Search

Fine-tuning pre-trained large language models (LLMs) with limited hardware presents challenges due to GPU memory constraints. Various distributed fine-tuning methods have been proposed to alleviate memory constraints on GPU. However, determining the most effective method for achieving rapid fine-tuning while preventing GPU out-of-memory issues in a given environment remains unclear. To address this challenge, we introduce LLMem, a solution that estimates the GPU memory consumption when applying distributed fine-tuning methods across multiple GPUs and identifies the optimal method. We conduct GPU memory usage estimation prior to fine-tuning, leveraging the fundamental structure of transformer-based decoder models and the memory usage distribution of each method. Experimental results show that LLMem accurately estimates peak GPU memory usage on a single GPU, with error rates of up to 1.6%. Additionally, it shows an average error rate of 3.0% when applying distributed fine-tuning methods to LLMs with more than a billion parameters on multi-GPU setups., Comment: 9 pages, 9 figures, accepted to IJCAI 2024

We introduce HyperCLOVA X, a family of large language models (LLMs) tailored to the Korean language and culture, along with competitive capabilities in English, math, and coding. HyperCLOVA X was trained on a balanced mix of Korean, English, and code data, followed by instruction-tuning with high-quality human-annotated datasets while abiding by strict safety guidelines reflecting our commitment to responsible AI. The model is evaluated across various benchmarks, including comprehensive reasoning, knowledge, commonsense, factuality, coding, math, chatting, instruction-following, and harmlessness, in both Korean and English. HyperCLOVA X exhibits strong reasoning capabilities in Korean backed by a deep understanding of the language and cultural nuances. Further analysis of the inherent bilingual nature and its extension to multilingualism highlights the model's cross-lingual proficiency and strong generalization ability to untargeted languages, including machine translation between several language pairs and cross-lingual inference tasks. We believe that HyperCLOVA X can provide helpful guidance for regions or countries in developing their sovereign LLMs., Comment: 44 pages; updated authors list and fixed author names

In Bernal-stacked bilayer graphene (BBG), the Landau levels give rise to an intimate connection between valley and layer degrees of freedom. Adding a moir\'e superlattice potential enriches the BBG physics with the formation of topological minibands - potentially leading to tunable exotic quantum transport. Here, we present magnetotransport measurements of a high-quality bilayer graphene-hexagonal boron nitride (hBN) heterostructure. The zero-degree alignment generates a strong moir\'e superlattice potential for the electrons in BBG and the resulting Landau fan diagram of longitudinal and Hall resistance displays a Hofstadter butterfly pattern with a high level of detail. We demonstrate that the intricate relationship between valley and layer degrees of freedom controls the topology of moir\'e-induced bands, significantly influencing the energetics of interacting quantum phases in the BBG superlattice. We further observe signatures of field-induced correlated insulators, helical edge states and clear quantizations of interaction-driven topological quantum phases, such as symmetry broken Chern insulators.

Most invariance-based self-supervised methods rely on single object-centric images (e.g., ImageNet images) for pretraining, learning features that invariant to geometric transformation. However, when images are not object-centric, the semantics of the image can be significantly altered due to cropping. Furthermore, as the model becomes insensitive to geometric transformations, it may struggle to capture location information. For this reason, we propose a Geometric Transformation Sensitive Architecture designed to be sensitive to geometric transformations, specifically focusing on four-fold rotation, random crop, and multi-crop. Our method encourages the student to be sensitive by predicting rotation and using targets that vary with those transformations through pooling and rotating the teacher feature map. Additionally, we use patch correspondence loss to encourage correspondence between patches with similar features. This approach allows us to capture long-term dependencies in a more appropriate way than capturing long-term dependencies by encouraging local-to-global correspondence, which occurs when learning to be insensitive to multi-crop. Our approach demonstrates improved performance when using non-object-centric images as pretraining data compared to other methods that train the model to be insensitive to geometric transformation. We surpass DINO[Caron et al.[2021b]] baseline in tasks including image classification, semantic segmentation, detection, and instance segmentation with improvements of 4.9 $Top-1 Acc$, 3.3 $mIoU$, 3.4 $AP^b$, and 2.7 $AP^m$. Code and pretrained models are publicly available at: https://github.com/bok3948/GTSA

This paper examines distributional properties and predictive performance of the estimated maximum agreement linear predictor (MALP) introduced in Bottai, Kim, Lieberman, Luta, and Pena (2022) paper in The American Statistician, which is the linear predictor maximizing Lin's concordance correlation coefficient (CCC) between the predictor and the predictand. It is compared and contrasted, theoretically and through computer experiments, with the estimated least-squares linear predictor (LSLP). Finite-sample and asymptotic properties are obtained, and confidence intervals are also presented. The predictors are illustrated using two real data sets: an eye data set and a bodyfat data set. The results indicate that the estimated MALP is a viable alternative to the estimated LSLP if one desires a predictor whose predicted values possess higher agreement with the predictand values, as measured by the CCC.

