ago
Alternatives and similar repositories for mxfp4-llm
Users that are interested in mxfp4-llm are comparing it to the libraries listed below
Sorting:
- ☆158Updated 11 months ago
- The evaluation framework for training-free sparse attention in LLMs☆110Updated 3 months ago
- ☆117Updated 3 weeks ago
- Boosting 4-bit inference kernels with 2:4 Sparsity☆93Updated last year
- Work in progress.☆79Updated 2 months ago
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆137Updated last year
- ☆163Updated 7 months ago
- Flash-Muon: An Efficient Implementation of Muon Optimizer☆229Updated 7 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆131Updated last year
- Repo for "LoLCATs: On Low-Rank Linearizing of Large Language Models"☆251Updated 11 months ago
- PB-LLM: Partially Binarized Large Language Models☆157Updated 2 years ago
- ☆133Updated 8 months ago
- KV cache compression for high-throughput LLM inference☆150Updated 11 months ago
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆176Updated last year
- Triton-based implementation of Sparse Mixture of Experts.☆262Updated 3 months ago
- Code for data-aware compression of DeepSeek models☆69Updated last month
- Fast and memory-efficient exact attention☆75Updated 10 months ago
- Code for studying the super weight in LLM☆120Updated last year
- Block Diffusion for Ultra-Fast Speculative Decoding☆432Updated this week
- ☆54Updated last year
- ShiftAddLLM: Accelerating Pretrained LLMs via Post-Training Multiplication-Less Reparameterization☆111Updated last year
- [ICLR'25] Fast Inference of MoE Models with CPU-GPU Orchestration☆258Updated last year
- An efficient implementation of the NSA (Native Sparse Attention) kernel☆128Updated 7 months ago
- Training-free Post-training Efficient Sub-quadratic Complexity Attention. Implemented with OpenAI Triton.☆148Updated 2 months ago
- ☆51Updated last year
- HALO: Hadamard-Assisted Low-Precision Optimization and Training method for finetuning LLMs. 🚀 The official implementation of https://arx…☆29Updated 11 months ago
- Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of…☆148Updated last year
- This repository contains the experimental PyTorch native float8 training UX☆227Updated last year
- ☆128Updated 2 years ago
- QuIP quantization☆61Updated last year