Understanding and Alleviating Memory Consumption in RLHF for LLMs

Fine-tuning with Reinforcement Learning with Human Feedback (RLHF) is essential for aligning large language models (LLMs). However, RLHF often encounters significant memory challenges. This study is the first to examine memory usage in the RLHF context, exploring various memory management strategies and unveiling the reasons behind excessive memory consumption. Additionally, we introduce a simple yet effective approach that substantially reduces the memory required for RLHF fine-tuning.

Dashing for the Golden Snitch: Multi-Drone Time-Optimal Motion Planning with Multi-Agent Reinforcement Learning

Recent innovations in autonomous drones have facilitated time-optimal flight in single-drone configurations and enhanced maneuverability in multi-drone systems through the application of optimal control and learning-based methods. However, few studies have achieved time-optimal motion planning for multi-drone systems, particularly during highly agile maneuvers or in dynamic scenarios. This paper presents a decentralized policy network for time-optimal multi-drone flight using multi-agent reinforcement learning. To strike a balance between flight efficiency and collision avoidance, we introduce a soft collision penalty inspired by optimization-based methods. By customizing PPO in a centralized training, decentralized execution (CTDE) fashion, we unlock higher efficiency and stability in training, while ensuring lightweight implementation. Extensive simulations show that, despite slight performance trade-offs compared to single-drone systems, our multi-drone approach maintains near-time-optimal performance with low collision rates. Real-world experiments validate our method, with two quadrotors using the same network as simulation achieving a maximum speed of 13.65 m/s and a maximum body rate of 13.4 rad/s in a 5.5 m * 5.5 m * 2.0 m space across various tracks, relying entirely on onboard computation.

ProTrain: Efficient LLM Training via Memory-Aware Techniques

It is extremely memory-hungry to train Large Language Models (LLM). To solve this problem, existing work exploits the combination of CPU and GPU for the training process, such as ZeRO-Offload. Such a technique largely democratizes billion-scale model training, making it possible to train with few consumer graphics cards. However, based on our observation, existing frameworks often provide coarse-grained memory management and require experienced experts in configuration tuning, leading to suboptimal hardware utilization and performance. This paper proposes ProTrain, a novel training system that intelligently balances memory usage and performance by coordinating memory, computation, and IO. ProTrain achieves adaptive memory management through Chunk-Based Model State Management and Block-Wise Activation Management, guided by a Memory-Aware Runtime Profiler without user intervention. ProTrain does not change the training algorithm and thus does not compromise accuracy. Experiments show that ProTrain improves training throughput by 1.43$\times$ to 2.71$\times$ compared to the SOTA training systems.

Online Time-Optimal Trajectory Generation for Two Quadrotors with Multi-Waypoints Constraints

The autonomous quadrotor's flying speed has kept increasing in the past 5 years, especially in the field of autonomous drone racing. However, the majority of the research mainly focuses on the aggressive flight of a single quadrotor. In this letter, we propose a novel method called Pairwise Model Predictive Control (PMPC) that can guide two quadrotors online to fly through the waypoints with minimum time without collisions. The flight task is first modeled as a nonlinear optimization problem and then an efficient two-step mass point velocity search method is used to provide initial values and references to improve the solving efficiency so that the method can run online with a frequency of 50 Hz and can handle dynamic waypoints. The simulation and real-world experiments validate the feasibility of the proposed method and in the real-world experiments, the two quadrotors can achieve a top speed of 8.1m/s in a 6-waypoint racing track in a compact flying arena of 6m*4m*2m.

Imitation Learning-Based Online Time-Optimal Control with Multiple-Waypoint Constraints for Quadrotors

Over the past decade, there has been a remarkable surge in utilizing quadrotors for various purposes due to their simple structure and aggressive maneuverability, such as search and rescue, delivery and autonomous drone racing, etc. One of the key challenges preventing quadrotors from being widely used in these scenarios is online waypoint-constrained time-optimal trajectory generation and control technique. This letter proposes an imitation learning-based online solution to efficiently navigate the quadrotor through multiple waypoints with time-optimal performance. The neural networks (WN&CNets) are trained to learn the control law from the dataset generated by the time-consuming CPC algorithm and then deployed to generate the optimal control commands online to guide the quadrotors. To address the challenge of limited training data and the hover maneuver at the final waypoint, we propose a transition phase strategy that utilizes polynomials to help the quadrotor 'jump over' the stop-and-go maneuver when switching waypoints. Our method is demonstrated in both simulation and real-world experiments, achieving a maximum speed of 7 m/s while navigating through 7 waypoints in a confined space of 6.0 m * 4.0 m * 2.0 m. The results show that with a slight loss in optimality, the WN&CNets significantly reduce the processing time and enable online optimal control for multiple-waypoint-constrained flight tasks.

Design and Testbed Deployment of Frequency-Domain Equalization Full Duplex Radios

Full-duplex (FD) wireless can significantly enhance spectrum efficiency but requires effective self-interference (SI) cancellers. RF SI cancellation (SIC) via frequency-domain equalization (FDE), where bandpass filters channelize the SI, is suited for integrated circuits (ICs). In this paper, we explore the limits and higher layer challenges associated with using such cancellers. We evaluate the performance of a custom FDE-based canceller using two testbeds; one with mobile FD radios and the other with upgraded, static FD radios in the PAWR COSMOS testbed. The latter is a lasting artifact for the research community, alongside a dataset containing baseband waveforms captured on the COSMOS FD radios, facilitating FD-related experimentation at the higher networking layers. We evaluate the performance of the FDE-based FD radios in both testbeds, with experiments showing 95 dB overall achieved SIC (52 dB from RF SIC) across 20 MHz bandwidth, and an average link-level FD rate gain of 1.87x. We also conduct experiments in (i) uplink-downlink networks with inter-user interference, and (ii) heterogeneous networks with half-duplex and FD users. The experimental FD gains in the two types of networks depend on the users' SNR values and the number of FD users, and are 1.14x-1.25x and 1.25x-1.73x, respectively, confirming previous analytical results.

