clinical-summarizer-qwen25-7b

<div align="center">

📝 Clinical Note Summarizer

Qwen2.5-7B fine-tuned for clinical note summarizer

Part of the Medical AI Fine-tuned Model Suite — 16 specialist models, one per task

</div>

---

TL;DR

Converts long, unstructured clinical notes into a structured SOAP format summary.

INPUT:  DISCHARGE SUMMARY: 67M admitted with worsening dyspnea and bilateral leg oedema. PMH: CHF (EF 35%), HTN, CKD stage 3...
OUTPUT: S: 67M with known CHF presents with worsening dyspnea and bilateral oedema.\nO: EF worsened to 28%. 3L diuresis achieved.\nA: Acute decompensated heart failure (I50.9), CKD stage 3 (N18.3).\nP: Furosemide 80mg OD, cardiology follow-up 2 weeks.

---

Architecture

                  +-------------------------+
  user prompt --> |  Qwen2.5-7B-Instruct  | --> base weights (frozen, 4-bit NF4)
                  |  + LoRA adapter (r=16)  | --> clinical-summarizer-qwen25-7b
                  +-------------------------+
                              |
                              v
                     structured output
                  (code / JSON / classification)

This repo contains only the LoRA adapter (~160MB), not the full merged weights. Load it on top of the base model as shown below — this keeps the download small and lets you swap adapters on one base model in memory.

---

Intended use

Reduce physician documentation burden by auto-generating SOAP summaries from raw visit notes.

Direct use

Paste a full clinical note, get back a Subjective/Objective/Assessment/Plan structured summary for the chart.

Downstream use

Pipe summaries into a clinical documentation improvement (CDI) review queue or a billing-code suggestion engine.

Out of scope

Generating notes from scratch without a real clinical encounter to summarize. Replacing physician sign-off on the final chart note.

This model is not a substitute for a certified medical professional's judgment. Output should be reviewed by a qualified person before being used in a clinical or billing decision.

---

Quickstart

Option A — Transformers + PEFT

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel
import torch

base_model = "unsloth/Qwen2.5-7B-Instruct"
adapter    = "AmareshHebbar/clinical-summarizer-qwen25-7b"

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(
    base_model,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
model = PeftModel.from_pretrained(model, adapter)

messages = [
    {"role": "system", "content": "You are a clinical documentation specialist. Summarize the following clinical note into a structured SOAP format."},
    {"role": "user", "content": "DISCHARGE SUMMARY: 67M admitted with worsening dyspnea and bilateral leg oedema. PMH: CHF (EF 35%), HTN, CKD stage 3..."},
]
inputs = tokenizer.apply_chat_template(messages, return_tensors="pt", add_generation_prompt=True).to(model.device)
outputs = model.generate(inputs, max_new_tokens=128, temperature=0.1, do_sample=True)
print(tokenizer.decode(outputs[0][inputs.shape[1]:], skip_special_tokens=True))

Expected output:

S: 67M with known CHF presents with worsening dyspnea and bilateral oedema.\nO: EF worsened to 28%. 3L diuresis achieved.\nA: Acute decompensated heart failure (I50.9), CKD stage 3 (N18.3).\nP: Furosemide 80mg OD, cardiology follow-up 2 weeks.

Option B — Unsloth (2x faster load + inference)

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="AmareshHebbar/clinical-summarizer-qwen25-7b",
    max_seq_length=512,
    load_in_4bit=True,
)
FastLanguageModel.for_inference(model)

messages = [
    {"role": "system", "content": "You are a clinical documentation specialist. Summarize the following clinical note into a structured SOAP format."},
    {"role": "user", "content": "Patient with 3-day history of fever and productive cough, found to have right lower lobe consolidation on CXR."},
]
prompt = tokenizer.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
outputs = model.generate(**inputs, max_new_tokens=128, temperature=0.1, do_sample=True)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[1]:], skip_special_tokens=True))

