Instructions to use nvidia/NVLM-D-72B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/NVLM-D-72B with Transformers:

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

pipe = pipeline("image-text-to-text", model="nvidia/NVLM-D-72B", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import NVLM_D
model = NVLM_D.from_pretrained("nvidia/NVLM-D-72B", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use nvidia/NVLM-D-72B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/NVLM-D-72B"
# 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/NVLM-D-72B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/nvidia/NVLM-D-72B

SGLang

How to use nvidia/NVLM-D-72B 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/NVLM-D-72B" \
    --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/NVLM-D-72B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

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/NVLM-D-72B" \
        --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/NVLM-D-72B",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use nvidia/NVLM-D-72B with Docker Model Runner:
```
docker model run hf.co/nvidia/NVLM-D-72B
```

GGUF wen?

by AIGUYCONTENT - opened Oct 3, 2024

Discussion

AIGUYCONTENT

Oct 3, 2024

many thanks

SimSim93

Oct 3, 2024

This one is highly unlikely. Even Llama 3.2 90B (vision) hasn't been adapted to gguf. See, Llama.cpp at it's core does no longer support multimodality. The wrappers (like ollama or the server version) sometimes do. But they need to adapt it ba hand. This why unfortunately so many great vision models aren't adapted to gguf.

By the way, I'd like it to be different. Just don't hold your breath for it to change immedietly.

mraudible1983

Oct 3, 2024

How many 4090 or 5090 do we need to run this?

AIGUYCONTENT

Oct 3, 2024

How many 4090 or 5090 do we need to run this?

You still have 2 kidneys?

AIGUYCONTENT

Oct 3, 2024

This one is highly unlikely. Even Llama 3.2 90B (vision) hasn't been adapted to gguf. See, Llama.cpp at it's core does no longer support multimodality. The wrappers (like ollama or the server version) sometimes do. But they need to adapt it ba hand. This why unfortunately so many great vision models aren't adapted to gguf.

By the way, I'd like it to be different. Just don't hold your breath for it to change immedietly.

Thanks. So I only have 136GB of VRAM. That means I cannot run the main model.

https://huggingface.co/models?search=nvlm

That leaves the fp8 (which I cannot run on a 3090) and the NF4—which appears to be a work in progress: https://huggingface.co/SeanScripts/NVLM-D-72B-nf4

Could you guys (or someone else) create a quant that would only need 70GB to 100 GB of VRAM to run?

nonetrix

Oct 6, 2024

I wanted to be able to run the NVIDIA model on my AMD GPU for lols :(

At least I have mistral nemo and nemotron mini I guess lol

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment