Search

With the commercialization of large language models (LLMs), weight-activation quantization has emerged to compress and accelerate LLMs, achieving high throughput while reducing inference costs. However, existing post-training quantization (PTQ) techniques for quantizing weights and activations of LLMs still suffer from non-negligible accuracy drops, especially on massive multitask language understanding. To address this issue, we propose Low-Rank Quantization (LRQ) $-$ a simple yet effective post-training weight quantization method for LLMs that reconstructs the outputs of an intermediate Transformer block by leveraging low-rank weight-scaling matrices, replacing the conventional full weight-scaling matrices that entail as many learnable scales as their associated weights. Thanks to parameter sharing via low-rank structure, LRQ only needs to learn significantly fewer parameters while enabling the individual scaling of weights, thus boosting the generalization capability of quantized LLMs. We show the superiority of LRQ over prior LLM PTQ works under (i) $8$-bit weight and per-tensor activation quantization, (ii) $4$-bit weight and $8$-bit per-token activation quantization, and (iii) low-bit weight-only quantization schemes. Our code is available at \url{https://github.com/onliwad101/FlexRound_LRQ} to inspire LLM researchers and engineers., Comment: Preprint

Multi-hop logical reasoning on knowledge graphs is a pivotal task in natural language processing, with numerous approaches aiming to answer First-Order Logic (FOL) queries. Recent geometry (e.g., box, cone) and probability (e.g., beta distribution)-based methodologies have effectively addressed complex FOL queries. However, a common challenge across these methods lies in determining accurate geometric bounds or probability parameters for these queries. The challenge arises because existing methods rely on linear sequential operations within their computation graphs, overlooking the logical structure of the query and the relation-induced information that can be gleaned from the relations of the query, which we call the context of the query. To address the problem, we propose a model-agnostic methodology that enhances the effectiveness of existing multi-hop logical reasoning approaches by fully integrating the context of the FOL query graph. Our approach distinctively discerns (1) the structural context inherent to the query structure and (2) the relation-induced context unique to each node in the query graph as delineated in the corresponding knowledge graph. This dual-context paradigm helps nodes within a query graph attain refined internal representations throughout the multi-hop reasoning steps. Through experiments on two datasets, our method consistently enhances the three multi-hop reasoning foundation models, achieving performance improvements of up to 19.5%. Our code is available at https://github.com/kjh9503/caqr., Comment: Accepted to ACL 2024 Findings

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 the realm of LiDAR-based perception, significant strides have been made, yet domain generalization remains a substantial challenge. The performance often deteriorates when models are applied to unfamiliar datasets with different LiDAR sensors or deployed in new environments, primarily due to variations in point cloud density distributions. To tackle this challenge, we propose a Density Discriminative Feature Embedding (DDFE) module, capitalizing on the observation that a single source LiDAR point cloud encompasses a spectrum of densities. The DDFE module is meticulously designed to extract density-specific features within a single source domain, facilitating the recognition of objects sharing similar density characteristics across different LiDAR sensors. In addition, we introduce a simple yet effective density augmentation technique aimed at expanding the spectrum of density in source data, thereby enhancing the capabilities of the DDFE. Our DDFE stands out as a versatile and lightweight domain generalization module. It can be seamlessly integrated into various 3D backbone networks, where it has demonstrated superior performance over current state-of-the-art domain generalization methods. Code is available at https://github.com/dgist-cvlab/MultiDensityDG., Comment: Accepted by ECCV 2024

