monitor-experiment

To check running machine learning experiments on a remote SSH screen, the Skill executes shell commands to list and hardcopy active screen logs. This retrieves finalized metrics and outputs directly from your remote execution environment.

Yes, you can pull Weights and Biases metrics from a remote Vast.ai instance. When enabled, the Skill calls the W&B API directly from the remote host to retrieve training curves, final summaries, and dashboard links.

The best way to summarize remote experiment results against a baseline is by reading JSON result files from the results directory. The Skill produces a numeric comparison table that interprets the deltas versus your correct baseline.

To monitor training progress across Modal apps and Vast.ai instances, the Skill checks running sessions and retrieves status logs. It validates outputs by locating and reading JSON result files to present concrete numbers.

Experiment monitoring works without Weights and Biases enabled by default. The Skill independently inspects screen logs and results directories for JSON outputs, applying W&B metric pulling only as an optional enrichment step.

Yes, you can send a Feishu notification with your experiment monitoring results. After generating a decision-ready comparison table with interpreted deltas, the Skill can send a configured Feishu notification to update your team.