Option C — vLLM (production serving, OpenAI-compatible)

vllm serve unsloth/Qwen2.5-7B-Instruct \
    --enable-lora \
    --lora-modules clinical-summarizer-qwen25-7b=AmareshHebbar/clinical-summarizer-qwen25-7b \
    --host 0.0.0.0 --port 8000 --dtype bfloat16

from openai import OpenAI

client = OpenAI(base_url="http://localhost:8000/v1", api_key="not-needed")
response = client.chat.completions.create(
    model="clinical-summarizer-qwen25-7b",
    messages=[
        {"role": "system", "content": "You are a clinical documentation specialist. Summarize the following clinical note into a structured SOAP format."},
        {"role": "user", "content": "Post-operative day 1 following laparoscopic cholecystectomy, tolerating oral intake, pain controlled."},
    ],
    temperature=0.1,
)
print(response.choices[0].message.content)

Option D — GGUF / llama.cpp (CPU / edge inference)

This repo ships LoRA adapter weights, not a pre-merged GGUF. To run on llama.cpp, merge first:

pip install unsloth
python -c "
from unsloth import FastLanguageModel
model, tokenizer = FastLanguageModel.from_pretrained('AmareshHebbar/clinical-summarizer-qwen25-7b', load_in_4bit=False)
model.save_pretrained_gguf('clinical-summarizer-qwen25-7b-gguf', tokenizer, quantization_method='q4_k_m')
"

---

Training details

Data

Trained on 30,000 examples extracted from 30k de-identified clinical notes (AGBonnet/augmented-clinical-notes) (source). No synthetic or LLM-generated training data — every example pairs real-world input with its authoritative output.

Full extraction pipeline documented on the dataset card.

Hyperparameters

Training compute

Fine-tuned with Unsloth for 2x faster training and reduced VRAM, using TRL's SFTTrainer. Full project: wandb.ai/amareshhebbar-/axiomapper.

---

Bias, risks & limitations

Data recency. Training data reflects a specific snapshot in time (CMS FY2026 / dataset publish date). Codes, rates, and rules referenced may become outdated as source authorities issue updates — always cross-check against the live authoritative source before high-stakes use.

Failure mode. Like any LLM, this model can produce a plausible-sounding but incorrect output, especially on rare, ambiguous, or highly compound real-world cases that fall outside the training distribution. It does not know when it's wrong.

Language. English-language input only (Hindi-medical model excepted, where Hindi system prompts are used but underlying clinical reasoning data is largely English-sourced).

Not a regulated medical device. This model has not been validated, cleared, or approved by any regulatory body (FDA, CDSCO, or equivalent) as a medical device or clinical decision support tool. It is a research/engineering artifact.

Misapplication risk. Do not use this model as the sole basis for a clinical, billing, or compliance decision affecting a real patient or claim. Do not deploy in an emergency triage context without a human-in-the-loop and clear escalation paths.

---

FAQ

Q: Can I merge the adapter into the base model for faster inference? Yes — use model.merge_and_unload() after loading with PEFT, or use Unsloth's save_pretrained_merged() method.

Q: Why QLoRA instead of full fine-tuning? The base model already has strong language and medical knowledge from pretraining. QLoRA adapts only ~0.5-1% of parameters, which is enough to specialize the output format and domain without the cost or overfitting risk of full fine-tuning.

Q: Can I fine-tune this further on my own data? Yes, this adapter can be used as a starting checkpoint for continued fine-tuning. Note this may require merging first depending on your training framework.

Q: Why is the output format so strict? Each task was trained on a fixed system prompt and consistent output structure. Following the documented system prompt closely (see Quickstart above) gives the most reliable results — deviating from it may produce inconsistent formatting.

Q: Does this model store or transmit my input data? No. Like any open-weight model, all inference happens locally on your own infrastructure (or wherever you deploy it) — nothing is sent back to the model author.

---

Troubleshooting

---

Related models in this suite

Full suite overview: AmareshHebbar/medical-ai-model-suite

---

Changelog

---

Citation

@misc{medicalai2026,
  author    = {Hebbar, Amaresh},
  title     = {Medical AI Fine-tuning Suite},
  year      = {2026},
  publisher = {HuggingFace},
  url       = {https://huggingface.co/AmareshHebbar}
}

Contact