ROCm / Megatron-LMLinks
Ongoing research training transformer models at scale
☆34Updated last week
Alternatives and similar repositories for Megatron-LM
Users that are interested in Megatron-LM are comparing it to the libraries listed below
Sorting:
- MAD (Model Automation and Dashboarding)☆30Updated last week
- ☆97Updated last year
- ☆54Updated this week
- Intel® Extension for DeepSpeed* is an extension to DeepSpeed that brings feature support with SYCL kernels on Intel GPU(XPU) device. Note…☆63Updated 5 months ago
- AMD RAD's multi-GPU Triton-based framework for seamless multi-GPU programming☆133Updated this week
- Microsoft Collective Communication Library☆66Updated last year
- ☆151Updated 11 months ago
- QuickReduce is a performant all-reduce library designed for AMD ROCm that supports inline compression.☆36Updated 3 months ago
- PyTorch bindings for CUTLASS grouped GEMM.☆135Updated 6 months ago
- A high-throughput and memory-efficient inference and serving engine for LLMs☆111Updated this week
- extensible collectives library in triton☆91Updated 8 months ago
- Accelerating MoE with IO and Tile-aware Optimizations☆351Updated this week
- NVIDIA Resiliency Extension is a python package for framework developers and users to implement fault-tolerant features. It improves the …☆239Updated this week
- Applied AI experiments and examples for PyTorch☆311Updated 3 months ago
- A high-throughput and memory-efficient inference and serving engine for LLMs☆85Updated this week
- Autonomous GPU Kernel Generation via Deep Agents☆187Updated this week
- Building the Virtuous Cycle for AI-driven LLM Systems☆98Updated this week
- An efficient GPU support for LLM inference with x-bit quantization (e.g. FP6,FP5).☆276Updated 5 months ago
- Fast and memory-efficient exact attention☆203Updated last week
- nnScaler: Compiling DNN models for Parallel Training☆120Updated 2 months ago
- Github mirror of trition-lang/triton repo.☆105Updated this week
- Ahead of Time (AOT) Triton Math Library☆84Updated last week
- Evaluating Large Language Models for CUDA Code Generation ComputeEval is a framework designed to generate and evaluate CUDA code from Lar…☆84Updated 3 weeks ago
- torchcomms: a modern PyTorch communications API☆309Updated this week
- Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.☆137Updated last month
- AI Tensor Engine for ROCm☆322Updated this week
- A hierarchical collective communications library with portable optimizations☆37Updated last year
- Automated Parallelization System and Infrastructure for Multiple Ecosystems☆81Updated last year
- Chimera: bidirectional pipeline parallelism for efficiently training large-scale models.☆69Updated 9 months ago
- A lightweight design for computation-communication overlap.☆196Updated 2 months ago