Your agent can do more. Find the skill that makes it happen.
Find your next superpower If you are an agent, refer to our llms.txt for full access.Guides selection and configuration of FSDP2, Megatron FSDP, and DDP in NeMo AutoModel. Covers associated parallelism settings.
Configures NeMo AutoModel job launches across interactive execution, Slurm clusters, and SkyPilot cloud environments.
Guides onboarding new architectures into NeMo AutoModel. Covers architecture discovery, implementation patterns, registration, and validation.
Explains how to develop NeMo AutoModel training and evaluation recipes. Covers YAML structure, builders, and execution flow.
Covers training runs in Megatron-LM and Megatron Bridge using mock or real datasets. Includes correlation testing, available recipes, and multi-GPU examples.
Covers converting single-node scripts into sbatch jobs using srun-native or uv run torch.distributed approaches. Includes containers, NCCL timeouts, MoE memory sizing, and interactive allocations.
Covers selective and full activation recomputation, trading additional computation for lower GPU memory use. Addresses activation memory failures and regressions involving recomputation settings and activation checkpointing.
Covers layer-level activation offloading and fractional optimizer state offloading. Includes use of HybridDeviceOptimizer in Megatron Bridge.
Covers local full-iteration CUDA graphs and Transformer Engine scoped graphs. Includes graph capture for attention, MLP, and MoE modules.
Covers overlap_moe_expert_parallel_comm and delay_wgrad_compute. Includes flex dispatcher backends such as DeepEP and HybridEP.
Provides operational guidance for hierarchical context parallelism. Covers configuration controls, relevant code locations, pitfalls, and verification.
Provides operational guidance for enabling Megatron FSDP. Includes configuration controls, relevant code locations, pitfalls, and verification.
Skills give your agent reusable instructions for a specific job. Pick one, read what it does, and bring it into your workflow.
A name only tells half the story. Search the full description and instructions to find the right fit.
Read the skill, visit its source, and see exactly what you’re adding to your agent.
Copy the install command from a skill page and run it in your project.
npx skillycli add owner/repo --skill name