
Turns a Jetson video encoding use case into a validated encoder recipe spec
- Open source
- 4K GitHub stars
- Design and media

Turns a Jetson video encoding use case into a validated encoder recipe spec

Sets up and verifies NVIDIA Video Codec SDK and PyNvVideoCodec on a Jetson device

Guides an agent through launching and debugging NeMo-RL training runs on Kubernetes

Triages a failing GitHub Actions job and files a bug issue for it

Runs Megatron-LM's autoformat and lint checks and applies its Python style rules

Launch distributed Megatron-LM training jobs on SLURM clusters with sbatch scripts

Splits a large PR into smaller ones that each need fewer CODEOWNERS reviewer groups

Guides choosing and configuring FSDP2, Megatron FSDP, DDP and parallelism settings in NeMo

Sets up NeMo AutoModel job launches on interactive, Slurm and SkyPilot runs

Guides an agent through adding new model architectures to NeMo AutoModel

Guides an agent through building and changing NeMo AutoModel training and eval recipes

Guides building and reviewing NeMo Fabric adapters that connect agent harnesses

Guides an agent through integrating NeMo Fabric's typed Python SDK into a consumer app

Runs Megatron-LM and Megatron Bridge training and checks their loss curves match

Guides an agent through multi-node Slurm sbatch jobs and debugging their failures

Guides selective and full activation recompute in Megatron Bridge to cut GPU memory use

Configures CPU offloading in Megatron Bridge to cut GPU memory during training

Guides validating and configuring CUDA graph capture in Megatron Bridge training runs

Guides validating and enabling MoE expert-parallel communication overlap

Guides an agent through enabling hierarchical context parallelism in Megatron-Bridge

Guides an agent through enabling Megatron FSDP sharded training in Megatron-Bridge

Guides an agent through reducing peak GPU memory in Megatron Bridge training runs

Guides tuning of MoE expert-parallel communication overlap in Megatron Bridge

Guides choice of MoE token dispatcher (alltoall, DeepEP, HybridEP) by hardware and EP size

Hardware-specific MoE training playbooks and candidate parallelism configs for Megatron

Guides long-context MoE training in Megatron Bridge: CP sizing, recompute and dispatcher

Guides an agent through evidence-gated MoE performance tuning in Megatron Bridge

Guidance for choosing FSDP or 3D parallelism when training MoE vision-language models

Guides choosing and combining TP, DP, PP, CP and EP degrees for Megatron Bridge training

Guides Megatron-Bridge sequence packing and long-context training setup and validation