clawhub

nemo-mbridge-perf-moe-optimization-workflow

Systematic workflow for MoE training optimization in Megatron Bridge, based on the Megatron-Core MoE paper. Covers the Three Walls framework, parallel folding, recompute strategy, dispatcher choice, and CUDA-graph bring-up.

copies: implexa run clawhub/nemo-mbridge-perf-moe-optimization-workflow

view source

installs

stars

karma

full SKILL.md lives at the source

we've indexed the metadata for this skill but the body is fetched on demand. click "view source" above to read the canonical SKILL.md on clawhub, or "run inline in claude" to apply it without leaving your session.

read on clawhub

related skills

semantically similar in the cross-vendor index

smithery

62% match

training-llms-megatron

Trains large language models (2B-462B parameters) using NVIDIA Megatron-Core with advanced parallelism strategies. Use when training models >1B parameters, need maximum GPU efficiency (47% MFU on H100

don't have the plugin yet? install it then click "run inline in claude" again.