Instructions to use Zero21/OncoScope with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Zero21/OncoScope with Transformers:

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

pipe = pipeline("text-generation", model="Zero21/OncoScope")
messages = [
    {"role": "user", "content": "Who are you?"},
]
pipe(messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Zero21/OncoScope", dtype="auto")

llama-cpp-python

How to use Zero21/OncoScope with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="Zero21/OncoScope",
	filename="oncoscope-gemma-3n-merged.Q8_0.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use Zero21/OncoScope with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Zero21/OncoScope:Q8_0
# Run inference directly in the terminal:
llama-cli -hf Zero21/OncoScope:Q8_0

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf Zero21/OncoScope:Q8_0
# Run inference directly in the terminal:
llama-cli -hf Zero21/OncoScope:Q8_0

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf Zero21/OncoScope:Q8_0
# Run inference directly in the terminal:
./llama-cli -hf Zero21/OncoScope:Q8_0

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf Zero21/OncoScope:Q8_0
# Run inference directly in the terminal:
./build/bin/llama-cli -hf Zero21/OncoScope:Q8_0

Use Docker

docker model run hf.co/Zero21/OncoScope:Q8_0

LM Studio
Jan

vLLM

How to use Zero21/OncoScope with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Zero21/OncoScope"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Zero21/OncoScope",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/Zero21/OncoScope:Q8_0

SGLang

How to use Zero21/OncoScope 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 "Zero21/OncoScope" \
    --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": "Zero21/OncoScope",
		"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 "Zero21/OncoScope" \
        --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": "Zero21/OncoScope",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Ollama
How to use Zero21/OncoScope with Ollama:
```
ollama run hf.co/Zero21/OncoScope:Q8_0
```

Unsloth Studio new

How to use Zero21/OncoScope with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Zero21/OncoScope to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for Zero21/OncoScope to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for Zero21/OncoScope to start chatting

Docker Model Runner
How to use Zero21/OncoScope with Docker Model Runner:
```
docker model run hf.co/Zero21/OncoScope:Q8_0
```

Lemonade

How to use Zero21/OncoScope with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull Zero21/OncoScope:Q8_0

Run and chat with the model

lemonade run user.OncoScope-Q8_0

List all available models

lemonade list

OncoScope Cancer Genomics Analysis Model

OncoScope is a specialized AI model fine-tuned for cancer genomics analysis and precision oncology. Built on Google's Gemma 3n architecture, this model provides expert-level analysis of cancer mutations, risk assessments, and therapeutic recommendations while maintaining complete privacy through on-device inference.

Model Details

Base Model: Google Gemma 3n 2B E4B Chat IT
Parameters: 6.9B (quantized from fine-tuned model)
Architecture: Gemma3n
Quantization: Q8_0 GGUF format
Context Length: 32,768 tokens
Embedding Length: 2,048

Key Features

Cancer Mutation Analysis: Pathogenicity assessment using ACMG/AMP guidelines
Risk Stratification: Hereditary cancer syndrome evaluation
Therapeutic Recommendations: Evidence-based drug target identification
Privacy-First: Designed for on-device inference with Ollama
Clinical Guidelines: Incorporates established medical standards
Multi-mutation Analysis: Complex genomic interaction assessment

Training Data

The model was fine-tuned on a curated dataset of 5,998 cancer genomics examples from:

ClinVar: Clinical variant database
COSMIC Top 50: Cancer mutation signatures
Expert-curated: Clinical oncology cases

Usage

With Ollama

Download the model files:
- oncoscope-gemma-3n-merged.Q8_0.gguf (6.8GB)
- Modelfile
Create the model:
```
ollama create oncoscope -f Modelfile
```

Run inference:

ollama run oncoscope "Analyze the clinical significance of BRCA1 c.5266dupC mutation"

Example Usage

ollama run oncoscope "Patient: 45-year-old female with family history of breast cancer. 
Mutation: BRCA1 c.68_69delAG (p.Glu23ValfsTer17). 
Please provide pathogenicity assessment and recommendations."

Example Response:

{
  "mutation_analysis": {
    "gene": "BRCA1",
    "variant": "c.68_69delAG",
    "protein_change": "p.Glu23ValfsTer17",
    "pathogenicity": "Pathogenic",
    "confidence_score": 0.95,
    "acmg_classification": "PVS1, PM2, PP3"
  },
  "clinical_significance": {
    "cancer_risk": "High",
    "associated_cancers": ["Breast", "Ovarian"],
    "lifetime_risk": {
      "breast_cancer": "55-85%",
      "ovarian_cancer": "15-40%"
    }
  },
  "recommendations": {
    "genetic_counseling": "Strongly recommended",
    "screening": "Enhanced surveillance starting age 25",
    "prevention": "Consider prophylactic surgery",
    "family_testing": "Cascade testing recommended"
  }
}

Model Capabilities

Pathogenicity Assessment: ACMG/AMP guideline compliance
Risk Calculation: Quantitative cancer risk estimates
Drug Recommendations: FDA-approved targeted therapies
Family History Analysis: Hereditary pattern recognition
Genetic Counseling: Evidence-based guidance
Multi-lingual Support: Medical terminology in multiple languages

Limitations

Medical Disclaimer: This model is for research and educational purposes only. Always consult qualified healthcare professionals for medical decisions.
Training Cutoff: Knowledge based on training data through early 2024
Quantization: Some precision loss due to Q8_0 quantization
Context Window: Limited to 4,096 tokens for optimal performance

Technical Specifications

Model Size: 6.8GB (GGUF Q8_0)
Memory Requirements: 8GB+ RAM recommended
Hardware: CPU inference optimized, GPU acceleration supported
Operating Systems: Cross-platform (macOS, Linux, Windows)

Performance

The model demonstrates expert-level performance on:

Variant pathogenicity classification (>90% accuracy vs. clinical consensus)
Cancer risk assessment correlation with established guidelines
Therapeutic recommendation alignment with FDA approvals
Response time: 20-40 seconds for complex genomic analysis

Privacy & Security

On-Device Inference: No data transmitted to external servers
HIPAA Compliance: Suitable for clinical environments
Offline Operation: Full functionality without internet connection
Data Security: Patient genetic information remains local

Citation

If you use this model in your research, please cite:

@misc{oncoscope2025,
  title={OncoScope: Privacy-First Cancer Genomics Analysis with Gemma 3n},
  author={Sheldon Aristide},
  year={2025},
  url={https://huggingface.co/Zero21/OncoScope}
}

License

This model is released under the Apache 2.0 license, consistent with the base Gemma model licensing.

Support & Contact

For questions, issues, or contributions:

GitHub: OncoScope Project
Issues: Please report bugs or feature requests via GitHub Issues

Disclaimer

This AI model is intended for research and educational purposes only. It should not be used as a substitute for professional medical advice, diagnosis, or treatment. Always seek the advice of qualified healthcare professionals regarding any medical condition or genetic testing decisions.

Downloads last month: 3

GGUF

Model size

7B params

Architecture

gemma3n

Hardware compatibility

8-bit

Model tree for Zero21/OncoScope

Base model

google/gemma-3n-E4B

Finetuned

google/gemma-3n-E4B-it

Quantized

unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit

Quantized

(1)

this model