Large Language Models (LLMs) have recently demonstrated remarkable success across various tasks. However, efficiently serving LLMs has been a challenge due to the large memory bottleneck, specifically in small batch inference settings (e.g. mobile devices). Weight-only quantization can be a promising approach, but sub-4 bit quantization remains a challenge due to large-magnitude activation outliers. To mitigate the undesirable outlier effect, we first propose per-IC quantization, a simple yet effective method that creates quantization groups within each input channel (IC) rather than the conventional per-output-channel (per-OC). Our method is motivated by the observation that activation outliers affect the input dimension of the weight matrix, so similarly grouping the weights in the IC direction can isolate outliers within a group. We also find that activation outliers do not dictate quantization difficulty, and inherent weight sensitivities also exist. With per-IC quantization as a new outlier-friendly scheme, we propose Adaptive Dimensions (AdaDim), a versatile quantization framework that can adapt to various weight sensitivity patterns. We demonstrate the effectiveness of AdaDim by augmenting prior methods such as Round-To-Nearest and GPTQ, showing significant improvements across various language modeling benchmarks for both base (up to +4.7% on MMLU) and instruction-tuned (up to +10% on HumanEval) LLMs. Code is available at https://github.com/johnheo/adadim-llm, Comment: ICLR 2024. 19 pages, 11 figures, 10 tables

The rapid development of 6G communications using terahertz (THz) electromagnetic waves has created a demand for highly sensitive THz nanoresonators capable of detecting these waves. Among the potential candidates, THz nanogap loop arrays show promising characteristics but require significant computational resources for accurate simulation. This requirement arises because their unit cells are 10 times smaller than millimeter wavelengths, with nanogap regions that are 1,000,000 times smaller. To address this challenge, we propose a rapid inverse design method using physics-informed machine learning, employing double deep Q-learning with an analytical model of the THz nanogap loop array. In about 39 hours on a middle-level personal computer, our approach identifies the optimal structure through 200,000 iterations, achieving experimental electric field enhancement of 32,000 at 0.2 THz, 300 % stronger than prior results. Our analytical model-based approach significantly reduces computational resources, offering a practical alternative to numerical simulation-based inverse design

Post-training quantization (PTQ) has been gaining popularity for the deployment of deep neural networks on resource-limited devices since unlike quantization-aware training, neither a full training dataset nor end-to-end training is required at all. As PTQ schemes based on reconstructing each layer or block output turn out to be effective to enhance quantized model performance, recent works have developed algorithms to devise and learn a new weight-rounding scheme so as to better reconstruct each layer or block output. In this work, we propose a simple yet effective new weight-rounding mechanism for PTQ, coined \emph{FlexRound}, based on element-wise division instead of typical element-wise addition such that FlexRound enables jointly learning a common quantization grid size as well as a different scale for each pre-trained weight. Thanks to the reciprocal rule of derivatives induced by element-wise division, FlexRound is inherently able to exploit pre-trained weights when updating their corresponding scales, and thus, flexibly quantize pre-trained weights depending on their magnitudes. We empirically validate the efficacy of FlexRound on a wide range of models and tasks. To the best of our knowledge, our work is the first to carry out comprehensive experiments on not only image classification and natural language understanding but also natural language generation. Moreover, we demonstrate, for the first time, that large language models can be efficiently quantized, with only a negligible impact on performance compared to half-precision baselines, achieved by reconstructing the output in a block-by-block manner. Our code is available at \url{https://github.com/onliwad101/FlexRound_LRQ}., Comment: Accepted to ICML 2023

Large language models (LLMs) face the challenges in fine-tuning and deployment due to their high memory demands and computational costs. While parameter-efficient fine-tuning (PEFT) methods aim to reduce the memory usage of the optimizer state during fine-tuning, the inherent size of pre-trained LLM weights continues to be a pressing concern. Even though quantization techniques are widely proposed to ease memory demands and accelerate LLM inference, most of these techniques are geared towards the deployment phase. To bridge this gap, this paper presents Parameter-Efficient and Quantization-aware Adaptation (PEQA) - a simple yet effective method that combines the advantages of PEFT with quantized LLMs. By updating solely the quantization scales, PEQA can be directly applied to quantized LLMs, ensuring seamless task transitions. Parallel to existing PEFT methods, PEQA significantly reduces the memory overhead associated with the optimizer state. Furthermore, it leverages the advantages of quantization to substantially reduce model sizes. Even after fine-tuning, the quantization structure of a PEQA-tuned LLM remains intact, allowing for accelerated inference on the deployment stage. We employ PEQA-tuning for task-specific adaptation on LLMs with up to 65 billion parameters. To assess the logical reasoning and language comprehension of PEQA-tuned LLMs, we fine-tune low-bit quantized LLMs using a instruction dataset. Our results show that even when LLMs are quantized to below 4-bit precision, their capabilities in language modeling, few-shot in-context learning, and comprehension can be resiliently restored to (or even improved over) their full-precision original performances with PEQA., Comment: Published at NeurIPS 2023. Camera-ready version

