Model CoverageLarge Language Models

Moonlight

View as Markdown

Moonlight is a Mixture-of-Experts language model from Moonshot AI trained using Muon optimizer. It uses the DeepseekV3ForCausalLM architecture with 16B total parameters and 3B activated per token.

TaskText Generation (MoE)
ArchitectureDeepseekV3ForCausalLM
Parameters16B total / 3B active
HF Orgmoonshotai

Available Models

  • Moonlight-16B-A3B: 16B total, 3B activated

Architecture

  • DeepseekV3ForCausalLM (same architecture as DeepSeek-V3)

Example HF Models

ModelHF ID
Moonlight 16B A3Bmoonshotai/Moonlight-16B-A3B

Example Recipes

RecipeDescription
moonlight_16b_te.yamlSFT — Moonlight 16B with Transformer Engine
moonlight_16b_te_packed_sequence.yamlSFT — Moonlight 16B with packed sequences

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/moonlight/moonlight_16b_te.yaml

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:26.06.00

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/moonlight/moonlight_16b_te.yaml

See the Installation Guide and LLM Fine-Tuning Guide.

Fine-Tuning

See the LLM Fine-Tuning Guide.

Hugging Face Model Cards