Alternatives and similar repositories for DeepSpeed-Kernels

Users that are interested in DeepSpeed-Kernels are comparing it to the libraries listed below

• stanford-futuredata / stk
  ☆115Updated last year
• IST-DASLab / Sparse-Marlin
  Boosting 4-bit inference kernels with 2:4 Sparsity
  ☆90Updated last year
• InternLM / turbomind
  ☆96Updated 9 months ago
• tgale96 / grouped_gemm
  PyTorch bindings for CUTLASS grouped GEMM.
  ☆139Updated 7 months ago
• hpcaitech / TensorNVMe
  A Python library transfers PyTorch tensors between CPU and NVMe
  ☆123Updated last year
• IBM / triton-dejavu
  Framework to reduce autotune overhead to zero for well known deployments.
  ☆92Updated 3 months ago
• anyscale / llm-continuous-batching-benchmarks
  ☆123Updated last year
• cchan / tccl
  extensible collectives library in triton
  ☆91Updated 9 months ago
• microsoft / AttentionEngine
  ☆116Updated 7 months ago
• tlc-pack / libflash_attn
  Standalone Flash Attention v2 kernel without libtorch dependency
  ☆113Updated last year
• wangsiping97 / FastGEMV
  High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.
  ☆123Updated last year
• triton-lang / kernels
  ☆100Updated last year
• usyd-fsalab / fp6_llm
  An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).
  ☆276Updated 5 months ago
• foundation-model-stack / foundation-model-stack
  🚀 Collection of components for development, training, tuning, and inference of foundation models leveraging PyTorch native components.
  ☆218Updated this week
• flashinfer-ai / cutlass-viz
  ☆65Updated 8 months ago
• NVIDIA / online-softmax
  Benchmark code for the "Online normalizer calculation for softmax" paper
  ☆104Updated 7 years ago
• microsoft / TileFusion
  TileFusion is an experimental C++ macro kernel template library that elevates the abstraction level in CUDA C for tile processing.
  ☆104Updated 6 months ago
• LeiWang1999 / AutoGPTQ.tvm
  GPTQ inference TVM kernel
  ☆41Updated last year
• feifeibear / Odysseus-Transformer
  Odysseus: Playground of LLM Sequence Parallelism
  ☆79Updated last year
• PipeFusion / PipeFusion
  A Suite for Parallel Inference of Diffusion Transformers (DiTs) on multi-GPU Clusters
  ☆53Updated last year
• AlibabaResearch / flash-llm
  Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity
  ☆231Updated 2 years ago
• meta-pytorch / applied-ai
  Applied AI experiments and examples for PyTorch
  ☆312Updated 4 months ago
• ankan-ban / llama_cu_awq
  llama INT4 cuda inference with AWQ
  ☆55Updated 11 months ago
• ademeure / DeeperGEMM
  DeeperGEMM: crazy optimized version
  ☆74Updated 8 months ago
• ROCm / aotriton
  Ahead of Time (AOT) Triton Math Library
  ☆85Updated last month
• SqueezeBits / QUICK
  QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference
  ☆118Updated last year
• Bruce-Lee-LY / decoding_attention
  Decoding Attention is specially optimized for MHA, MQA, GQA and MLA using CUDA core for the decoding stage of LLM inference.
  ☆46Updated 7 months ago
• Infini-AI-Lab / MagicDec
  [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding
  ☆136Updated last year
• meta-pytorch / BackendBench
  Ship correct and fast LLM kernels to PyTorch
  ☆130Updated this week
• meta-pytorch / KernelAgent
  Autonomous GPU Kernel Generation via Deep Agents
  ☆202Updated this week