tspeterkim / mixed-precision-from-scratch
Mixed precision training from scratch with Tensors and CUDA
☆21Updated 10 months ago
Alternatives and similar repositories for mixed-precision-from-scratch:
Users that are interested in mixed-precision-from-scratch are comparing it to the libraries listed below
- Fast Matrix Multiplications for Lookup Table-Quantized LLMs☆232Updated 3 weeks ago
- The simplest implementation of recent Sparse Attention patterns for efficient LLM inference.☆58Updated last month
- Cataloging released Triton kernels.☆204Updated 2 months ago
- Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.☆104Updated this week
- Fast low-bit matmul kernels in Triton☆267Updated this week
- CUDA and Triton implementations of Flash Attention with SoftmaxN.☆68Updated 9 months ago
- ring-attention experiments☆127Updated 5 months ago
- Applied AI experiments and examples for PyTorch☆249Updated this week
- A minimal implementation of vllm.☆36Updated 7 months ago
- Experiment of using Tangent to autodiff triton☆78Updated last year
- Collection of kernels written in Triton language☆114Updated last month
- ☆76Updated 8 months ago
- Tree Attention: Topology-aware Decoding for Long-Context Attention on GPU clusters☆123Updated 3 months ago
- ☆95Updated 9 months ago
- Code for studying the super weight in LLM☆91Updated 3 months ago
- ☆191Updated this week
- ☆151Updated last year
- [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding☆109Updated 3 months ago
- extensible collectives library in triton☆84Updated 6 months ago
- Boosting 4-bit inference kernels with 2:4 Sparsity☆70Updated 6 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
- This repository contains the experimental PyTorch native float8 training UX☆222Updated 7 months ago
- Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking cache compression methods built on top of…☆122Updated 7 months ago
- PyTorch bindings for CUTLASS grouped GEMM.☆74Updated 4 months ago
- a minimal cache manager for PagedAttention, on top of llama3.☆73Updated 6 months ago
- ☆86Updated last year
- ArcticTraining is a framework designed to simplify and accelerate the post-training process for large language models (LLMs)☆51Updated this week
- Make triton easier☆47Updated 9 months ago