Instructions to use Div97/orena-focus-segment-fullvis-w64 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Div97/orena-focus-segment-fullvis-w64 with Transformers:

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

pipe = pipeline("image-text-to-text", model="Div97/orena-focus-segment-fullvis-w64")
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)

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoProcessor, AutoModelForMultimodalLM

processor = AutoProcessor.from_pretrained("Div97/orena-focus-segment-fullvis-w64")
model = AutoModelForMultimodalLM.from_pretrained("Div97/orena-focus-segment-fullvis-w64", 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=256)
print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:]))

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Div97/orena-focus-segment-fullvis-w64 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Div97/orena-focus-segment-fullvis-w64"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Div97/orena-focus-segment-fullvis-w64",
		"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/Div97/orena-focus-segment-fullvis-w64

SGLang

How to use Div97/orena-focus-segment-fullvis-w64 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 "Div97/orena-focus-segment-fullvis-w64" \
    --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": "Div97/orena-focus-segment-fullvis-w64",
		"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 "Div97/orena-focus-segment-fullvis-w64" \
        --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": "Div97/orena-focus-segment-fullvis-w64",
		"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 Div97/orena-focus-segment-fullvis-w64 with Docker Model Runner:
```
docker model run hf.co/Div97/orena-focus-segment-fullvis-w64
```

Div97 commited on Sep 8

Commit

b7b0405

verified ·

1 Parent(s): 449d778

Document detailed method and data terms

Browse files

Files changed (1) hide show

README.md +44 -16

README.md CHANGED Viewed

@@ -1,5 +1,7 @@
 ---
-license: apache-2.0
 base_model: Qwen/Qwen3-VL-4B-Instruct
 library_name: transformers
 pipeline_tag: image-text-to-text
@@ -16,31 +18,57 @@ Merged bfloat16 checkpoint used by **DISCOVR-SEGMENT**, submitted to the ORena
 SAVE FOCUS 2026 SEGMENT track by team **Incision Impossible**.
 - Base: `Qwen/Qwen3-VL-4B-Instruct`
 - Merge: PEFT `merge_and_unload`
 - Serving budget: 64 frames
 - Model SHA-256:
   `622fd66547b2ad88f9fcf9c74a22450f44b4c88cef8fcf1a9b464de2a51dcff3`
-- Source code:
-  `https://github.com/mdivyanshu97/orena-focus-segment`
-The model was fine-tuned for surgical foreign-object visual question
-answering. It is intended for research and challenge reproduction, not
-clinical use. No patient videos or raw annotations are distributed in this
-repository.
 ## Inference method
-DISCOVR-SEGMENT uses a fixed 64-frame budget with question-dependent temporal
-routing:
-- ordinary questions sample the complete segment;
 - questions containing timestamps use up to two ±30-second evidence windows;
 - single-timestamp questions receive a second 64-frame pass in a 30-second
   window around the first prediction;
-- temporal frames receive an absolute-time overlay;
-- output is cleaned to the inferred answer format and written with one response
-  per request.
-The full method and architecture diagram are documented in the
-[`METHOD.md`](https://github.com/mdivyanshu97/orena-focus-segment/blob/main/METHOD.md)
-source release.

 ---
+license: other
+license_name: mixed-apache-2.0-and-third-party-data-terms
+license_link: https://github.com/mdivyanshu97/orena-focus-segment/blob/main/NOTICE
 base_model: Qwen/Qwen3-VL-4B-Instruct
 library_name: transformers
 pipeline_tag: image-text-to-text
 SAVE FOCUS 2026 SEGMENT track by team **Incision Impossible**.
 - Base: `Qwen/Qwen3-VL-4B-Instruct`
+- Adaptation: language, vision-encoder, and merger LoRA
 - Merge: PEFT `merge_and_unload`
 - Serving budget: 64 frames
 - Model SHA-256:
   `622fd66547b2ad88f9fcf9c74a22450f44b4c88cef8fcf1a9b464de2a51dcff3`
+- Selected challenge image:
+  `b74f595d-06d4-488a-b76b-217544cf8e55`
+Source and documentation:
+- [GitHub source](https://github.com/mdivyanshu97/orena-focus-segment)
+- [Concise implementation method](https://github.com/mdivyanshu97/orena-focus-segment/blob/main/METHOD.md)
+- [Detailed challenge method draft](https://github.com/mdivyanshu97/orena-focus-segment/blob/main/METHOD_DESCRIPTION.md)
+- [Training provenance](https://github.com/mdivyanshu97/orena-focus-segment/blob/main/TRAINING.md)
+## Training summary
+FullVis-W64 was trained in two stages:
+1. one-epoch surgical scene-literacy warm-up on a 238,925-row SSG-VQA /
+   CholecT45 manifest, with language rank 16, vision rank 16, and merger rank
+   64; and
+2. three scheduled epochs of continued all-track FOCUS fine-tuning on 34,290
+   HeiCo and LapChole training rows with a maximum of 64 frames.
+The detailed document records exact data counts and hashes, frame generation,
+optimizer settings, prompt construction, selected artifact hashes, known
+historical dataset-revision differences, and official pre-evaluation metrics.
 ## Inference method
+DISCOVR-SEGMENT uses question-dependent temporal routing:
+- ordinary questions sample 64 frames across the complete segment;
 - questions containing timestamps use up to two ±30-second evidence windows;
 - single-timestamp questions receive a second 64-frame pass in a 30-second
   window around the first prediction;
+- temporal frames receive an absolute-procedure-time overlay; and
+- output is normalized to the inferred answer format.
+## Intended use and limitations
+The checkpoint is intended for non-commercial research and challenge
+reproduction. It is not a medical device and must not be used for clinical
+decision making. No patient videos or raw challenge annotations are
+distributed here.
+## License and data terms
+The Qwen3-VL base model and released source code use Apache-2.0. Training also
+used SSG-VQA, whose repository specifies CC BY-NC-SA 4.0 for non-commercial
+scientific research, plus challenge datasets governed by their respective
+owners. The `license: other` metadata reflects these mixed terms; it does not
+replace any source-dataset license or access agreement.