Although code generation for Convolution Neural Network (CNN) models has been extensively studied, performing efficient data slicing and parallelization for highly-constrai\-ned Multicore Neural Processor Units (NPUs) is still a challenging problem. Given the size of convolutions' input/output tensors and the small footprint of NPU on-chip memories, minimizing memory transactions while maximizing parallelism and MAC utilization are central to any effective solution. This paper proposes a TensorFlow XLA/LLVM compiler optimization pass for Multicore NPUs, called Tensor Slicing Optimization (TSO), which: (a) maximizes convolution parallelism and memory usage across NPU cores; and (b) reduces data transfers between host and NPU on-chip memories by using DRAM memory burst time estimates to guide tensor slicing. To evaluate the proposed approach, a set of experiments was performed using the NeuroMorphic Processor (NMP), a multicore NPU containing 32 RISC-V cores extended with novel CNN instructions. Experimental results show that TSO is capable of identifying the best tensor slicing that minimizes execution time for a set of CNN models. Speed-ups of up to 21.7\% result when comparing the TSO burst-based technique to a no-burst data slicing approach. To validate the generality of the TSO approach, the algorithm was also ported to the Glow Machine Learning framework. The performance of the models were measured on both Glow and TensorFlow XLA/LLVM compilers, revealing similar results.

Here, we optimized the annealing temperature of HZO/TiN thin film heterostructure via multiscale analysis of remnant polarization, crystallographic phase, minimum ferroelectric domain size, and average grain size. We found that the remnant polarization was closely related to the relative amount of the orthorhombic phase whereas the minimum domain size was to the relative amount of the monoclinic phase. The minimum domain size was obtained at the annealing temperature of 500$^\cird$C while the optimum remnant polarization and capacitance at the annealing temperature of 600$^\circ$C. We conclude that the minimum domain size is more important than the sheer magnitude of remnant polarization considering the retention and fatigue of switchable polarization in nanoscale ferroelectric devices. Our results are expected to contribute to the development of ultra-low-power logic transistors and next-generation non-volatile memory devices., Comment: 28 pages, 11 figures, 1 table

Under strong laser fields, electrons in solids radiate high-harmonic fields by travelling through quantum pathways in Bloch bands in the sub-laser-cycle timescales. Understanding these pathways in the momentum space through the high-harmonic radiation can enable an all-optical ultrafast probe to observe coherent lightwave-driven processes and measure electronic structures as recently demonstrated for semiconductors. However, such demonstration has been largely limited for semimetals because the absence of the bandgap hinders an experimental characterization of the exact pathways. In this study, by combining electrostatic control of chemical potentials with HHG measurement, we resolve quantum pathways of massless Dirac fermions in graphene under strong laser fields. Electrical modulation of HHG reveals quantum interference between the multi-photon interband excitation channels. As the light-matter interaction deviates beyond the perturbative regime, elliptically polarized laser fields efficiently drive massless Dirac fermions via an intricate coupling between the interband and intraband transitions, which is corroborated by our theoretical calculations. Our findings pave the way for strong-laser-field tomography of Dirac electrons in various quantum semimetals and their ultrafast electronics with a gate control., Comment: 33 pages, 4 figures, Soonyoung Cha and Minjeong Kim contributed equally to this work, To whom correspondence should be addressed: jonghwankim@postech

We consider an expected-value ranking and selection (R&S) problem where all k solutions' simulation outputs depend on a common parameter whose uncertainty can be modeled by a distribution. We define the most probable best (MPB) to be the solution that has the largest probability of being optimal with respect to the distribution and design an efficient sequential sampling algorithm to learn the MPB when the parameter has a finite support. We derive the large deviations rate of the probability of falsely selecting the MPB and formulate an optimal computing budget allocation problem to find the rate-maximizing static sampling ratios. The problem is then relaxed to obtain a set of optimality conditions that are interpretable and computationally efficient to verify. We devise a series of algorithms that replace the unknown means in the optimality conditions with their estimates and prove the algorithms' sampling ratios achieve the conditions as the simulation budget increases. Furthermore, we show that the empirical performances of the algorithms can be significantly improved by adopting the kernel ridge regression for mean estimation while achieving the same asymptotic convergence results. The algorithms are benchmarked against a state-of-the-art contextual R&S algorithm and demonstrated to have superior empirical performances.

Mobile devices run deep learning models for various purposes, such as image classification and speech recognition. Due to the resource constraints of mobile devices, researchers have focused on either making a lightweight deep neural network (DNN) model using model pruning or generating an efficient code using compiler optimization. Surprisingly, we found that the straightforward integration between model compression and compiler auto-tuning often does not produce the most efficient model for a target device. We propose CPrune, a compiler-informed model pruning for efficient target-aware DNN execution to support an application with a required target accuracy. CPrune makes a lightweight DNN model through informed pruning based on the structural information of subgraphs built during the compiler tuning process. Our experimental results show that CPrune increases the DNN execution speed up to 2.73x compared to the state-of-the-art TVM auto-tune while satisfying the accuracy requirement., Comment: 18 pages, 9 figures, accepted to ECCV 2022

