iclementine / optimize_softmaxLinks
Optimize softmax in triton in many cases
☆21Updated last year
Alternatives and similar repositories for optimize_softmax
Users that are interested in optimize_softmax are comparing it to the libraries listed below
Sorting:
- ☆106Updated 4 months ago
- Examples of CUDA implementations by Cutlass CuTe☆236Updated 3 months ago
- Implement Flash Attention using Cute.☆96Updated 9 months ago
- ☆56Updated 2 months ago
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆75Updated last year
- ☆43Updated last year
- This project is about convolution operator optimization on GPU, include GEMM based (Implicit GEMM) convolution.☆39Updated last week
- ☆135Updated 10 months ago
- 使用 cutlass 实现 flash-attention 精简版,具有教学意义☆49Updated last year
- ☆17Updated last year
- ☆109Updated 6 months ago
- ☆139Updated last year
- ☆150Updated 8 months ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆118Updated 4 months ago
- A GPU-optimized system for efficient long-context LLMs decoding with low-bit KV cache.☆60Updated last month
- ☆69Updated 8 months ago
- ☆153Updated 9 months ago
- From Minimal GEMM to Everything☆47Updated this week
- 使用 CUDA C++ 实现的 llama 模型推理框架☆62Updated 10 months ago
- Optimize GEMM with tensorcore step by step☆32Updated last year
- llm theoretical performance analysis tools and support params, flops, memory and latency analysis.☆108Updated 2 months ago
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆94Updated 3 weeks ago
- A Easy-to-understand TensorOp Matmul Tutorial☆378Updated last year
- Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.☆40Updated 7 months ago
- ☆98Updated last year
- [ICML 2025] Official PyTorch implementation of "FlatQuant: Flatness Matters for LLM Quantization"☆170Updated this week
- hands on model tuning with TVM and profile it on a Mac M1, x86 CPU, and GTX-1080 GPU.☆50Updated 2 years ago
- ☆140Updated last year
- 🤖FFPA: Extend FlashAttention-2 with Split-D, ~O(1) SRAM complexity for large headdim, 1.8x~3x↑🎉 vs SDPA EA.☆219Updated last month
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆186Updated 8 months ago