Instructions to use erjui/CheXagent-2-3b-csrrg-impression with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use erjui/CheXagent-2-3b-csrrg-impression with Transformers:

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

pipe = pipeline("image-text-to-text", model="erjui/CheXagent-2-3b-csrrg-impression", 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 AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("erjui/CheXagent-2-3b-csrrg-impression", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("erjui/CheXagent-2-3b-csrrg-impression", trust_remote_code=True, 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 = 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 erjui/CheXagent-2-3b-csrrg-impression with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "erjui/CheXagent-2-3b-csrrg-impression"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "erjui/CheXagent-2-3b-csrrg-impression",
		"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/erjui/CheXagent-2-3b-csrrg-impression

SGLang

How to use erjui/CheXagent-2-3b-csrrg-impression 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 "erjui/CheXagent-2-3b-csrrg-impression" \
    --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": "erjui/CheXagent-2-3b-csrrg-impression",
		"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 "erjui/CheXagent-2-3b-csrrg-impression" \
        --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": "erjui/CheXagent-2-3b-csrrg-impression",
		"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 erjui/CheXagent-2-3b-csrrg-impression with Docker Model Runner:
```
docker model run hf.co/erjui/CheXagent-2-3b-csrrg-impression
```

erjui commited on Oct 28, 2025

Commit

b924d73

verified ·

1 Parent(s): 47cd87d

Upload README.md with huggingface_hub

Browse files

Files changed (1) hide show

README.md +3 -34

README.md CHANGED Viewed

@@ -90,7 +90,7 @@ This model performs **Contextualized Structured Radiology Report Generation (CSR
 - Max images per sample: 2
 **Hardware:**
-- GPU: NVIDIA A100 (or equivalent)
 - Training framework: HuggingFace Transformers + PEFT
 ## Usage
@@ -161,37 +161,6 @@ IMPRESSION:
 2. Recommend repeat dedicated AP and lateral chest radiograph, or CT for further evaluation.
 ```
-## Evaluation
-The model is evaluated using standard medical NLG metrics with additional temporal reasoning evaluation:
-- **RadGraph F1**: Measures clinical entity and relation extraction accuracy
-- **BLEU**: N-gram overlap with reference reports
-- **ROUGE** (ROUGE-1, ROUGE-2, ROUGE-L): Recall-oriented metrics
-- **BERTScore**: Semantic similarity using contextual embeddings
-- **CheXbert F1**: Clinical accuracy for pathology classification
-- **Temporal consistency**: Evaluation of comparison statements and longitudinal reasoning
-For detailed evaluation results, see the paper: [Automated Structured Radiology Report Generation with Rich Clinical Context](https://arxiv.org/abs/2510.00428)
-## Limitations
-- Trained exclusively on chest X-ray images (not applicable to other imaging modalities)
-- Performance may vary on images from different institutions or imaging protocols
-- May not capture all rare pathologies or edge cases
-- Requires expert radiologist review before clinical use
-- Temporal reasoning accuracy depends on quality of prior study information
-- Should not be used as the sole diagnostic tool
-## Ethical Considerations
-- **Medical AI Responsibility**: This model generates medical text and must be used responsibly
-- **Human Oversight Required**: All outputs should be reviewed by qualified radiologists
-- **Data Privacy**: Ensure compliance with HIPAA, GDPR, and local healthcare regulations
-- **Bias and Fairness**: Model trained on specific datasets may have biases; validate on diverse populations
-- **Clinical Validation**: Requires thorough validation before any clinical deployment
-- **Longitudinal Data Handling**: Ensure proper patient consent and data governance for temporal data
 ## Citation
 If you use this model, please cite:
@@ -208,8 +177,8 @@ If you use this model, please cite:
 Also cite the base model:
 ```bibtex
 @article{chen2024chexagent,
-  title={CheXagent: Towards a Foundation Model for Chest X-Ray Interpretation},
-  author={Chen, Zhihong and Varma, Maya and Delbrouck, Jean-Benoit and Paschali, Magdalini and Blankemeier, Louis and Van Veen, Dave and Valanarasu, Jeya Maria Jose and Youssef, Alaa and Cohen, Joseph Paul and Reis, Eduardo Pontes and Tsai, Emily B. and Johnston, Andrew and Olsen, Cameron and Abraham, Tanishq Mathew and Gatidis, Sergios and Chaudhari, Akshay S. and Lungren, Matthew P.},
   journal={arXiv preprint arXiv:2401.12208},
   year={2024}
 }

 - Max images per sample: 2
 **Hardware:**
+- GPU: NVIDIA H100
 - Training framework: HuggingFace Transformers + PEFT
 ## Usage
 2. Recommend repeat dedicated AP and lateral chest radiograph, or CT for further evaluation.
 ```
 ## Citation
 If you use this model, please cite:
 Also cite the base model:
 ```bibtex
 @article{chen2024chexagent,
+  title={Chexagent: Towards a foundation model for chest x-ray interpretation},
+  author={Chen, Zhihong and Varma, Maya and Delbrouck, Jean-Benoit and Paschali, Magdalini and Blankemeier, Louis and Van Veen, Dave and Valanarasu, Jeya Maria Jose and Youssef, Alaa and Cohen, Joseph Paul and Reis, Eduardo Pontes and others},
   journal={arXiv preprint arXiv:2401.12208},
   year={2024}
 }