Instructions to use nvidia/Ising-Calibration-1-35B-A3B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use nvidia/Ising-Calibration-1-35B-A3B with Transformers:

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

pipe = pipeline("image-text-to-text", model="nvidia/Ising-Calibration-1-35B-A3B")
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 AutoProcessor, AutoModelForImageTextToText

processor = AutoProcessor.from_pretrained("nvidia/Ising-Calibration-1-35B-A3B")
model = AutoModelForImageTextToText.from_pretrained("nvidia/Ising-Calibration-1-35B-A3B")
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?"}
        ]
    },
]
inputs = processor.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(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps

vLLM

How to use nvidia/Ising-Calibration-1-35B-A3B with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "nvidia/Ising-Calibration-1-35B-A3B"
# 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/Ising-Calibration-1-35B-A3B",
		"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/Ising-Calibration-1-35B-A3B

SGLang

How to use nvidia/Ising-Calibration-1-35B-A3B 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/Ising-Calibration-1-35B-A3B" \
    --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/Ising-Calibration-1-35B-A3B",
		"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/Ising-Calibration-1-35B-A3B" \
        --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/Ising-Calibration-1-35B-A3B",
		"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/Ising-Calibration-1-35B-A3B with Docker Model Runner:
```
docker model run hf.co/nvidia/Ising-Calibration-1-35B-A3B
```

373× GPU speedup for portfolio optimization using Ising Calibration (QAOA + QPU validated)

by VenkatMadhav - opened 28 days ago

Discussion

VenkatMadhav

28 days ago

•

edited 28 days ago

Sharing an early external use case of Ising Calibration 1,
applied 4 days after launch to a new domain.
https://github.com/maddykws/quantum-portfolio-ising-claude

Headline results:

373× GPU speedup on NVIDIA L4 via CUDA-Q (measured at N=20 qubits, nvidia vs cpu target, identical kernel code)
+3.1% median Sharpe vs naive classical portfolio selection across 56 quarterly S&P 500 windows (75% win rate, 2010–2024)
1.4 million × sampling efficiency vs random combination search — median 3 shots to near-optimal portfolio
Validated on Rigetti Cepheus-1-108Q QPU via Amazon Braket ($1.44 total, parameters transferred from GPU simulation)
−4.8% vs classically-optimised top-N — disclosed honestly as the hard baseline

How I used Ising Calibration:

I sent Ising Calibration 1 bar charts of QAOA measurement
distributions from a portfolio optimisation circuit — not
QPU calibration outputs. The model correctly identified low
measurement concentration as poor QAOA convergence, flagging
underperforming windows before downstream portfolio results
confirmed the problem.

Small QCalEval-style comparison across 10 windows:

Pipeline	Circuit quality	Financial insight
Ising Calibration alone	3.9 / 5	1.0 / 5
LLM orchestration alone	1.0 / 5	3.9 / 5
Combined	3.9 / 5	4.0 / 5

Suggests complementary, non-overlapping capabilities — even
outside the intended QPU calibration domain.

Honest framing:

This is a hybrid workflow validation, not a quantum advantage
claim. Applying Ising Calibration to portfolio QAOA circuits
is a domain extension experiment — early external use, not
validated production. The model's circuit interpretation
capability appears to transfer to non-calibration quantum
experiment outputs, but I would be interested in the team's
view on whether that generalisation is expected.

Reproducible end-to-end:

Full pipeline runs on Google Colab L4 GPU for approximately
$3. One-click Colab badge in README. Two free API keys
(NVIDIA NIM + Anthropic).

GitHub: https://github.com/maddykws/quantum-portfolio-ising-claude

Filed a scope question on the NVIDIA Ising-Calibration
GitHub repo and posted to the NVIDIA Developer Forum with
additional figures. Happy to share more detail with anyone
interested.

Thanks for shipping Ising — great to see NVIDIA open-source
a calibration VLM family.

maddykws

VenkatMadhav changed discussion title from Ising Calibration applied to QAOA circuits: 373× GPU speedup, validated across 56 quarterly portfolio windows, QPU validated to 373× GPU speedup for portfolio optimization using Ising Calibration (QAOA + QPU validated) 28 days ago

ShuxiangCaoNV

NVIDIA org 25 days ago

Thanks for the benchmarking and for sharing the reproducible repo! While not specifically designed for this use case, we’re glad to see Ising Calibration’s diagnostic capabilities applied to QAOA and other experiments.

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