Harnessing Your DRAM and SSD for Sustainable and Accessible LLM Inference with Mixed-Precision and Multi-level Caching

Although Large Language Models (LLMs) have demonstrated remarkable capabilities, their massive parameter counts and associated extensive computing make LLMs' deployment the main part of carbon emission from nowadays AI applications. Compared to modern GPUs like H$100$, it would be significantly carbon-sustainable if we could leverage old-fashioned GPUs such as M$40$ (as shown in Figure 1, M$40$ only has one third carbon emission of H$100$'s) for LLM servings. However, the limited High Bandwidth Memory (HBM) available on such GPU often cannot support the loading of LLMs due to the gigantic model size and intermediate activation data, making their serving challenging. For instance, a LLaMA2 model with $70$B parameters typically requires $128$GB for inference, which substantially surpasses $24$GB HBM in a $3090$ GPU and remains infeasible even considering the additional $64$GB DRAM. To address this challenge, this paper proposes a mixed-precision with a model modularization algorithm to enable LLM inference on outdated hardware with resource constraints. (The precision denotes the numerical precision like FP16, INT8, INT4) and multi-level caching (M2Cache).) Specifically, our M2Cache first modulizes neurons in LLM and creates their importance ranking. Then, it adopts a dynamic sparse mixed-precision quantization mechanism in weight space to reduce computational demands and communication overhead at each decoding step. It collectively lowers the operational carbon emissions associated with LLM inference. Moreover, M2Cache introduces a three-level cache management system with HBM, DRAM, and SSDs that complements the dynamic sparse mixed-precision inference. To enhance communication efficiency, M2Cache maintains a neuron-level mixed-precision LRU cache in HBM, a larger layer-aware cache in DRAM, and a full model in SSD.

Flex-MoE: Modeling Arbitrary Modality Combination via the Flexible Mixture-of-Experts

Multimodal learning has gained increasing importance across various fields, offering the ability to integrate data from diverse sources such as images, text, and personalized records, which are frequently observed in medical domains. However, in scenarios where some modalities are missing, many existing frameworks struggle to accommodate arbitrary modality combinations, often relying heavily on a single modality or complete data. This oversight of potential modality combinations limits their applicability in real-world situations. To address this challenge, we propose Flex-MoE (Flexible Mixture-of-Experts), a new framework designed to flexibly incorporate arbitrary modality combinations while maintaining robustness to missing data. The core idea of Flex-MoE is to first address missing modalities using a new missing modality bank that integrates observed modality combinations with the corresponding missing ones. This is followed by a uniquely designed Sparse MoE framework. Specifically, Flex-MoE first trains experts using samples with all modalities to inject generalized knowledge through the generalized router ($\mathcal{G}$-Router). The $\mathcal{S}$-Router then specializes in handling fewer modality combinations by assigning the top-1 gate to the expert corresponding to the observed modality combination. We evaluate Flex-MoE on the ADNI dataset, which encompasses four modalities in the Alzheimer's Disease domain, as well as on the MIMIC-IV dataset. The results demonstrate the effectiveness of Flex-MoE highlighting its ability to model arbitrary modality combinations in diverse missing modality scenarios. Code is available at https://github.com/UNITES-Lab/flex-moe.

Glider: Global and Local Instruction-Driven Expert Router

The availability of performant pre-trained models has led to a proliferation of fine-tuned expert models that are specialized to particular domains. This has enabled the creation of powerful and adaptive routing-based "Model MoErging" methods with the goal of using expert modules to create an aggregate system with improved performance or generalization. However, existing MoErging methods often prioritize generalization to unseen tasks at the expense of performance on held-in tasks, which limits its practical applicability in real-world deployment scenarios. We observe that current token-level routing mechanisms neglect the global semantic context of the input task. This token-wise independence hinders effective expert selection for held-in tasks, as routing decisions fail to incorporate the semantic properties of the task. To address this, we propose, Global and Local Instruction Driven Expert Router (GLIDER) that integrates a multi-scale routing mechanism, encompassing a semantic global router and a learned local router. The global router leverages LLM's advanced reasoning capabilities for semantic-related contexts to enhance expert selection. Given the input query and LLM, the router generates semantic task instructions that guide the retrieval of the most relevant experts across all layers. This global guidance is complemented by a local router that facilitates token-level routing decisions within each module, enabling finer control and enhanced performance on unseen tasks. Our experiments using T5-based models for T0 and FLAN tasks demonstrate that GLIDER achieves substantially improved held-in performance while maintaining strong generalization on held-out tasks. We also perform ablations experiments to dive deeper into the components of GLIDER. Our experiments highlight the importance of our multi-scale routing that leverages LLM-driven semantic reasoning for MoErging methods.

