실험실 도구
Dalva Autoresearch
Monitor, analyze, and guide deep learning training experiments using Dalva's CLI tools. Use this skill when the user is running or planning to run training experiments and needs to track metrics, inspect model performance, compare runs, read training configurations, analyze tabular data (predictions, hyperparameter sweeps, layer statistics), or make autonomous decisions about training adjustments. Triggers on: training monitoring, experiment tracking, loss curves, metric analysis, training loops, hyperparameter tuning, model convergence, overfitting detection, run comparison, experiment management, ML experiment dashboard, training progress, val loss, learning rate scheduling, checkpoint decisions, early stopping, autonomous research, auto-research, or any scenario where an agent needs visibility into what's happening inside a training run. Also use when the user mentions Dalva, dalva query, experiment observability, or wants an automated agent to watch and react to training.