MuhammadSanan99989/safescan-phi3-mini-intent

SafeScan Phi-3 Mini — Intent Routing Model

Fine-tuned version of microsoft/Phi-3-mini-4k-instruct for the SafeScan mobile security utility app (Flutter).

What it does

Given a natural language security query, returns a structured JSON object routing the request to the correct SafeScan module:

{"intent": "wifi_check", "module": "WifiSecurityScan", "action": "navigate_to_wifi_security_scan", "parameters": {"network": null}}

Training Details

Property	Value
Base model	microsoft/Phi-3-mini-4k-instruct
Method	QLoRA (4-bit NF4)
LoRA rank	16
LoRA alpha	32
Dataset size	1,250 samples
Epochs	3
Learning rate	2e-4
Optimizer	paged_adamw_8bit
Max seq length	512

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
import torch

model = AutoModelForCausalLM.from_pretrained(
    "MuhammadSanan99989/safescan-phi3-mini-intent",
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True,
)
tokenizer = AutoTokenizer.from_pretrained("MuhammadSanan99989/safescan-phi3-mini-intent", trust_remote_code=True)

pipe = pipeline("text-generation", model=model, tokenizer=tokenizer)

prompt = "<|user|>\nCheck if my WiFi is secure<|end|>\n<|assistant|>"
result = pipe(prompt, max_new_tokens=128, do_sample=False)
print(result[0]["generated_text"][len(prompt):])
# Output: {"intent": "wifi_check", "module": "WifiSecurityScan", ...}