Accelerating the discovery of low-energy structure configurations: a computational approach that integrates first-principles calculations, Monte Carlo sampling, and Machine Learning

Finding Minimum Energy Configurations (MECs) is essential in fields such as physics, chemistry, and materials science, as they represent the most stable states of the systems. In particular, identifying such MECs in multi-component alloys considered candidate PFMs is key because it determines the most stable arrangement of atoms within the alloy, directly influencing its phase stability, structural integrity, and thermo-mechanical properties. However, since the search space grows exponentially with the number of atoms considered, obtaining such MECs using computationally expensive first-principles DFT calculations often results in a cumbersome task. To escape the above compromise between physical fidelity and computational efficiency, we have developed a novel physics-based data-driven approach that combines Monte Carlo sampling, first-principles DFT calculations, and Machine Learning to accelerate the discovery of MECs in multi-component alloys. More specifically, we have leveraged well-established Cluster Expansion (CE) techniques with Local Outlier Factor models to establish strategies that enhance the reliability of the CE method. In this work, we demonstrated the capabilities of the proposed approach for the particular case of a tungsten-based quaternary high-entropy alloy. However, the method is applicable to other types of alloys and enables a wide range of applications.

MHRC: Closed-loop Decentralized Multi-Heterogeneous Robot Collaboration with Large Language Models

The integration of large language models (LLMs) with robotics has significantly advanced robots' abilities in perception, cognition, and task planning. The use of natural language interfaces offers a unified approach for expressing the capability differences of heterogeneous robots, facilitating communication between them, and enabling seamless task allocation and collaboration. Currently, the utilization of LLMs to achieve decentralized multi-heterogeneous robot collaborative tasks remains an under-explored area of research. In this paper, we introduce a novel framework that utilizes LLMs to achieve decentralized collaboration among multiple heterogeneous robots. Our framework supports three robot categories, mobile robots, manipulation robots, and mobile manipulation robots, working together to complete tasks such as exploration, transportation, and organization. We developed a rich set of textual feedback mechanisms and chain-of-thought (CoT) prompts to enhance task planning efficiency and overall system performance. The mobile manipulation robot can adjust its base position flexibly, ensuring optimal conditions for grasping tasks. The manipulation robot can comprehend task requirements, seek assistance when necessary, and handle objects appropriately. Meanwhile, the mobile robot can explore the environment extensively, map object locations, and communicate this information to the mobile manipulation robot, thus improving task execution efficiency. We evaluated the framework using PyBullet, creating scenarios with three different room layouts and three distinct operational tasks. We tested various LLM models and conducted ablation studies to assess the contributions of different modules. The experimental results confirm the effectiveness and necessity of our proposed framework.

AnyDesign: Versatile Area Fashion Editing via Mask-Free Diffusion

Fashion image editing aims to modify a person's appearance based on a given instruction. Existing methods require auxiliary tools like segmenters and keypoint extractors, lacking a flexible and unified framework. Moreover, these methods are limited in the variety of clothing types they can handle, as most datasets focus on people in clean backgrounds and only include generic garments such as tops, pants, and dresses. These limitations restrict their applicability in real-world scenarios. In this paper, we first extend an existing dataset for human generation to include a wider range of apparel and more complex backgrounds. This extended dataset features people wearing diverse items such as tops, pants, dresses, skirts, headwear, scarves, shoes, socks, and bags. Additionally, we propose AnyDesign, a diffusion-based method that enables mask-free editing on versatile areas. Users can simply input a human image along with a corresponding prompt in either text or image format. Our approach incorporates Fashion DiT, equipped with a Fashion-Guidance Attention (FGA) module designed to fuse explicit apparel types and CLIP-encoded apparel features. Both Qualitative and quantitative experiments demonstrate that our method delivers high-quality fashion editing and outperforms contemporary text-guided fashion editing methods.

Beyond Over-smoothing: Uncovering the Trainability Challenges in Deep Graph Neural Networks

The drastic performance degradation of Graph Neural Networks (GNNs) as the depth of the graph propagation layers exceeds 8-10 is widely attributed to a phenomenon of Over-smoothing. Although recent research suggests that Over-smoothing may not be the dominant reason for such a performance degradation, they have not provided rigorous analysis from a theoretical view, which warrants further investigation. In this paper, we systematically analyze the real dominant problem in deep GNNs and identify the issues that these GNNs towards addressing Over-smoothing essentially work on via empirical experiments and theoretical gradient analysis. We theoretically prove that the difficult training problem of deep MLPs is actually the main challenge, and various existing methods that supposedly tackle Over-smoothing actually improve the trainability of MLPs, which is the main reason for their performance gains. Our further investigation into trainability issues reveals that properly constrained smaller upper bounds of gradient flow notably enhance the trainability of GNNs. Experimental results on diverse datasets demonstrate consistency between our theoretical findings and empirical evidence. Our analysis provides new insights in constructing deep graph models.

