OpenClaw · Skills: AI & machine learning

NVIDIA operational guide for turning on Megatron FSDP for Megatron Bridge.

NVIDIA guide for lowering peak GPU memory during Megatron Bridge training.

NVIDIA guide for hiding MoE expert communication behind compute on Megatron Bridge.

NVIDIA guide for choosing an MoE token dispatcher on Megatron Bridge.

NVIDIA reference of dated MoE training playbooks sorted by GPU hardware and model.

NVIDIA guide for long-context MoE training on Megatron Bridge.

NVIDIA workflow for optimizing MoE performance on Megatron Bridge, gated by evidence.

NVIDIA guide for training MoE vision-language models on Megatron Bridge.

NVIDIA operational guide for selecting and mixing parallelism strategies on Megatron Bridge.

NVIDIA guide for sequence packing and long-context runs on Megatron Bridge.

NVIDIA operational guide for overlapping tensor, data, and pipeline parallel communication.

NVIDIA skill for recommending and customizing Megatron Bridge recipes for a given model and hardware.

NVIDIA guide to Megatron Bridge resiliency: fault tolerance, stragglers, restarts, and preemption.

NVIDIA skill for debugging missing or wrong NeMo Relay runtime behavior after installation.

NVIDIA skill for newcomers trying NeMo Relay with the simplest supported quick start.

NVIDIA skill for choosing and running the right NeMo Relay install.

NVIDIA skill for wrapping tool and LLM calls in NeMo Relay scopes.

NVIDIA skill for keeping NeMo Relay scope stacks separate across concurrent work.

NVIDIA skill for adding typed wrappers, domain types, and provider codecs to NeMo Relay.

NVIDIA skill for migrating code and docs from NeMo Flow to Relay across languages.