Nemo Mbridge Perf Moe Optimization Workflow
nvidia/skills/skills/nemo-mbridge-perf-moe-optimization-workflowby nvidiacf5224d14250Apache-2.03.5K starsListed Oct 8, 2026Updated Oct 8, 2026Repository updated today
Evidence-gated workflow for MoE performance optimization in Megatron Bridge. Covers measurement contracts, the Three Walls framework, parallel folding, profiling, matched A/B tuning, and final validation.
Only the file list is public. File contents are available once the skill is installed in a workspace.
| Path | Size | Type |
|---|---|---|
| BENCHMARK.md | 4.8 KB | text/markdown |
| card.yaml | 7.6 KB | application/yaml |
| evals/evals.json | 3.6 KB | application/json |
| skill-card.md | 4.3 KB | text/markdown |
| SKILL.md | 12.7 KB | text/markdown |
| skill.oms.sig | 4.7 KB | text/plain |
Source and attribution
Source:nvidia/skillsinskills/nemo-mbridge-perf-moe-optimization-workflowat commitcf5224d
License: Apache-2.0
Content belongs to its original authors. SourceWeft indexes it from a public repository.
More from nvidia/skills
Tilegym Cutile Python
nvidia
Guides writing, debugging and optimizing cuTile Python GPU kernels, with examples, validation loops and optional multi-agent orchestration.
Tilegym Monkey Patch Kernels To Transformers
nvidia
Integrates TileGym cuTile kernels into Hugging Face transformers models via non-intrusive monkey-patching and auto-generates new kernels.
Tilegym Converting Cutile To Julia
nvidia
Converts cuTile Python GPU kernels to cuTile.jl Julia equivalents, with rules, examples and a static validator.
Tilegym Adding Cutile Kernel
nvidia
Guides adding a new cuTile GPU kernel operator to TileGym, from dispatch registration through tests and benchmarks.
Physicsnemo Discover
nvidia
Guides discovery of PhysicsNeMo model families, datapipes and examples for a SciML task, citing live repo paths.
Nemoclaw User Guide
nvidia
Routes an AI agent to NemoClaw's canonical Markdown docs and docs MCP server to answer setup and operations questions.