SUPREME: A Multi-GPU Framework for Reproducible Image Unlearning Method Evaluation

2026-05-29 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Computer Vision & Pattern Recognition · Depth: Expert, quick

Summary

SUPREME is an open-source, multi-GPU framework designed to enhance the reproducible evaluation of machine unlearning methods for image classification. Published on 2026-05-29, this framework addresses the computational expense of evaluating unlearning techniques across multiple seeds, a limitation of existing single-GPU solutions. SUPREME distributes training, unlearning, and evaluation stages across multiple GPUs, supporting various accelerators and precision modes. Its architecture includes a registry-based design for easily integrating new methods, metrics, models, and unlearning scenarios. The framework's utility is demonstrated on the Pins Face Recognition dataset, utilizing ResNet18 and ViT models under both full-class and random-sample unlearning scenarios, evaluated across ten seeds.

Key takeaway

For MLOps engineers or AI researchers evaluating machine unlearning methods, SUPREME offers a critical solution to accelerate rigorous testing. Its multi-GPU architecture significantly reduces the computational burden of running multi-seed evaluations, enabling more comprehensive reproducibility studies. You should consider integrating this open-source framework to streamline your unlearning method validation, especially when working with large image datasets and complex models like ResNet18 or ViT.

Key insights

SUPREME is a multi-GPU framework accelerating reproducible evaluation of image unlearning methods.

Principles

• Reproducibility requires multi-seed evaluation.
• Distributed computing scales unlearning assessment.
• Modular design supports extensibility.

Method

SUPREME distributes training, unlearning, and evaluation stages across multiple GPUs, using a registry-based design to integrate new components and scenarios.

In practice

• Evaluate unlearning on Pins Face Recognition.
• Test ResNet18 and ViT models.
• Compare full-class and random-sample unlearning.

Topics

• Machine Unlearning
• Multi-GPU Frameworks
• Image Classification
• Model Evaluation
• Reproducibility
• ResNet18
• ViT

Code references

• pedroandreou/supreme-unlearning

Best for: Research Scientist, Computer Vision Engineer, Machine Learning Engineer, AI Scientist, MLOps Engineer

Related on AIssential

• Object detection with Amazon Nova 2 Lite — Amazon Nova 2 Lite enables zero-shot object detection via natural language prompts, simplifying computer vision deployment.
• Honey, I Shrunk the Arc de Triomphe! — A new dataset and method mitigate scale-collapse in monocular depth estimation for distant, unconstrained scenes.
• TROPHIES: Temporal Reconstruction of Places, Humans, and Cameras from Multi-view Videos — TROPHIES unifies human, scene, and camera reconstruction from multi-view videos into a globally consistent 4D space.
• Edge-aware Decoding for Neural Asymmetric Routing — Neural asymmetric routing models improve by explicitly exposing transition-level edge information in the decoder, resolving representation-decision mismatch.
• LALE: Lightweight-Transformer Architecture for Land-Cover Estimation — LALE efficiently segments remote sensing imagery by combining ConvMixers for local detail and transformers for global context in a bifurcated encoder.
• Beyond Low-Rank: Low-Rank Sparse Prompting via Spiking Neural Network and Prompt Factorization — LoRSP uses spiking neural networks and low-rank factorization to create sparse, efficient visual prompts for vision models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.