What problem does it solve?
This Skill turns Nemotron model customization requests into the right repo-native workflow, so you can avoid guessing which training, data prep, conversion, evaluation, or optimization step to use.
Core Features & Use Cases
- Workflow routing: Selects the correct Nemotron step for curation, translation, SFT, PEFT, pretraining, RL alignment, conversion, optimization, benchmarking, or evaluation.
- Pipeline composition: Chains multiple steps by matching artifacts such as JSONL, packed Parquet, bin/idx shards, and checkpoints.
- Guardrailed execution: Applies repo rules for tokenizer alignment, checkpoint conversion boundaries, hosted service requirements, and safe remote execution.
- Use Case: A developer wants to translate a dataset, fine-tune a model, then evaluate the result; this Skill identifies the required steps and how their outputs connect.
Quick Start
Ask for the exact Nemotron step or pipeline you need, along with your input paths, model or checkpoint, output location, and whether you want a local dry run or remote execution.