yanring / Megatron-MoE-ModelZooLinks
Best practices for training DeepSeek, Mixtral, Qwen and other MoE models using Megatron Core.
☆102Updated 2 weeks ago
Alternatives and similar repositories for Megatron-MoE-ModelZoo
Users that are interested in Megatron-MoE-ModelZoo are comparing it to the libraries listed below
Sorting:
- Official repository for DistFlashAttn: Distributed Memory-efficient Attention for Long-context LLMs Training☆215Updated last year
- PyTorch bindings for CUTLASS grouped GEMM.☆152Updated last month
- PyTorch bindings for CUTLASS grouped GEMM.☆123Updated 4 months ago
- Allow torch tensor memory to be released and resumed later☆144Updated this week
- Bridge Megatron-Core to Hugging Face/Reinforcement Learning☆130Updated last week
- Utility scripts for PyTorch (e.g. Memory profiler that understands more low-level allocations such as NCCL)☆55Updated 3 weeks ago
- ByteCheckpoint: An Unified Checkpointing Library for LFMs☆249Updated 2 months ago
- [ICLR 2025] COAT: Compressing Optimizer States and Activation for Memory-Efficient FP8 Training☆241Updated last month
- Triton-based implementation of Sparse Mixture of Experts.☆242Updated this week
- [ICML 2024] Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference☆339Updated 2 months ago
- Train speculative decoding models effortlessly and port them smoothly to SGLang serving.☆412Updated this week
- Zero Bubble Pipeline Parallelism☆428Updated 5 months ago
- A collection of memory efficient attention operators implemented in the Triton language.☆279Updated last year
- Applied AI experiments and examples for PyTorch☆296Updated last month
- Implementation for FP8/INT8 Rollout for RL training without performence drop.☆249Updated last week
- nnScaler: Compiling DNN models for Parallel Training☆118Updated 2 weeks ago
- Pipeline Parallelism Emulation and Visualization☆67Updated 3 months ago
- Code for paper: [ICLR2025 Oral] FlexPrefill: A Context-Aware Sparse Attention Mechanism for Efficient Long-Sequence Inference☆143Updated 4 months ago
- GEAR: An Efficient KV Cache Compression Recipefor Near-Lossless Generative Inference of LLM☆168Updated last year
- ☆129Updated 4 months ago
- Training library for Megatron-based models☆109Updated this week
- ☆147Updated 7 months ago
- ☆113Updated last year
- ☆282Updated 2 months ago
- Code associated with the paper **Draft & Verify: Lossless Large Language Model Acceleration via Self-Speculative Decoding**☆202Updated 7 months ago
- Estimate MFU for DeepSeekV3☆24Updated 9 months ago
- [ICLR 2025] PEARL: Parallel Speculative Decoding with Adaptive Draft Length☆117Updated 5 months ago
- 🚀 Collection of components for development, training, tuning, and inference of foundation models leveraging PyTorch native components.☆213Updated this week
- Flash-LLM: Enabling Cost-Effective and Highly-Efficient Large Generative Model Inference with Unstructured Sparsity☆221Updated 2 years ago
- 16-fold memory access reduction with nearly no loss☆105Updated 6 months ago