NVIDIA-Merlin / HierarchicalKV
HierarchicalKV is a part of NVIDIA Merlin and provides hierarchical key-value storage to meet RecSys requirements. The key capability of HierarchicalKV is to store key-value feature-embeddings on high-bandwidth memory (HBM) of GPUs and in host memory. It also can be used as a generic key-value storage.
☆133Updated 2 months ago
Related projects ⓘ
Alternatives and complementary repositories for HierarchicalKV
- TePDist (TEnsor Program DISTributed) is an HLO-level automatic distributed system for DL models.☆90Updated last year
- A high-performance framework for training wide-and-deep recommender systems on heterogeneous cluster☆157Updated 7 months ago
- NCCL Fast Socket is a transport layer plugin to improve NCCL collective communication performance on Google Cloud.☆112Updated last year
- Paella: Low-latency Model Serving with Virtualized GPU Scheduling☆57Updated 6 months ago
- PyTorch distributed training acceleration framework☆34Updated this week
- An Efficient Pipelined Data Parallel Approach for Training Large Model☆70Updated 3 years ago
- Easy Parallel Library (EPL) is a general and efficient deep learning framework for distributed model training.☆263Updated last year
- MSCCL++: A GPU-driven communication stack for scalable AI applications☆250Updated this week
- A fast communication-overlapping library for tensor parallelism on GPUs.☆224Updated 3 weeks ago
- ☆73Updated last year
- PET: Optimizing Tensor Programs with Partially Equivalent Transformations and Automated Corrections☆114Updated 2 years ago
- Efficient and easy multi-instance LLM serving☆213Updated this week
- A baseline repository of Auto-Parallelism in Training Neural Networks☆142Updated 2 years ago
- ☆138Updated 2 weeks ago
- Dynamic Memory Management for Serving LLMs without PagedAttention☆238Updated last week
- NCCL Profiling Kit☆112Updated 4 months ago
- gossip: Efficient Communication Primitives for Multi-GPU Systems☆58Updated 2 years ago
- ☆31Updated last year
- TiledCUDA is a highly efficient kernel template library designed to elevate CUDA C’s level of abstraction for processing tiles.☆154Updated this week
- ☆51Updated last year
- Automated Parallelization System and Infrastructure for Multiple Ecosystems☆75Updated this week
- A home for the final text of all TVM RFCs.☆101Updated last month
- Artifact of OSDI '24 paper, ”Llumnix: Dynamic Scheduling for Large Language Model Serving“☆57Updated 5 months ago
- ☆196Updated last year
- ☆79Updated 2 months ago
- ☆33Updated 2 months ago
- RDMA and SHARP plugins for nccl library☆162Updated last week
- A low-latency & high-throughput serving engine for LLMs☆245Updated 2 months ago
- Synthesizer for optimal collective communication algorithms☆99Updated 7 months ago
- PArametrized Recommendation and Ai Model benchmark is a repository for development of numerous uBenchmarks as well as end to end nets for…☆124Updated this week