Alternatives and similar repositories for Megatron-DeepSpeed:

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

• deepspeedai / DeepSpeed-Kernels
  ☆59Updated 2 weeks ago
• IST-DASLab / QIGen
  Repository for CPU Kernel Generation for LLM Inference
  ☆25Updated last year
• pytorch-labs / tritonbench
  Tritonbench is a collection of PyTorch custom operators with example inputs to measure their performance.
  ☆89Updated this week
• stanford-futuredata / stk
  ☆100Updated 5 months ago
• IST-DASLab / SparseFinetuning
  Repository for Sparse Finetuning of LLMs via modified version of the MosaicML llmfoundry
  ☆40Updated last year
• vedantroy / gpu_kernels
  ☆25Updated last year
• SqueezeBits / QUICK
  QUICK: Quantization-aware Interleaving and Conflict-free Kernel for efficient LLM inference
  ☆116Updated 11 months ago
• IST-DASLab / Sparse-Marlin
  Boosting 4-bit inference kernels with 2:4 Sparsity
  ☆64Updated 5 months ago
• tgale96 / grouped_gemm
  PyTorch bindings for CUTLASS grouped GEMM.
  ☆64Updated 3 months ago
• wangsiping97 / FastGEMV
  High-speed GEMV kernels, at most 2.7x speedup compared to pytorch baseline.
  ☆97Updated 7 months ago
• INT-FlashAttention2024 / INT-FlashAttention
  ☆61Updated 3 weeks ago
• tridao / flash-attention-wheels
  ☆46Updated last year
• facebookresearch / MODel_opt
  Memory Optimizations for Deep Learning (ICML 2023)
  ☆62Updated 11 months ago
• Infini-AI-Lab / MagicDec
  [ICLR2025] Breaking Throughput-Latency Trade-off for Long Sequences with Speculative Decoding
  ☆107Updated 2 months ago
• MDK8888 / vllmini
  A minimal implementation of vllm.
  ☆33Updated 6 months ago
• graphcore-research / out-of-the-box-fp8-training
  Demo of the unit_scaling library, showing how a model can be easily adapted to train in FP8.
  ☆43Updated 7 months ago
• Bruce-Lee-LY / decoding_attention
  Decoding Attention is specially optimized for multi head attention (MHA) using CUDA core for the decoding stage of LLM inference.
  ☆29Updated 3 months ago
• IBM / triton-dejavu
  Framework to reduce autotune overhead to zero for well known deployments.
  ☆61Updated 3 weeks ago
• usyd-fsalab / fp6_llm
  An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).
  ☆234Updated 3 months ago
• OpenNLPLab / LASP
  Linear Attention Sequence Parallelism (LASP)
  ☆77Updated 8 months ago
• Raincleared-Song / sparse_gpu_operator
  GPU operators for sparse tensor operations
  ☆30Updated 11 months ago
• microsoft / TileFusion
  TileFusion is a highly efficient kernel template library designed to elevate the level of abstraction in CUDA C for processing tiles.
  ☆56Updated this week
• ScalingIntelligence / CATS
  ☆23Updated 3 months ago
• PipeFusion / PipeFusion
  A Suite for Parallel Inference of Diffusion Transformers (DiTs) on multi-GPU Clusters
  ☆39Updated 6 months ago
• lfsszd / CS-Drafting
  Cascade Speculative Drafting
  ☆28Updated 10 months ago
• casper-hansen / AutoAWQ_kernels
  ☆65Updated 2 months ago
• feifeibear / ChituAttention
  Quantized Attention on GPU
  ☆34Updated 2 months ago
• siyan-zhao / prepacking
  The source code of our work "Prepacking: A Simple Method for Fast Prefilling and Increased Throughput in Large Language Models"
  ☆59Updated 4 months ago
• feifeibear / Odysseus-Transformer
  Odysseus: Playground of LLM Sequence Parallelism
  ☆64Updated 8 months ago
• tile-ai / tvm
  Open deep learning compiler stack for cpu, gpu and specialized accelerators
  ☆18Updated 2 weeks ago