API Reference

Auto-generated reference documentation for the nemo_automodel Python package.

This reference is built from docstrings in the source code. Browse the sidebar to explore modules, classes, and functions.

Key Modules

Module	Description
`nemo_automodel._transformers`	Hugging Face bridge (`NeMoAutoModelForCausalLM`, `NeMoAutoModelForImageTextToText`, model registry, capability flags, and device-mesh infrastructure)
`nemo_automodel._diffusers`	Diffusion pipeline wrapper (`NeMoAutoDiffusionPipeline`)
`nemo_automodel.components.models`	27+ model families (LLM, VLM, MoE, retrieval) and their state-dict adapters
`nemo_automodel.components.datasets`	LLM, VLM, and diffusion data pipelines
`nemo_automodel.components.distributed`	FSDP2, HSDP, DDP utilities and device-mesh helpers
`nemo_automodel.components.checkpoint`	Async DCP and SafeTensors checkpointing
`nemo_automodel.components.quantization`	FP8, QAT, calibration
`nemo_automodel.components._peft`	LoRA, QLoRA adapters
`nemo_automodel.components.launcher`	Interactive, SkyPilot, and NeMo-Run job launchers; Slurm jobs use the repository’s `slurm.sub` script
`nemo_automodel.cli`	`uv run automodel <config.yaml> [--nproc-per-node N] [--key.subkey=override ...]` entry point

Extension APIs

Model Parallelizer API: Add a model-owned parallelization sidecar while keeping topology, dispatch, and default TP/CP/EP/FSDP behavior in shared infrastructure.