Bruce-Lee-LY / decoding_attentionLinks
Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.
☆40Updated 2 months ago
Alternatives and similar repositories for decoding_attention
Users that are interested in decoding_attention are comparing it to the libraries listed below
Sorting:
- ☆61Updated 3 months ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆100Updated 3 months ago
- A practical way of learning Swizzle☆25Updated 6 months ago
- Multiple GEMM operators are constructed with cutlass to support LLM inference.☆19Updated 3 weeks ago
- Performance of the C++ interface of flash attention and flash attention v2 in large language model (LLM) inference scenarios.☆40Updated 5 months ago
- ☆92Updated 4 months ago
- Quantized Attention on GPU☆44Updated 9 months ago
- ☆97Updated 11 months ago
- ☆50Updated 3 months ago
- Implement Flash Attention using Cute.☆94Updated 8 months ago
- ☆77Updated 3 months ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆111Updated 11 months ago
- A GPU-optimized system for efficient long-context LLMs decoding with low-bit KV cache.☆56Updated 2 weeks ago
- GPTQ inference TVM kernel☆40Updated last year
- ☆91Updated this week
- SpInfer: Leveraging Low-Level Sparsity for Efficient Large Language Model Inference on GPUs☆52Updated 4 months ago
- ☆81Updated 7 months ago
- ☆31Updated 6 months ago
- DeeperGEMM: crazy optimized version☆71Updated 3 months ago
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆94Updated 2 weeks ago
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆74Updated last year
- QuTLASS: CUTLASS-Powered Quantized BLAS for Deep Learning☆67Updated last month
- llama INT4 cuda inference with AWQ☆54Updated 7 months ago
- ☆98Updated 3 months ago
- Tile-based language built for AI computation across all scales☆46Updated this week
- Triton adapter for Ascend. Mirror of https://gitee.com/ascend/triton-ascend☆64Updated this week
- ☆12Updated 5 months ago
- [DAC'25] Official implement of "HybriMoE: Hybrid CPU-GPU Scheduling and Cache Management for Efficient MoE Inference"☆66Updated 2 months ago
- Persistent dense gemm for Hopper in `CuTeDSL`☆15Updated 2 weeks ago
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆94Updated last month