Nemotron-H#

NVIDIA Nemotron-H is a hybrid Mamba-2 / transformer architecture that interleaves selective state space layers with standard attention layers for improved efficiency on long sequences.

Task

Text Generation

Architecture

NemotronHForCausalLM

Parameters

9B – 30B

HF Org

nvidia

Available Models#

  • NVIDIA-Nemotron-Nano-9B-v2: 9B hybrid model

  • NVIDIA-Nemotron-Nano-12B-v2: 12B hybrid model

  • NVIDIA-Nemotron-3-Nano-30B-A3B-BF16: 30B total, 3B activated (sparse MoE + Mamba-2)

Architecture#

  • NemotronHForCausalLM

Example HF Models#

Model

HF ID

Nemotron-Nano 9B v2

nvidia/NVIDIA-Nemotron-Nano-9B-v2

Nemotron-Nano 12B v2

nvidia/NVIDIA-Nemotron-Nano-12B-v2

Nemotron-3-Nano 30B A3B

nvidia/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16

Example Recipes#

Recipe

Description

nemotron_nano_9b_squad.yaml

SFT — Nemotron-Nano 9B on SQuAD

nemotron_nano_9b_squad_peft.yaml

LoRA — Nemotron-Nano 9B on SQuAD

nemotron_nano_v3_hellaswag.yaml

SFT — Nemotron-3-Nano 30B on HellaSwag

nemotron_nano_v3_hellaswag_peft.yaml

LoRA — Nemotron-3-Nano 30B on HellaSwag

Try with NeMo AutoModel#

1. Install (full instructions):

pip install nemo-automodel

2. Clone the repo to get the example recipes:

git clone https://github.com/NVIDIA-NeMo/Automodel.git
cd Automodel

3. Run the recipe from inside the repo:

automodel --nproc-per-node=8 examples/llm_finetune/nemotron/nemotron_nano_9b_squad.yaml
Run with Docker

1. Pull the container and mount a checkpoint directory:

docker run --gpus all -it --rm \
  --shm-size=8g \
  -v $(pwd)/checkpoints:/opt/Automodel/checkpoints \
  nvcr.io/nvidia/nemo-automodel:26.02.00

2. Navigate to the AutoModel directory (where the recipes are):

cd /opt/Automodel

3. Run the recipe:

automodel --nproc-per-node=8 examples/llm_finetune/nemotron/nemotron_nano_9b_squad.yaml

See the Installation Guide and LLM Fine-Tuning Guide.

Fine-Tuning#

See the LLM Fine-Tuning Guide.

Hugging Face Model Cards#