NVIDIA / online-softmax
Benchmark code for the "Online normalizer calculation for softmax" paper
☆91Updated 6 years ago
Alternatives and similar repositories for online-softmax:
Users that are interested in online-softmax are comparing it to the libraries listed below
- High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.☆106Updated 9 months ago
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆63Updated 8 months ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆108Updated 7 months ago
- ☆92Updated 7 months ago
- Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.☆35Updated last month
- Several optimization methods of half-precision general matrix vector multiplication (HGEMV) using CUDA core.☆61Updated 7 months ago
- ☆102Updated last month
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆181Updated 2 months ago
- ☆198Updated 9 months ago
- llama INT4 cuda inference with AWQ☆54Updated 3 months ago
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆91Updated 3 weeks ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆73Updated 3 weeks ago
- ☆55Updated 2 weeks ago
- ☆60Updated this week
- ☆88Updated 3 weeks ago
- play gemm with tvm☆90Updated last year
- ☆68Updated 3 months ago
- ☆68Updated 3 months ago
- ☆30Updated last year
- PyTorch bindings for CUTLASS grouped GEMM.☆81Updated 5 months ago
- Implement Flash Attention using Cute.☆76Updated 4 months ago
- Examples of CUDA implementations by Cutlass CuTe☆159Updated 2 months ago
- ☆82Updated last month
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆82Updated this week
- ☆115Updated 4 months ago
- ☆78Updated 5 months ago
- 📚FFPA(Split-D): Yet another Faster Flash Attention with O(1) GPU SRAM complexity large headdim, 1.8x~3x↑🎉 faster than SDPA EA.☆169Updated 2 weeks ago
- An extension library of WMMA API (Tensor Core API)☆96Updated 9 months ago
- 使用 cutlass 实现 flash-attention 精简版,具有教学意义☆39Updated 8 months ago
- ☆43Updated last week