How to use from
Unsloth StudioInstall 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 nemanisri/security-guardian-gemma4 to start chattingUsing HuggingFace Spaces for Unsloth
# No setup required# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for nemanisri/security-guardian-gemma4 to start chattingLoad model with FastModel
pip install unsloth
from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
model_name="nemanisri/security-guardian-gemma4",
max_seq_length=2048,
)Quick Links
SecurityGuardian — Gemma 4 E2B Fine-tuned
A fine-tuned Gemma 4 E2B model that explains mobile security threats in plain language. Runs completely offline.
Built for Gemma 4 Good Hackathon 2026
Training
- Base model: gemma-4-E2B-it
- Method: LoRA (Unsloth)
- Steps: 60
- Final loss: 0.021
Example
Input: ALERT: App 'FreeFlashlight' requested ACCESS_FINE_LOCATION, READ_CONTACTS, SEND_SMS, RECORD_AUDIO
Output: This is a suspicious request. A flashlight app only needs to turn on your screen light. RECOMMENDATION: Do not install this app.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support
Install Unsloth Studio (macOS, Linux, WSL)
# Run unsloth studio unsloth studio -H 0.0.0.0 -p 8888 # Then open http://localhost:8888 in your browser # Search for nemanisri/security-guardian-gemma4 to start chatting