To adopt convolutional neural networks (CNN) for a range of resource-constrained targets, it is necessary to compress the CNN models by performing quantization, whereby precision representation is converted to a lower bit representation. To overcome problems such as sensitivity of the training dataset, high computational requirements, and large time consumption, post-training quantization methods that do not require retraining have been proposed. In addition, to compensate for the accuracy drop without retraining, previous studies on post-training quantization have proposed several complementary methods: calibration, schemes, clipping, granularity, and mixed-precision. To generate a quantized model with minimal error, it is necessary to study all possible combinations of the methods because each of them is complementary and the CNN models have different characteristics. However, an exhaustive or a heuristic search is either too time-consuming or suboptimal. To overcome this challenge, we propose an auto-tuner known as Quantune, which builds a gradient tree boosting model to accelerate the search for the configurations of quantization and reduce the quantization error. We evaluate and compare Quantune with the random, grid, and genetic algorithms. The experimental results show that Quantune reduces the search time for quantization by approximately 36.5x with an accuracy loss of 0.07 ~ 0.65% across six CNN models, including the fragile ones (MobileNet, SqueezeNet, and ShuffleNet). To support multiple targets and adopt continuously evolving quantization works, Quantune is implemented on a full-fledged compiler for deep learning as an open-sourced project., Comment: 13 page, 9 figures, Accepted in Future Generation Computer Systems

An MRI-only adaptive radiotherapy (ART) workflow is desirable for managing interfractional changes in anatomy, but producing synthetic CT (sCT) data through paired data-driven deep learning (DL) for abdominal dose calculations remains a challenge due to the highly variable presence of intestinal gas. We present the preliminary dosimetric evaluation of our novel approach to sCT reconstruction that is well suited to handling intestinal gas in abdominal MRI-only ART. We utilize a paired data DL approach enabled by the intensity projection prior, in which well-matching training pairs are created by propagating air from MRI to corresponding CT scans. Evaluations focus on two classes: patients with (1) little involvement of intestinal gas, and (2) notable differences in intestinal gas presence between corresponding scans. Comparisons between sCT-based plans and CT-based clinical plans for both classes are made at the first treatment fraction to highlight the dosimetric impact of the variable presence of intestinal gas. Class 1 patients ($n=13$) demonstrate differences in prescribed dose coverage of the PTV of $1.3 \pm 2.1\%$ between clinical plans and sCT-based plans. Mean DVH differences in all structures for Class 1 patients are found to be statistically insignificant. In Class 2 ($n=20$), target coverage is $13.3 \pm 11.0\%$ higher in the clinical plans and mean DVH differences are found to be statistically significant. Significant deviations in calculated doses arising from the variable presence of intestinal gas in corresponding CT and MRI scans may limit the effectiveness of adaptive dose escalation efforts. We have proposed a paired data-driven DL approach to sCT reconstruction for accurate dose calculations in abdominal ART enabled by the creation of a clinically unavailable training data set with well-matching representations of intestinal gas.

It is a typical standard assumption in the density deconvolution problem that the characteristic function of the measurement error distribution is non-zero on the real line. While this condition is assumed in the majority of existing works on the topic, there are many problem instances of interest where it is violated. In this paper we focus on non--standard settings where the characteristic function of the measurement errors has zeros, and study how zeros multiplicity affects the estimation accuracy. For a prototypical problem of this type we demonstrate that the best achievable estimation accuracy is determined by the multiplicity of zeros, the rate of decay of the error characteristic function, as well as by the smoothness and the tail behavior of the estimated density. We derive lower bounds on the minimax risk and develop optimal in the minimax sense estimators. In addition, we consider the problem of adaptive estimation and propose a data-driven estimator that automatically adapts to unknown smoothness and tail behavior of the density to be estimated., Comment: 32 pages

Monolayer and few-layer phosphorene are anisotropic quasi-two-dimensional (quasi-2D) van der Waals (vdW) semiconductors with a linear-dichroic light-matter interaction and a widely-tunable direct-band gap in the infrared frequency range. Despite recent theoretical predictions of strongly-bound excitons with unique properties, it remains experimentally challenging to probe the excitonic quasiparticles due to the severe oxidation during device fabrication. In this study, we report observation of strongly-bound excitons and trions with highly-anisotropic optical properties in intrinsic bilayer phosphorene, which are protected from oxidation by encapsulation with hexagonal boron nitride (hBN), in a field-effect transistor (FET) geometry. Reflection contrast and photoluminescence spectroscopy clearly reveal the linear-dichroic optical spectra from anisotropic excitons and trions in the hBN-encapsulated bilayer phosphorene. The optical resonances from the exciton Rydberg series indicate that the neutral exciton binding energy is over 100 meV even with the dielectric screening from hBN. The electrostatic injection of free holes enables an additional optical resonance from a positive trion (charged exciton) ~ 30 meV below the optical bandgap of the charge-neutral system. Our work shows exciting possibilities for monolayer and few-layer phosphorene as a platform to explore many-body physics and novel photonics and optoelectronics based on strongly-bound excitons with two-fold anisotropy., Comment: 20 pages, 4 figures, Sangho Yoon, Taeho Kim contributed equally to this work