Looking for rigorous courses on energy-efficient ML

Has anyone taken a program that goes beyond pruning to cover end-to-end efficiency — data pipeline profiling, carbon-aware scheduling, and hardware-aware training? I’m evaluating options for Q2 and need hands-on labs (PyTorch on A100s) with quantifiable outcomes like kWh/epoch reductions and throughput-per-watt improvements.

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​‌‌‍‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠‌‌⁠⁠‌⁠‌​‌‍⁠⁠‌⁠​​‌‍‍‌‌‍​⁠​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​‍​‍‌‍⁠‍‌‍‌‌‌⁠‌⁠​‍​‍​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​‍​⁠​​​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌‍⁠​‌‍‍⁠​‍⁠‌‌‍‍‍‌⁠​​​⁠​⁠‌‍‍​‌‍‌​‌‌​​​⁠‌‍‌​⁠⁠‌⁠​⁠‌‍⁠⁠‌‌‍‌​⁠‍​‌⁠​‍​‍​‍‌⁠⁠‌

I pair NVIDIA DLI PyTorch labs with DCGM power + CodeCarbon for ‘kWh/epoch’. https://codecarbon.io — are you cloud or on-prem?

‌⁠‍⁠​‍​‍‌⁠‌​​‍​‍​⁠‍‍​‍​‍‌‍​‌‌‍‍‌​‍​‍​‍⁠​​‍​‍‌‍‍⁠​‍​‍​⁠‍‍​‍​‍‌⁠​‍‌‍‌‌‌⁠​​‌‍⁠​‌⁠‍‌​‍​‍​‍⁠​​‍​‍‌‍‍‌‌‍‌​​‍​‍​⁠‍‍​⁠​‌​⁠​‍​⁠​​​‍⁠​​‍​‍‌‍‌​​‍​‍​⁠‍‍​‍​‍​⁠​‍​⁠​​​⁠​‍​⁠‌‌​⁠​‌​⁠​‍​⁠​‍​⁠​⁠​‍​‍​‍⁠​​‍​‍‌‍‍​​‍​‍​⁠‍‍​‍​‍‌​‌​‌⁠‍​‌‌‌⁠‌‌​‍‌​⁠‌‌‌‍‌‌‍⁠⁠‌⁠‍‌‌‍​‌‌‌‍‌​‍⁠‌‌​​⁠​⁠​​‌‍‌‍‌⁠​‍‌⁠‌‌​‍​‍‌⁠⁠‌