Instructions to use arunimas1107/gpt-oss-medical with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use arunimas1107/gpt-oss-medical with PEFT:

from peft import PeftModel
from transformers import AutoModelForCausalLM

base_model = AutoModelForCausalLM.from_pretrained("openai/gpt-oss-20b")
model = PeftModel.from_pretrained(base_model, "arunimas1107/gpt-oss-medical")

Transformers

How to use arunimas1107/gpt-oss-medical with Transformers:

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

pipe = pipeline("text-generation", model="arunimas1107/gpt-oss-medical")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("arunimas1107/gpt-oss-medical", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use arunimas1107/gpt-oss-medical with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "arunimas1107/gpt-oss-medical"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "arunimas1107/gpt-oss-medical",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/arunimas1107/gpt-oss-medical

SGLang

How to use arunimas1107/gpt-oss-medical 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 "arunimas1107/gpt-oss-medical" \
    --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": "arunimas1107/gpt-oss-medical",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

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 "arunimas1107/gpt-oss-medical" \
        --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": "arunimas1107/gpt-oss-medical",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Docker Model Runner
How to use arunimas1107/gpt-oss-medical with Docker Model Runner:
```
docker model run hf.co/arunimas1107/gpt-oss-medical
```

Model Card for Medical GPT-OSS-20B LoRA Adapter

Model Details

Model Description

This model is a LoRA adapter fine-tuned on openai/gpt-oss-20b using the PEFT library. It is optimized for medical domain tasks such as question answering, summarization, and knowledge retrieval in healthcare contexts. The adapter modifies the base model with efficient fine-tuning techniques while retaining the general-purpose reasoning capabilities of the underlying 20B parameter model.

Developed by: Arunima Surendran
Funded by: E2E Cloud
Shared by: Arunima Surendran
Model type: Large Language Model with LoRA adapter (20B base)
Language(s): English
License: Apache 2.0
Finetuned from model: openai/gpt-oss-20b
Finetuned using: NVIDIA 2xH100 for 12+ hours on E2E Cloud TIR Instance

Model Sources

Repository: arunimakanavu/gpt-oss-medical

Requirements

torch>=2.0.0
transformers @ git+https://github.com/huggingface/transformers.git
datasets>=2.12.0
peft>=0.10.0
accelerate>=0.22.0
bitsandbytes>=0.41.0
sentencepiece>=0.1.99

Uses

Direct Use

Medical Q&A
Clinical text summarization
Educational content generation in healthcare

Downstream Use

Integrating into RAG pipelines with domain-specific medical knowledge bases
Deployment in medical chatbots (for informational purposes only)

Out-of-Scope Use

Direct clinical decision-making or diagnostic tools without human oversight
High-stakes medical applications without proper validation

Bias, Risks, and Limitations

The model may generate:

Inaccurate or hallucinated medical information
Biased outputs due to limitations in training data
Text not suitable for unsupervised clinical decision-making

Recommendations

Users should:

Treat outputs as assistive, not authoritative
Always cross-verify with trusted medical sources
Avoid using the model for patient-facing diagnosis without professional review

How to Get Started with the Model

from transformers import AutoModelForCausalLM, AutoTokenizer, PeftModel

base_model = "openai/gpt-oss-20b"
adapter_path = "./gpt-oss-medical"

model = AutoModelForCausalLM.from_pretrained(base_model, device_map="auto", torch_dtype="auto")
model = PeftModel.from_pretrained(model, adapter_path)

tokenizer = AutoTokenizer.from_pretrained(base_model)

Training Details

Training Data

FreedomIntelligence/medical-o1-reasoning-SFT
Domain-specific medical text corpus
Filtered for quality and relevance

Training Procedure

Fine-tuned using LoRA on top of GPT-OSS-20B
Mixed precision training (bf16)

Training Hyperparameters

Training regime: bf16 mixed precision
Compute: NVIDIA 2xH200 GPUs
Training time: 12+ hours

Evaluation

Testing Data, Factors & Metrics

Domain: medical Q&A and summarization
Metrics: Perplexity, BLEU/ROUGE for summarization, accuracy for Q&A

Results

BERTScore = 0.834

Environmental Impact

Hardware Type: NVIDIA 2xH200 (E2E TIR platform)
Hours used: 12+
Cloud Provider: E2E Networks

Technical Specifications

Model Architecture and Objective

Base: GPT-OSS-20B (20 billion parameters)
Adapter: LoRA (low-rank fine-tuning)

Compute Infrastructure

Hardware: NVIDIA 2xH100 (E2E TIR)
Software: PyTorch, Transformers, PEFT

Citation

BibTeX:

@misc{gptoss20b-medical,
  title = {Medical GPT-OSS-20B LoRA Adapter},
  author = {Arunima Surendran},
  year = {2025},
  url = {https://github.com/arunimakanavu/gpt-oss-medical}
}

Model Card Authors

Arunima Surendran

Framework versions

PEFT 0.17.0

Downloads last month: 3

Model tree for arunimas1107/gpt-oss-medical

Base model

openai/gpt-oss-20b

Adapter

(383)

this model