Instructions to use nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoTokenizer, AutoModelForMultimodalLM

tokenizer = AutoTokenizer.from_pretrained("nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8", trust_remote_code=True)
model = AutoModelForMultimodalLM.from_pretrained("nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8", trust_remote_code=True)
messages = [
    {"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
	messages,
	add_generation_prompt=True,
	tokenize=True,
	return_dict=True,
	return_tensors="pt",
).to(model.device)

outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8

SGLang

How to use nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8 with Docker Model Runner:
```
docker model run hf.co/nvidia/NVIDIA-Nemotron-3-Nano-4B-FP8
```

NVIDIA-Nemotron-3-Nano-4B-FP8 / model-00002-of-00002.json

bkartal

fix_safetensors_structure (#1)

9b479fa 3 months ago

raw

history blame contribute delete

13.1 kB

	{
	"metadata": {
	"total_size": 1285003864
	},
	"weight_map": {
	"backbone.layers.26.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.26.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.26.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.27.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.27.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.28.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.28.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.29.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.29.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.30.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.30.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.31.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.31.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.32.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.32.mixer.q_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.32.mixer.k_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.32.mixer.v_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.32.mixer.o_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.33.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.33.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.34.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.34.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.35.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.35.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.36.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.36.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.37.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.37.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.38.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.38.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.39.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.39.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.40.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.A_log": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.D": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.dt_bias": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.conv1d.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.conv1d.bias": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.in_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.in_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.in_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.out_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.out_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.40.mixer.out_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.41.norm.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.up_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.up_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.up_proj.input_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.down_proj.weight": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.down_proj.weight_scale": "model-00002-of-00010.safetensors",
	"backbone.layers.41.mixer.down_proj.input_scale": "model-00002-of-00010.safetensors"
	}
	}