inclusionAI / asystem-amemLinks
A NCCL extension library, designed to efficiently offload GPU memory allocated by the NCCL communication library.
☆80Updated last month
Alternatives and similar repositories for asystem-amem
Users that are interested in asystem-amem are comparing it to the libraries listed below
Sorting:
- ☆340Updated 2 weeks ago
- A lightweight design for computation-communication overlap.☆213Updated 3 weeks ago
- NVSHMEM‑Tutorial: Build a DeepEP‑like GPU Buffer☆153Updated 4 months ago
- Automated Parallelization System and Infrastructure for Multiple Ecosystems☆82Updated last year
- Pipeline Parallelism Emulation and Visualization☆75Updated last week
- DeeperGEMM: crazy optimized version☆73Updated 8 months ago
- DeepXTrace is a lightweight tool for precisely diagnosing slow ranks in DeepEP-based environments.☆86Updated last month
- ☆65Updated 8 months ago
- nnScaler: Compiling DNN models for Parallel Training☆123Updated 3 months ago
- DLSlime: Flexible & Efficient Heterogeneous Transfer Toolkit☆91Updated last week
- Microsoft Collective Communication Library☆66Updated last year
- ☆47Updated last year
- ☆74Updated last year
- ☆157Updated last year
- Sequence-level 1F1B schedule for LLMs.☆38Updated 4 months ago
- Efficient Long-context Language Model Training by Core Attention Disaggregation☆79Updated 3 weeks ago
- A prefill & decode disaggregated LLM serving framework with shared GPU memory and fine-grained compute isolation.☆123Updated 3 weeks ago
- ☆96Updated 9 months ago
- Utility scripts for PyTorch (e.g. Make Perfetto show some disappearing kernels, Memory profiler that understands more low-level allocatio…☆80Updated 4 months ago
- Toolchain built around the Megatron-LM for Distributed Training☆84Updated last month
- Efficient Compute-Communication Overlap for Distributed LLM Inference☆68Updated 2 months ago
- Thunder Research Group's Collective Communication Library☆47Updated 6 months ago
- a size profiler for cuda binary☆69Updated this week
- High-performance distributed data shuffling (all-to-all) library for MoE training and inference☆107Updated 3 weeks ago
- Allow torch tensor memory to be released and resumed later☆202Updated last week
- A simple calculation for LLM MFU.☆64Updated 4 months ago
- ☆104Updated last year
- High performance Transformer implementation in C++.☆148Updated last year
- Stateful LLM Serving☆94Updated 10 months ago
- Distributed MoE in a Single Kernel [NeurIPS '25]☆183Updated this week