Instructions to use thefertilityplan/ivf-bench-qwen9b-vlm-orpo with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use thefertilityplan/ivf-bench-qwen9b-vlm-orpo with Transformers:

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

pipe = pipeline("image-text-to-text", model="thefertilityplan/ivf-bench-qwen9b-vlm-orpo")
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, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("thefertilityplan/ivf-bench-qwen9b-vlm-orpo")
model = AutoModelForMultimodalLM.from_pretrained("thefertilityplan/ivf-bench-qwen9b-vlm-orpo", device_map="auto")
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 Settings

vLLM

How to use thefertilityplan/ivf-bench-qwen9b-vlm-orpo with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "thefertilityplan/ivf-bench-qwen9b-vlm-orpo"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "thefertilityplan/ivf-bench-qwen9b-vlm-orpo",
		"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/thefertilityplan/ivf-bench-qwen9b-vlm-orpo

SGLang

How to use thefertilityplan/ivf-bench-qwen9b-vlm-orpo 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 "thefertilityplan/ivf-bench-qwen9b-vlm-orpo" \
    --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": "thefertilityplan/ivf-bench-qwen9b-vlm-orpo",
		"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 "thefertilityplan/ivf-bench-qwen9b-vlm-orpo" \
        --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": "thefertilityplan/ivf-bench-qwen9b-vlm-orpo",
		"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 thefertilityplan/ivf-bench-qwen9b-vlm-orpo with Docker Model Runner:
```
docker model run hf.co/thefertilityplan/ivf-bench-qwen9b-vlm-orpo
```

Correct the card for the sighted-judge rerun and withdraw the ablation figure

by andrewmonostate - opened Aug 5

base: refs/heads/main

←

from: refs/pr/4

Discussion Files changed

+58

-29

andrewmonostate

The Fertility Plan org Aug 5

Hugging Face CLI

The evaluation behind this card was rerun with the judge shown the same embryo
image the model was shown. Every score changes, and two claims on the live card
are no longer supported.

What changed

All scores are regenerated under the sighted judge. Morphology grounding falls
for every system (this model 3.56 to 3.14), because the blind judge had been
accepting descriptions it could not check against the image.
The card said "the judge never sees the embryo image". That is no longer true
and the sentence is removed.
The card put the image's contribution at 0.07 to 0.23 points. That figure came
from the blind judge and is withdrawn; the experiment it came from cannot
support a quantitative claim in either direction.
Second place is now reported as conditional. A cross-judge pass with Claude
Sonnet 4.6 over the same 824 held-out judgements puts this model fifth and Opus
4.6 second, reversing the margin from +0.15 to -0.49. Each judge is most
generous to its own model family.
The gain over the base model, which is the claim that survives both judges, is
stated as +24.6% under GPT-5.4 and +15.5% under Sonnet 4.6.
Two conditions that were not equal across systems are now disclosed: the
generation budget, and the decoding and serving stack differing between this
model and the base it is compared against.

Numbers match the released artifacts; the repository's verification suite checks
177 of them.

Correct the card for the sighted-judge rerun and withdraw the ablation figure05d28a2b

andrewmonostate changed pull request status to merged Aug 5

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