Qwen3-0.6B megakernel: 527 tok/s decode on RTX 3090 (3.8x faster than PyTorch)
☆137Feb 10, 2026Updated 7 months ago
Alternatives and similar repositories for MegaQwen
Users that are interested in MegaQwen are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.
Sorting:
- Aggressive decode optimizations for Qwen3-0.6B on RTX 5090☆62Feb 25, 2026Updated 6 months ago
- ☆23Aug 29, 2026Updated 2 weeks ago
- ☆20Apr 26, 2026Updated 4 months ago
- Pure Triton kernels for Qwen3.5-27B inference on NVIDIA B200☆123Feb 28, 2026Updated 6 months ago
- LLM Inference via Triton (Flexible & Modular): Focused on Kernel Optimization using CUBIN binaries, Starting from gpt-oss Model☆119Apr 28, 2026Updated 4 months ago
- End-to-end encrypted email - Proton Mail • AdSpecial offer: 40% Off Yearly / 80% Off First Month. All Proton services are open source and independently audited for security.
- Kernels, of the mega variety :)☆822May 26, 2026Updated 3 months ago
- This repositories contains the reference implementation for the Sparse Delta Memory paper.More precisely, it contains the model definitio…☆38Jul 9, 2026Updated 2 months ago
- Humming is a high-performance, lightweight, and highly flexible JIT (Just-In-Time) compiled GEMM kernel library specifically designed for…☆229Updated this week
- Region-level profiling for CUDA kernels with trace, NVBit, CUPTI, NSys, and an interactive Explorer.☆129Apr 17, 2026Updated 4 months ago
- 存放一些 CUDA 编程相关的博客文件。☆23Oct 16, 2025Updated 10 months ago
- Optimize GEMM with tensorcore step by step☆40Dec 17, 2023Updated 2 years ago
- mKernel: fast multi-node, multi-GPU fused kernels☆270Updated this week
- High-performance LLM operator library built on TileLang.☆181Updated this week
- Companion code for The Physics of LLM Inference book☆28Apr 21, 2026Updated 4 months ago
- Bare Metal GPUs on DigitalOcean Gradient AI • AdPurpose-built for serious AI teams training foundational models, running large-scale inference, and pushing the boundaries of what's possible.
- MoE training for Me and You and maybe other people☆399Mar 15, 2026Updated 5 months ago
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆118Aug 4, 2026Updated last month
- Mirage Persistent Kernel: Compiling LLMs into a MegaKernel☆2,490Updated this week
- Repository for the CUDA H100 Course☆78Apr 12, 2026Updated 5 months ago
- Fastest kernels written from scratch☆621Aug 15, 2026Updated 3 weeks ago
- A PyTorch implementation of the GPT-OSS-20B architecture. All components are coded from scratch: RoPE with YaRN, RMSNorm, SwiGLU with cla…☆238Dec 2, 2025Updated 9 months ago
- ☆56Dec 19, 2025Updated 8 months ago
- ☆202Aug 9, 2026Updated last month
- Benchmark tests supporting the TiledCUDA library.☆19Nov 19, 2024Updated last year
- Managed hosting for WordPress and PHP on Cloudways • AdManaged hosting for WordPress, Magento, Laravel, or PHP apps, on multiple cloud providers. Deploy in minutes on Cloudways by DigitalOcean.
- Github mirror of trition-lang/triton repo.☆195Updated this week
- CODA: Rewriting Transformer Blocks as GEMM-Epilogue Programs☆250Updated this week
- A pure-Python implementation of the Nvidia CuTe layout algebra intended to be approachable and easy to learn.☆242Jun 29, 2026Updated 2 months ago
- FlashTile is a CUDA Tile IR compiler that is compatible with NVIDIA's tileiras, targeting SM70 through SM121 NVIDIA GPUs.☆60Feb 6, 2026Updated 7 months ago
- Slow-Fast Inference: Training-free Inference Acceleration via Within-sentence Support Stability☆19Sep 2, 2026Updated last week
- Winner 🏆 (Agent-only) MLSys 2026 - FlashInfer AI Kernel Generation Contest for the DeepSeek Sparse Attention (DSA) track with an average…☆159Updated this week
- My submission for the GPUMODE/AMD fp8 mm challenge☆29Jun 4, 2025Updated last year
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆158May 10, 2025Updated last year
- A series of high-performance GEMM (General Matrix Multiply) implementations Iteratively optimised for H100 GPUs in Pure CUDA.☆82Feb 18, 2026Updated 6 months ago
- 1-Click AI Models by DigitalOcean Gradient • AdDeploy popular AI models on DigitalOcean Gradient GPU virtual machines with just a single click. Zero configuration with optimized deployments.
- ☆103Updated this week
- Exploring how optimizations for GEMMs work☆36Feb 28, 2026Updated 6 months ago
- An agent harness that compiles a model into one provably-correct, self-retargeting CUDA megakernel and self-tunes it past cuBLAS at batch…☆138Jun 29, 2026Updated 2 months ago
- ☆15Apr 28, 2026Updated 4 months ago
- Row-wise block scaling for fp8 quantization matrix multiplication. Solution to GPU mode AMD challenge.☆19Feb 9, 2026Updated 7 months ago
- ☆78Apr 1, 2026Updated 5 months ago
- EleutherAI ML Performance reading group repository (slides, meeting recordings, annotated papers)☆37Mar 20, 2026Updated 5 months ago