---
license: mit
base_model: microsoft/Phi-3-mini-4k-instruct
library_name: transformers
pipeline_tag: text-generation
tags:
- phi3
- lora
- hr-assistant
- fine-tuned
- text-generation
language:
- en
inference: true
---

# Phi-3 HR Assistant

This model is a fine-tuned version of [microsoft/Phi-3-mini-4k-instruct](https://huggingface.co/microsoft/Phi-3-mini-4k-instruct) using LoRA (Low-Rank Adaptation) for HR policy assistance.

## Model Details

- **Base Model**: microsoft/Phi-3-mini-4k-instruct
- **Fine-tuning Method**: LoRA (r=8, alpha=16, dropout=0.05)
- **Task**: HR Policy Question Answering
- **Language**: English
- **Model Size**: ~3.8B parameters

## Quick Start

```python
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch

# Load model and tokenizer
model = AutoModelForCausalLM.from_pretrained(
    "SK0988/phi3-hr-assistant",
    torch_dtype=torch.float16,
    device_map="auto",
    trust_remote_code=True
)
tokenizer = AutoTokenizer.from_pretrained("SK0988/phi3-hr-assistant")

# Generate HR response
def ask_hr_question(question):
    prompt = f'''You are the company's HR Helpdesk assistant. Answer HR policy questions accurately.

### Human:
{question}

### Assistant:
'''
    
    inputs = tokenizer(prompt, return_tensors="pt")
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=200,
            do_sample=False,
            pad_token_id=tokenizer.eos_token_id
        )
    
    response = tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
    return response.strip()

# Example usage
question = "What is the leave policy for permanent employees?"
answer = ask_hr_question(question)
print(answer)
```

## API Usage

```python
# Using Hugging Face Inference API
import requests

API_URL = "https://api-inference.huggingface.co/models/SK0988/phi3-hr-assistant"
headers = {"Authorization": "Bearer YOUR_HF_TOKEN"}

def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()

# Ask HR question
output = query({
    "inputs": "What is the maternity leave policy?",
    "parameters": {"max_new_tokens": 200}
})
print(output)
```

## Training Details

- **Training Data**: HR policies and Q&A pairs
- **Training Method**: Supervised Fine-tuning with LoRA
- **Target Modules**: q_proj, k_proj, v_proj, o_proj
- **Training Steps**: 1000
- **Base Model**: microsoft/Phi-3-mini-4k-instruct

## Use Cases

- HR policy inquiries
- Leave policy questions
- Travel allowance information
- Certification reimbursement queries
- General HR assistance

## Limitations

- Model responses should be verified against current HR policies
- Not suitable for sensitive HR decisions without human oversight
- May require additional context for complex policy questions
- Responses are based on training data and may not reflect latest policy changes

## Ethical Considerations

- This model is designed for HR assistance only
- Should not be used for discriminatory purposes
- Human oversight recommended for important decisions
- Ensure compliance with local employment laws

## Model Performance

The model has been fine-tuned specifically for HR-related queries and shows good performance on:
- Leave policy questions
- Travel allowance inquiries
- Certification reimbursement
- General HR procedures

## Citation

If you use this model, please cite:

```
@misc{phi3-hr-assistant,
  title={Phi-3 HR Assistant},
  author={SK0988},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/SK0988/phi3-hr-assistant}
}
```