CRE-2025-0140: Stable Diffusion WebUI CUDA Out of Memory Detection

Overview

This pull request implements a Critical Runtime Event (CRE) detection rule for Stable Diffusion AUTOMATIC1111 WebUI CUDA out of memory errors during batch processing. The rule identifies critical failures where GPU memory exhaustion causes complete generation pipeline crashes.

Problem Statement

Stable Diffusion WebUI systems can experience critical CUDA out of memory errors during batch image generation, creating scenarios where:

• GPU memory allocation fails during tensor operations
• Entire batch processing pipelines crash mid-execution
• Generated images are lost without proper error handling
• API endpoints fail with memory-related errors
• Production workflows are interrupted requiring manual intervention

CRE Playground Links

Playground Link

test.log

• Rule File: rules/cre-2025-0140/sd-webui-oom.yaml
• Test Logs: rules/cre-2025-0140/test.log

Demo Environment

https://github.com/Sahelisaha04/n8n-cre-demo (invitation send)

https://github.com/user-attachments/assets/da3f289b-76ab-4931-a7a1-858b2067ea45

docker compose -f docker-compose-simple.yml up log-generator
# Test CUDA OOM detection (should detect failure)
cat tests/test.log | preq -r rules/cre-2025-0140/sd-webui-oom.yaml -d


# Test with freshly generated logs
cat tests/generated_failure.log | preq -r rules/cre-2025-0140/sd-webui-oom.yaml -d

References

• AUTOMATIC1111 WebUI Optimizations
• PyTorch CUDA Memory Management
• Stable Diffusion Memory Issues

Fixes #130 /claim #130