Modern research computing cyberinfrastructure, such as ACCESS-CI and NAIRR Pilot, offers GPU resources across geographically distributed clusters to accommodate the increasing needs of scientific deep learning (DL) workloads. Even for high-performance computing (HPC) experts, configuring environments and managing DL workloads across supercomputers remain significant barriers. To address these obstacles, we present Diamond, an open-source platform to simplify and streamline the DL lifecycle on HPC. Diamond provides an intuitive graphical interface that abstracts system-level complexity, enabling users to develop, debug, and deploy DL models with minimal overhead. We identify several challenges in building such a platform, including portability, security, and usability, and propose effective architectural solutions to each. Notably, Diamond enables users to share and reuse DL workload environments across systems and collaborators, reducing redundant setup efforts. Experimental results demonstrate that Diamond reduces the time to first successful deployment by an average of 68%, compared to manual configuration with command lines. The Diamond service is available at https://diamondhpc.ai.

While deep learning (DL) has emerged as a powerful technique, its benefits must be carefully considered in relation to computational costs. Specifically, although DL methods have achieved strong performance in log anomaly detection, they often require extended time for log preprocessing, model training, and model inference, hindering their adoption in online distributed cloud systems that require rapid deployment of log anomaly detection service.

Reliable control schemes are critical to ensuring converter operation in microgrids. This work proposes a robust adaptive sliding mode control for the three-level neutral-point-clamped power converter with multidisturbances. Specifically, an adaptive observer-based proportional (P) voltage controller is proposed to accurately and quickly regulate dc-voltage in real-time identifying unknown equivalent dc-loads, compensating for the active power reference. To track the reference, a disturbance observer-based integral sliding mode controller (ISMC) is adopted to dramatically enhance the power tracking performance in case of parameter mismatches and bias caused by current path changes and switch mode noise. In addition, a sliding mode observer coupled withPcontrol is established to balance dc-link (two) capacitors, rejecting harmonic injection and power ripple behaviors. Experimental data confirm that the proposed control scheme outperforms super-twisting observer-based ISMC, super-twisting algorithm, and proportional–integral control schemes in the transient/steady state operations in terms of dynamic response, grid current harmonic distortions, and robustness.

Three-level neutral-point-clamped (3L-NPC) power converters are necessary interfaces to form micro-energy systems. Naturally, designing a suitable control scheme, featuring superior dynamics, strong robustness, and simple structure, is a promising solution to guarantee more efficient operation of the converter. This article proposes a robust high-quality current control strategy for the 3L-NPC power converter in the stationary$\alpha \beta$frame. A super-twisting algorithm coupled with a Luenberger observer current controller is proposed to deal with the poor sinusoidal current tracking issue due to the existing inductance/grid frequency deviations and the disturbance of the sinusoidal dynamic nature. Additionally, an extended sliding mode disturbance observer-based proportional control is built to dramatically enhance the voltage regulation performance, in the case of capacitance deviations and unknown dc-loads. Experimental data confirm the effectiveness of the proposed solution outperforms the conventional proportional-resonant/-integral control in terms of accurate tracking current/voltage, antidisturbance, and grid current total harmonic distortion.