IST-DASLab / gemm-fp8Links
High Performance FP8 GEMM Kernels for SM89 and later GPUs.
☆20Updated 9 months ago
Alternatives and similar repositories for gemm-fp8
Users that are interested in gemm-fp8 are comparing it to the libraries listed below
Sorting:
- Persistent dense gemm for Hopper in `CuTeDSL`☆15Updated 3 months ago
- Multiple GEMM operators are constructed with cutlass to support LLM inference.☆20Updated 3 months ago
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆78Updated last year
- Standalone Flash Attention v2 kernel without libtorch dependency☆112Updated last year
- Tutorials of Extending and importing TVM with CMAKE Include dependency.☆16Updated last year
- ☆83Updated 9 months ago
- ☆64Updated 6 months ago
- ☆33Updated 9 months ago
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆96Updated last month
- SpInfer: Leveraging Low-Level Sparsity for Efficient Large Language Model Inference on GPUs☆59Updated 7 months ago
- High Performance Int8 GEMM Kernels for SM80 and later GPUs.☆17Updated 7 months ago
- Implement Flash Attention using Cute.☆96Updated 10 months ago
- Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.☆45Updated 4 months ago
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆124Updated 6 months ago
- GPTQ inference TVM kernel☆39Updated last year
- llama INT4 cuda inference with AWQ☆55Updated 9 months ago
- Quantized Attention on GPU☆44Updated 11 months ago
- LLM Inference with Microscaling Format☆32Updated 11 months ago
- ☆50Updated 5 months ago
- ☆14Updated last week
- A Triton JIT runtime and ffi provider in C++☆29Updated this week
- A practical way of learning Swizzle☆31Updated 9 months ago
- Official implementation of the EMNLP23 paper: Outlier Suppression+: Accurate quantization of large language models by equivalent and opti…☆47Updated 2 years ago
- A GPU-optimized system for efficient long-context LLMs decoding with low-bit KV cache.☆61Updated 2 weeks ago
- Framework to reduce autotune overhead to zero for well known deployments.☆85Updated last month
- ☆47Updated last year
- Several optimization methods of half-precision general matrix vector multiplication (HGEMV) using CUDA core.☆68Updated last year
- DeeperGEMM: crazy optimized version☆72Updated 6 months ago
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆100Updated 4 months ago
- ☆19Updated last year