training-llms-megatron

SKILLWorkflowcommunity
v0.0.0Orchestra-ResearchMITUpdated 2mo agoSource →

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), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used f

Community-submitted skill. Not yet reviewed by the Forge team. Full prompt content may not be available.Request review →
12kRepo stars
1Clients
1Formats
2mo agoLast update
Skill
AuthorOrchestra-Research
Version0.0.0
LicenseMIT
CategoryWorkflow
Formatsskill.md
PromptOpen (see Prompt tab)
Compatibility
Claude✓ Supported
Cursor—
Copilot—
ChatGPT—
Gemini—
About

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), or require tensor/pipeline/sequence/context/expert parallelism. Production-ready framework used for Nemotron, LLaMA, DeepSeek.

Keywords
skillclaude