Mew: Multiplexed Immunofluorescence Image Analysis through an Efficient Multiplex Network

Recent advancements in graph-based approaches for multiplexed immunofluorescence (mIF) images have significantly propelled the field forward, offering deeper insights into patient-level phenotyping. However, current graph-based methodologies encounter two primary challenges: (1) Cellular Heterogeneity, where existing approaches fail to adequately address the inductive biases inherent in graphs, particularly the homophily characteristic observed in cellular connectivity and; (2) Scalability, where handling cellular graphs from high-dimensional images faces difficulties in managing a high number of cells. To overcome these limitations, we introduce Mew, a novel framework designed to efficiently process mIF images through the lens of multiplex network. Mew innovatively constructs a multiplex network comprising two distinct layers: a Voronoi network for geometric information and a Cell-type network for capturing cell-wise homogeneity. This framework equips a scalable and efficient Graph Neural Network (GNN), capable of processing the entire graph during training. Furthermore, Mew integrates an interpretable attention module that autonomously identifies relevant layers for image classification. Extensive experiments on a real-world patient dataset from various institutions highlight Mew's remarkable efficacy and efficiency, marking a significant advancement in mIF image analysis. The source code of Mew can be found here: \url{https://github.com/UNITES-Lab/Mew}

LDP: A Local Diffusion Planner for Efficient Robot Navigation and Collision Avoidance

The conditional diffusion model has been demonstrated as an efficient tool for learning robot policies, owing to its advancement to accurately model the conditional distribution of policies. The intricate nature of real-world scenarios, characterized by dynamic obstacles and maze-like structures, underscores the complexity of robot local navigation decision-making as a conditional distribution problem. Nevertheless, leveraging the diffusion model for robot local navigation is not trivial and encounters several under-explored challenges: (1) Data Urgency. The complex conditional distribution in local navigation needs training data to include diverse policy in diverse real-world scenarios; (2) Myopic Observation. Due to the diversity of the perception scenarios, diffusion decisions based on the local perspective of robots may prove suboptimal for completing the entire task, as they often lack foresight. In certain scenarios requiring detours, the robot may become trapped. To address these issues, our approach begins with an exploration of a diverse data generation mechanism that encompasses multiple agents exhibiting distinct preferences through target selection informed by integrated global-local insights. Then, based on this diverse training data, a diffusion agent is obtained, capable of excellent collision avoidance in diverse scenarios. Subsequently, we augment our Local Diffusion Planner, also known as LDP by incorporating global observations in a lightweight manner. This enhancement broadens the observational scope of LDP, effectively mitigating the risk of becoming ensnared in local optima and promoting more robust navigational decisions.

Human-like object concept representations emerge naturally in multimodal large language models

The conceptualization and categorization of natural objects in the human mind have long intrigued cognitive scientists and neuroscientists, offering crucial insights into human perception and cognition. Recently, the rapid development of Large Language Models (LLMs) has raised the attractive question of whether these models can also develop human-like object representations through exposure to vast amounts of linguistic and multimodal data. In this study, we combined behavioral and neuroimaging analysis methods to uncover how the object concept representations in LLMs correlate with those of humans. By collecting large-scale datasets of 4.7 million triplet judgments from LLM and Multimodal LLM (MLLM), we were able to derive low-dimensional embeddings that capture the underlying similarity structure of 1,854 natural objects. The resulting 66-dimensional embeddings were found to be highly stable and predictive, and exhibited semantic clustering akin to human mental representations. Interestingly, the interpretability of the dimensions underlying these embeddings suggests that LLM and MLLM have developed human-like conceptual representations of natural objects. Further analysis demonstrated strong alignment between the identified model embeddings and neural activity patterns in many functionally defined brain ROIs (e.g., EBA, PPA, RSC and FFA). This provides compelling evidence that the object representations in LLMs, while not identical to those in the human, share fundamental commonalities that reflect key schemas of human conceptual knowledge. This study advances our understanding of machine intelligence and informs the development of more human-like artificial cognitive systems.