Ionut-Vlad Modoranu

15 papers A* 5Journal 10
YearRankTypeTitle / Venue / Authors
2026 J jnl
DASH: Faster Shampoo via Batched Block Preconditioning and Efficient Inverse-Root Solvers.
CoRR
Ionut-Vlad Modoranu, Philip Zmushko, Erik Schultheis, Mher Safaryan, Dan Alistarh
2026 J jnl
LoRDO: Distributed Low-Rank Optimization with Infrequent Communication.
CoRR
Andrej Jovanovic, Alex Iacob, Mher Safaryan, Ionut-Vlad Modoranu, Lorenzo Sani, William F. Shen, Xinchi Qiu, Dan Alistarh, Nicholas D. Lane
2025 A* conf
LDAdam: Adaptive Optimization from Low-Dimensional Gradient Statistics.
ICLR
Thomas Robert, Mher Safaryan, Ionut-Vlad Modoranu, Dan Alistarh
2025 J jnl
Optimizers Qualitatively Alter Solutions And We Should Leverage This.
CoRR
Razvan Pascanu, Clare Lyle, Ionut-Vlad Modoranu, Naima Elosegui Borras, Dan Alistarh, Petar Velickovic, Sarath Chandar, Soham De, James Martens
2025 J jnl
SVD-Free Low-Rank Adaptive Gradient Optimization for Large Language Models.
CoRR
Ionut-Vlad Modoranu, Mher Safaryan, Erik Schultheis, Dan Alistarh
2025 J jnl
Unified Scaling Laws for Compressed Representations.
CoRR
Andrei Panferov, Alexandra Volkova, Ionut-Vlad Modoranu, Vage Egiazarian, Mher Safaryan, Dan Alistarh
2024 A* conf
Error Feedback Can Accurately Compress Preconditioners.
ICML
Ionut-Vlad Modoranu, Aleksei Kalinov, Eldar Kurtic, Elias Frantar, Dan Alistarh
2024 J jnl
LDAdam: Adaptive Optimization from Low-Dimensional Gradient Statistics.
CoRR
Thomas Robert, Mher Safaryan, Ionut-Vlad Modoranu, Dan Alistarh
2024 A* conf
MicroAdam: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence.
NeurIPS
Ionut-Vlad Modoranu, Mher Safaryan, Grigory Malinovsky, Eldar Kurtic, Thomas Robert, Peter Richtárik, Dan Alistarh
2024 J jnl
MicroAdam: Accurate Adaptive Optimization with Low Space Overhead and Provable Convergence.
CoRR
Ionut-Vlad Modoranu, Mher Safaryan, Grigory Malinovsky, Eldar Kurtic, Thomas Robert, Peter Richtárik, Dan Alistarh
2024 A* conf
The Iterative Optimal Brain Surgeon: Faster Sparse Recovery by Leveraging Second-Order Information.
NeurIPS
Diyuan Wu, Ionut-Vlad Modoranu, Mher Safaryan, Denis Kuznedelev, Dan Alistarh
2024 J jnl
The Iterative Optimal Brain Surgeon: Faster Sparse Recovery by Leveraging Second-Order Information.
CoRR
Diyuan Wu, Ionut-Vlad Modoranu, Mher Safaryan, Denis Kuznedelev, Dan Alistarh
2023 J jnl
Error Feedback Can Accurately Compress Preconditioners.
CoRR
Ionut-Vlad Modoranu, Aleksei Kalinov, Eldar Kurtic, Dan Alistarh
2021 A* conf
A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference.
ICLR
Sanghyun Hong, Yigitcan Kaya, Ionut-Vlad Modoranu, Tudor Dumitras
2020 J jnl
A Panda? No, It's a Sloth: Slowdown Attacks on Adaptive Multi-Exit Neural Network Inference.
CoRR
Sanghyun Hong, Yigitcan Kaya, Ionut-Vlad Modoranu, Tudor Dumitras