EpiLoRA: Fusing Many LoRA Adapters into Low-Precision GEMMs for Serving and RL
cs.DC, cs.CL, cs.LG
Submitted: 2026-11-02
Updated: 2026-11-02
Code: https://github.com/ingambe/trl
Terminology
Sources
- Compress then Serve: Serving Thousands of LoRA Adapters with Little Overhead
- FLUX: Fast Software-based Communication Overlap On GPUs Through Kernel Fusion
- Towards Full Pipeline FP8 Reinforcement Learning for LLMs
- InfiniLoRA: Disaggregated Multi-LoRA Serving for Large Language Models
- Punica: Multi-Tenant LoRA Serving
- AReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning
- QaRL: Rollout-Aligned Quantization-Aware RL for Fast and Stable Training under Training--Inference Mismatch
- OpenRLHF: An Easy-to-use, Scalable and High-performance RLHF Framework
- SparseRL-Sync: Lossless Weight Synchronization with ~100x Less Communication
- QeRL: Beyond Efficiency -- Quantization-enhanced Reinforcement Learning for LLMs
- LoRA-Switch: Boosting the Efficiency of Dynamic LLM Adapters via System-Algorithm Co-design
- SVDQuant: Absorbing Outliers by Low-Rank Components for 4-Bit Diffusion Models
- CaraServe: CPU-Assisted and Rank-Aware LoRA Serving for Generative LLM Inference
- Scale-QLoRA: Code-Invariant Adapter Merging for Native 4-bit Microscaling LLMs
- The Llama 3 Herd of Models
- Score Centering Stabilizes Off-policy Reinforcement Learning
- Empower Vision Applications with LoRA LMM
- Understanding and Exploiting Weight Update Sparsity for Communication-Efficient Distributed RL
- FlexLLM: Token-Level Co-Serving of LLM Inference and Finetuning with SLO Guarantees
- FP8 Formats for Deep Learning
Related papers
- Programmable Probabilistic Computer with 1,000,000 p-bits
- BatchGen: An Architecture for Scalable and Efficient Batch Inference
- Seamless Reconfiguration for DAG-BFT (Technical Report)
- ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving
- iScheduler: Reinforcement Learning-Driven Continual Optimization for Large-Scale Resource Investment Problems
- TIRx: A Unified Tile Primitive ML Compiler Across Heterogeneous Backends