Ascend / torchairLinks
☆14Updated this week
Alternatives and similar repositories for torchair
Users that are interested in torchair are comparing it to the libraries listed below
Sorting:
- ☆60Updated 2 months ago
- DeeperGEMM: crazy optimized version☆69Updated last month
- Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.☆38Updated 2 weeks ago
- Standalone Flash Attention v2 kernel without libtorch dependency☆110Updated 9 months ago
- ☆86Updated 3 months ago
- GPTQ inference TVM kernel☆40Updated last year
- 方便扩展的Cuda算子理解和优化框架,仅用在学习使用☆15Updated last year
- ⚡️Write HGEMM from scratch using Tensor Cores with WMMA, MMA and CuTe API, Achieve Peak⚡️ Performance.☆80Updated last month
- A practical way of learning Swizzle☆20Updated 4 months ago
- ☆96Updated 9 months ago
- Compare different hardware platforms via the Roofline Model for LLM inference tasks.☆100Updated last year
- A prefill & decode disaggregated LLM serving framework with shared GPU memory and fine-grained compute isolation.☆87Updated last month
- ☆49Updated last month
- ☆19Updated 8 months ago
- Pipeline Parallelism Emulation and Visualization☆43Updated 2 weeks ago
- TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.☆90Updated 2 weeks ago
- We invite you to visit and follow our new repository at https://github.com/microsoft/TileFusion. TiledCUDA is a highly efficient kernel …☆183Updated 4 months ago
- A standalone GEMM kernel for fp16 activation and quantized weight, extracted from FasterTransformer☆92Updated 3 weeks ago
- Triton adapter for Ascend. Mirror of https://gitee.com/ascend/triton-ascend☆54Updated this week
- Summary of system papers/frameworks/codes/tools on training or serving large model☆57Updated last year
- A simple calculation for LLM MFU.☆38Updated 3 months ago
- Benchmark code for the "Online normalizer calculation for softmax" paper☆94Updated 6 years ago
- A lightweight design for computation-communication overlap.☆143Updated this week
- 使用 cutlass 仓库在 ada 架构上实现 fp8 的 flash attention☆71Updated 10 months ago
- Odysseus: Playground of LLM Sequence Parallelism☆70Updated last year
- Sequence-level 1F1B schedule for LLMs.☆28Updated last week
- Multiple GEMM operators are constructed with cutlass to support LLM inference.☆18Updated 8 months ago
- Quantized Attention on GPU☆44Updated 7 months ago
- ☆71Updated last month
- ☆29Updated 4 months ago