Repeerc / flash-attention-v2-RDNA3-minimalLinks
a simple Flash Attention v2 implementation with ROCM (RDNA3 GPU, roc wmma), mainly used for stable diffusion(ComfyUI) in Windows ZLUDA environments.
☆50Updated last year
Alternatives and similar repositories for flash-attention-v2-RDNA3-minimal
Users that are interested in flash-attention-v2-RDNA3-minimal are comparing it to the libraries listed below
Sorting:
- Fast and memory-efficient exact attention☆208Updated this week
- ☆79Updated last year
- ☆163Updated 6 months ago
- A safetensors extension to efficiently store sparse quantized tensors on disk☆233Updated last week
- Development repository for the Triton language and compiler☆140Updated last week
- QQQ is an innovative and hardware-optimized W4A8 quantization solution for LLMs.☆154Updated 4 months ago
- Model Compression Toolbox for Large Language Models and Diffusion Models☆732Updated 5 months ago
- ☆49Updated last month
- Ahead of Time (AOT) Triton Math Library☆87Updated this week
- QuTLASS: CUTLASS-Powered Quantized BLAS for Deep Learning☆161Updated 2 months ago
- AI Tensor Engine for ROCm☆341Updated this week
- A high-throughput and memory-efficient inference and serving engine for LLMs☆113Updated this week
- ☆61Updated 6 months ago
- PyTorch half precision gemm lib w/ fused optional bias + optional relu/gelu☆77Updated last year
- An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).☆276Updated 6 months ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆113Updated last year
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆144Updated 8 months ago
- Low overhead tracing library and trace visualizer for pipelined CUDA kernels☆129Updated last month
- High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.☆124Updated last year
- 🎬 3.7× faster video generation E2E 🖼️ 1.6× faster image generation E2E ⚡ ColumnSparseAttn 9.3× vs FlashAttn‑3 💨 ColumnSparseGEMM 2.5× …☆99Updated 4 months ago
- ☆33Updated 11 months ago
- ☆85Updated 11 months ago
- Efficient implementation of DeepSeek Ops (Blockwise FP8 GEMM, MoE, and MLA) for AMD Instinct MI300X☆74Updated last month
- Kernel Library Wheel for SGLang☆17Updated this week
- [ICML 2024] BiLLM: Pushing the Limit of Post-Training Quantization for LLMs☆229Updated last year
- ☆56Updated last week
- A CUDA kernel for NHWC GroupNorm for PyTorch☆22Updated last year
- Fast Hadamard transform in CUDA, with a PyTorch interface☆272Updated 3 months ago
- 🤖FFPA: Extend FlashAttention-2 with Split-D, ~O(1) SRAM complexity for large headdim, 1.8x~3x↑🎉 vs SDPA EA.☆245Updated 2 months ago
- ☆64Updated this week