nemo-mbridge-perf-moe-dispatcher-selection
Pass
Audited by Gen Agent Trust Hub on Jun 18, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues detected. The skill provides technical guidance and empirical benchmarks for optimizing Mixture-of-Experts (MoE) models in high-performance computing environments.
- [COMMAND_EXECUTION]: The skill provides examples of command-line flags (e.g.,
--moe-deepep-num-sms 20and--moe-hybridep-num-sms 16) used for tuning SM counts in training frameworks. These are documented as configuration parameters and do not represent a security risk.
Audit Metadata