An official website of the United States government
Word-wise Fully Homomorphic Encryption (FHE) schemes, such as CKKS, are gaining significant traction due to their ability to provide post-quantum-resistant, privacy preserving approximate computing—an especially desirable feature in the Machine-Learning-as-a-Service (MLaaS) paradigm. In this work, we introduce FIDESlib, the first open-source server-side CKKS GPU library that is fully interoperable with well-established client-side OpenFHE operations. Unlike other existing open-source GPU libraries, FIDESlib provides the first implementation featuring heavily optimized GPU kernels for all CKKS primitives, including bootstrapping. Our library also integrates robust benchmarking and testing, ensuring it remains adaptable to further optimization. Comparing our scheme against Phantom (the previously top open-source CKK library, we show that FIDESlib offers superior performance and scalability. For bootstrapping, FIDESlib achieves no less than 70× speedup over the AVX-optimized OpenFHE implementation. FIDESlib is available on Github.
more »
« less
This content will become publicly available on February 1, 2026
gsplat: An Open-Source Library for Gaussian Splatting
gsplat is an open-source library designed for training and developing Gaussian Splat- ting methods. It features a front-end with Python bindings compatible with the Py- Torch library and a back-end with highly optimized CUDA kernels. gsplat o↵ers nu- merous features that enhance the optimization of Gaussian Splatting models, which in- clude optimization improvements for speed, memory, and convergence times. Experimen- tal results demonstrate that gsplat achieves up to 10% less training time and 4⇥ less memory than the original Kerbl et al. (2023) implementation. Utilized in several re- search projects, gsplat is actively maintained on GitHub. Source code is available at https://github.com/nerfstudio-project/gsplat under Apache License 2.0. We wel- come contributions from the open-source community.
more »
« less
- Award ID(s):
- 2235013
- PAR ID:
- 10579936
- Publisher / Repository:
- JMLR
- Date Published:
- Journal Name:
- Journal of machine learning research
- ISSN:
- 1532-4435
- Subject(s) / Keyword(s):
- gaussian splatting 3D reconstruction novel view synthesis PyTorch CUDA
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
This work introduces TopoBench, an open-source library designed to standardize benchmarking and accelerate research in topological deep learning (TDL). TopoBench decomposes TDL into a sequence of independent modules for data generation, loading, transforming and processing, as well as model training, optimization and evaluation. This modular organization provides flexibility for modifications and facilitates the adaptation and optimization of various TDL pipelines. A key feature of TopoBench is its support for transformations and lifting across topological domains. Mapping the topology and features of a graph to higher-order topological domains, such as simplicial and cell complexes, enables richer data representations and more fine-grained analyses. The applicability of TopoBench is demonstrated by benchmarking several TDL architectures across diverse tasks and datasets.more » « less
-
Large language models (LLMs) are notoriously memory-intensive during training, particularly with the popular AdamW optimizer. This memory burden necessitates using more or higher-end GPUs or reducing batch sizes, limiting training scalability and throughput. To address this, various memory-efficient optimizers have been proposed to reduce optimizer memory usage. However, they face critical challenges: (i) reliance on costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial optimizer memory overhead to maintain competitive performance. In this work, we identify that AdamW's learning rate adaptation rule can be effectively coarsened as a structured learning rate update. Based on this insight, we propose Approximated Gradient Scaling for Memory-Efficient LLM Optimization (APOLLO), which approximates learning rate scaling using an auxiliary low-rank optimizer state based on pure random projection. This structured learning rate update rule makes APOLLO highly tolerant to further memory reductions while delivering comparable pre-training performance. Even its rank-1 variant, APOLLO-Mini, achieves superior pre-training performance compared to AdamW with SGD-level memory costs. Extensive experiments demonstrate that the APOLLO series performs on-par with or better than AdamW, while achieving greater memory savings by nearly eliminating the optimization states of AdamW. These savings provide significant system-level benefits: (1) Enhanced Throughput: 3x throughput on an 8xA100-80GB setup compared to AdamW by supporting 4x larger batch sizes. (2) Improved Model Scalability: Pre-training LLaMA-13B with naive DDP on A100-80GB GPUs without system-level optimizations. (3) Low-End GPU Friendly Pre-training: Pre-training LLaMA-7B on a single GPU using less than 12 GB of memory with weight quantization.more » « less
-
Large language models (LLMs) are notoriously memory-intensive during training, particularly with the popular AdamW optimizer. This memory burden necessitates using more or higher-end GPUs or reducing batch sizes, limiting training scalability and throughput. To address this, various memory-efficient optimizers have been proposed to reduce optimizer memory usage. However, they face critical challenges: (i) reliance on costly SVD operations; (ii) significant performance trade-offs compared to AdamW; and (iii) still substantial optimizer memory overhead to maintain competitive performance. In this work, we identify that AdamW's learning rate adaptation rule can be effectively coarsened as a structured learning rate update. Based on this insight, we propose Approximated Gradient Scaling for Memory-Efficient LLM Optimization (APOLLO), which approximates learning rate scaling using an auxiliary low-rank optimizer state based on pure random projection. This structured learning rate update rule makes APOLLO highly tolerant to further memory reductions while delivering comparable pre-training performance. Even its rank-1 variant, APOLLO-Mini, achieves superior pre-training performance compared to AdamW with SGD-level memory costs. Extensive experiments demonstrate that the APOLLO series performs on-par with or better than AdamW, while achieving greater memory savings by nearly eliminating the optimization states of AdamW. These savings provide significant system-level benefits: (1) Enhanced Throughput: 3x throughput on an 8xA100-80GB setup compared to AdamW by supporting 4x larger batch sizes. (2) Improved Model Scalability: Pre-training LLaMA-13B with naive DDP on A100-80GB GPUs without system-level optimizations. (3) Low-End GPU Friendly Pre-training: Pre-training LLaMA-7B on a single GPU using less than 12 GB of memory with weight quantization.more » « less
-
Quantum computing has become widely available to researchers via cloud-hosted devices with different technologies using a multitude of software development frameworks. The vertical stack behind such solutions typically features quantum language abstraction and high-level translation frameworks that tend to be open source, down to pulse-level programming. However, the lower-level mapping to the control electronics, such as controls for laser and microwave pulse generators, remains closed source for contemporary commercial cloud-hosted quantum devices. One exception is the ARTIQ (Advanced Real-Time Infrastructure for Quantum physics) open-source library for trapped-ion control electronics. This stack has been complemented by the Duke ARTIQ Extensions (DAX) to provide modularity and better abstraction. It, however, remains disconnected from the wealth of features provided by popular quantum computing languages. This paper contributes QisDAX, a bridge between Qiskit and DAX that fills this gap. QisDAX provides interfaces for Python programs written using IBM's Qiskit and transpiles them to the DAX abstraction. This allows users to generically interface to the ARTIQ control systems accessing trapped-ion quantum devices. Consequently, the algorithms expressed in Qiskit become available to an open-source quantum software stack. This provides the first open-source, end-to-end, full-stack pipeline for remote submission of quantum programs for trapped-ion quantum systems in a non-commercial setting.more » « less
