SPLASH: Switching Parallel Layouts of Attention with Seamless Handoff for LLM Serving
cs.DC, cs.AI
Submitted: 2026-09-29
Updated: 2026-09-29
Code: https://github.com/ict-agent/SPLASH-sglang
Sources
- PipeLive: Efficient Live In-place Pipeline Parallelism Reconfiguration for Dynamic LLM Serving
- Amoeba: Runtime Tensor Parallel Transformation for LLM Inference Services
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
- DeepSeek-V3 Technical Report
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Fast Transformer Decoding: One Write-Head is All You Need
- Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism
- Kimi k1.5: Scaling Reinforcement Learning with LLMs
- Moebius: Serving Mixture-of-Expert Models with Seamless Runtime Parallelism Switch
- ReMP: Low-Downtime Runtime Model-Parallelism Reconfiguration for LLM Serving
- Agentic AI Workload Characteristics
- Kimi Linear: An Expressive, Efficient Attention Architecture
Related papers
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- SAMM: Sharded Automated Market Maker
- InferScale: GPU-Native KV Injection for Personalized LLM Serving
- Vigil: Accountable Liveness against Selective Silence
- Steelhead: Interleaving Partially Synchronous and Asynchronous Commit Rules on a Shared DAG
- Democratizing MoE inference on commodity GPUs with CoMoE