README.md

---
license: mit
tags:
- unsloth
- phi-3
- command-line
- cli
- lora
- fine-tuned
base_model:
- unsloth/Phi-3-mini-4k-instruct-bnb-4bit
fine-tuned with:
- https://github.com/unslothai/unsloth
LoRA config:
- r=8, alpha=128, dropout=0
---

# 🛠️ Finetuned Phi-3 CLI Assistant

This model is a fine-tuned version of [unsloth/Phi-3-mini-4k-instruct-bnb-4bit](https://huggingface.co/unsloth/Phi-3-mini-4k-instruct-bnb-4bit), trained on a command-line Q&A dataset for Linux/Git/tar/curl/grep/etc.

## 🔍 Use Cases

- Terminal helpbots
- CLI Q&A agents
- Developer assistants

## 🧠 How to Use

```python
from unsloth import FastLanguageModel
from transformers import AutoTokenizer
import torch
import gradio as gr

# 🔃 Load model and tokenizer from Hugging Face
model, tokenizer = FastLanguageModel.from_pretrained(
    model_name="sreebhargav/finetuned-phi3-cli",  # Your HF model path
    max_seq_length=2048,
    load_in_4bit=True,
    device_map="auto"
)
FastLanguageModel.for_inference(model)

# 🔍 CLI Assistant function
def cli_assistant(prompt):
    messages = [{"role": "user", "content": prompt}]
    inputs = tokenizer.apply_chat_template(
        messages,
        return_tensors="pt",
        tokenize=True,
        add_generation_prompt=True
    ).to(model.device)

    outputs = model.generate(
        input_ids=inputs,
        max_new_tokens=256,
        temperature=0.7,
        top_p=0.9,
        do_sample=True,
        eos_token_id=tokenizer.eos_token_id
    )

    decoded = tokenizer.decode(outputs[0], skip_special_tokens=True)
    return decoded.split("### Output:\n")[-1].strip()

# 🚀 Launch Gradio demo
gr.Interface(
    fn=cli_assistant,
    inputs=gr.Textbox(lines=2, placeholder="Ask about a Linux/Git/Bash command..."),
    outputs=gr.Textbox(label="🧠 AI Response"),
    title="🧠 CLI Assistant - Phi-3 Mini + Unsloth",
    description="Ask your command-line questions. This model was fine-tuned with QLoRA using Unsloth."
).launch(share=True)