microsoft / tokenweaveLinks
Efficient Compute-Communication Overlap for Distributed LLM Inference
☆22Updated 2 weeks ago
Alternatives and similar repositories for tokenweave
Users that are interested in tokenweave are comparing it to the libraries listed below
Sorting:
- Artifact for "Apparate: Rethinking Early Exits to Tame Latency-Throughput Tensions in ML Serving" [SOSP '24]☆25Updated 7 months ago
- ☆25Updated last year
- Vector search with bounded performance.☆36Updated last year
- DISB is a new DNN inference serving benchmark with diverse workloads and models, as well as real-world traces.☆52Updated 10 months ago
- ☆37Updated 2 months ago
- Official resporitory for "IPDPS' 24 QSync: Quantization-Minimized Synchronous Distributed Training Across Hybrid Devices".☆20Updated last year
- FractalTensor is a programming framework that introduces a novel approach to organizing data in deep neural networks (DNNs) as a list of …☆28Updated 6 months ago
- Tacker: Tensor-CUDA Core Kernel Fusion for Improving the GPU Utilization while Ensuring QoS☆27Updated 5 months ago
- An Attention Superoptimizer☆22Updated 5 months ago
- SOTA Learning-augmented Systems☆36Updated 3 years ago
- Stateful LLM Serving☆76Updated 4 months ago
- ☆64Updated last year
- Microsoft Collective Communication Library☆64Updated 7 months ago
- ☆54Updated 4 years ago
- A source-to-source compiler for optimizing CUDA dynamic parallelism by aggregating launches☆15Updated 6 years ago
- ☆13Updated 2 years ago
- NEO is a LLM inference engine built to save the GPU memory crisis by CPU offloading☆44Updated last month
- ☆49Updated last month
- High performance RDMA-based distributed feature collection component for training GNN model on EXTREMELY large graph☆54Updated 3 years ago
- ☆24Updated 2 years ago
- Aims to implement dual-port and multi-qp solutions in deepEP ibrc transport☆55Updated 2 months ago
- A memory profiler for NVIDIA GPUs to explore memory inefficiencies in GPU-accelerated applications.☆25Updated 9 months ago
- A GPU-accelerated DNN inference serving system that supports instant kernel preemption and biased concurrent execution in GPU scheduling.☆42Updated 3 years ago
- TiledLower is a Dataflow Analysis and Codegen Framework written in Rust.☆14Updated 7 months ago
- Artifact of ASPLOS'23 paper entitled: GRACE: A Scalable Graph-Based Approach to Accelerating Recommendation Model Inference☆18Updated 2 years ago
- ☆16Updated 2 years ago
- GVProf: A Value Profiler for GPU-based Clusters☆51Updated last year
- ☆37Updated 7 months ago
- ☆39Updated 2 years ago
- [ACM EuroSys 2023] Fast and Efficient Model Serving Using Multi-GPUs with Direct-Host-Access☆57Updated last year