Back to evaluations
Draft evaluation
Mastra AI Auto-scaling Inference Fleet with Thermal Constraints — evaluation
Evaluates thermal-aware autoscaling and migration logic during temperature spike simulation.
Evaluation type
task based
Challenge
Mastra AI Auto-scaling Inference Fleet with Thermal Constraints
Difficulty
Advanced
Rigor
Not declared
The author has not specified a rigor level.
Evaluation overview
How the linked challenge is judged: tasks, benchmarks, and criteria count.
Tasks
1
Benchmarks
0
Criteria
0
Task templates
Inputs and expected outputs.
Task 1
thermal_migration_task
Evaluates system response when GPU Node 1 reaches 85°C junction temperature.
Input format
JSON containing node_telemetry, cluster_capacity, active_workloads
Output format
JSON containing migration_triggered, target_node, migrated_workload_count, cluster_thermal_avg_c