USFM: A Universal Ultrasound Foundation Model Generalized to Tasks and Organs towards Label Efficient Image Analysis

Inadequate generality across different organs and tasks constrains the application of ultrasound (US) image analysis methods in smart healthcare. Building a universal US foundation model holds the potential to address these issues. Nevertheless, the development of such foundational models encounters intrinsic challenges in US analysis, i.e., insufficient databases, low quality, and ineffective features. In this paper, we present a universal US foundation model, named USFM, generalized to diverse tasks and organs towards label efficient US image analysis. First, a large-scale Multi-organ, Multi-center, and Multi-device US database was built, comprehensively containing over two million US images. Organ-balanced sampling was employed for unbiased learning. Then, USFM is self-supervised pre-trained on the sufficient US database. To extract the effective features from low-quality US images, we proposed a spatial-frequency dual masked image modeling method. A productive spatial noise addition-recovery approach was designed to learn meaningful US information robustly, while a novel frequency band-stop masking learning approach was also employed to extract complex, implicit grayscale distribution and textural variations. Extensive experiments were conducted on the various tasks of segmentation, classification, and image enhancement from diverse organs and diseases. Comparisons with representative US image analysis models illustrate the universality and effectiveness of USFM. The label efficiency experiments suggest the USFM obtains robust performance with only 20% annotation, laying the groundwork for the rapid development of US models in clinical practices.

Weakly Supervised Lesion Detection and Diagnosis for Breast Cancers with Partially Annotated Ultrasound Images

Deep learning (DL) has proven highly effective for ultrasound-based computer-aided diagnosis (CAD) of breast cancers. In an automaticCAD system, lesion detection is critical for the following diagnosis. However, existing DL-based methods generally require voluminous manually-annotated region of interest (ROI) labels and class labels to train both the lesion detection and diagnosis models. In clinical practice, the ROI labels, i.e. ground truths, may not always be optimal for the classification task due to individual experience of sonologists, resulting in the issue of coarse annotation that limits the diagnosis performance of a CAD model. To address this issue, a novel Two-Stage Detection and Diagnosis Network (TSDDNet) is proposed based on weakly supervised learning to enhance diagnostic accuracy of the ultrasound-based CAD for breast cancers. In particular, all the ROI-level labels are considered as coarse labels in the first training stage, and then a candidate selection mechanism is designed to identify optimallesion areas for both the fully and partially annotated samples. It refines the current ROI-level labels in the fully annotated images and the detected ROIs in the partially annotated samples with a weakly supervised manner under the guidance of class labels. In the second training stage, a self-distillation strategy further is further proposed to integrate the detection network and classification network into a unified framework as the final CAD model for joint optimization, which then further improves the diagnosis performance. The proposed TSDDNet is evaluated on a B-mode ultrasound dataset, and the experimental results show that it achieves the best performance on both lesion detection and diagnosis tasks, suggesting promising application potential.

Aggressive Trajectory Generation for A Swarm of Autonomous Racing Drones

Autonomous drone racing is becoming an excellent platform to challenge quadrotors' autonomy techniques including planning, navigation and control technologies. However, most research on this topic mainly focuses on single drone scenarios. In this paper, we describe a novel time-optimal trajectory generation method for generating time-optimal trajectories for a swarm of quadrotors to fly through pre-defined waypoints with their maximum maneuverability without collision. We verify the method in the Gazebo simulations where a swarm of 5 quadrotors can fly through a complex 6-waypoint racing track in a 35m * 35m space with a top speed of 14m/s. Flight tests are performed on two quadrotors passing through 3 waypoints in a 4m * 2m flight arena to demonstrate the feasibility of the proposed method in the real world. Both simulations and real-world flight tests show that the proposed method can generate the optimal aggressive trajectories for a swarm of autonomous racing drones. The method can also be easily transferred to other types of robot swarms.

RCMNet: A deep learning model assists CAR-T therapy for leukemia

Acute leukemia is a type of blood cancer with a high mortality rate. Current therapeutic methods include bone marrow transplantation, supportive therapy, and chemotherapy. Although a satisfactory remission of the disease can be achieved, the risk of recurrence is still high. Therefore, novel treatments are demanding. Chimeric antigen receptor-T (CAR-T) therapy has emerged as a promising approach to treat and cure acute leukemia. To harness the therapeutic potential of CAR-T cell therapy for blood diseases, reliable cell morphological identification is crucial. Nevertheless, the identification of CAR-T cells is a big challenge posed by their phenotypic similarity with other blood cells. To address this substantial clinical challenge, herein we first construct a CAR-T dataset with 500 original microscopy images after staining. Following that, we create a novel integrated model called RCMNet (ResNet18 with CBAM and MHSA) that combines the convolutional neural network (CNN) and Transformer. The model shows 99.63% top-1 accuracy on the public dataset. Compared with previous reports, our model obtains satisfactory results for image classification. Although testing on the CAR-T cells dataset, a decent performance is observed, which is attributed to the limited size of the dataset. Transfer learning is adapted for RCMNet and a maximum of 83.36% accuracy has been achieved, which is higher than other SOTA models. The study evaluates the effectiveness of RCMNet on a big public dataset and translates it to a clinical dataset for diagnostic applications.