Violetta Shevchenko

Taming Curvature: Architecture Warm-up for Stable Transformer Training

SENTINEL: StagewisE iNtegriTy verification for pIpeliNe parallEL decentralized training

AsyncMesh: Fully Asynchronous Optimization for Data and Pipeline Parallelism

SRSR: Enhancing Semantic Accuracy in Real-World Image Super-Resolution with Spatially Re-Focused Text-Conditioning

Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization

Mixtures of Subspaces for Bandwidth Efficient Context Parallel Training