Monitor WandB training metrics periodically and alert on anomalies.
npx skills add https://github.com/jkfee/Auto-Research --skill training-check-jkfee
Please help me install this Agent Skill. Skill: training-check Source: https://github.com/jkfee/Auto-Research/tree/main/skills/training-check Command: npx skills add https://github.com/jkfee/Auto-Research --skill training-check-jkfee
This skill automates monitoring of WandB metrics during ML training to detect issues early and prevent wasted GPU hours.
Configure WandB path and schedule a recurring check that alerts when training metrics diverge.
High-intent search queries and answers about installing and using this skill.
This skill automates WandB metric monitoring during active ML training to detect issues early. It periodically retrieves metrics, tracks loss and eval data, and alerts you when anomalies appear to prevent wasted GPU hours.
Yes, you can schedule periodic health checks using CronCreate-based scheduling with configurable reset rules. This adapts the check frequency for your machine learning training runs automatically.
Codex evaluates ambiguous signals during WandB metric monitoring by providing automated escalation guidance. It applies judgment to unclear metric data retrieved during active training to determine if anomalies exist.
Yes, this skill integrates with Watchdog for process health monitoring. It combines WandB training metric evaluation with Watchdog to track the overall health of your machine learning training process.
The best way to prevent wasted GPU hours when training loss diverges is to automate WandB metric monitoring. This skill detects diverging training loss early and alerts you to stop the run.
Yes, you need WandB API access to automate training health checks. The skill implements periodic checks using the WandB API to retrieve training metrics and evaluate model health during active runs.
raja21068Monitor WandB training metrics to detect NaN, divergence, or degraded evaluation.
satsuki-64Monitor WandB metrics during training to detect anomalies and decide whether to continue or halt.
ZHLX2005Monitor GitHub open issues periodically and trigger flutter-work-flow after closing each.
prashsubOrchestrate Databricks observability setup with Lakehouse Monitoring, anomaly detection, dashboards, and SQL alerts.