There are growing interests in adapting large-scale language models using parameter-efficient fine-tuning methods. However, accelerating the model itself and achieving better inference efficiency through model compression has not been thoroughly explored yet. Model compression could provide the benefits of reducing memory footprints, enabling low-precision computations, and ultimately achieving cost-effective inference. To combine parameter-efficient adaptation and model compression, we propose AlphaTuning consisting of post-training quantization of the pre-trained language model and fine-tuning only some parts of quantized parameters for a target task. Specifically, AlphaTuning works by employing binary-coding quantization, which factorizes the full-precision parameters into binary parameters and a separate set of scaling factors. During the adaptation phase, the binary values are frozen for all tasks, while the scaling factors are fine-tuned for the downstream task. We demonstrate that AlphaTuning, when applied to GPT-2 and OPT, performs competitively with full fine-tuning on a variety of downstream tasks while achieving >10x compression ratio under 4-bit quantization and >1,000x reduction in the number of trainable parameters., Comment: Findings of EMNLP 2022

Recent advances in self-supervised learning and the Transformer architecture have significantly improved natural language processing (NLP), achieving remarkably low perplexity. However, the growing size of NLP models introduces a memory wall problem during the generation phase. To mitigate this issue, recent efforts have focused on quantizing model weights to sub-4-bit precision while preserving full precision for activations, resulting in practical speed-ups during inference on a single GPU. However, these improvements primarily stem from reduced memory movement, which necessitates a resource-intensive dequantization process rather than actual computational reduction. In this paper, we introduce LUT-GEMM, an efficient kernel for quantized matrix multiplication, which not only eliminates the resource-intensive dequantization process but also reduces computational costs compared to previous kernels for weight-only quantization. Furthermore, we proposed group-wise quantization to offer a flexible trade-off between compression ratio and accuracy. The impact of LUT-GEMM is facilitated by implementing high compression ratios through low-bit quantization and efficient LUT-based operations. We show experimentally that when applied to the OPT-175B model with 3-bit quantization, LUT-GEMM substantially accelerates token generation latency, achieving a remarkable 2.1$\times$ improvement on a single GPU when compared to OPTQ, which relies on the costly dequantization process., Comment: ICLR 2024

Federated Learning (FL) has received a significant amount of attention in the industry and research community due to its capability of keeping data on local devices. To aggregate the gradients of local models to train the global model, existing works require that the global model and the local models are the same. However, Internet of Things (IoT) devices are inherently diverse regarding computation speed and onboard memory. In this paper, we propose an FL framework targeting the heterogeneity of IoT devices. Specifically, local models are compressed from the global model, and the gradients of the compressed local models are used to update the global model. We conduct preliminary experiments to illustrate that our framework can facilitate the design of IoT-aware FL.

Novelty detection using deep generative models such as autoencoder, generative adversarial networks mostly takes image reconstruction error as novelty score function. However, image data, high dimensional as it is, contains a lot of different features other than class information which makes models hard to detect novelty data. The problem gets harder in multi-modal normality case. To address this challenge, we propose a new way of measuring novelty score in multi-modal normality cases using orthogonalized latent space. Specifically, we employ orthogonal low-rank embedding in the latent space to disentangle the features in the latent space using mutual class information. With the orthogonalized latent space, novelty score is defined by the change of each latent vector. Proposed algorithm was compared to state-of-the-art novelty detection algorithms using GAN such as RaPP and OCGAN, and experimental results show that ours outperforms those algorithms., Comment: Accepted to AAAI 2021 Workshop: Towards Robust, Secure and Efficient Machine Learning